Epidermal growth factor receptor tyrosine kinase (EGFR-TK) mutation testing in adults with locally advanced or metastatic non-small-cell lung cancer: a systematic review and cost-effectiveness analysis

Targeted mutation detection tests

The different targeted tests look for different numbers and combinations of EGFR mutations and are able to detect different levels of mutation. For example, a sample may contain a high proportion of tumour cells, but only a low proportion of these may harbour mutations, and a low proportion of mutation, although detectable by some tests, may not be clinically significant. Thus tests may differ in their ability to accurately select patients who are likely to benefit from chemotherapy with TKIs. EGFR mutations are known to be restricted to four exons (18–21), with deletions in exon 19 and point mutations in exon 21 accounting for > 90%. 6,7,13 Observational studies have linked deletions in exon 19, point mutations at codons 858 and 861 of exon 21, and point mutations at codon 719 of exon 18 to tumours that are responsive to treatment with gefitinib. 13,15

The licensed indication for the TKIs gefitinib and erlotinib is treatment of locally advanced or metastatic NSCLC in patients who are previously untreated and whose tumours test positive for EGFR mutations. NICE Technology Appraisal 192 recommends gefitinib as an option for the first-line treatment of people with locally advanced or metastatic NSCLC if they test positive for an EGFR mutation. 1 The mutation test used in the trial that informed NICE Technology Appraisal 192 was version 1 of the Therascreen^® EGFR polymerase chain reaction (PCR) Kit (Qiagen, Venlo, the Netherlands); it should be noted that this version is no longer being marketed and has been superseded by version 2, the Therascreen^® EGFR RGQ PCR Kit (Qiagen, Venlo, the Netherlands). NICE Technology Appraisal 258 recommends erlotinib as an option for the first-line treatment of people with locally advanced or metastatic NSCLC if they test positive for an EGFR mutation. 2 Trials used in this assessment were conducted only in patients whose tumours were EGFR mutation positive, and used a direct sequencing approach to select patients with exon 19 deletions or exon 21 L858R point mutations for inclusion. 2,16

The Therascreen EGFR RGQ PCR Kit is a molecular diagnostic kit for detection of the 29 most common EGFR mutations against a background of wild-type genomic deoxyribonucleic acid (DNA). It uses real-time PCR on the Rotor-Gene Q 5plex HRM Instrument (a real-time PCR cycler). All versions of the Therascreen EGFR PCR Kit and the Therascreen^® EGFR Pyro Kit (Qiagen, Venlo, the Netherlands) will be included in the assessment. The mutations detected by the currently available Therascreen EGFR RGQ PCR Kit include 19 deletions in exon 19, T790M, L858R, L861Q, G719X (Therascreen detects the presence of these mutations but does not distinguish between them), S768I, and three insertions in exon 20; version 1 of the Therascreen EGFR PCR Kit, as used in the studies included in this assessment but no longer available, detected the same mutations. A version of the Therascreen EGFR PCR Kit that did not detect the resistance mutation T790M was previously marketed by Qiagen but this version is no longer available and was not used in any of the studies included in this review. Versions 1 and 2 of the Therascreen EGFR PCR Kit, referred to in this assessment, may therefore be considered equivalent. The Therascreen EGFR RGQ PCR kit includes all reagents needed to perform a PCR-based assay, where specific areas of DNA containing mutations are targeted by amplification refractory mutation system (ARMS) primers and Scorpions technology is used to detect amplifications of those specific areas of DNA. The test uses DNA isolated from formalin-fixed and paraffin-embedded (FFPE) tissue obtained from lung biopsy. The Therascreen EGFR RGQ PCR Kit uses a two-step procedure. The first step is performance of the control assay to assess the total DNA in a sample. The second step is to complete the mutation assay for the presence or absence of mutated DNA.

The cobas^® EGFR Mutation Testing Kit (Roche Molecular Systems, Inc., Branchburg, NJ, USA) is a CE-marked real-time PCR test for the detection of 41 EGFR mutations [G719X (G719S/G719A/G719C) in exon 18, 29 deletions and complex mutations in exon 19, T790M in exon 20, S768I in exon 20, five insertions in exon 20, L858R point mutation in exon 21]. The first step is to process the tumour tissue using the cobas^® DNA Sample Preparation Kit. The second step is PCR amplification and detection of EGFR mutations using complementary primer pairs and fluorescently labelled probes. The PCR is run using the cobas^® z 480 analyser (Roche Molecular Systems, Inc., Branchburg, NJ, USA), which automates amplification and detection. cobas^® 4800 software (Roche Molecular Systems, Inc., Branchburg, NJ, USA) provides automated test result reporting.

Pyrosequencing methods are usually set up to detect specific EGFR-TK mutations and are sometimes used to look for point mutations alongside fragment length analysis to detect deletions and insertions. The process involves first extracting DNA from the sample and amplifying it using PCR. The PCR product is then cleaned up before the pyrosequencing reaction. The reaction involves the sequential addition of nucleotides to the mixture. A series of enzymes incorporate nucleotides into the complementary DNA strand, generate light proportional to the number of nucleotides added and degrade unincorporated nucleotides. The DNA sequence is determined from the resulting pyrogram trace.

Fragment length analysis can be used to detect deletions in exon 19 and insertions in exon 20. DNA is first extracted from the sample and then amplified and labelled with fluorescent dye using PCR. Amplified DNA is mixed with size standards and is analysed using capillary electrophoresis. The fluorescence intensity is monitored as a function of time, and analysis software can determine the size of the fragments. The presence or absence of a deletion/insertion can then be reported.

Mutation screening tests

Direct sequencing is used to screen for all EGFR mutations (known and novel) in exons 18–21. This process is known as ‘comprehensive testing’ and has been considered as the routine method for detecting EGFR mutations; however, it requires larger tumour samples than other methods. Randomised controlled trials (RCTs) comparing the effectiveness of erlotinib with standard chemotherapy, in participants whose tumours were EGFR mutation positive, selected participants using direct sequencing to identify mutations in exon 19 or 21. A comparison of version 1 of the Therascreen EGFR PCR Kit with direct sequencing reported that Therascreen was ‘more sensitive’, i.e. some EGFR mutations were detected, which were not identified by direct sequencing. This was ascribed to low density of tumour cells in the sample. 17 Other mutation screening methods include single-strand confirmation polymorphism, high-resolution melt (HRM) analysis and next-generation sequencing.

For single-strand conformation polymorphism, DNA is first extracted from the sample and amplified using PCR. The PCR product is then prepared for analysis by heat denaturing and analysed using capillary electrophoresis under non-denaturing conditions. Sequence variations (single point mutations and other small changes) are detected through electrophoretic mobility differences.

High-resolution melt analysis detects all mutations, known and novel. The DNA is first extracted from the sample and amplified using PCR. The HRM reaction is then performed. This involves a precise warming of the DNA, during which the two strands of DNA ‘melt’ apart. Fluorescent dye that binds only to double-stranded DNA is used to monitor the process. A region of DNA with a mutation will ‘melt’ at a different temperature to the same region of DNA without a mutation. These changes are documented as melt curves and the presence or absence of a mutation can be reported.

Next-generation sequencing can also be used to identify all mutations. As with Sanger sequencing, there is much variation in the methodology used. The concept is similar to Sanger sequencing; however, the sample DNA is first fragmented into a library of small segments that can be sequenced in parallel reactions.

Diagnosis and staging of lung cancer

Guidance from NICE on the diagnosis and treatment of lung cancer was updated in 2011. 18 Patients referred for suspected lung cancer should initially undergo urgent chest radiography. If the chest radiograph is suggestive of lung cancer a contrast-enhanced computed tomography (CT) scan of the thorax, upper abdomen and lower neck is performed. Patients can then undergo a variety of diagnostic and staging investigations, which should be selected to provide the most information with the least risk to the patient. Most pathways in the diagnostic algorithm include biopsy for histological confirmation and tissue typing (e.g. to confirm if NSCLC is adenocarcinoma, squamous cell carcinoma, adenosquamous carcinoma or large cell carcinoma). The mediastinal lymph nodes are assessed for malignancy using positron emission tomography (PET)-CT, or endobronchial ultrasound (EBUS)-guided transbronchial needle aspiration (TBNA), or endoscopic ultrasound (EUS)-guided fine-needle aspiration (FNA), or non-ultrasound-guided TBNA. Patients with clinical and/or radiological features of advanced/metastatic disease may undergo further imaging (e.g. PET/CT or MRI) with possible biopsy of the most accessible site. 18

Where biopsy is undertaken, DNA extraction and mutation analysis may be carried out on the biopsy tissue, after pathological examination, to determine whether the tumour is EGFR mutation positive or negative. NICE clinical guidance recommends that adequate samples are taken without unacceptable risk to the patient to permit tumour subtyping and measurement of predictive markers. 18 For the 32,347 cases of lung cancer recorded in the 2010 NLCA data, the median [interquartile range (IQR)] percentage of patients receiving a histological–cytological diagnosis was 76.0% (70.5–83.6%) across NHS Trusts in England and Wales. NLCA data for 2010 reported a median of 20.0% (IQR 13.1–28.9%) of patients with NSCLC with unspecified histology for NHS trusts in England and Wales. 5 This assessment will assume that, in line with current clinical guidance, biopsy is undertaken in all patients for whom it is considered possible and clinically appropriate. However, the proportion of patients in whom the biopsy sample is inadequate is an important consideration for this assessment, as it represents a requirement for additional mutation testing, possible additional invasive procedures (in order to obtain an adequate sample) and associated additional costs.

Treatment of non-small cell lung cancer

Once NSCLC has been confirmed, NICE clinical guidance recommends that chemotherapy should be offered to people with stage III or IV (locally and regionally advanced or metastatic) NSCLC and a good performance status (PS) [World Health Organization (WHO) 0, 1 or Karnofsky score 80–100] with the aim of improving survival, disease control (DC) and quality of life. Treatment with curative intent is not possible for these patients. First-line chemotherapy should be a combination of a single third-generation drug (docetaxel, gemcitabine, paclitaxel or vinorelbine) and a platinum drug (carboplatin or cisplatin). People who are unable to tolerate a platinum combination may be offered single-agent chemotherapy with a third-generation drug. 18 Pemetrexed in combination with cisplatin is recommended as a first-line treatment for patients with locally advanced or metastatic NSCLC if the histology of the tumour has been confirmed as adenocarcinoma or large cell tumour. 19 The most recent data for England and Wales (NLCA 2011) suggest that the median proportion of patients with stage III or IV NSCLC receiving chemotherapy was 51.5% (IQR 48.2–64%); however, the case ascertainment rate for this measure was < 50%. 5

The NICE Technology Appraisal 192 recommends the EGFR-TKI gefitinib as an option for the first-line treatment of people with locally advanced or metastatic NSCLC who test positive for EGFR mutation. 1 NICE Technology Appraisal 258 recommends erlotinib as an option for the first-line treatment of people with locally advanced or metastatic NSCLC if they test positive for an EGFR mutation. 2 NICE guidance does not currently include any recommendations on the type of diagnostic tests used to identify EGFR mutations, and there is no consensus on which testing method should be preferred for clinical decision-making. 12

Measuring response to treatment

In 1979 the WHO and the International Union Against Cancer introduced criteria for the classification of the response of solid tumours to treatment. 20 These criteria were an early attempt to standardise reporting of response outcomes and were widely adopted. However, some problems with their use have subsequently developed: there has been variation in the methods used for incorporating into response assessments the change in size of measurable lesions, as defined by WHO; the minimum lesion size and number of lesions to be recorded have also varied; the definitions of progressive disease (PD) have sometimes been related to change in a single lesion and sometimes to change in overall tumour load (sum of the measurements of all lesions); and there has been confusion around how to use three-dimensional measures from new technologies, such as CT and magnetic resonance imaging (MRI), in the context of WHO criteria. 21 The Response Evaluation Criteria in Solid Tumours (RECIST) group is a collaborative initiative that was initiated to review the WHO criteria. The RECIST criteria use the same categories [complete response (CR), partial response (PR), stable disease (SD) and PD]. 21 RECIST guidance states that ‘CT and MRI are the best currently available and most reproducible methods for measuring target lesions selected for response assessment’ and that imaging-based evaluation is generally preferable to clinical examination. It is suggested that follow-up assessments every 6–8 weeks is a ‘reasonable norm’. 21 Taking into account the longest diameter for only all target lesions, the RECIST criteria, as they are applicable to this assessment, can be summarised as follows:21

• CR Disappearance of all target lesions and no new lesions.

• PR At least 30% decrease in the sum of the longest diameter of target lesions, taking the sum of the baseline diameters as the reference, and no new lesions.

• PD At least a 20% increase in the sum of the longest diameter of target lesions, taking the smallest sum of the longest diameters recorded since treatment started as the reference, or appearance of one or more new lesions.

• SD Neither sufficient shrinkage to be classified as PR or sufficient increase to be classified as PD, taking the smallest sum of the longest diameters recorded since treatment started as the reference, and no new lesions.

Best overall response is defined as the best response recorded from the start of treatment to disease progression. 21

This assessment compares the performance and cost-effectiveness of EGFR mutation testing options currently available in the NHS in England and Wales to identify previously untreated adults with locally advanced or metastatic NSCLC who may benefit from first-line treatment with EGFR inhibitors (gefitinib or erlotinib).

Search strategy

Search strategies were based on target condition and intervention, as recommended in the CRD guidance for undertaking reviews in health care and the Cochrane Handbook for diagnostic test accuracy reviews. 22,25

Candidate search terms were identified from target references, browsing database thesauri (e.g. MEDLINE MeSH and EMBASE Emtree), existing reviews identified during the rapid appraisal process and initial scoping searches. These scoping searches were used to generate test sets of target references, which informed text mining analysis of high-frequency subject indexing terms using EndNote X4 reference management software (Thomson Reuters, CA, USA). Strategy development involved an iterative approach testing candidate text and indexing terms across a sample of bibliographic databases and aimed to reach a satisfactory balance of sensitivity and specificity.

The following databases were searched for relevant studies from 2000 to August 2012:

• MEDLINE (OvidSP) (2000 to July 2012 week 1)

• MEDLINE In-Process & Other Non-Indexed Citations and Daily Update (OvidSP) (up to 17 July 2012)

• EMBASE (OvidSP) (2000 to 2012 week 28)

• Cochrane Database of Systematic Reviews (CDSR) (internet) (2000 to 2012/Issue 7)

• Cochrane Central Register of Controlled Trials (CENTRAL) (internet) (2000 to 2012/Issue 7)

• Database of Abstracts of Reviews of Effects (DARE) (via The Cochrane Library) (2000 to 2012/Issue 3)

• Health Technology Assessment database (HTA) (via The Cochrane Library) (2000 to 2012/Issue 3)

• Science Citation Index (SCI) (Web of Science) (2000 to 18 July 2012)

• Latin American and Caribbean Health Sciences Literature (LILACS) (internet) (2000 to 6 July 2012) http://regional.bvsalud.org/php/index.php?lang=en

• BIOSIS Previews (Web of Knowledge) (2000 to 24 August 2012)

• NIHR Health Technology Assessment programme (internet) (2000 to 18 July 2012)

• PROSPERO (International Prospective Register of Systematic Reviews) (internet) (up to 19 July 2012) www.crd.york.ac.uk/prospero/

Completed and ongoing trials were identified by searches of the following resources:

• National Institutes of Health (NIH) ClinicalTrials.gov (2000 to 19 July 2012) (internet) www.clinicaltrials.gov/

• Current Controlled Trials (2000 to 30 August 2012) (internet) www.controlled-trials.com/

• WHO International Clinical Trials Registry Platform (ICTRP) (2000 to 30 August 2012) (internet) www.who.int/ictrp/en/

Searches were undertaken to identify studies of EGFR-TK mutation testing in NSCLC. The main EMBASE strategy for each set of searches was independently peer reviewed by a second information specialist, using the Peer Review of Electronic Search Strategies Evidence-Based Checklist (PRESS-EBC). 26 Search strategies were developed specifically for each database and the keywords associated with NSCLC were adapted according to the configuration of each database. Searches took into account generic and other product names for the intervention. No restrictions on language or publication status were applied. Limits were applied to remove animal studies. Full search strategies are reported in Appendix 1.

Electronic searches were undertaken for the following conference abstracts:

• American Society of Clinical Oncology (ASCO) conference proceedings (2007 to 2012) (internet) www.asco.org/ASCOv2/Meetings/Abstracts

• European Society of Medical Oncology (ESMO) conference proceedings (2007 to 2012) (internet) www.esmo.org/no_cache/education/abstracts-and-virtual-meetings.html

  • 2008 33rd ESMO Congress, Stockholm – http://annonc.oxfordjournals.org/content/vol19/suppl_8/

  • 2009 European Cancer Congress (ECCO) 15 and 34th ESMO Multidisciplinary Congress – www.ejcancer.info

  • 2010 35th ESMO Congress, Milan – http://annonc.oxfordjournals.org/content/21/suppl_8

  • 2011 ECCO 16 and 36th ESMO Multidisciplinary Congress, Brussels – www.ejcancer.info/issues

  • 2012 37th ESMO Congress, Vienna – http://annonc.oxfordjournals.org/content/23/suppl_9

• World Conference on Lung Cancer (International Association for the Study of Lung Cancer) (2007 to 2012) (internet) http://iaslc.org/

  • 14th World Conference on Lung Cancer – http://journals.lww.com/jto/toc/2011/06001

  • 13th World Conference on Lung Cancer – http://journals.lww.com/jto/Citation/2009/09001/Abstracts.1.aspx

  • 12th World Conference on Lung Cancer – http://journals.lww.com/jto/toc/2007/08001

Identified references were downloaded in EndNote X4 software for further assessment and handling.

References in retrieved articles were checked for additional studies. The final list of included papers was also checked on PubMed for retractions, errata and related citations. 27–29

Inclusion and exclusion criteria

Separate inclusion criteria were developed for each of the three clinical effectiveness questions; these are summarised in Table 2.

Inclusion screening and data extraction

Two reviewers (MW and PW) independently screened the titles and abstracts of all reports identified by searches and any discrepancies were discussed and resolved by consensus. Full copies of all studies that were deemed potentially relevant were obtained and the same two reviewers independently assessed these for inclusion; any disagreements were resolved by consensus. Details of studies excluded at the full paper screening stage are presented in Appendix 5.

Studies provided by the manufacturers of Therascreen (Qiagen) and cobas EGFR Mutation Testing Kit were first checked against the project reference database in EndNote X4; any studies not already identified by our searches were screened for inclusion following the process described above.

Data were extracted on the following: study design/details, participant details (e.g. tumour stage, histological diagnosis, PS, smoking status, ethnicity), EGFR mutation test(s) and mutations targeted, clinical outcomes, test performance outcome measures (against treatment response as reference standard), details of specific mutations identified by outcome measure (where reported) and test failure rates. Data were extracted by one reviewer, using a piloted, standard data extraction form, and checked by a second reviewer (MW and PW); any disagreements were resolved by consensus. Full data extraction tables are provided in Appendix 2.

Quality assessment

The risk of bias in included RCTs was assessed using the Cochrane Collaboration’s tool for assessing risk of bias in randomised trials. 30 Studies used to derive accuracy data, for the ability of EGFR mutation tests to predict treatment response, were assessed using QUADAS-2. 31 The version of QUADAS-2 used in this report did not include assessment of applicability because both the index test and study population were tightly defined by our inclusion criteria, and clinical outcome measures were treated as the reference standard. Studies that provided both accuracy data and data on the effectiveness of treatment with TKIs following testing were assessed using both tools. Risk of bias assessments were undertaken by one reviewer and checked by a second reviewer (MW and PW), and any disagreements were resolved by consensus.

The results of the risk of bias assessments were summarised and presented in tables and graphs in the results of the systematic review, and are presented in full, by study, in Appendix 3.

Survey of laboratories providing epidermal growth factor receptor mutation testing

We conducted a web-based survey (October 2012) to gather data on the technical performance characteristics of EGFR mutation tests. We sent an e-mail invitation to NHS laboratories participating in the UK NEQAS pilot scheme for EGFR mutation testing, which had responded to a request to provide information to NICE at the start of this assessment. We used the SurveyMonkey (SurveyMonkey, Palo Alto, CA, USA) online software to run the survey. We structured the survey into sections on:

• laboratory details

• EGFR testing methods

• logistics

• technical methods

• costs.

Where possible we used multiple choice options with tick boxes to make the survey quick and easy to complete. A copy of the survey is provided in Appendix 4.

Methods of analysis/synthesis

The results of studies included in this review were summarised by research question (see Chapter 1), i.e. studies providing technical information on EGFR mutation testing in NHS laboratories in England and Wales (see What are the technical performance characteristics of the different epidermal growth factor receptor mutation tests?, below), studies providing information on the accuracy of EGFR mutation tests for predicting response to TKI treatment (see What is the accuracy of epidermal growth factor receptor mutation testing, using any test, for predicting response to treatment with tyrosine kinase inhibitors?, below), and studies reporting information on how clinical outcomes may vary according to which test is used to select patients for TKI treatment (see How do outcomes from treatment with epidermal growth factor receptor inhibitors vary according to which test is used to select patients for treatment?, below). We planned to use a bivariate/hierarchical summary receiver operating characteristic (SROC) random-effects model to generate summary estimates and an SROC curve for test accuracy data,32–34 and a DerSimonian and Laird random-effects model to generate summary estimates of treatment effects. However, because the review identified a relatively small number of studies with between-study variation in participant characteristics, methods used to test for EGFR mutations and mutations targeted, we did not consider meta-analyses to be appropriate and have provided a structured narrative synthesis.

For all studies that provided data on accuracy for the prediction of response to treatment with TKIs, the absolute numbers of true-positive (TP), false-negative (FN), false-positive (FP) and true-negative (TN) test results, as well as sensitivity and specificity values, with 95% confidence intervals (CIs) are presented in results tables, for each reference standard response [e.g. objective response (OR), DC] reported. Where reported, data on the numbers of failed EGFR mutation tests and reasons for failure were also included in the results tables. The results of individual studies were plotted in the receiver operating characteristic (ROC) plane to illustrate the trade-off between sensitivity and specificity, and for ease of comparison between test methods; separate plots were provided for each reference standard response. For RCTs providing information on how clinical outcomes may vary according to which test is used to select patients for TKI treatment, hazard ratios (HRs), with 95% CIs, are provided for survival outcome measures [progression-free survival (PFS), overall survival (OS)] and relative risks (RRs), with 95% CIs, are reported for tumour response outcomes (OR and DC). The results of individual studies were illustrated in forest plots. Between-study clinical heterogeneity was assessed qualitatively. There were insufficient studies to assess heterogeneity statistically, such as the chi-squared test and I²-statistic. 35

This report contains reference to confidential information provided as part of the NICE appraisal process. This information has been removed from the report and the results, discussions and conclusions of the report do not include the confidential information. These sections are clearly marked in the report.

What are the technical performance characteristics of the different epidermal growth factor receptor mutation tests?

Epidermal growth factor receptor mutation test methods (see Figure 2 and Table 3)

The Therascreen EGFR PCR Kit was the most commonly used EGFR mutation test (Figure 2 and Table 3), with six laboratories using this test. A combination of fragment length analysis and pyrosequencing was used in three laboratories, and Sanger sequencing in two; other tests were each used in single laboratories. Most laboratories that used the Therascreen EGFR PCR Kit cited ease of use (n = 5) and/or proportion of tumour cells required (n = 5) as their reasons for choosing this method; three laboratories also cited mutation coverage and two cited cost. All laboratories that used fragment length analysis cited cost as a reason for their choice of this method; one also cited proportion of tumour cells required, mutation coverage and flexibility of method; another also cited ease of use; and the third claimed that accuracy was high. The two laboratories that used Sanger sequencing both cited mutation coverage as a reason for choice, and one also cited cost and ease of use; both used a second testing option for samples with insufficient tumour cells or for verification of mutations. Although only three laboratories completed the questionnaire separately for more than one test, 11 laboratories answered the question on reason for using more than one EGFR testing method. Reasons for this included insufficient tumour cells (n = 3), verification of mutations (n = 5), validating a new method (n = 1), ‘back up technique in case kits are made unavailable’, ‘methods are complementary and detect different mutations’ and ‘coverage of mutations and simplicity, cost’. Of the laboratories that completed the questionnaire more than once, one used the Therascreen EGFR PCR test but is also developing and validating a new next-generation sequencing method, which it thinks may be cheaper and target more mutations. The second used Sanger sequencing and Roche cobas, and cites verification of mutations and insufficient tumour cell as its reason for using multiple tests. The third used Sanger sequencing, TaqMan/real-time PCR/EntroGen and fragment length analysis, and also cites verification of mutations and insufficient tumour cell as its reason for using multiple tests. Two further laboratories indicated that they use a combination of pyrosequencing and fragment length analysis as complementary tests that detect different mutations; laboratories using fragment length analysis always do so as part of a strategy that involves more than one test.

FIGURE 2.

Epidermal growth factor receptor mutation test used in NHS laboratories in England and Wales participating in the UK NEQAS pilot scheme for EGFR mutation testing.

TABLE 3 - Details of EGFR mutation tests used in NHS laboratories in England and Wales participating in the UK NEQAS pilot scheme for EGFR mutation testing

Scoping reported this strategy as ‘Sanger sequencing (exons 18–21) followed by fragment length analysis (exon 19 deletions)/PCR (to detect L858R) of negative samples’.

Scoping reported this strategy as ‘Sanger sequencing (exons 18–21) of samples with > 30% of tumour cells and cobas EGFR Mutation Testing Kit for samples with < 30% of tumour cells’.

EGFR mutation test used	Reasons for choosing test	Mutations targeted
Qiagen Therascreen EGFR PCR Kit	Ease of use	28/29 mutations in Therascreen Kit
	Proportion of tumour cells required; ease of use; ‘We had a trainee project comparing several different methods. Qiagen picked up more mutations than Sanger (more sensitive), and was very easy to use’
	Proportion of tumour cells required; mutation coverage
	Proportion of tumour cells required; mutation coverage; ease of use
	Cost; proportion of tumour cells required; ease of use
	Cost; proportion of tumour cells required; ease of use; mutation coverage
Fragment length analysis and pyrosequencing	Cost; proportion of tumour cells required; mutation coverage; not a black box method so easily modified if required	All exon 18–21 mutations
	Cost; ‘Sensitivity is greater than Sanger and specificity is good. Equipment for pyrosequencing is in house and is a platform used reliable for many molecular pathology investigations’	Exon 19 deletions Insertions in exon 20 Exon 21 – L858R mutation Targeted exon 18–21 mutations; 12 mutations in total but other mutations may be detected if they are within the same region
Sanger sequencing and/or fragment length analysis/TaqMan/real-time PCR [used for verification of mutations, or where sample contains insufficient tumour cells for Sanger sequencing (< 30%)]a	Sanger sequencing: cost; ease of use; mutation coverage; fits in with laboratory high throughput sequencing pipeline so samples will be processed quickly Fragment length analysis: cost; ease of use TaqMan/real-time PCR: cost; ease of use	Sanger sequencing: all exon 18–21 mutations Fragment length analysis: exon 19 deletions TaqMan/real-time PCR: Exon 21 – L858R mutation
Sanger sequencing and/or Roche cobas (used for verification of mutations, or where sample contains insufficient tumour cells for Sanger sequencing (< 30%)b	Sanger sequencing: mutation coverage Roche cobas: proportion of tumour cells required	Sanger sequencing: all exon 18–21 mutations Roche cobas: 41 mutations in cobas kit
Next-generation sequencing, stated ‘in process of developing and validation’	Cost; proportion of tumour cells required; mutation coverage; capacity to test multiple genes/samples/patients	Potentially all
HRM analysis	Mutation coverage; ease of use	All exon 18–21 mutations
Single-strand conformation analysis	Cost; ease of use; the vast majority of cases (90%) are EGFR wild type, therefore an easy method that reliably detects wild-type cases with ease of analysis seems cost-effective	All exon 18–21 mutations
Pyrosequencing	Cost; mutation coverage	Exon 19 deletions Insertions in exon 20 Exon 21 – L858R mutation

Epidermal growth factor receptor mutation test logistics (see Figure 3 and Table 4)

The number of samples screened for EGFR mutations in a typical week varied by laboratory from less than five (six laboratories) to > 20 (three laboratories). The batch size ranged from less than 3 to 10 samples (Figure 3 and Table 4). Only laboratories with five or fewer samples screened per week ran batches of three or fewer. Only one laboratory had a batch size of 10 and this laboratory screened > 20 samples per week; all other laboratories had batch sizes of between five and eight. For the Therascreen EGFR PCR test, all batch sizes were five or seven. The frequency at which the laboratories ran the test ranged from daily to every other week, although the laboratory that ran the test every other week stated that they would match demand. Three laboratories stated that they waited for a minimum batch size (five to seven samples), although one of these stated that they would match demand.

FIGURE 3.

Summary of logistic information. (a) In a typical week, how many samples do you screen for EGFR mutations?; (b) what is you average batch size (number of samples)?; (c) how often do you run the EGFR mutation test?; (d) on average, how long does it take from receiving a sample at the laboratory to sending a result back to the clinician?

TABLE 4 - Laboratory throughput by EGFR mutation test

Scoping reported this strategy as ‘Sanger sequencing (exons 18–21) followed by fragment length analysis (exon 19 deletions)/PCR (to detect L858R) of negative samples’.

Scoping reported this strategy as ‘Sanger sequencing (exons 18–21) of samples with > 30% of tumour cells and cobas EGFR Mutation Testing Kit for samples with < 30% of tumour cells’.

EGFR mutation test	Samples per week	Batch size	Frequency of test	Wait for batch size?	Time from receiving test to returning result to clinician
Qiagen Therascreen EGFR PCR Kit	> 20	7	Daily	No	3–5 days
	> 20	7	Three to four times per week	Yes	3–5 days
	11–15	7	Weekly	No	8–10 days
	6–10	7	Weekly plus further run when required	No	3–5 days
	≤ 5	5	Weekly	No	24–48 hours
	≤ 5	5	Every other week	Yes, but will match demand	6–7 days
Fragment length analysis and pyrosequencing	6–10	5	Two to three times per week	No	6–7 days
	6–10	5	Two to three times per week	No	6–7 days
Sanger sequencing and/or fragment length analysis/TaqMan/real-time PCR [used for verification of mutations, or where sample contains insufficient tumour cells for Sanger sequencing (< 30%)]a	≤ 5	1–3	Daily	No	6–7 days
Sanger sequencing and/or Roche cobas [used for verification of mutations, or where sample contains insufficient tumour cells for Sanger sequencing (< 30%)]b	16–20	6	Two to three times per week	Yes	6–7 days
	16–20	6	Two to three times per week	No	3–5 days
HRM analysis	11–15	7	Two to three times per week	No	3–5 days
Next-generation sequencing	≤ 5	5	Weekly	No	3–5 days
Pyrosequencing	16–20	6–8	Two to three times per week	No	6–7 days
Single-strand conformation analysis	> 20	10	Two to three times per week	No	3–5 days

The majority of laboratories had a turnaround time from receiving the sample to reporting the result to the clinician of 3–5 or 6–7 days with only one laboratory having a time of 24–48 hours and one having a time of 8–10 days. The laboratory with the shortest turnaround time was one that used the Therascreen EGFR PCR test, and which tested fewer than five samples per week. The laboratory with the longest turnaround time was also a laboratory that used Therascreen EGFR PCR, but had a higher throughput of 11–15 samples per week. Neither of these two laboratories waited for a minimum batch size before running the test.

What is the accuracy of epidermal growth factor receptor mutation testing, using any test, for predicting response to treatment with tyrosine kinase inhibitors?

Six studies – two RCTs41,48,49 and four cohort studies43–46 – provided data on the accuracy of EGFR mutation testing for predicting response to treatment in patients with stage IIIB or IV NSCLC when they are treated with TKIs. Three studies were conducted in patients treated with gefitinib,41,46,48 and three were conducted in patients treated with erlotinib. 43–45 These studies are particularly useful as they provide full information on the extent to which EGFR mutation tests are able to discriminate between patients who will benefit from TKI treatment and those who will not. We defined TPs as those patients with an EGFR mutation who have a positive response to TKI treatment. Where presence or absence of OR was the reference standard, a positive response was defined as best observed response = CR or PR. Where presence or absence of DC was the reference standard, a positive response was defined as best observed response = CR, PR or SD. FPs were defined as those patients with an EGFR mutation who did not have a positive response to TKI treatment (SD or PD) for the reference standard OR, or disease progression for the reference standard DC; FNs were defined as those without an EGFR mutation who had a positive response to TKI treatment; and TNs were defined as those without EGFR mutation who did not have a positive response to TKI treatment. Full definitions of CR, PR, SD and PD are provided in Chapter 2 (see Measuring response to treatment).

Epidermal growth factor receptor mutation test accuracy

The Therascreen EGFR PCR Kit appeared to have the best overall performance for discriminating between patients who are likely to benefit from TKI treatment and those who are not. The sensitivity and specificity estimates for OR were 99% (95% CI 94% to 100%) and 69% (95% CI 60% to 77%), respectively. 49 As might be expected, the specificity was higher when a lower threshold (DC) was used to define response to treatment and, conversely, sensitivity was higher when a higher threshold (OR) was used to define response to treatment (see Table 7). It should be noted that although initial examination may appear to indicate a better performance for the Therascreen EGFR PCR Kit, only one data set was available for this test and no studies reported direct comparisons between tests conducted in the same population. Figure 4 illustrates the results for all studies reporting accuracy data with the Therascreen EGFR PCR Kit study (IPASS) indicated in green. Four of the five studies that used direct sequencing methods to identify EGFR mutations reported high estimates of specificity (> 80%) for OR, and specificities ranged from 60% to 80%. 41,43–45 Three of these studies also assessed DC; specificities remained high (> 90%), whereas sensitivity estimates were very low (≤ 35%). 43–45 The remaining direct sequencing study reported low sensitivity (66%) and specificity (50%) for DC, and low specificity (61%) with high sensitivity (84%) for OR. All direct sequencing studies had small sample sizes, reflected in the wide CIs around sensitivity and specificity estimates. There were no clear common participant characteristics across studies that reported similar sensitivity or specificity estimates for DC or OR. All test accuracy results are summarised in Table 7. It is possible that the lower specificity values observed in two studies46,49 may be explained, at least in part, by the classification of resistance mutations as a positive result for EGFR mutation testing. The four direct sequencing studies that reported high specificity estimates for DC and/or OR41,43–45 either stated that patients whose tumours showed resistance or non-sensitising mutations were classified as EGFR mutation negative, or did not identify any patients whose tumours showed these types of mutation (Table 8). Although the number of resistance mutations identified was generally small, their potential effect on specificity estimates was magnified by the very small sample size in most studies. Data relating best response to individual mutations appeared to indicate that there may be a less favourable response to TKIs in patients with T790M or other exon 20 mutations (see Table 8); [Commercial-in-confidence (CiC) information has been removed]. The most commonly observed mutations were exon 19 deletions and the exon 21 point mutation L858R, and most patients with these mutations achieved a minimum response of SD. Two studies did not report sufficient information to derive best response data by mutation type and both of these studies identified only exon 19 deletions and exon 21 point mutation L858R. 41,45 One study reported a CR in three patients whose tumours were positive for EGFR mutations and no CRs in patients whose tumours were negative for EGFR mutations;49 none of the other studies reported any CRs.

TABLE 7 - Accuracy of EGFR mutation testing for the prediction of response to treatment with TKIs

NR, not reported.

Calculated values.

Study	EGFR test and mutations targeted	Non-evaluable samples	DC						OR
			TP	FP	FN	TN	Sensitivity (95% CI)	Specificity (95% CI)	TP	FP	FN	TN	Sensitivity (95% CI)	Specificity (95% CI)
Fukuoka (IPASS) (2011)48,49	Therascreen EGFR PCR Kit (version 1)	386/609 unknown mutation status (number with insufficient sample quality NR)	121	10	36	47	77 (70 to 83)a	83 (70 to 91)a	94	37	1	82	99 (94 to 100)a	69 (60 to 77)
Giaccone (2006)43	Direct sequencing (nested PCR) of all exon 18–21 mutations	24/53 no sample available, no samples of insufficient quality reported	5	0	12	12	29 (10 to 56)a	100 (74 to 100)a	4	1	1	23	80 (28 to 100)a	96 (79 to 100)a
Han (First-SIGNAL) (2012)41	Direct sequencing (PCR) of all exon 19–21 mutations	53/159 unknown mutation status (number with insufficient sample quality NR)	NR	NR	NR	NR	NR	NR	22	4	7	20	76 (57 to 90)a	83 (63 to 95)a
Jackman (2007)44	Direct sequencing (34 samples), or WAVE-HS (Transgenomic, Omaha, NE, USA) (nine samples) for inadequate samples (< 50% of tumour cells) of all exon 18–24 mutations	4/80 no sample available, 26/80 samples of insufficient quality	9	0	17	11	35 (15 to 56)a	100 (72 to 100)a	3	6	2	26	60 (15 to 95)a	81 (64 to 93)a
Pallis (2012)45	Direct sequencing (PCR) of all exon 18–21 mutations	13/49 no sample available, no samples of insufficient quality reported	8	1	16	11	33 (16 to 55)a	92 (62 to 100)a	6	3	4	23	60 (26 to 88)a	89 (70 to 98)
Yang (2008)46	Direct sequencing (PCR) of all exon 18–21 mutations	16/106 EGFR mutation status not successfully determined, no details reported	47	5	24	5	66 (54 to 71)a	50 (19 to 81)a	38	14	7	22	84 (71 to 94)a	61 (44 to 77)a

FIGURE 4.

Receiver operating characteristic plane plots comparing EGFR mutation testing methods for the prediction of response to treatment with TKIs. Plots show sensitivity and specificity estimates with 95% CIs.

TABLE 8 - Best response to treatment by mutation type in EGFR mutation-positive patients treated with TKIs

Study	EGFR mutation	n	Best response
			CR	PR	SD	PD
CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed
	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed
	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed
	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed
	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed
	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed
	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed
	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed	CiC information has been removed
Giaccone (2006)43	Exon 19 deletion only	5	0	4	1	0
Jackman (2007)44	Exon 19 deletion only	3	0	2	1	0
	Exon 21 L858R only	5	0	1	4	0
	Exon 19 deletion and exon 21 L861Q	1	0	0	1	0
Yang (2008)46	Exon 19 deletion only	20	0	19	0	1
	Exon 21 L858R only	22	0	17	5	0
	Exon 21 L861R	1	0	0	1	0
	Exon 21 L858R and H850D	2	0	0	1	1
	Exon 21 L861Q and R831H	1	0	0	1	0
	Exon 20 SVD 786–770 insertions	3	0	1	0	2
	Exon 21 L858R and exon 20 S768I	1	0	1	0	0
	Exon 21 L858R and exon 20 T790M	1	0	0	0	1
	Exon 21 L861Q and exon 20 R776H	1	0	0	1	0

The IPASS trial, which used version 1 of the Therascreen EGFR PCR Kit, reported the minimum quantity of DNA required to detect 1% for each mutation targeted (1.5 ng for all mutations, except insertions that required 3.0 ng). 49 No direct sequencing study reported information on the limit of detection of the EGFR mutation test method used. Two studies specified a minimum proportion of tumour cells as a sample quality prerequisite for testing: these were 50% of tumour cells44 and 80% of tumour cells,45 respectively. Details of non-evaluable samples were generally poorly reported; any information reported is presented in Table 7.

QUADAS-2 assessments

All studies in this section were rated as ‘low’ risk of bias for the ‘index test’ and ‘reference standard’ domains of the quality assessment tool. 41,43–46,49,50 The two RCTs – IPASS48,49 and First-SIGNAL41 – were rated at ‘low’ risk of bias for participant selection; none of the other studies reported details of participant selection and consequently all were rated as ‘unclear’ risk of bias for this domain. Three studies had a ‘high’ risk of bias rating for any domain. 41,44,46 All of these were for the ‘flow and timing’ domain. For two cohorts the ‘high’ risk of bias rating arose because patients who were not evaluable for response were excluded from the analysis and these patients were judged to represent a significant proportion of the study population. 44,46 One RCT was rated as at ‘high’ risk of bias for the ‘flow and timing’ domain because only a small proportion of trial participants were assessed for tumour EGFR mutation status, no reasons were reported for why participants were not assessed, and no information was available to assess possible differences between those with known mutation status and those without. The results of QUADAS-2 assessments are summarised in Table 9 and Figure 5, and full QUADAS-2 assessments for each study are provided in Appendix 3.

TABLE 9 - QUADAS-2 results for studies assessing the accuracy of EGFR mutation testing methods for the prediction of response to treatment with TKIs

☺, low risk of bias; ?, unclear risk of bias; ☹, high risk of bias.

Study	Risk of bias
	Patient selection	Index test	Reference standard	Flow and timing
Fukuoka (IPASS) (2011)48,49	☺	☺	☺	☺
Giaccone (2006)43	?	☺	☺	☺
Han (First-SIGNAL) (2012)41	☺	☺	☺	?
Jackman (2007)44	?	☺	☺	☹
Pallis (2012)45	?	☺	☺	☺
Yang (2008)46	?	☺	☺	☹

FIGURE 5.

Summary of QUADAS-2 results.

How do outcomes from treatment with epidermal growth factor receptor inhibitors vary according to which test is used to select patients for treatment?

Five RCTs provided data on the clinical effectiveness of TKIs compared with standard chemotherapy in patients with stage IIIB or IV NSCLC, whose tumours tested positive for EGFR mutations,16,40,41,47,49 and one additional study36 reported data for a subgroup of patients from the EURTAC trial,40 whose samples had been re-analysed using a different EGFR mutation testing method (cobas EGFR Mutation Test). The trials compared the TKIs gefitinib or erlotinib with various single agent or combination standard chemotherapy regimens (Table 10). Three of the trials included only patients with EGFR mutation-positive tumours,16,40,47 and the remaining two trials (IPASS and First-SIGNAL) included chemotherapy-naive patients with stage IIIB or IV NSCLC, and reported a subgroup analysis for patients who had received EGFR mutation testing. 41,48

Clinical outcomes in patients with epidermal growth factor receptor mutation-positive tumours who were treated with tyrosine kinase inhibitors compared with those treated with standard chemotherapy

All studies in this section reported improvements in OR and improvements or trends towards improvement in PFS for patients with EGFR mutation-positive tumours who were treated with TKIs compared with those treated with standard chemotherapy. There were no clear differences in treatment effect, regardless of which EGFR mutation test (selective for activating mutations exon 19 deletions and exon 21 L858R, or targeting a wider range of mutations) was used to select patients (Figures 6 and 7). Based on subgroup analyses conducted within the trials, three trials reported no significant difference in the HR for PFS between patients with exon 19 deletions and those with the exon 21 mutation L858R. 40,47,48 However, the IPASS study also noted that, although the OR rate was higher in patients with exon 19 deletions who were treated with gefitinib (84.8%) than in those who were treated with standard chemotherapy (43.2%), there was no significant difference between the two treatment groups for patients with the exon 21 mutation L858R (OR rates were 60.9% and 53.2% for the gefitinib and standard chemotherapy groups, respectively). 48 One trial also reported that HRs for PFS did not differ significantly between patients with and without previous surgery, radiotherapy or adjuvant/neoadjuvant chemotherapy, by age, gender or PS; subgroup analyses by smoking status indicated that the treatment effect in favour of gefitinib was significant only in patients who had never smoked [HR 0.24 (95% CI 0.15 to 0.39)]. 40 One further trial noted that HRs for PFS appeared similar across all clinical subgroups (age, gender, PS, disease stage, histology and smoking status). 48 However, the authors noted that the trial was not powered to detect differences between subgroups. Where reported, the median PFS for participants with EGFR mutation-positive tumours in the TKI group was 9.7 months (95% CI 8.4 to 12.3 months),40 10.8 months47 and 13.1 months (95% CI 10.6 to 16.5 months). 16 The corresponding PFS values in the standard chemotherapy groups were 5.2 (95% CI 4.3 to 5.8) months,40 5.4 months47 and 4.6 months (95% CI 4.2 to 5.4 months). 16 The OR rates for participants with EGFR mutation-positive tumours in the TKI groups were 71% (94/132),49 58% (50/86),40 74% (84/114)47 and 83% (68/82). 16 The corresponding OR rates in the standard chemotherapy groups were 47% (61/129),49 15% (13/87),40 31% (35/114)47 and 36% (26/72). 16 Where DC was used as the outcome measure, the observed benefits of TKI treatment were generally more marginal, but there were no clear differences between studies using different EGFR mutation testing methods (Figure 8). Three studies reported OS40,47,48 but none found a significant difference between patients treated with TKIs and those treated with standard chemotherapy (see Table 10). Four studies reported data on the number of patients with CR as the best observed response; the numbers of CR were small in all cases (two,40 two,16 three48 and five47 patients in the TKI groups, and one49 patient in one standard chemotherapy group).

FIGURE 6.

Progression-free survival in patients with EGFR mutation-positive tumours who were treated with TKIs compared with those treated with standard chemotherapy.

FIGURE 7.

Objective response in patients with EGFR mutation-positive tumours who were treated with TKIs compared with those treated with standard chemotherapy.

FIGURE 8.

Disease control in patients with EGFR mutation-positive tumours who were treated with TKIs compared with those treated with standard chemotherapy.

Risk of bias

All of the studies in this section were rated as ‘low’ or ‘unclear’ risk of bias for randomisation, allocation concealment and selective outcome reporting. All studies were rated as ‘high’ risk of bias for blinding of study participants and personnel; blinding of study participants and personnel was not possible in these trials because of the different routes of administration used for the treatment and comparator arms [oral TKI vs. intravenous (i.v.) standard chemotherapy]. However, only one study was rated as ‘high’ risk of bias for blinding of outcome assessors;16 three studies reported independent outcome assessment40,41,47 and the remaining study did not report details of outcome assessor blinding. 48,49 With the exception of the OPTIMAL16 and First-SIGNAL41 trials, all studies were rated as ‘low’ risk of bias for incomplete reporting of outcome data; all other studies either reported ITT analyses40,48 or very small numbers of withdrawals (< 2% of the total study population). 47 For risk of bias assessment, EURTAC trial40 and the re-analysis of the EURTAC trial were treated as one study. The results of risk of bias assessments are summarised in Table 11 and Figure 9, and full risk of bias assessments for each study are provided in Appendix 3.

TABLE 11 - Risk of bias assessments for RCTs providing data on how the effectiveness of TKIs varies according to which EGFR mutation test is used to select patients for treatment

?, unclear risk of bias; ☹, high risk of bias; ☺, low risk of bias.

Study	Risk of bias
	Randomisation	Allocation concealment	Participant and personnel blinding	Outcome assessor blinding	Incomplete outcome data	Selective outcome reporting
Fukuoka (IPASS) (2011)48,49	?	?	☹	?	☺	☺
Han (First-SIGNAL) (2012)41	?	?	☹	☺	☹	☺
Maemondo (NEJSG) (2010)47	?	?	☹	☺	☺	☺
Rosell (EURTAC) (2012)40/Benlloch (2012)36	☺	?	☹	☺	☺	☺
Zhou (OPTIMAL) (2011)16	☺	☺	☹	☹	☹	☺

FIGURE 9.

Summary of risk of bias assessments.

Search strategy

Searches were undertaken to identify cost-effectiveness studies of EGFR-TK testing in NSCLC. As with the clinical effectiveness searching, the main EMBASE strategy for each set of searches was independently peer reviewed by a second information specialist, using the PRESS-EBC checklist. 26 Search strategies were developed specifically for each database, and searches took into account generic and other product names for the intervention. All search strategies are reported in Appendix 1.

The following databases were searched for relevant studies from 2000 to present:

• MEDLINE (OvidSP) (2000 to September 2012 week 4)

• MEDLINE In-Process Citations and Daily Update (OvidSP) (2000 to 29 August 2012)

• EMBASE (OvidSP) (2000–2012 week 34)

• NHS Economic Evaluation Database (NHS EED) (via The Cochrane Library) (2000–2012/Issue 3)

• Health Economic Evaluation Database (HEED) (Wiley) (2000 to 30 August 2012)

  • http://onlinelibrary.wiley.com/book/10.1002/9780470510933

• Science Citation Index (SCI) (Web of Science) (2000 to 29 August 2012)

Additional searches were undertaken to update the Resource Utilisation searches in the manufacturer’s submission for STA 192. 51 For this work, the following resources were searched:

• MEDLINE (OvidSP) (2000 to September 2012 week 4)

• Medline In-Process & Other Non-Indexed Citations and Daily Update (OvidSP) (2000 to 29 August 2012)

• EMBASE (OvidSP) (2000–2012 week 40)

• NHS Economic Evaluation Database (NHS EED) (Internet) (2009 to 30 August 2012)

  • www.crd.york.ac.uk/crdweb/

• Cumulative Index to Nursing and Allied Health Literature (CINAHL) (EBSCOhost) (2009 to 24 August 2012).

Identified references were downloaded in EndNote X4 software for further assessment and handling.

References in retrieved articles were checked for additional studies.

Inclusion criteria

Studies reporting a full economic analysis that related explicitly to the test–treat combination of EGFR mutation testing and treatment with EGFR-TKIs were eligible for inclusion. Specifically, one of the comparators included EGFR mutation testing and for this comparator the treatment decision was guided by the test result; patients whose tumour was EGFR mutation negative were also included in the treatment pathway.

Results

The search retrieved 606 references. Studies were independently assessed for inclusion by two health economists (BR and AvA) and any disagreements were resolved by discussion. After initial screening of titles and abstracts four studies remained, all of which were published as conference abstracts only. During the course of the assessment we identified two additional studies: one published as a conference abstract only and one published as a full paper and a conference abstract. In total, six studies were included, of which only one was published as a full paper. A summary of the full paper by Borget et al. 52 is provided in Table 12, with a quality checklist based on Drummond et al. 53 shown in Table 13. A condensed summary of the conference abstracts is provided in Table 14.

Borget et al. 52 developed a Markov model to compare three hypothetical strategies for second-line treatment with erlotinib in patients with NSCLC for whom at least one platinum-based chemotherapy regimen had failed and who were eligible for erlotinib or chemotherapy.

The three hypothetical strategies were:

1. no patient selection, all patients receive erlotinib

2. clinically guided, patients with favourable clinical features (female, never smokers, with adenocarcinoma) receive erlotinib, others receive docetaxel

3. biologically guided, patients with known EGFR mutations received erlotinib, others receive docetaxel.

Clinical inputs were derived from IPD in the ERMETIC study59 and the GFPC0506 study. 60 Utilities were derived from population-based studies of advanced NSCLC performed in the UK. 61 Total costs included the following categories: chemotherapy drugs, erlotinib, supportive treatments (including treatment for adverse events), transfusion and hospitalisation for any reason, costs after progression and palliative care.

Total quality-adjusted life-years (QALYs) were 0.478, 0.558 and 0.559 for the no selection, clinically guided and biologically guided strategies, respectively. The respective total costs were €21,025, €16,005 and €15,210. The no selection strategy was both the least effective and the most expensive. The biologically and clinically guided strategies had comparable effectiveness, but the biologically guided strategy was slightly less expensive. Results were robust in the sensitivity analyses.

Although this study was of good quality, it does not match our decision problem, as it concerns second-line use of EGFR-TKIs, whereas this assessment concerns first-line treatment with TKIs. The conference abstracts identified all concern the first-line use of TKIs, but do not provide sufficient information to be of use. However, as all were relatively recent, more informative full publications may follow.

Epidermal growth factor receptor tyrosine kinase mutation tests considered in the model

In the health-economic analysis, the cost-effectiveness of different methods for EGFR-TK mutation testing to decide between standard chemotherapy and anti-EGFR-TKIs in patients with locally advanced or metastatic NSCLC was assessed. A range of methods for EGFR-TK mutation testing are currently used in NHS laboratories in England and Wales.

Ideally, the performance of these tests would be assessed against an objective measure of the true presence/absence of a clinically relevant EGFR-TK mutation (the ‘reference standard’). Comparative effectiveness of treatment (TKI vs. chemotherapy) conditional upon the true or false presence/absence of the EGFR-TK mutation could then be determined. However, each different testing method targets a different range of mutations and has different limits of detection (lowest proportion of mutation detectable in tumour cells) and the exact combination of mutation type and level that will provide optimal treatment selection remains unclear. For this reason, assessment of test performance based on comparison with a conventional ‘reference standard’ is currently not possible. In this situation, an alternative way to determine the relative value of diagnostic methods for EGFR-TK mutation testing is to use studies that report on the comparative treatment effect in patients with different EGFR mutation status (positive, negative or unknown) as defined using different EGFR mutation tests. As outlined in the previous chapter, information on comparative effectiveness (PFS and OS) of TKI and chemotherapy in patients with mutation-positive, mutation-negative and mutation-unknown tumours was available for only the Therascreen EGFR PCR Kit48,49 and in patients with mutation-positive and mutation-negative tumours for one type of direct sequencing (direct sequencing of all exon 19–21 mutations). 41 A major assumption underlying the use of these data in the health-economic modelling, however, is that the difference in comparative treatment effect between the two treatments (e.g. TKI vs. chemotherapy) is solely attributable to the use of different mutation tests. Although direct sequencing of all exon 19–21 mutations is not listed in the scope, it was included in the analyses because of a lack of effectiveness and/or survival data on other direct sequencing methods.

In the absence of evidence on the comparative treatment effect in patients with different EGFR mutation status as defined using different EGFR mutation tests, one could consider the accuracy of different EGFR mutation tests for the prediction of response to treatment with TKIs; in this case, response to treatment with TKIs serves as a clinical reference standard. This type of accuracy data was available for two other direct sequencing tests (direct sequencing of all exon 18–21 mutations46 and direct sequencing or WAVE-HS (Transgenomic, Omaha, NE, USA) for inadequate samples (< 50% of tumour cells of all exon 18–24 mutations44). These studies provided no data on the relative PFS and/or OS separately for patients with mutation-positive and mutation-negative tumours. Therefore, evidence available on the relative PFS and OS for mutation positives and mutation negatives, as observed for direct sequencing of all exon 19–21 mutations, was ‘linked’ to direct sequencing of all exon 18–21 mutations46 and direct sequencing or WAVE-HS for inadequate samples (< 50% of tumour cells). Again, although the test strategy direct sequencing or WAVE-HS for inadequate samples (< 50% of tumour cells) is not listed in the scope, it was included in the analysis because it was the only test for which information on the proportion of patients with unknown mutations status was available.

For the remaining EGFR mutation tests listed in the scope, no accuracy data or information to predict (relative) treatment response, PFS or OS in mutation-positive patients (after treatment with TKIs), mutation-negative patients or patients with unknown mutation status (after treatment with doublet chemotherapy) were available. As a result, for the remaining tests, it was only possible to make a comparison based on differences in technical performance and test costs retrieved from the online survey of NHS laboratories in England and Wales (see Chapter 3, What are the technical performance characteristics of the different epidermal growth factor receptor mutation tests?), while assuming equal prognostic value across tests. The latter assumption was not based on evidence of equality but rather absence of any reliable evidence to model a difference in prognostic value for these tests.

Based on the information available to us, three analyses were performed:

• ‘Evidence on comparative effectiveness available’ analysis Therascreen EGFR PCR Kit compared with direct sequencing of all exon 19–21 mutations in order to estimate cost and QALYs using the observed response to treatment and relative PFS and OS data. Information on relative (HR of TKI vs. chemotherapy) PFS and OS in patients with mutation-positive tumours and patients with mutation-negative tumours is not available for other tests. Therefore, in this analysis direct sequencing of all exon 19–21 mutations was used as the closest approximation available to the comparator listed in the scope (direct sequencing of all exon 18–21 mutations).

• ‘Linked evidence’ analysis In this analysis, besides Therascreen EGFR PCR Kit compared with direct sequencing of all exon 19–21 mutations, two other direct sequencing tests [direct sequencing of all exon 18–21 mutations and direct sequencing or WAVE-HS for inadequate samples (< 50% of tumour cells)] for which accuracy data to predict response to treatment were available were included. This was based on the assumption that for the last two direct sequencing methods, the relative PFS and OS for mutation positives and mutation negatives correlate perfectly with relative PFS and OS, as observed for direct sequencing of all exon 19–21 mutations.

• ‘Assumption of equal prognostic value’ analysis For all tests for which information on cost and/or technical performance was available from the online survey. This includes the tests for which neither comparative effectiveness nor response data were available. In this analysis we assessed whether the tests were likely to be cost-effective given an assumption of equal prognostic value and test-specific information on cost and failure rate only. The equal prognostic value assigned was based on data for the Therascreen EGFR PCR Kit, as this was the only test for which prognostic data were available on patients with positive, negative and unknown mutation status. In addition, other tests used in NHS laboratories in England and Wales were considered to have technical characteristics (low limit of detection and similar proportion of tumour cells required for analysis), which were more similar to this test than to direct sequencing methods and would therefore be more likely to have similar prognostic value to the Therascreen EGFR PCR Kit than to direct sequencing. The following tests were included in this analysis:

  • Therascreen EGFR PCR Kit

  • Direct sequencing of exons 19–21

  • Direct sequencing or WAVE-HS for samples with insufficient tumour cells

  • Direct sequencing of exons 18–21

  • Fragment length analysis combined with pyrosequencing

  • Sanger sequencing and fragment length analysis/PCR of negative samples

  • Roche cobas test

  • HRM analysis

  • single-strand conformation analysis

  • Sanger sequencing or Roche cobas for samples with insufficient tumour cells

  • Sanger sequencing or Therascreen for samples with insufficient tumour cells

  • next-generation sequencing

  • Therascreen and Pyrosequencing Kit.

Direct sequencing of all exon 18–21 mutations was taken as the comparator in the ‘linked evidence’ and ‘assumption of equal prognostic value’ analyses.

Consistency with related assessments

This assessment does not update the appraisal of gefitinib for the first-line treatment of locally advanced or metastatic NSCLC. 1 In order to ensure consistency between the modelling approach used in Technology Appraisal 192 and the assessment of the cost-effectiveness of different methods for EGFR-TK mutation testing in this report, the assessment group received the health-economic model submitted by AstraZeneca for Technology Appraisal 192. This model calculates the expected cost-effectiveness of gefitinib compared with doublet chemotherapy for the first-line treatment of patients with locally advanced or metastatic NSCLC with a positive EGFR mutation test based on Therascreen EGFR PCR Kit. This model, together with the amendments suggested and made by the External Review Group (ERG), was used to inform the development of a de novo model in which the long-term consequences of using different EGFR mutation tests were assessed not only in patients with a positive EGFR mutation test, but also in patients with a negative test result, or an unknown test result. The assessment group tested the consistency between the de novo model, the AstraZeneca model, and the amendments made by the ERG. We compared the results of patients with a positive EGFR mutation test using Therascreen EGFR PCR Kit with the initial manufacturer’s submission. Subsequently, the ERG amendments were incorporated and ICERs from the de novo model were compared with ICERs, as reported in the final appraisal determination of Technology Assessment 1921 (see Appendix 6 for results). Furthermore, the health-economic analysis did not assess any differences between different TKIs.

Model structure

In the health-economic model the mean expected costs, life-years (LYs) and QALYs were calculated for each alternative.

The health-economic analysis considers the long-term consequences of technical performance and accuracy of the different tests/test combinations followed by treatment with either standard chemotherapy or a TKI in patients with NSCLC. For this purpose a decision tree and a Markov model were developed. The decision tree was used to model the test result (positive, negative or unknown) and the treatment decision. Patients with a positive test result receive an anti-EGFR-TKI. It is assumed that patients with a negative test result or unknown EGFR mutation status will receive doublet chemotherapy (pemetrexed and cisplatin), as the negative consequences of treatment with TKIs in FPs are greater than the negative consequences of treatment with doublet chemotherapy in FNs. 49 The decision tree is shown in Figure 10.

FIGURE 10.

Decision tree structure. CTX, chemotherapy.

The long-term consequences in terms of costs and QALYs were estimated using a Markov model with a cycle time of 21 days (resembling the duration of one cycle of chemotherapy) and a time horizon of 6 years. Health states in the Markov model are progression free (subdivided into ‘response’ and ‘SD’), disease progression and death. In the progression-free state, patients are on treatment (either TKI or doublet chemotherapy). In each cycle these patients are subdivided over the ‘SD’ and ‘response’ states, based on the OR rate, in order to account for a difference in quality of life between those states. In addition, disutilities and costs associated with treatment related characteristics (i.v. or oral therapy) are modelled. For adverse events of treatment, disutilities and costs were applied for a single cycle in the model. The Markov model structure is shown in Figure 11. The model is described in more detail in NICE Technology Appraisal 192. 51

FIGURE 11.

Markov model structure.

Model parameters

Estimates for model input parameters were retrieved from the industry submission for NICE Technology Appraisal 192,51 the assessment of the clinical effectiveness of different EGFR mutation tests (see Chapter 3, What is the accuracy of epidermal growth factor receptor mutation testing, using any test, for predicting response to treatment with tyrosine kinase inhibitors? and How do outcomes from treatment with epidermal growth factor receptor inhibitors vary according to which test is used to select patients for treatment?), an online survey of NHS laboratories in England and Wales (see Chapter 3, What are the technical performance characteristics of the different EGFR mutation tests?) and the Personal Social Services Research Unit (PSSRU). 62

Survival

As was the case in NICE Technology Appraisal 192, two separate Weibull models were used to estimate cycle-dependent transitions for PFS and OS while on doublet chemotherapy for positive, negative and unknown mutation status. Figure 12 provides a schematic representation of the modelling approach.

FIGURE 12.

Modelling of overall and PFS.

For testing using the Therascreen EGFR PCR Kit, PFS and OS were modelled using the Weibull regression models based on the IPASS study49 and a HR for TKI (based on a meta-analysis and mixed-treatment comparison) used in NICE Technology Appraisal 192. 51 The Weibull regression models have separate lambda and alpha parameters for patients with mutation-positive, -unknown and -negative tumours, and are based on treatment with doublet chemotherapy (Table 18).

TABLE 18 - Weibull models used to model survival on paclitaxel and carboplatin after use of the Therascreen EGFR PCR Kit.a (CiC information has been removed.)

To estimate PFS and OS for patients treated with TKIs after a positive test result using the Therascreen EGFR PCR Kit, a HR of 0.43 (95% CI 0.34 to 0.53) was applied to the Weibull function for mutation positives. This HR was modelled using a log-normal distribution.

For direct sequencing of all exon 19–21 mutations, PFS and OS for mutation positives after EGFR-TKI and negatives after doublet chemotherapy were modelled using Kaplan–Meier curves extracted from the First-SIGNAL trial. 41 The corresponding SEs were calculated using the Peto method. 64 In the First-SIGNAL trial, mutation-negative patients were treated with gemcitabine and cisplatin. 41 The PFS and OS estimates obtained for these mutation-negative patients were adjusted (HR = 1.087 for PFS, and HR = 1.087 for OS) to correspond with treatment with paclitaxel and carboplatin. 51 PFS and OS for patients with tumours of unknown mutation status were based on the IPASS Weibull model for unknown mutations, as these were not reported in the First-SIGNAL trial.

Consistent with the use of pemetrexed and cisplatin as doublet chemotherapy, the HRs reported in Table 19 were used to recalculate PFS and OS for both comparators. Accordingly, OR rate presented in Table 17 was recalculated to correspond with pemetrexed and cisplatin. These HRs and odds ratios were retrieved from the updated mixed-treatment comparison from NICE Technology Appraisal 192.

TABLE 19 - Hazard ratios and odds ratios for paclitaxel and carboplatin compared with pemetrexed and cisplatin (updated mixed-treatment comparison)65

	Estimate	Lower 95% CI	Upper 95% CI	Distribution
HRs progression free and OS
PFS	0.88	0.74	1.05	Log-normal
OS	0.78	0.65	0.93	Log-normal
Odds ratios
OR rate	1.64	1.15	2.27	Log-normal
Neutropenia	0.46	0.07	1.62	Log-normal
Febrile neutropenia	0.19	0.01	0.84	Log-normal
Fatigue	2.62	1.30	4.65	Log-normal
Nausea and vomiting	10.92	1.11	41.94	Log-normal
Diarrhoea	1.00	–	–	Fixed
Hair loss (grade 2)	1.00	–	–	Fixed
Anaemia	1.62	0.54	3.75	Log-normal

The PFS and OS curves for patients tested with the Therascreen EGFR PCR Kit and with direct sequencing of all exon 19–21 mutations for the ‘evidence on comparative effectiveness available’ analysis are presented in Figures 13 and 14.

FIGURE 13.

Progression-free survival for patients tested with the Therascreen EGFR PCR Kit49 and with direct sequencing of all exon 19–20 mutations. 41 (CiC information has been removed.)

FIGURE 14.

Overall survival for patients tested with the Therascreen EGFR PCR Kit49 and with direct sequencing of all exon 19–20 mutations. 41 (CiC information has been removed.)

In the ‘linked evidence’ analysis, PFS and OS for patients tested with direct sequencing of all exon 18–21 mutations and direct sequencing or WAVE-HS for inadequate samples (< 50% of tumour cells) were assumed equal to the PFS and OS as described above for direct sequencing of all exon 19–21 mutations. PFS and OS for patients tested with the Therascreen EGFR PCR Kit and with direct sequencing of all exon 19–21 mutations in the ‘linked evidence’ analysis was equal to the estimates used in the ‘evidence on comparative effectiveness available’ analysis.

Overview of main model assumptions

The main assumptions in the health-economic analyses were:

• The differences in relative treatment response, PFS and OS reported in the First-SIGNAL trial41 and those reported for the IPASS trial49 are attributable solely to the different tests used (Therascreen EGFR PCR Kit and direct sequencing of all exon 19–21 mutations, respectively) to distinguish between patients whose tumours are EGFR mutation positive (and receive TKI treatment) and patients whose tumours are EGFR mutation negative (and receive doublet chemotherapy) (‘evidence of comparative effectiveness available’ and ‘linked evidence’ analyses).

• To calculate the sensitivity and specificity of the tests, required to calculate the proportion of positive and negative test results (see Table 15), patients who tested positive were categorised as FP if no treatment response was observed after TKI, whereas patients were categorised as TP if treatment response was observed TKI. Similarly, patients who tested negative were categorised as FN if treatment response was observed after TKI, whereas patients were categorised as TN if no treatment response was observed after TKI (all analyses).

• The proportion of patients with unknown mutation status relative to the number of patients for whom a tissue sample was available in the trials44,49 provides a realistic approximation of the proportion of patients with an unknown test result in clinical practice (all analyses).

• The OR rate, PFS and OS in patients with an unknown test result as reported in the IPASS trial49 is generalisable to direct sequencing methods (‘evidence of comparative effectiveness available’ and ‘linked evidence’ analyses).

• The probability of an unknown test result as reported in the study by Jackman et al. 44 [direct sequencing or WAVE-HS for inadequate samples (< 50% of tumour cells) is generalisable to other direct sequencing methods (‘linked evidence’ analysis)].

• The OR rate in patients with a negative test result as reported in the First-SIGNAL trial41 is generalisable to other direct sequencing methods (‘linked evidence’ analysis).

• PFS and OS in patients with a positive or negative test result reported in the First-SIGNAL trial41 (direct sequencing of exons 19–21) are generalisable to other direct sequencing methods (exons 18–21) (‘linked evidence’ analysis). In other words, we assumed there is no clinical significance in testing exon 18 mutations.

Sensitivity analyses

For analyses 1 and 2, in a sensitivity analysis the costs reported in Table 24 were updated. For all three analyses, in a sensitivity analysis the proportion of unknown patients was based on the results of the online survey instead of the literature (see Table 5).

‘Evidence on comparative effectiveness available’ analysis

The probabilistic results of the ‘evidence on comparative effectiveness available’ analysis are shown in Table 26. It should be noted that this analysis is based on a number of assumptions outlined above (see Model analyses), of which the following two are particularly problematic:

• The proportion of patients with a positive or negative test result after the use of these tests in the NHS population was estimated based on the proportion of EGFR mutation-positive patients in England and Wales, the proportion of patients with an unknown test result, and test accuracy for the prediction of treatment response derived from two separate trials. 41,49

• The differences in relative treatment response, PFS and OS between the results of First-SIGNAL41 that were used to model EGFR mutation testing with direct sequencing of all exon 19–21 mutations and the results of the IPASS trial48,49 that were used to model EGFR mutation testing with the Therascreen EGFR PCR Kit, are assumed to be solely attributable to the different tests used to distinguish between patients who are EGFR mutation positive (and receive TKI treatment) and patients who are EGFR mutation negative (and receive doublet chemotherapy).

In this analysis, the Therascreen EGFR PCR Kit was both less effective and less costly than direct sequencing of all exons 19–21 at an ICER of £32,167. The lower costs and QALYs for the Therascreen EGFR PCR Kit can be explained by the fact that patients whose tumours are mutation negative do worse on OS in the IPASS trial48,49 than in First-SIGNAL,41 whereas for mutation-positive patients the outcome is similar, and for unknowns it is the same (by assumption) – see Figures 13 and 14. Therefore, on average, with the Therascreen EGFR PCR Kit strategy patients have shorter survival, and therefore fewer QALYs than testing with direct sequencing of all exons 19–21. The apparent shorter survival also reduces costs. The cost-effectiveness acceptability curve (Figure 15) shows that at a threshold value of £32,500 direct sequencing of all exons 19–21 becomes the preferred strategy.

FIGURE 15.

Cost-effectiveness acceptability curve for ‘evidence on comparable effectiveness available’ analysis, base case.

Results were robust for changed assumptions in the sensitivity analyses, in the sense that testing with Therascreen EGFR PCR Kit was always less effective and less expensive. The ICERs amounted to £34,555 (unknowns from survey) and £32,196 (updated costs). The cost-effectiveness acceptability curves for the sensitivity analyses are presented in Appendix 7.

‘Linked evidence’ analysis

The ‘linked evidence’ analysis includes four tests, i.e. all tests for which either evidence on relative effectiveness or accuracy was available. Table 27 shows the probabilistic results of this analysis.

This analysis was also based on a number of assumptions, including those described above (see Model analyses and ‘Evidence on comparative effectiveness available’ analysis). The following additional assumption should be particularly noted:

• For direct sequencing of all exon 18–21 mutations and direct sequencing or WAVE-HS for inadequate samples (< 50% of tumour cells), the relative PFS and OS for mutation positives and mutation negatives correlate perfectly with relative PFS and OS as observed for direct sequencing of all exon 19–21 mutations in the First-SIGNAL trial. 41

In the base-case analysis, compared with direct sequencing of all exon 18–21 mutations, the Therascreen EGFR PCR Kit was less costly and less effective at an ICER of £31,849 per QALY lost. Direct sequencing of all exon 19–21 mutations and direct sequencing or WAVE-HS for inadequate samples were both more expensive and more effective than the comparator. For thresholds of < £33,500, testing with the Therascreen EGFR PCR Kit is the preferred strategy, then direct sequencing of all exon 18–21 mutations is preferred up to a threshold of £39,000, at which direct sequencing of all exon 19–21 mutations has the highest probability of being cost-effective (Figure 16). The sensitivity analyses (see Appendix 7) show that these findings are quite robust in the sense that compared with direct sequencing of all exon 18–21 mutations the Therascreen EGFR PCR Kit is always less expensive and less effective, and the remaining two tests are more effective and more expensive.

FIGURE 16.

Cost-effectiveness acceptability curve for ‘linked evidence’ analysis.

‘Assumption of equal prognostic value’ analysis

The ‘assumption of equal prognostic value’ analysis included all tests for which information on cost and/or technical performance was available from the online survey of NHS laboratories in England and Wales. This includes the tests for which neither comparative effectiveness nor response data were available. Therefore, this analysis assessed whether the tests were likely to be cost-effective given an assumption of equal prognostic value (based on the prognostic value of testing with the Therascreen EGFR PCR Kit, as this was the only test for which prognostic data were available on patients with positive, negative and unknown tumour EGFR mutation status) and test-specific information on cost only. As a result, the strategies differ only with respect to costs. As shown in Table 28, Sanger sequencing or Roche cobas for samples with insufficient tumour cells is the least expensive strategy, and fragment length analysis combined with pyrosequencing is the most expensive strategy. However, the difference between the costs of these strategies amounts to only £477 (< 1% of total strategy costs).

TABLE 28 - Probabilistic results for ‘assumption of equal prognostic value’ analysis, base case

Although this test was not listed in the scope, it was included in the analyses as discussed above (see Epidermal growth factor receptor tyrosine kinase mutation tests considered in this model).

Strategy	Costs (95% CI)	Incremental costs compared with direct sequencing (exons 18–21)
Sanger sequencing or Roche cobas for samples with insufficient tumour cells	CiC information has been removed	–£15
Sanger sequencing and fragment length analysis/PCR of negative samples	CiC information has been removed	–£11
Sanger sequencing or Therascreen EGFR PCR Kit for samples with insufficient tumour cells	CiC information has been removed	–£9
Roche cobas	CiC information has been removed	–£9
HRM analysis	CiC information has been removed	–£3
Direct sequencing of exons 19–21 a	CiC information has been removed	£0
Direct sequencing of exons 18–21	CiC information has been removed
Single-strand conformation analysis	CiC information has been removed	£1
Direct sequencing or WAVE-HS a	CiC information has been removed	£1
Therascreen EGFR PCR Kit	CiC information has been removed	£5
Fragment length analysis combined with pyrosequencing	CiC information has been removed	£33

In a sensitivity analysis the proportion of patients with tumours of unknown mutation status were taken from the online survey of NHS laboratories in England and Wales, rather than based on the literature. As a result, in this sensitivity analysis a difference in health outcomes (QALYs) is modelled. The results in Table 29 show that this assumption has some impact on the relative costs and effects of the strategies, in the sense that single-strand conformation analysis is now the most costly. This is caused by the fact that the percentage of failures as reported in the survey is the highest for single-strand conformation analysis (10%, n = 1), whereas for Sanger sequencing and fragment length analysis/PCR it is 0% (n = 1). A higher failure rate will, in its turn, lead to a lower proportion of patients with mutation-positive and mutation-negative tumours and, therefore, on average, to higher costs. This is because patients with an unknown mutation status are more costly than the average of the patients with a known (positive or negative) mutation status. The cost-effectiveness acceptability curve is presented in Figure 17.

TABLE 29 - Probabilistic results for ‘assumption of equal prognostic value’ analysis, sensitivity analyses: unknown based on survey

Although this test was not listed in the scope, it was included in the analyses as discussed above (see Epidermal growth factor receptor tyrosine kinase mutation tests considered in this model).

Strategy	Costs	QALYs	Compared with direct sequencing of all exon 18–21 mutations			Compared with next best strategy
			Incremental costs	Incremental QALYs	ICER	Comparator	Incremental costs	Incremental QALYs	ICER
Sanger sequencing and fragment length analysis/PCR of negative samples	CiC information has been removed	0.871	–£226	–0.007	£33,437
HRM analysis	CiC information has been removed	0.871	–£211	–0.007	£31,848	Sanger sequencing and fragment length analysis/PCR of negative samples	£14	0.000	Extended dominance
Sanger sequencing or Therascreen EGFR PCR Kit for samples with insufficient tumour cells	CiC information has been removed	0.877	–£40	–0.001	£45,629	Sanger sequencing and fragment length analysis/PCR of negative samples	£186	0.006	Extended dominance
Therascreen EGFR PCR Kit	CiC information has been removed	0.877	–£26	–0.001	£24,977	Sanger sequencing and fragment length analysis/PCR of negative samples	£200	0.006	Extended dominance
Sanger Sequencing or Roche cobas for samples with insufficient tumour cells	CiC information has been removed	0.878	–£18	0.000	Dominated	Sanger sequencing and fragment length analysis/PCR of negative samples	£207	0.007	£30,602
Direct sequencing or WAVE-HS a	CiC information has been removed	0.878	£0	0.000	Dominated	Sanger sequencing or Roche cobas for samples with insufficient tumour cells	£18	0.000	Dominated
Direct sequencing of exons 18–21	CiC information has been removed	0.878				Sanger sequencing or Roche cobas for samples with insufficient tumour cells	£18	0.000	Dominated
Direct sequencing of exons 19–21 a	CiC information has been removed	0.878	£0	0.000	£615,549	Sanger sequencing or Roche cobas for samples with insufficient tumour cells	£19	0.000	Dominated
Roche cobas	CiC information has been removed	0.879	£15	0.001	£19,501	Sanger sequencing or Roche cobas for samples with insufficient tumour cells	£33	0.001	Extended dominance
Fragment length analysis combined with pyrosequencing	CiC information has been removed	0.879	£62	0.001	£79,807	Sanger sequencing or Roche cobas for samples with insufficient tumour cells	£81	0.001	Extended dominance
Single-strand conformation analysis	CiC information has been removed	0.886	£264	0.008	£31,080	Sanger sequencing or Roche cobas for samples with insufficient tumour cells	£283	0.008	£33,338

FIGURE 17.

Cost-effectiveness acceptability curve for ‘assumption of equal prognostic value’ analysis, sensitivity analysis: unknown based on survey.

Clinical effectiveness

There was no strong evidence that any one EGFR mutation test had greater accuracy than any other test, although there was a suggestion that Therascreen EGFR PCR Kit may be more accurate than direct sequencing for predicting response to treatment with TKIs. Eleven studies were included in the review; these evaluated the Therascreen EGFR PCR Kit (version 1), direct sequencing, cobas EGFR Mutation Testing Kit, fragment length analysis, and Sanger sequencing. Six studies (two RCTs and four cohort studies) provided data on the accuracy of EGFR mutation testing for predicting response to treatment with TKIs in patients with stage IIIB or IV NSCLC. Five RCTs, including two that also provided accuracy data, reported data on the clinical effectiveness of TKIs compared with standard chemotherapy in patients with stage IIIB or IV NSCLC with EGFR mutation-positive tumours; one additional study reported data for a subgroup of patients from one of these RCTs, whose biopsy samples had been re-analysed using a different EGFR mutation testing method. The remaining study was included as a supplement to the survey of laboratories in England and Wales that currently provide EGFR mutation testing, and did not report any data on clinical outcomes.

The survey of laboratories providing EGFR mutation testing indicated that the Therascreen EGFR PCR Kit was the single most commonly used method (6 out of 13 respondents). Reasons cited by respondents for their choice of the Therascreen EGFR PCR Kit were proportion of tumour cells required; ease of use; cost; and mutations covered. There was no clear indication that choice of test method was related to volume of throughput. Most respondents reported turnaround times – from receipt of sample to reporting to the clinician – of between 3 and 7 days. The only laboratory to report a turnaround time of < 3 days (24–48 hours) used the Therascreen EGFR PCR Kit. All respondents reported turnaround times of less than the 10 working day maximum recommended by the European EGFR Workshop Group. 12 With the exception of those whose testing strategy included direct sequencing methods, all respondents reported a minimum requirement for testing at or < 10% of tumour cells, with some of the laboratories that used the Therascreen EGFR PCR Kit reporting minimum requirements as low as 1%. Although most respondents included costs in their reasons for choosing a particular test, it is worth noting that a relatively narrow range of costs was reported across all tests (£110–190), with a similar level of variation apparent within a single test, Therascreen EGFR PCR Kit (£120–190). When contacted by NICE, UK NEQAS stated:

Error rates are not always method related and it is not always possible to obtain data from all the labs committing critical genotyping errors. Therefore, any data which could be provided would be skewed with processing and reporting issues rather than being method related. There has been no correlation between any method used for EGFR testing and errors since we started providing the scheme in 2010.

Studies that provided data on test accuracy assessed the Therascreen EGFR PCR Kit (version 1) or direct sequencing methods (exons 18 or 19 to exons 21 or 24). No studies were identified that reported accuracy data for any other EGFR mutation testing method. The Therascreen EGFR PCR Kit appeared to have the best overall performance for discriminating between patients who are likely to benefit from TKI treatment and those who are not. The sensitivity and specificity estimates for OR were 99% (95% CI 94% to 100%) and 69% (95% CI 60% to 77%), respectively, with specificity increasing and sensitivity decreasing where a lower response threshold (DC) was used. 49 Four of the five direct sequencing studies reported high estimates of specificity (> 80%) for OR, with sensitivities ranging from 60% to 80%. 41,43–45 Three of these studies also assessed DC and reported high specificities (> 90%) and very low sensitivities (≤ 35%). 43–45 The remaining direct sequencing study reported low sensitivity (66%) and specificity (50%) for DC and low specificity (61%) with high sensitivity (84%) for OR.

There were no clear common participant characteristics, across studies, which reported similar sensitivity or specificity estimates for DC or OR. Specificity estimates may have been affected by the way in which resistance mutations were classified; the three direct sequencing studies that reported high specificity estimates either stated that patients whose tumours showed resistance or non-sensitising mutations were classified as EGFR mutation negative or did not identify any patients with tumours showing these types of mutation. Although the number of resistance mutations identified was generally small, their potential effect on specificity estimates was magnified by the very small sample size in most studies. The most commonly observed mutations were exon 19 deletions and the exon 21 point mutation L858R; most patients in the included studies who had these mutations achieved a minimum response of SD when treated with TKIs. Large database studies provide some support for the idea that mutations in exon 20, and in particular the mutation T790M, may be associated with a lack of response to TKIs (see Clinical effectiveness, below). A second possible explanation may be that the Therascreen EGFR PCR Kit has a lower limit of detection, i.e. it is able to detect EGFR mutations at a lower abundance (fewer cancer cells carrying the mutation) than direct sequencing methods. A lower limit of detection would be beneficial only if it could be shown that patients whose tumours have a lower abundance of EGFR mutation benefit from treatment with TKIs, and the apparent improved diagnostic performance of the Therascreen EGFR PCR Kit, compared with direct sequencing methods, indicates that this may be the case. However, none of the studies identified by this review reported data on the relationship between abundance of EGFR mutation and response to first-line TKI treatment in patients with stage IIIB or IV NSCLC.

The five RCTs included in this review compared the TKIs gefitinib or erlotinib with various single agent or combination standard chemotherapy regimens and reported data on PFS. Three of the trials included only patients with EGFR mutation-positive tumours,40,47 and the remaining two trials (IPASS and First-SIGNAL) included chemotherapy naive patients with stage IIIB or IV NSCLC and reported a subgroup analysis for patients who had received EGFR mutation testing using the Therascreen EGFR PCR Kit (version 1)48,49 or direct sequencing. 41 Though derived from a subgroup analysis of tested patients, data from these trials were most the most complete available, in that they provided information on the effectiveness of TKIs compared with standard chemotherapy in both test-positive and test-negative patients. The results of the IPASS subgroup analyses indicated that PFS was significantly longer for patients receiving gefitinib than for those receiving standard chemotherapy in the EGFR mutation-positive subgroup [HR 0.48 (95% CI 0.36 to 0.64)] and significantly shorter for patients receiving gefitinib than for those receiving standard chemotherapy in the EGFR mutation-negative subgroup [HR 2.85 (95% CI 2.05 to 3.98)]. This trial formed the basis of the technology appraisal that informed NICE guidance TA 192 on gefitinib for the first-line treatment of locally advanced or metastatic NSCLC. 1 The results of the First-SIGNAL trial indicated a trend towards longer PFS for patients receiving gefitinib than for those receiving standard chemotherapy in the EGFR mutation-positive subgroup [HR 0.54 (95% CI 0.27 to 1.10)] and a trend towards significantly shorter PFS for patients receiving gefitinib than for those receiving standard chemotherapy in the EGFR mutation-negative subgroup [HR 1.42 (95% CI 0.82 to 2.47)]. 41 The remaining trials provided information on only the effectiveness of TKIs compared with standard chemotherapy in patients with EGFR mutation-positive tumours; HRs for PFS ranged from 0.48 (95% CI 0.36 to 0.64) to 0.16 (95% CI 0.10 to 0.26). The included trials used various methods to assess EGFR mutation status. Two trials used direct sequencing methods, but limited the definition of positive EGFR mutation status to the presence of an ‘activating mutation’ (exon 19 deletions or exon 21 mutation L858R). These two trials were included in the technology appraisal that informed NICE guidance TA 258 on erlotinib for the first-line treatment of locally advanced or metastatic EGFR-TK mutation-positive NSCLC. 2 The re-analysis of samples from one of these trials and the two remaining trials used EGFR mutation tests that targeted a wider range of mutations, including resistance mutations. Overall, there were no clear differences in any measure of TKI treatment effect (PFS, OR or DC), regardless of which EGFR mutation test (selective for activating mutations exon 19 deletions and exon 21 L858R, or targeting a wider range of mutations) was used to select patients. No study reported a significant difference in TKI treatment effect between patients with exon 19 deletions and those with the exon 20 mutation L858R. One additional trial, the Western Japan Oncology Group study, was included in TA 258 but did not meet the inclusion criteria for our review, as it focused on the treatment of patients with postoperative recurrence with or without postoperative adjuvant chemotherapy; patients with stage IIIB or IV NSCLC were also included but no separate data were reported for them. 70 EGFR mutation testing in this study also targeted exon 19 deletions and the exon 21 mutation L858R, and used a combination of fragment analysis and direct sequencing methods; the reported treatment effect of TKI (gefitinib) compared with standard chemotherapy (cisplatin plus docetaxel) was similar to that seen in the trials included in our review [PFS HR 0.49 (95% CI 0.34 to 0.71)]. 70

The estimates of the effectiveness of first-line treatment with TKIs, compared with standard chemotherapy, in patients with advanced NSCLC whose tumours tested positive for an EGFR mutation reported by studies included in this review, were consistent with pooled estimates reported in recent systematic reviews. Three systematic reviews had inclusion criteria that matched ours in terms of population intervention and comparator but which did not specify reporting of EGFR testing methods. All three reviews reported pooled HRs, which indicated increased PFS in patients with EGFR mutation-positive tumours who were treated with TKIs compared with those treated with standard chemotherapy [HR 0.43 (95% CI 0.32 to 0.58),71 HR 0.37 (95% CI 0.27 to 0.52)72 and HR 0.45 (95% CI 0.36 to 0.58)73]. Two reviews also reported significantly higher OR rates [RR 5.68 (95% CI 3.17 to 10.18)]72 and HR 2.08 (95% CI 1.75 to 2.46)73 for patients treated with TKIs, and no significant difference in OS between the two treatment groups. 72,73

Cost-effectiveness

The review of economic analyses of different methods for EGFR TK mutation testing to decide between standard chemotherapy or EGFR-TKIs for first-line treatment of patients with locally advanced or metastatic NSCLC found one full paper52 and five conference abstracts. 54–58 The full paper did not fit the decision problem, as it concerned second-line use of anti-EGFR-TKIs. Although the conference abstracts were all about first-line use of TKIs, they did not provide enough specific information to be of use; future full publications may provide more information.

In the health-economic analysis, the cost-effectiveness of different methods for EGFR-TK mutation testing to decide between standard chemotherapy or EGFR-TKIs for first-line treatment of patients with locally advanced or metastatic NSCLC was assessed. In light of the scarce evidence that was available, three analyses were performed: ‘evidence on comparative effectiveness available’, ‘linked evidence’, and ‘assumption of equal prognostic value’. Direct sequencing of all exon 18–21 mutations, the comparator, could only be included in the last two analyses.

In the ‘evidence on comparative effectiveness available’ analysis, testing with the Therascreen EGFR PCR Kit was compared with direct sequencing of all exon 19–21 mutations in order to estimate lifetime cost and QALYs using the observed response to treatment and the available relative PFS and OS data. The results of this analysis suggested that direct sequencing of all exon 19–21 mutations was both more effective and more costly than testing with the Therascreen EGFR PCR Kit at an ICER of £32,167 per QALY gained. The sensitivity analyses all resulted in similar outcomes. The key drivers behind this result were the differences in the proportion of patients with EGFR mutation positive, unknown mutation and mutation-negative tumours, and differences in OR, PFS and OS. In particular, the predicted OS for mutation-negative patients differed substantially between the studies using the Therascreen EGFR PCR Kit48,49 and the study which used direct sequencing of all exon 19–2141 (see Figure 13). OS for mutation negatives after testing using the Therascreen EGFR PCR Kit was substantially lower than for testing using direct sequencing of all exon 19–21 mutations, whereas PFS was similar. As a result, testing using the Therascreen EGFR PCR Kit appeared less effective in terms of QALYs but was also less costly as the gained LYs for direct sequencing of all exon 19–21 mutations were mainly spent in the relatively expensive disease progression health state.

It should be noted that this analysis was based on a number of assumptions, of which the following two are particularly problematic:

• The proportion of patients with a positive or negative test result after the use of these tests in the NHS population was estimated based on the proportion of EGFR mutation-positive patients in England and Wales, the proportion of patients with an unknown test result and test accuracy for the prediction of treatment response derived from two separate trials. 41,49

• The differences in relative treatment response, PFS and OS between the results of the First-SIGNAL trial,41 that were used to model direct sequencing of all exon 19–21 mutations, and the results of the IPASS trial,48,49 that were used to model testing using the Therascreen EGFR PCR Kit, are solely attributable to the different tests used to distinguish between patients whose tumours are EGFR mutation positive (and who receive TKI treatment) and patients whose tumours are EGFR mutation negative (and who receive doublet chemotherapy).

The results of the ‘evidence on comparative effectiveness available’ analysis should therefore be interpreted on the condition that these assumptions hold. Moreover, the uncertainty presented surrounding the results is an underestimation of the true uncertainty, as the uncertainty associated with the assumptions was not parameterised and is therefore not reflected in the probabilistic sensitivity analyses.

In the ‘linked evidence’ analysis, two other direct sequencing tests [direct sequencing of all exon 18–21 mutations and direct sequencing or WAVE-HS for inadequate samples (< 50% of tumour cells)] for which accuracy data to predict response to treatment with TKIs were available were also included in the analysis. The results of this analysis showed that, compared with direct sequencing of all exons 18–21 mutations, the Therascreen EGFR PCR Kit was less effective and less costly (ICER £31,849), whereas the other tests were more effective and more expensive (ICERs £35,634 and £38,251). Sensitivity analyses did not show any substantial changes to these results. However, it should be noted that this analysis is also based on a number of substantive assumptions, including those described for the ‘evidence on comparative effectiveness available’ analysis. The following additional assumption should also be noted:

• For direct sequencing of all exon 18–21 mutations and direct sequencing or WAVE-HS for inadequate samples (< 50% of tumour cells), the relative PFS and OS for mutation positives and mutation negatives correlates perfectly with relative PFS and OS as observed for direct sequencing of all exon 19–21 mutations in the First-SIGNAL trial. 41

The same caveat for the interpretation of the results and surrounding uncertainty as explained above for the ‘evidence on comparative effectiveness available’ analysis applies to the interpretation of the results of the ‘linked evidence’ analysis.

The ‘assumption of equal prognostic value’ analysis included all tests for which information on cost and/or technical performance were available from the online survey of NHS laboratories in England and Wales. This included the tests for which neither comparative effectiveness nor response data were available. Therefore, in this analysis, the costs of the tests were assessed given an assumption of equal prognostic value and test-specific information on costs only. For this purpose, the prognostic value of all tests was based on the Therascreen EGFR PCR Kit, as this was the only test for which prognostic data were available on patients with positive, negative and unknown tumour mutation status. In addition, other tests used in NHS laboratories in England and Wales were considered to have technical characteristics (low limit of detection and similar proportion of tumour cells required for analysis) that were more similar to this test than to direct sequencing methods and would therefore be more likely to have similar prognostic value to the Therascreen EGFR PCR Kit than to direct sequencing. The results of the ‘assumption of equal prognostic value’ analysis indicated that the strategies were almost equal, i.e. the lowest total strategy cost was (CiC information has been removed) (Sanger sequencing or Roche cobas) compared with (CiC information has been removed) for the most expensive strategy (fragment length analysis combined with pyrosequencing). The sensitivity analysis, where the number of unknowns was based on results from the online survey of NHS laboratories in England and Wales, instead of being assumed equal based on literature, showed a slightly larger range of costs (CiC information has been removed) and a small range in QALYs (0.871 to 0.886) for the included mutation tests.

Clinical effectiveness

Extensive literature searches were conducted in an attempt to maximise retrieval of relevant studies. These included electronic searches of a variety of bibliographic databases, as well as screening of clinical trials registers and conference abstracts to identify unpublished studies. Because of the known difficulties in identifying test accuracy studies using study design-related search terms,74 and potential need to include non-RCTs, search strategies were developed to maximise sensitivity at the expense of reduced specificity. Thus, large numbers of citations were identified and screened, many of which did not meet the inclusion criteria of the review.

The possibility of publication bias remains a potential problem for all systematic reviews. Considerations may differ for systematic reviews of test accuracy studies. It is relatively simple to define a positive result for studies of treatment, for example a significant difference which favours treatment between the treatment and control groups. This is not the case for test accuracy studies, which measure agreement between index test and reference standard. It would seem likely that studies finding greater agreement (high estimates of sensitivity and specificity) will be published more often. This distinction may be less applicable to studies in this review which provided accuracy data, as in all cases these studies aimed to assess the effectiveness of treatment with TKIs in different patient groups rather than being primarily focused upon test performance. Our review included small numbers of clinically heterogeneous studies, both for the accuracy of EGFR mutation testing to predict response to treatment with TKIs and for the relative effectiveness of TKIs in populations selected using different EGFR mutation test methods. We were therefore unable to undertake any meta-analyses or formal assessment of publication bias. However, our search strategy included a variety of routes to identify unpublished studies and resulted in the inclusion of a number of conference abstracts.

Clear inclusion criteria were specified in the protocol for this review and the one protocol modification that occurred during the assessment has been highlighted in the protocol. The eligibility of studies for inclusion is therefore transparent. In addition, we have provided specific reasons for excluding all of the studies considered potentially relevant at initial citation screening (see Appendix 5). The review process followed recommended methods to minimise the potential for error and/or bias;22 studies were independently screened for inclusion by two reviewers and data extraction and quality assessment were carried out by one reviewer and checked by a second (MW and PW). Any disagreements were resolved by consensus.

Studies included in this review were assessed for risk of bias using published tools appropriate to study design and/or the type of data extracted. Studies that provided data on the accuracy of EGFR mutation testing to predict response to treatment with TKIs were assessed using a modification of the QUADAS-2 tool. 31 QUADAS-2 is structured into four key domains covering participant selection, index test, reference standard and the flow of patients through the study (including timing of tests). Each domain is rated for risk of bias (low, high or unclear); the participant selection, index test and reference standard domain are also separately rated for concerns regarding the applicability of the study to the review question (low, high or unclear). The version of QUADAS-2 used in this report did not include assessment of applicability because both the index test and study population were tightly defined by our inclusion criteria and clinical outcome measures were treated as the reference standard. Studies that provided data on the effectiveness of treatment with TKIs, compared with standard chemotherapy, in patients with EGFR mutation-positive tumours were all RCTs or subgroup analyses from RCTs. These studies were therefore assessed using the Collaboration’s tool for assessing risk of bias in randomised trials. 25,30 The results of the risk of bias assessment are reported, in full, for all included studies (see Appendix 3) and in summary in the results (see Chapter 3, What is the accuracy of epidermal growth factor receptor mutation testing, using any test, for predicting response to treatment with tyrosine kinase inhibitors? and How do outcomes from treatment with epidermal growth factor receptor inhibitors vary according to which test is used to select patients for treatment?). The main potential sources of bias identified were exclusion of withdrawals from the analyses (for both studies providing data on the accuracy of EGFR mutation tests to predict response to TKIs and RCTs of TKIs in patients with EGFR mutation-positive tumours) and blinding of participants and personnel in treatment trials, which was not possible owing to the different delivery modes of intervention and comparator drugs.

All of the studies included in this review have some limitations in respect of their ability to address the overall aim of comparing the clinical effectiveness of different EGFR mutation tests to determine which patients may benefit from treatment with TKIs and which should receive standard chemotherapy. The IPASS48,49 and First-SIGNAL41 trials represent the closest approximation to the ideal study in that they provide full information on the comparative treatment effect (TKI vs. standard chemotherapy) for both patients with EGFR mutation-positive and EGFR mutation-negative tumours, for which mutation status was defined using the Therascreen EGFR PCR Kit (version 1) and direct sequencing, respectively. However, data were derived from subgroup analyses of patients included in the original trial who had received EGFR mutation testing and, in the case of the First-SIGNAL study, this subgroup included a small number of participants and was poorly described. 41 Because methods of testing EGFR mutation status differ both in terms of the mutations targeted and limit of detection (the lowest proportion of tumour cells with a mutation that can be detected), the definition of EGFR mutation positive varies according to which test is used. All testing methods are essentially reference standard methods for classifying mutation status, as defined by the specific test characteristics. It is therefore not useful to compare tests solely in terms of their ability to detect particular combinations of mutations. The essential clinical question is ‘which testing method is best at classifying patients, such that the maximum treatment effect is achieved both for mutation-positive patients who receive TKIs and mutation-negative patients who receive standard chemotherapy?’ To fully address this question, IPASS-type data would be required for patients with mutation-positive tumours and patients with mutation-negative tumours, as defined by each proposed classification method (i.e. each different EGFR test). Following the IPASS trial and subsequent NICE recommendations,1,75 obtaining these data may be problematic, as it could be argued that a trial for which patients are randomised to TKI or standard chemotherapy regardless of tumour EGFR mutation status would be unethical. Additionally, once the principle had been established that TKIs are more effective in EGFR mutation-positive patients, subsequent trials have tended to focus on assessing the effectiveness of various TKIs in populations with EGFR mutation-positive tumours; trials are not primarily concerned with the method used to establish mutation status. An alternative approach to this problem is provided by studies that report sufficient data to calculate the accuracy of different EGFR mutation tests for predicting response to treatment with TKIs. These studies provide information on the extent to which different EGFR mutation tests are able to discriminate between patients who will respond to TKI treatment and those who will not; treatment response data are reported for patients with EGFR mutation-positive and EGFR mutation-negative tumours. However, we were able to identify only four studies of this type: all used direct sequencing methods, three pre-dated the IPASS trial, and three had very small sample sizes, which were reflected in the wide CIs around sensitivity and specificity estimates. In addition, no study reported data for more than one EGFR mutation test, hence any apparent differences in test performance observed between studies may have arisen as a result of differences in study populations. Trials that compared the effectiveness of TKIs with that of standard chemotherapy in patients with advanced NSCLC whose tumours tested positive for EGFR mutations were also included in this review. These trials were included with the aim of providing some indication on how the favourable TKI treatment effect seen in patients with mutation-positive tumours in the IPASS trial may vary according to how these patients are selected (which EGFR mutation test is used). However, it should be noted that differences between these studies, other than the way in which positive EGFR mutation status is defined, particularly in relation to the baseline participant characteristics, may contribute to any differences in treatment effects observed. In addition, these trials can provide no information about the relative effectiveness of TKIs and standard chemotherapy in patients whose tumours are classified as EGFR mutation negative by tests other than the Therascreen EGFR PCR Kit. Some trials reported the results of subgroup analyses to assess possible variation in treatment effect (e.g. smoking history, tumour histology); however, trials were generally not powered to detect any difference in treatment effect between subgroups.

This assessment assumes equivalent treatment effects for the two TKIs (gefitinib and erlotinib), which are recommended by NICE as first-line treatments for patients with advanced, EGFR mutation-positive NSCLC. 1,75 This assumption is supported by the conclusion of the appraisal committee in NICE guidance 258 that ‘there was insufficient evidence to suggest a difference in clinical effectiveness between erlotinib and gefitinib’. 75 No RCTs directly comparing gefitinib and erlotinib have been identified and the results of indirect treatment comparisons vary. 75,76 Our review identified one retrospective Taiwanese study comparing gefitinib and erlotinib, which did not meet our inclusion criteria. This study included 224 patients, with known tumour EGFR mutation status, who had received TKI treatment (124 gefitinib and 100 erlotinib) but was not restricted to first-line treatment; no significant difference between the two treatments was observed for either PFs or OR rate. 77

Cost-effectiveness

A de novo probabilistic model was developed to assess the cost-effectiveness of different methods for EGFR-TK mutation testing to decide between standard chemotherapy or EGFR-TKIs for first-line treatment of patients with locally advanced or metastatic NSCLC. In order to be consistent with related assessments/appraisals, we first ensured that the results for patients with an EGFR positive mutation tumour using the Therascreen EGFR PCR Kit in the de novo model were similar to the results of these patients in the initial manufacturer’s model used in NICE Technology Appraisal 192. 51 Subsequently, the ERG amendments were incorporated and ICERs from the de novo model were compared with ICERs as reported in the final appraisal determination of STA 19251 (see Appendix 6 for results).

Test failures and costs were based on information obtained from the online survey of NHS laboratories in England and Wales. These real-life data provided an important source of information, which is likely to be representative of clinical practice.

In the assessment of economic value of different tests, a link has to be established between test accuracy, clinical value (e.g. treatment response, PFS, OS) and relative cost-effectiveness. Ideally, the performance of EGFR mutation tests would be assessed against an objective measure of the true presence/absence of a clinically relevant EGFR-TK mutation (the ‘reference standard’), and comparative effectiveness of treatment (TKI vs. chemotherapy) conditional upon the true or false presence/absence of the EGFR-TK mutation would be determined. However, each different testing method targets a different range of mutations and has different limits of detection (lowest proportion of mutation detectable in tumour cells) and the exact combination of mutation type and level that will provide optimal treatment selection remains unclear. For this reason, assessment of test performance based on comparison with a conventional ‘reference standard’ is not currently possible. In this situation, an alternative way to determine the relative value of diagnostic methods for EGFR-TK mutation testing is to use studies that report on the comparative treatment effect in patients with different EGFR mutation status (positive, negative, or unknown) as defined using different EGFR mutation tests. Thus, OR on anti-EGFR-TKIs was assumed to correlate perfectly with the ‘true’ presence/absence of the EGFR-TK mutation. The use of alternative measures of EGFR-TK mutation in the assessment of cost-effectiveness might impact the proportion of mutation positives and negatives (see Tables 15 and 16) and thus might substantially impact the assessment of cost-effectiveness (in either direction) as this is one of the key drivers of cost-effectiveness. In the absence of an objective measure of the ‘true’ presence/absence of a clinically significant EGFR-TK mutation (i.e. which mutations, present at what levels, as defined by which testing method, will result in differential treatment effects for TKIs vs. standard chemotherapy), the current cost-effectiveness assessment is, at best, an approximation of the ‘true’ cost-effectiveness of test-guided treatments.

Evidence on the comparative treatment effect in patients with different tumour EGFR mutation status (positive, negative or unknown) as defined using different tests was available for only two tests (the Therascreen EGFR PCR Kit and direct sequencing of all exon 19–21 mutations). A major assumption underpinning our analyses was that the differences in OR, PFS and OS observed in the two included studies from which these data were derived41,48,49 can be solely ascribed to differences in test performance. In practice this assumption would seem unlikely to hold true. These differences could also be caused by differences in participant characteristics, differences in the standard chemotherapy regimen or differences in treatment strategies following progression that may affect OS, all of which were apparent between these two studies.

It was not part of the scope of this assessment to update the appraisal of gefitinib for the first-line treatment of locally advanced or metastatic NSCLC (NICE Technology Appraisal 192). 1 However, the ERG’s report for NICE Technology Appraisal 19265 noted that the cost-effectiveness of the ‘EGFR mutation test + TKI treatment if positive and doublet chemotherapy if negative’ strategy compared with the ‘doublet chemotherapy without EGFR mutation testing’ strategy is conditional upon the accuracy of the mutation test used to distinguish between patients who receive TKI treatment and patients who receive doublet chemotherapy. This is a simplification of the issue because, as described previously, each EGFR mutation testing method identifies a subtly different combination of type and level of EGFR mutation, and the clinical significance of these different combinations is largely unknown. It is particularly problematic if a test defines positive mutation status for a type and/or level of mutation that is not clinically significant (associated with response to treatment with TKIs), as the patients thus ‘falsely’ identified as having mutation-positive tumours will experience a loss of survival time and quality of life due to not receiving the most effective treatment option for them, while still experiencing treatment related adverse events; the costs of treatment are also considerably increased. The effects of this might even outweigh the relative gains of TKI treatment compared with doublet chemotherapy for those patients correctly selected for TKI treatment. Therefore, the economic evaluation of TKI treatment should not be seen as an assessment of the relative value of the drug in isolation from the mutation test used to select eligible patients, but as an assessment of a specific ‘mutation test’–’treatment’ combination, which may not be valid if other methods for mutation testing are used. For this assessment, this means that the results described are partial in the sense that the ‘doublet chemotherapy without EGFR mutation testing’ strategy was not taken into account.

Clinical effectiveness

As discussed above (see Strengths and limitations of assessment, Clinical effectiveness), one key consideration when selecting an EGFR mutation testing method is the variation between tests in limit of detection (i.e. the minimum percentage of mutation in tumour cells required to produce a positive result). A lower limit of detection can enhance the ability of laboratories to produce results from poor-quality samples. Similarly, methods of specimen handling, for example laser-capture microdissection, may affect the ability of the EGFR mutation testing method to detect mutations present at a low level. However, it should not be assumed that a lower limit of detection will necessarily result in a more clinically effective test, as it is possible that TKIs may be less effective in patients with a low proportion of tumour cells harbouring mutation. Discussions with clinical experts suggest that there is ongoing uncertainty around this issue as quantitative results of EGFR mutation testing are not routinely reported. None of the studies that met the inclusion criteria for this review reported any data on variation in treatment effect with the proportion of tumour cells having EGFR mutations. A Chinese study, which did not meet our inclusion criteria, assessed tissue bank tumour samples from NSCLC patients who had been treated with gefitinib at any stage during the course of their disease. 78 This study analysed samples using both direct DNA sequencing and the Therascreen EGFR PCR Kit: samples that were positive by both methods were classified as having a high abundance of EGFR mutations; samples that were positive using the Therascreen EGFR PCR Kit and negative on direct sequencing were classified as having a low abundance of EGFR mutations; and samples that were negative on both tests were classified as wild type. The results of this study were mixed: median PFS was significantly longer in both the high abundance [11.3 months (95% CI 7.4 to 15.2 months)] and low abundance [6.9 months (95% CI 5.5 to 8.4 months)] groups compared with wild type [2.1 months (95% CI 1.0 to 3.2 months)]; however, for other outcome measures (OR rate and OS) benefits were limited to the high-abundance group. 78 It should also be noted that this study provides no information on the relative effectiveness of standard chemotherapy in these patient groups.

A further area of uncertainty concerns the clinical value of detecting rare mutations and possible resistance mutations. The majority of the evidence on the effectiveness of first-line treatment with TKIs in patients with EGFR mutation-positive NSCLC has been derived from patients with exon 19 deletions or the exon 21 mutation L858R. This is unsurprising, as these account for > 90% of all EGFR mutations. 6,7,13 The additional clinical value of using tests that target a wider range of mutations remains uncertain, as the low frequency of most EGFR mutations makes it very difficult to adequately assess treatment effects in patients with mutations other than exon 19 deletions or L858R. Some of the studies in our review that provided data on the accuracy of EGFR testing in predicting response to treatment with TKIs reported response data by individual mutation; these data appeared to indicate that there may be a less favourable response to TKIs in patients with T790M or other exon 20 mutations (see Table 8); however, these data were very limited. There are a number of registry studies that did not meet our inclusion criteria, but which have reported some information of clinical response in patients with different EGFR mutations. Murray et al. compiled a database of 202 articles, which provided data on 2,548 NSCLC patients (disease stage and previous treatment not specified) who had been treated with TKIs. 79 This study reported an OR rate of 86% for patients with a mutation in exon 19 compared with 33% for those with a mutation in exon 20; subgroup analysis indicated that a mutation in exon 20, in the absence of T790M, was associated with an OR rate of 68% (similar to that for mutations in exon 18 or 21). 79 Of the 115 different mutations for which response data were available, only 13 demonstrated PD as a response, of which eight were located in exon 20. However, as noted by the authors, some caution is required in interpreting these data, as the two most common mutations account for > 90%, with T790M occurring in only around 2% of patients. 79 An observational study conducted in 15 of 28 French National Cancer Institute laboratories identified 1048 EGFR mutations from 10,117 patients with NSCLC who were tested. 80 Of these, 108 were rare mutations (48 in exon 18 and 60 in exon 20); 36 of these patients received a TKI and were evaluable for response. The best response was progression in 18 patients, stabilisation in 11 patients and PR in seven patients. 80 REASON, a large registry study of > 4000 patients at 151 centres in Germany, aims to generate data on EGFR mutation status and clinical response to TKIs in patients with stage IIIB or stage IV data; however, to date, this study has been published only as a conference abstract, with no data for specific mutations. 81 A similar programme, EGFR FASTnet, exists in Italy, although again we have not been able to identify any publication that reports mutation-specific response data. 82,83 Both programmes are supported by AstraZeneca.

The clinical significance of rare mutations and the possible increased risk of ‘false-positives’ associated with the use of EGFR mutation tests that are able to detect very low levels of mutation were both highlighted as areas requiring further research by the European EGFR Workshop Group in a 2009 multidisciplinary consensus meeting on the implementation of EGFR mutation testing. 12

As with the issue of rare mutations, there is uncertainty regarding the clinical effectiveness of identifying EGFR mutations in non-adenocarcinoma NSCLC. The majority of the evidence on the effectiveness of first-line treatment with TKIs in patients with EGFR mutation-positive NSCLC has been derived from patients with adenocarcinomas. All but one41 of the studies included in our review included small numbers of patients with other histological diagnoses, but none reported separate data for these patients. We identified one retrospective analysis of patients with advanced NSCLC and known EGFR mutation status (determined by direct sequencing), which did not meet our inclusion criteria but which reported comparative data on 12 patients with non-adenocarcinoma and 269 with EGFR mutation-positive adenocarcinoma who were treated with TKIs. 84 OR and DC rates were lower in patients with non-adenocarcinoma than in those with adenocarcinoma (50% vs. 78% and 75% vs. 89%, respectively), and PFS was also significantly longer in the adenocarcinoma group [11.27 months (95% CI 9.87 to 12.67 months) vs. 3.67 months (95% CI 1.34 to 5.99) months]. 84 Similar results were reported for a systematic review which compared data for 33 EGFR mutation-positive non-adenocarcinoma NSCLC patients treated with gefitinib from 15 studies with adenocarcinoma patients from the same studies. 85 Although it appears that patients with non-adenocarcinoma NSCLC, which is positive for EGFR mutations, may derive less benefit from treatment with TKIs than those with adenocarcinomas, it should be noted that this question was outside the scope of our review and the studies discussed above do not provide any information on the relative effectiveness of TKIs and standard chemotherapy regimens in this group of patients.

A wide variety of EGFR mutation test methods are currently used by accredited NHS laboratories in England and Wales; however, for the majority of these methods, no studies were identified that could provide data linking the results of EGFR testing to the effectiveness treatment. Therefore, the potential clinical effects of using different EGFR tests to make decisions on first-line treatment in patients with stage IIIB or IV remains uncertain. The available data were for version 1 of the Therascreen EGFR PCR Kit and for direct sequencing methods targeting various mutations. Version 1 of the Therascreen kit is no longer being actively marketed by Qiagen and equivalent data are not available for its replacement – the Therascreen EGFR RGQ PCR Kit – or for the Therascreen EGFR Pyro Kit. However, it may be reasonable to assume equivalent diagnostic performance for all three products, as both versions of the Therascreen EGFR PCR Kit target the same mutations and the Therascreen EGFR Pyro Kit targets a similar set of mutations, with the addition of one further exon 19 deletion and the exon 21 mutation L861R and the loss of three exon 20 insertions. 86,87 All three methods have a low limit of detection (≤ 5%). 86,87 The Therascreen EGFR Pyro Kit can also produce quantitative results. 87 No data are currently available for next-generation sequencing; a method for this is currently being developed and validated by one NHS laboratory but next-generation sequencing is not yet in routine clinical use in any of NHS laboratories in England and Wales that responded to our survey.

Cost-effectiveness

Major assumptions were made in order to be able to model the relative cost-effectiveness of different EGFR mutation tests. It was assumed that the differences in relative treatment response, PFS and OS between the results of First-SIGNAL trial41 and the results of the IPASS trial48,49 were solely attributable to the different mutation tests used (the Therascreen EGFR PCR Kit and direct sequencing of all exons 19–21, respectively) to distinguish between patients whose tumours are EGFR mutation positive and those whose tumours are EGFR mutation negative (‘evidence of comparative effectiveness’ and ‘linked evidence’ analyses). As described in the previous section, it is highly questionable whether this assumption would hold. Furthermore, in order to calculate the proportion of patients with a positive and negative test result, patients who tested positive were categorised as FP if no treatment response was observed after TKI, whereas patients were categorised as TP if treatment response was observed after TKI. Similarly, patients who tested negative were categorised as FN if treatment response was observed after TKI, whereas patients were categorised as TN if no treatment response was observed after TKI. Ideally, the categorisation of true/false positives/negatives should be based on an objective measure of the true presence/absence of a clinically relevant EGFR-TK mutation. However, as previously described, the uncertainty around the exact definition of a clinically relevant mutation is such that this is not currently possible. It was also assumed that the proportion of patients with unknown mutation status relative to the number of patients for whom a tissue sample was available, as reported in the trials included in the systematic review,44,49 provides a realistic approximation of the proportion of patients with an unknown test result in clinical practice. Outcomes in patients with unknown tumour mutation status were only reported in the IPASS trial. 48,49 These results were used to model the outcomes in patients with an unknown test result for the other testing methods considered in this assessment, assuming that the OR rate, PFS and OS in patients with an unknown test result after use of the Therascreen EGFR PCR Kit, as reported in the IPASS trial,48,49 were generalisable to the direct sequencing methods. In the ‘linked evidence’ analysis, information from Jackman et al. 44 [direct sequencing or WAVE-HS for inadequate samples (< 50% of tumour cells)] and the First-SIGNAL trial41 were used to model the other direct sequencing methods if information was missing, thus assuming this information was generalisable to the other direct sequencing methods. The extent to which these results are actually generalisable to testing methods other than the Therascreen EGFR PCR Kit is unknown.

Moreover, as this model was partially based on the evidence and model structure used in the appraisal of gefitinib for the first-line treatment of locally advanced or metastatic NSCLC (NICE Technology Appraisal 192),1,51 the assumptions underlying that appraisal also apply to this assessment; for instance, assumptions regarding the applicability of the findings in the trials to the population in England and Wales.

Finally, it should be emphasised that the uncertainty resulting from the above mentioned assumptions was not parameterised in the model and is therefore not reflected in the probabilistic sensitivity analyses and hence cost-effectiveness acceptability curves.

EMBASE (OvidSP): 2000 to 2012 week 28

Searched 18 July 2012.

1. erlotinib/ or (Erlotinib or Nsc-718781 or nsc718781 or osi-774 or osi774 or r-1415 or r1415 or tarceva or cp-358774 or cp358774 or 183321-74-6 or 183319 69 9).ti,ab,ot,hw,rn. (11,968)

2. gefitinib/ or (Gefitinib or Geftinat or Geftib or iressa or zd-1839 or zd1839 or 184475-35-2).ti,ab,ot,hw,rn. (13,035)

3. or/1-2 (18,405)

4. lung non small cell cancer/ (45,170)

5. (nsclc or nsclcs).ti,ab,ot,hw. (22,339)

6. (lung$ adj3 (adeno-carcinoma$ or adenocarcinom$)).ti,ab,ot. (9347)

7. ((non-small cell or large cell) adj3 lung$).ti,ab,ot. (35,098)

8. (lclc or lclcs).ti,ab,ot,hw. (56)

9. or/4-8 (59,672)

10. Receptor, Epidermal Growth Factor/ (34,579)

11. (epidermal growth factor receptor$ or epidermis growth factor receptor$ or transforming growth factor alpha receptor$).ti,ab,ot. (24,964)

12. ((tgf-alpha or urogastrone) adj2 receptor$).ti,ab,ot. (183)

13. ((erbB1 or erbB-1 or erbB) adj1 (protein$ or receptor$)).ti,ab,ot. (1421)

14. (EGFR or EGFRTK).ti,ab,ot. (30,350)

15. EGF receptor$.ti,ab,ot. (8985)

16. (Cobas adj3 EGFR).af. (0)

17. (Cobas adj3 epidermal growth factor).ti,ab,ot. (0)

18. (thera?screen$ or therascreen$).af. (46)

19. or/10-18 (56,110)

20. 3 and 9 and 19 (4768)

21. lung non small cell cancer/di [Diagnosis] (5261)

22. diagnostic test/ (53,292)

23. diagnosis/ (875,184)

24. differential diagnosis/ (295,658)

25. laboratory diagnosis/ (40,591)

26. laboratory test/ (100,888)

27. diagnos$.ti,ab,ot. (1,925,228)

28. (test or tests or testing or tested).ti,ab,ot. (2,207,310)

29. ((lab or labs or laborator$) adj2 (procedure$ or exam$)).ti,ab,ot. (15,288)

30. or/21-29 (4,581,699)

31. 9 and 19 and 30 (2035)

32. animal/ or animal experiment/ (3,398,728)

33. (rat or rats or mouse or mice or murine or rodent or rodents or hamster or hamsters or pig or pigs or porcine or rabbit or rabbits or animal or animals or dogs or dog or cats or cow or bovine or sheep or ovine or monkey or monkeys).mp. (5,489,895)

34. or/32-33 (5,489,895)

35. exp human/ or human experiment/ (13,717,180)

36. 34 not (34 and 35) (4,418,831)

37. 20 or 31 (5626)

38. 37 not 36 (5547)

39. limit 38 to yr=“2000 -Current” (5500)

40. limit 39 to embase (4910)

41. remove duplicates from 40 (4897)

MEDLINE (OvidSP): 2000 to July 2012 week 1

Searched 18 July 2012.

1. Quinazolines/ (11,462)

2. (Erlotinib or Nsc-718781 or nsc718781 or osi-774 or osi774 or r-1415 or r1415 or tarceva or cp-358774 or cp358774 or 183321-74-6 or 183319 69 9).ti,ab,ot,hw,rn. (2563)

3. (Gefitinib or Geftinat or Geftib or iressa or zd-1839 or zd1839 or 184475-35-2).ti,ab,ot,hw,rn. (3588)

4. or/1-3 (12,590)

5. Carcinoma, Non-Small-Cell Lung/ (26,828)

6. (nsclc or nsclcs).ti,ab,ot,hw. (14,105)

7. (lung$ adj3 (adeno-carcinoma$ or adenocarcinom$)).ti,ab,ot. (6982)

8. ((non-small cell or large cell) adj3 lung$).ti,ab,ot. (24,330)

9. (lclc or lclcs).ti,ab,ot,hw. (42)

10. or/5-9 (37,809)

11. Receptor, Epidermal Growth Factor/ (25,521)

12. epidermal growth factor receptor$.ti,ab,ot. (20,251)

13. epidermis growth factor receptor$.ti,ab,ot. (0)

14. transforming growth factor alpha receptor$.ti,ab,ot. (10)

15. ((tgf-alpha or urogastrone) adj2 receptor$).ti,ab,ot. (243)

16. ((erbB1 or erbB-1 or erbB) adj1 (protein$ or receptor$)).ti,ab,ot. (1223)

17. EGFR.ti,ab,ot. (19,756)

18. EGFRTK.ti,ab,ot. (10)

19. EGF receptor$.ti,ab,ot. (8278)

20. (Cobas adj3 EGFR).af. (0)

21. (Cobas adj3 epidermal growth factor).ti,ab,ot. (0)

22. (thera?screen$ or therascreen$).af. (13)

23. or/11-22 (38,939)

24. 4 and 10 and 23 (2059)

25. Carcinoma, Non-Small-Cell Lung/di [Diagnosis] (1966)

26. “diagnostic techniques and procedures”/ or diagnostic tests, routine/ (7996)

27. clinical laboratory techniques/ or molecular diagnostic techniques/ (19,825)

28. Diagnosis/ (16,321)

29. Diagnosis, Differential/ (355,501)

30. diagnos$.ti,ab,ot. (1,397,222)

31. (test or tests or testing or tested).ti,ab,ot. (1,683,480)

32. ((lab or labs or laborator$) adj2 (procedure$ or exam$)).ti,ab,ot. (10,488)

33. or/25-32 (3,069,443)

34. 10 and 23 and 33 (887)

35. 24 or 34 (2529)

36. animals/ not (animals/ and humans/) (3,660,877)

37. 35 not 36 (2499)

38. limit 37 to yr=“2000 -Current” (2463)

39. remove duplicates from 38 (2318)

MEDLINE In-Process Citations (OvidSP): 2000 to 17 July 2012

MEDLINE Daily Update (OvidSP): 2000 to 17 July 2012

Searched 18 July 2012.

1. Quinazolines/ (31)

2. (Erlotinib or Nsc-718781 or nsc718781 or osi-774 or osi774 or r-1415 or r1415 or tarceva or cp-358774 or cp358774 or 183321-74-6 or 183319 69 9).ti,ab,ot,hw,rn. (278)

3. (Gefitinib or Geftinat or Geftib or iressa or zd-1839 or zd1839 or 184475-35-2).ti,ab,ot,hw,rn. (233)

4. or/1-3 (432)

5. Carcinoma, Non-Small-Cell Lung/ (78)

6. (nsclc or nsclcs).ti,ab,ot,hw. (1275)

7. (lung$ adj3 (adeno-carcinoma$ or adenocarcinom$)).ti,ab,ot. (468)

8. ((non-small cell or large cell) adj3 lung$).ti,ab,ot. (1905)

9. (lclc or lclcs).ti,ab,ot,hw. (6)

10. or/5-9 (2407)

11. Receptor, Epidermal Growth Factor/ (57)

12. epidermal growth factor receptor$.ti,ab,ot. (1221)

13. epidermis growth factor receptor$.ti,ab,ot. (0)

14. transforming growth factor alpha receptor$.ti,ab,ot. (0)

15. ((tgf-alpha or urogastrone) adj2 receptor$).ti,ab,ot. (5)

16. ((erbB1 or erbB-1 or erbB) adj1 (protein$ or receptor$)).ti,ab,ot. (48)

17. EGFR.ti,ab,ot. (1697)

18. EGFRTK.ti,ab,ot. (1)

19. EGF receptor$.ti,ab,ot. (195)

20. (Cobas adj3 EGFR).af. (0)

21. (Cobas adj3 epidermal growth factor).ti,ab,ot. (0)

22. (thera?screen$ or therascreen$).af. (2)

23. or/11-22 (2307)

24. 4 and 10 and 23 (163)

25. Carcinoma, Non-Small-Cell Lung/di [Diagnosis] (7)

26. “diagnostic techniques and procedures”/ or diagnostic tests, routine/ (23)

27. clinical laboratory techniques/ or molecular diagnostic techniques/ (86)

28. Diagnosis/ (1)

29. Diagnosis, Differential/ (316)

30. diagnos$.ti,ab,ot. (71,695)

31. (test or tests or testing or tested).ti,ab,ot. (101,066)

32. ((lab or labs or laborator$) adj2 (procedure$ or exam$)).ti,ab,ot. (550)

33. or/25-32 (160,664)

34. 10 and 23 and 33 (103)

35. 24 or 34 (219)

36. animals/ not (animals/ and humans/) (3555)

37. 35 not 36 (219)

38. limit 37 to yr=“2000 -Current” (219)

39. remove duplicates from 38 (215)

Cochrane Database of Systematic Reviews (CDSR) (Wiley): Issue 7, 2012

Cochrane Central Register of Controlled Trials (CENTRAL) (Wiley): Issue 7, 2012

Database of Abstracts of Reviews of Effects (DARE) (Wiley): Issue 3, 2012

Health Technology Assessment Database (HTA) (Wiley): Issue 3, 2012

Search limited to 2000 to 2012.

Searched 18 July 2012.

#1 MeSH descriptor Quinazolines, this term only (612)

#2 (Erlotinib or Nsc-718781 or nsc718781 or osi-774 or osi774 or r-1415 or r1415 or tarceva or cp-358774 or cp358774 or 183321-74-6 or 183319 69 9):ti,ab,kw (130)

#3 (Gefitinib or Geftinat or Geftib or iressa or zd-1839 or zd1839 or 184475-35-2):ti,ab,kw (171)

#4 (#1 OR #2 OR #3) (738)

#5 MeSH descriptor Carcinoma, Non-Small-Cell Lung, this term only (1952)

#6 (nsclc or nsclcs or lclc or lclcs):ti,ab (2101)

#7 (lung* NEAR/3 (adeno-carcinoma* or adenocarcinom*)):ti,ab,kw (73)

#8 ((non-small NEXT cell) NEAR/3 lung*):ti,ab,kw (3584)

#9 ((large NEXT cell) NEAR/3 lung*):ti,ab,kw (4)

#10 (#5 OR #6 OR #7 OR #8 OR #9) (3812)

#11 MeSH descriptor Receptor, Epidermal Growth Factor, this term only (264)

#12 (epidermal NEXT growth NEXT factor NEXT receptor*):ti,ab,kw (405)

#13 (epidermis NEXT growth NEXT factor NEXT receptor*):ti,ab,kw (0)

#14 (transforming NEXT growth NEXT factor NEXT alpha NEXT receptor*):ti,ab,kw (0)

#15 (tgf-alpha NEAR/2 receptor*):ti,ab,kw (1)

#16 (urogastrone NEAR/2 receptor*):ti,ab,kw (0)

#17 ((erbB1 or erbB-1 or erbB) NEAR/2 (protein* or receptor*)):ti,ab,kw (292)

#18 (EGFR or EGFRTK):ti,ab,kw (446)

#19 (EGF NEXT receptor*):ti,ab,kw (23)

#20 (Cobas NEAR/3 EGFR) (0)

#21 (Cobas NEAR/3 (epidermal NEXT growth NEXT factor)) (0)

#22 (thera-screen* or therascreen*) (0)

#23 (#11 OR #12 OR #13 OR #14 OR #15 OR #16 OR #17 OR #18 OR #19 OR #20 OR #21 OR #22) 921

#24 (#4 AND #10 AND #23) (103)

#25 MeSH descriptor Carcinoma, Non-Small-Cell Lung, this term only with qualifier: DI (72)

#26 MeSH descriptor Diagnostic Techniques and Procedures, this term only (95)

#27 MeSH descriptor Diagnostic Tests, Routine, this term only (251)

#28 MeSH descriptor Clinical Laboratory Techniques, this term only (111)

#29 MeSH descriptor Molecular Diagnostic Techniques, this term only (33)

#30 MeSH descriptor Diagnosis, this term only (73)

#31 MeSH descriptor Diagnosis, Differential, this term only (1330)

#32 diagnos*:ti,ab,kw (70,823)

#33 (test or tests or testing or tested):ti,ab,kw (127,012)

#34 ((lab or labs or laborator*) NEAR/2 (procedure* or exam*)):ti,ab,kw (605)

#35 (#25 OR #26 OR #27 OR #28 OR #29 OR #30 OR #31 OR #32 OR #33 OR #34) (174,193)

#36 (#10 AND #23 AND #35) (38)

#37 (#24 OR #36), from 2000 to 2012 (116)

CDSR search retrieved 0 references.

CENTRAL search retrieved 96 references.

DARE search retrieved 7 references.

HTA search retrieved 11 references.

PROSPERO (International Prospective Register of Systematic Reviews) (Internet): up to 19 July 2012

www.crd.york.ac.uk/prospero/

Searched 19 July 2012.

Displayed all records (n = 650) and browsed the titles for the following terms:

Latin American and Caribbean Health Sciences (LILACS): 2000 to 6 July 2012

http://regional.bvsalud.org/php/index.php?lang=en

Searched 19 July 2012.

Spanish and Portuguese translations of MeSH terms identified using the DECS (Health Sciences Descriptors) thesaurus: http://decs.bvs.br/I/homepagei.htm

Date limit applied within EndNote Library.

Clinicaltrials.gov (Internet)

http://clinicaltrials.gov/ct2/search/advanced

Limited 1 January 2000 to 19 July 2012.

Searched 19 July 2012.

Advanced search option – search terms box:

metaRegister of Controlled Trials (mRCT) (Internet)

www.controlled-trials.com/

Up to 30 August 2012.

Searched 30 August 2012.

WHO International Clinical Trials Registry Platform (ICTRP) (Internet)

www.who.int/ictrp/en/

Limited to 1 January 2000 to 30 August 2012.

Searched 30 August 2012.

Advanced search option:

BIOSIS Previews (Web of Knowledge): 2000 to 24 August 2012

Searched 30 July 2012.

Advanced search (Lemmatization off):

# 30 44 #28 not #29

Databases=BIOSIS Previews Timespan=2000–2012

# 29 5,773,678 TS=(cat or cats or dog or dogs or animal or animals or rat or rats or hamster or hamster or feline or ovine or canine or bovine or sheep OR macaque* OR monkey*)

# 28 1954 #21 or #27

# 27 889 #7 and #20 and #26

# 26 1933,480 #22 or #23 or #24 or #25

# 25 45,681 TS=((laborator* NEAR procedure*) or (laborator* NEAR exam*))

# 24 99 TS=((labs NEAR procedure*) or (labs NEAR exam*))

# 23 685 TS=((lab NEAR procedure*) or (lab NEAR exam*))

# 22 1,913,268 TS=(diagnos* OR test or tests or testing or tested)

# 21 1411 #3 and #7 and #20

# 20 34,254 #8 or #9 or #10 or #11 or #12 or #13 or #14 or #15 or #16 or #17 or #18 or #19

# 19 13 TS=(thera-screen* or therascreen*)

# 18 0 TS=(Cobas NEAR epidermal NEAR growth NEAR factor)

# 17 0 TS=(Cobas NEAR EGFR)

# 16 7196 TS=(EGF NEAR receptor*)

# 15 20,895 TS=(EGFR or EGFRTK)

# 14 3256 TS=((erbB1 or erbB-1 or erbB) NEAR (protein* or receptor*))

# 13 0 TS=(urogastrone NEAR receptor*)

# 12 700 TS=(tgf-alpha NEAR receptor*)

# 11 1 TS=(transform NEAR growth NEAR factor NEAR alpha NEAR receptor*)

# 10 1962 TS=(transforming NEAR growth NEAR factor NEAR alpha NEAR receptor*)

# 9 62 TS=(epidermis NEAR growth NEAR factor NEAR receptor*)

# 8 21,660 TS=(epidermal NEAR growth NEAR factor NEAR receptor*)

# 7 27,387 #4 or #5 or #6

# 6 19,225 TS=((non-small NEAR cell NEAR lung*) or (large NEAR cell NEAR lung*))

# 5 10,230 TS=((lung* NEAR adeno-carcinoma*) OR (lung NEAR adenocarcinom*))

# 4 8560 TS=(nsclc or nsclcs or lclc or lclcs)

# 3 4669 #1 or #2

# 2 3230 TS=(Gefitinib or Geftinat or Geftib or iressa or zd-1839 or zd1839 or 184475-35-2)

# 1 2401 TS=(Erlotinib or Nsc-718781 or nsc718781 or osi-774 or osi774 or r-1415 or r1415 or tarceva or cp-358774 or cp358774 or 183321-74-6 or 183319-69-9)

American Society of Clinical Oncology (ASCO) Conference Proceedings: 2007 to 2012

www.asco.org/ASCOv2/Meetings/Abstracts

Searched 26 October 2012.

Searched 2007–12 Annual Meetings:

ESMO Conference Proceedings (European Society of Medical Oncology): 2007 to 2012

www.esmo.org/no_cache/education/abstracts-and-virtual-meetings.html

Searched 31 October 2012.

• 2008 33rd ESMO Congress, Stockholm: http://annonc.oxfordjournals.org/content/vol19/suppl_8/

• 2009 ECCO 15 and 34th ESMO Multidisciplinary Congress: www.ejcancer.info

• 2010 35th ESMO Congress, Milan: http://annonc.oxfordjournals.org/content/21/suppl_8

• 2011 ECCO 16 and 36th ESMO Multidisciplinary Congress, Brussels: www.ejcancer.info/issues

• 2012 37th ESMO Congress, Vienna - http://annonc.oxfordjournals.org/content/23/suppl_9

World Conference on Lung Cancer (International Association for the Study of Lung Cancer): 2007 to 2012

http://iaslc.org/

Searched 30 October 2012.

• 14th World Conference on Lung Cancer: http://journals.lww.com/jto/toc/2011/06001

• 13th World Conference on Lung Cancer: http://journals.lww.com/jto/Citation/2009/09001/Abstracts.1.aspx

• 12th World Conference on Lung Cancer: http://journals.lww.com/jto/toc/2007/08001

PubMed Related Citations search undertaken for included studies

Results sorted by Link Ranking.

www.ncbi.nlm.nih.gov/pubmed/

Searched 24 October 2012.

Of 30 included studies, 12 references were indexed on PubMed. For each reference, the first 20 related citations were retrieved by carrying out a Related Citations search using PubMed’s similarity matching algorithm. These records were downloaded for screening. All related citations were checked against the EndNote Library to remove duplicate, and only new unique references were imported and screened.

Review of cost-effectiveness literature

NHS Economic Evaluation Database (NHS EED) (Wiley): 2000 to 2012, Issue 3

Searched 30 August 2012.

#1 MeSH descriptor Quinazolines, this term only (613)

#2 (Erlotinib or Nsc-718781 or nsc718781 or osi-774 or osi774 or r-1415 or r1415 or tarceva or cp-358774 or cp358774 or 183321-74-6 or 183319 69 9):ti,ab,kw (131)

#3 (Gefitinib or Geftinat or Geftib or iressa or zd-1839 or zd1839 or 184475-35-2):ti,ab,kw (171)

#4 (#1 OR #2 OR #3) (739)

#5 MeSH descriptor Carcinoma, Non-Small-Cell Lung, this term only (1953)

#6 (nsclc or nsclcs or lclc or lclcs):ti,ab (2101)

#7 (lung* NEAR/3 (adeno-carcinoma* or adenocarcinom*)):ti,ab,kw (73)

#8 ((non-small NEXT cell) NEAR/3 lung*):ti,ab,kw (3585)

#9 ((large NEXT cell) NEAR/3 lung*):ti,ab,kw (4)

#10 (#5 OR #6 OR #7 OR #8 OR #9) (3813)

#11 MeSH descriptor Receptor, Epidermal Growth Factor, this term only (265)

#12 (epidermal NEXT growth NEXT factor NEXT receptor*):ti,ab,kw (406)

#13 (epidermis NEXT growth NEXT factor NEXT receptor*):ti,ab,kw (0)

#14 (transforming NEXT growth NEXT factor NEXT alpha NEXT receptor*):ti,ab,kw (0)

#15 (tgf-alpha NEAR/2 receptor*):ti,ab,kw (1)

#16 ( urogastrone NEAR/2 receptor*):ti,ab,kw (0)

#17 ((erbB1 or erbB-1 or erbB) NEAR/2 (protein* or receptor*)):ti,ab,kw (292)

#18 (EGFR or EGFRTK):ti,ab,kw (449)

#19 (EGF NEXT receptor*):ti,ab,kw (23)

#20 (Cobas NEAR/3 EGFR) (0)

#21 (Cobas NEAR/3 (epidermal NEXT growth NEXT factor)) (0)

#22 (thera-screen* or therascreen*) (0)

#23 (#11 OR #12 OR #13 OR #14 OR #15 OR #16 OR #17 OR #18 OR #19 OR #20 OR #21 OR #22) (924)

#24 (#4 OR #23) (1476)

#25 (#10 AND #24), from 2000 to 2012 (8)

EMBASE (OvidSP): 2000 to 2012 week 28

Searched 30 August 2012.

1. health-economics/ (31,839)

2. exp economic-evaluation/ (188,273)

3. exp health-care-cost/ (180,330)

4. exp pharmacoeconomics/ (156,985)

5. or/1-4 (433,309)

6. (econom$ or cost or costs or costly or costing or price or prices or pricing or pharmacoeconomic$).ti,ab. (521,624)

7. (expenditure$ not energy).ti,ab. (20,859)

8. (value adj2 money).ti,ab. (1141)

9. budget$.ti,ab. (21,476)

10. or/6-9 (543,342)

11. 5 or 10 (796,287)

12. letter.pt. (796,544)

13. editorial.pt. (414,244)

14. note.pt. (527,749)

15. or/12-14 (1,738,537)

16. 11 not 15 (716,763)

17. (metabolic adj cost).ti,ab. (768)

18. ((energy or oxygen) adj cost).ti,ab. (2933)

19. ((energy or oxygen) adj expenditure).ti,ab. (17,921)

20. or/17-19 (20,863)

21. 16 not 20 (712,137)

22. exp animal/ (1,796,019)

23. exp animal-experiment/ (1,636,900)

24. nonhuman/ (3,899,172)

25. (rat or rats or mouse or mice or hamster or hamsters or animal or animals or dog or dogs or cat or cats or bovine or sheep).ti,ab,sh. (4,729,791)

26. or/22-25 (6,695,652)

27. exp human/ (13,830,628)

28. exp human-experiment/ (303,941)

29. 27 or 28 (13,832,063)

30. 26 not (26 and 29) (5,273,705)

31. 21 not 30 (661,610)

32. erlotinib/ or (Erlotinib or Nsc-718781 or nsc718781 or osi-774 or osi774 or r-1415 or r1415 or tarceva or cp-358774 or cp358774 or 183321-74-6 or 183319 69 9).ti,ab,ot,hw,rn. (12,196)

33. gefitinib/ or (Gefitinib or Geftinat or Geftib or iressa or zd-1839 or zd1839 or 184475-35-2).ti,ab,ot,hw,rn. (13,183)

34. or/32-33 (18,694)

35. lung non small cell cancer/ (45,891)

36. (nsclc or nsclcs).ti,ab,ot,hw. (22,788)

37. (lung$ adj3 (adeno-carcinoma$ or adenocarcinom$)).ti,ab,ot. (9502)

38. ((non-small cell or large cell) adj3 lung$).ti,ab,ot. (35,647)

39. (lclc or lclcs).ti,ab,ot,hw. (56)

40. or/35-39 (60,634)

41. Receptor, Epidermal Growth Factor/ (35,039)

42. (epidermal growth factor receptor$ or epidermis growth factor receptor$ or transforming growth factor alpha receptor$).ti,ab,ot. (25,326)

43. ((tgf-alpha or urogastrone) adj2 receptor$).ti,ab,ot. (183)

44. ((erbB1 or erbB-1 or erbB) adj1 (protein$ or receptor$)).ti,ab,ot. (1443)

45. (EGFR or EGFRTK).ti,ab,ot. (31,087)

46. EGF receptor$.ti,ab,ot. (9045)

47. (Cobas adj3 EGFR).af. (0)

48. (Cobas adj3 epidermal growth factor).ti,ab,ot. (0)

49. (thera?screen$ or therascreen$).af. (50)

50. or/41-49 (57,139)

51. 34 or 50 (66,682)

52. 31 and 40 and 51 (743)

53. limit 52 to yr=“2000 -Current” (743)

54. remove duplicates from 53 (736)

55. limit 54 to embase (703)

Costs filter:

Centre for Reviews and Dissemination. NHS EED Economics Filter: EMBASE (Ovid) weekly search [Internet]. York: Centre for Reviews and Dissemination; 2010 (cited 17 March 2011). Available from: www.york.ac.uk/inst/crd/intertasc/nhs_eed_strategies.html

MEDLINE (OvidSP): 2000 to August 2012 week 4

Searched 30 August 2012:

1. economics/ (26,369)

2. exp “costs and cost analysis”/ (167,172)

3. economics, dental/ (1844)

4. exp “economics, hospital”/ (18,137)

5. economics, medical/ (8482)

6. economics, nursing/ (3868)

7. economics, pharmaceutical/ (2362)

8. (economic$ or cost or costs or costly or costing or price or prices or pricing or pharmacoeconomic$).ti,ab. (369,952)

9. (expenditure$ not energy).ti,ab. (15,358)

10. (value adj1 money).ti,ab. (18)

11. budget$.ti,ab. (15,574)

12. or/1-11 (487,655)

13. ((energy or oxygen) adj cost).ti,ab. (2460)

14. (metabolic adj cost).ti,ab. (652)

15. ((energy or oxygen) adj expenditure).ti,ab. (14,385)

16. or/13-15 (16,851)

17. 12 not 16 (483,875)

18. letter.pt. (757,777)

19. editorial.pt. (305,167)

20. historical article.pt. (285,776)

21. or/18-20 (1,335,091)

22. 17 not 21 (457,810)

23. animals/ not (animals/ and humans/) (3,680,958)

24. 22 not 23 (430,529)

25. Quinazolines/ (11,624)

26. (Erlotinib or Nsc-718781 or nsc718781 or osi-774 or osi774 or r-1415 or r1415 or tarceva or cp-358774 or cp358774 or 183321-74-6 or 183319 69 9).ti,ab,ot,hw,rn. (2631)

27. (Gefitinib or Geftinat or Geftib or iressa or zd-1839 or zd1839 or 184475-35-2).ti,ab,ot,hw,rn. (3648)

28. or/25-27 (12,777)

29. Carcinoma, Non-Small-Cell Lung/ (27,182)

30. (nsclc or nsclcs).ti,ab,ot,hw. (14,335)

31. (lung$ adj3 (adeno-carcinoma$ or adenocarcinom$)).ti,ab,ot. (7087)

32. ((non-small cell or large cell) adj3 lung$).ti,ab,ot. (24,664)

33. (lclc or lclcs).ti,ab,ot,hw. (42)

34. or/29-33 (38,312)

35. Receptor, Epidermal Growth Factor/ (25,843)

36. epidermal growth factor receptor$.ti,ab,ot. (20,568)

37. epidermis growth factor receptor$.ti,ab,ot. (0)

38. transforming growth factor alpha receptor$.ti,ab,ot. (10)

39. ((tgf-alpha or urogastrone) adj2 receptor$).ti,ab,ot. (243)

40. ((erbB1 or erbB-1 or erbB) adj1 (protein$ or receptor$)).ti,ab,ot. (1239)

41. EGFR.ti,ab,ot. (20,234)

42. EGFRTK.ti,ab,ot. (10)

43. EGF receptor$.ti,ab,ot. (8326)

44. (Cobas adj3 EGFR).af. (0)

45. (Cobas adj3 epidermal growth factor).ti,ab,ot. (0)

46. (thera?screen$ or therascreen$).af. (14)

47. or/35-46 (39,620)

48. 28 or 47 (47,729)

49. 24 and 34 and 48 (90)

50. limit 49 to yr=“2000 -Current” (90)

51. remove duplicates from 50 (87)

Costs filter:

Centre for Reviews and Dissemination. NHS EED Economics Filter: MEDLINE (Ovid) monthly search [Internet]. York: Centre for Reviews and Dissemination; 2010 (cited 28 September 2010]). Available from: www.york.ac.uk/inst/crd/intertasc/nhs_eed_strategies.html

MEDLINE In-Process Citations (OvidSP): 2000 to 29 August 2012

MEDLINE Daily Update (OvidSP): 2000 to 29 August 2012

Searched 30 August 2012:

1. economics/ (1)

2. exp “costs and cost analysis”/ (111)

3. economics, dental/ (0)

4. exp “economics, hospital”/ (5)

5. economics, medical/ (1)

6. economics, nursing/ (0)

7. economics, pharmaceutical/ (0)

8. (economic$ or cost or costs or costly or costing or price or prices or pricing or pharmacoeconomic$).ti,ab. (29,272)

9. (expenditure$ not energy).ti,ab. (821)

10. (value adj1 money).ti,ab. (2)

11. budget$.ti,ab. (1501)

12. or/1-11 (30,841)

13. ((energy or oxygen) adj cost).ti,ab. (155)

14. (metabolic adj cost).ti,ab. (53)

15. ((energy or oxygen) adj expenditure).ti,ab. (652)

16. or/13-15 (838)

17. 12 not 16 (30,593)

18. letter.pt. (17,056)

19. editorial.pt. (10,671)

20. historical article.pt. (96)

21. or/18-20 (27,818)

22. 17 not 21 (30,243)

23. animals/ not (animals/ and humans/) (1498)

24. 22 not 23 (30,207)

25. Quinazolines/ (16)

26. (Erlotinib or Nsc-718781 or nsc718781 or osi-774 or osi774 or r-1415 or r1415 or tarceva or cp-358774 or cp358774 or 183321-74-6 or 183319 69 9).ti,ab,ot,hw,rn. (259)

27. (Gefitinib or Geftinat or Geftib or iressa or zd-1839 or zd1839 or 184475-35-2).ti,ab,ot,hw,rn. (223)

28. or/25-27 (402)

29. Carcinoma, Non-Small-Cell Lung/ (30)

30. (nsclc or nsclcs).ti,ab,ot,hw. (1282)

31. (lung$ adj3 (adeno-carcinoma$ or adenocarcinom$)).ti,ab,ot. (455)

32. ((non-small cell or large cell) adj3 lung$).ti,ab,ot. (1908)

33. (lclc or lclcs).ti,ab,ot,hw. (6)

34. or/29-33 (2376)

35. Receptor, Epidermal Growth Factor/ (23)

36. epidermal growth factor receptor$.ti,ab,ot. (1197)

37. epidermis growth factor receptor$.ti,ab,ot. (0)

38. transforming growth factor alpha receptor$.ti,ab,ot. (0)

39. ((tgf-alpha or urogastrone) adj2 receptor$).ti,ab,ot. (6)

40. ((erbB1 or erbB-1 or erbB) adj1 (protein$ or receptor$)).ti,ab,ot. (44)

41. EGFR.ti,ab,ot. (1666)

42. EGFRTK.ti,ab,ot. (1)

43. EGF receptor$.ti,ab,ot. (189)

44. (Cobas adj3 EGFR).af. (0)

45. (Cobas adj3 epidermal growth factor).ti,ab,ot. (0)

46. (thera?screen$ or therascreen$).af. (2)

47. or/35-46 (2255)

48. 28 or 47 (2406)

49. 24 and 34 and 48 (12)

50. limit 49 to yr=“2000 -Current” (12)

51. remove duplicates from 50 (12)

Costs filter:

Centre for Reviews and Dissemination. NHS EED Economics Filter: MEDLINE (Ovid) monthly search [Internet]. York: Centre for Reviews and Dissemination; 2010 [cited 28.9.10]. Available from: www.york.ac.uk/inst/crd/intertasc/nhs_eed_strategies.html

Health Economic Evaluation Database (HEED) (Internet): up to 30 August 2012

http://onlinelibrary.wiley.com/book/10.1002/9780470510933

Searched 30 August 2012:

Compound search, (all data), unable to limit by date:

Erlotinib OR Nsc-718781 OR nsc718781 OR osi-774 OR osi774 OR r-1415 OR r1415 OR tarceva OR cp-358774 OR cp358774 OR 183321-74-6 OR 183319 69 9 OR Gefitinib OR Geftinat OR Geftib OR iressa OR zd-1839 OR zd1839 OR 184475-35-2

AND

lung* OR NSCLC OR NSCLCS OR LCLC OR LCLCS

N=41

(Therascreen OR Thera-screen OR Cobas OR EGF OR EGFR OR EGFRTK OR TGF OR epidermal OR erbb OR ERBB1 OR urogastrone)

AND

(lung* OR NSCLC OR NSCLCS OR LCLC OR LCLCS)

N=8

HEED search retrieved 49 records.

Science Citation Index (SCI) (Web of Knowledge): 2000 to 29 August 2012

Search limited to 2000 to 2012.

Searched 30 August 2012.

Advanced search (Lemmatization off):

# 32 146 #11 and #15 and #31

# 31 40,025 #30 OR #29 OR #28

# 30 6378 TS=(Gefitinib or Geftinat or Geftib or iressa or zd-1839 or zd1839 or 184475-35-2)

# 29 3959 TS=(Erlotinib or Nsc-718781 or nsc718781 or osi-774 or osi774 or r-1415 or r1415 or tarceva or cp-358774 or cp358774 or 183321-74-6 or 183319-69-9)

# 28 36,741 #16 or #17 or #18 or #19 or #20 or #21 or #22 or #23 or #24 or #25 or #26 or #27

# 27 13 TS=(thera-screen* or therascreen*)

# 26 0 TS=(Cobas NEAR epidermal NEAR growth NEAR factor)

# 25 0 TS=(Cobas NEAR EGFR)

# 24 8720 TS=(EGF NEAR receptor*)

# 23 21,367 TS=(EGFR or EGFRTK)

# 22 3781 TS=((erbB1 or erbB-1 or erbB) NEAR (protein* or receptor*))

# 21 13 TS=(urogastrone NEAR receptor*)

# 20 702 TS=(tgf-alpha NEAR receptor*)

# 19 0 TS=(transform NEAR growth NEAR factor NEAR alpha NEAR receptor*)

# 18 819 TS=(transforming NEAR growth NEAR factor NEAR alpha NEAR receptor*)

# 17 68 TS=(epidermis NEAR growth NEAR factor NEAR receptor*)

# 16 20,767 TS=(epidermal NEAR growth NEAR factor NEAR receptor*)

# 15 34,161 #12 or #13 or #14

# 14 24,966 TS=((non-small NEAR cell NEAR lung*) or (large NEAR cell NEAR lung*))

# 13 8029 TS=((lung* NEAR adeno-carcinoma*) OR (lung NEAR adenocarcinom*))

# 12 15,691 TS=(nsclc or nsclcs or lclc or lclcs)

# 11 484,626 #9 not #10

# 10 1,160,972 TS=(cat or cats or dog or dogs or animal or animals or rat or rats or hamster or hamster or feline or ovine or canine or bovine or sheep OR macaque* OR monkey*)

# 9 508,156 #4 not #8

# 8 26,623 #5 or #6 or #7

# 7 14,802 TS=((energy or oxygen) NEAR expenditure)

# 6 1295 TS=(metabolic NEAR cost)

# 5 11,720 TS=((energy or oxygen) NEAR cost)

# 4 521,849 #1 or #2 or #3

# 3 796 TS=(value NEAR money)

# 2 10,358 TS=(expenditure* not energy)

# 1 517,471 TS=(economic* or cost or costs or costly or costing or price or prices or pricing or pharmacoeconomic* or budget*)

EMBASE (OvidSP): January 2009 to August 2012 week 34

Searched 30 August 2012:

1. Socioeconomics/ (102,445)

2. Cost benefit analysis/ (61,757)

3. Cost-effectiveness analysis/ (82,283)

4. Cost of illness/ (13,206)

5. Cost control/ (42,633)

6. Economic aspect/ (99,388)

7. Financial management/ (96,915)

8. Health care cost/ (111,972)

9. Health care financing/ (10,847)

10. Health economics/ (31,839)

11. Hospital cost/ (12,121)

12. (fiscal or financial or finance or funding).tw. (88,152)

13. Cost minimization analysis/ (2109)

14. (cost adj estimate$).mp. (1709)

15. (cost adj variable$).mp. (135)

16. (unit adj cost$).mp. (1997)

17. or/1-16 (603,136)

18. Carboplatin/ or carboplatin.mp. or Paraplatin.mp. (38,879)

19. Cisplatin/ or Cisplatin.mp. or Platinol.mp. (115,027)

20. Paclitaxel/ or Paclitaxel.mp. or Taxol.mp. (58,541)

21. Topotecan/ or Topotecan.mp. or Hycamtin.mp. (7943)

22. (irinotecan or Campto).mp. (21,091)

23. (docetax?l or Taxotere).mp. (27,947)

24. (vinorelbine or Navelbine).mp. (11,861)

25. (gemcitabine or Gemzar).mp. (27,119)

26. (zactima or ZD6474).mp. (615)

27. (bevacizumab or Avastin).mp. (23,392)

28. (pemetrexed or Alimta).mp. (4688)

29. (erlotinib or Tarceva).mp. (12,217)

30. (bortezomib or Velcade).mp. (11,513)

31. (vinflunine or Javlor).mp. (456)

32. (cetuximab or Erbitux).mp. (12,952)

33. (gefitinib or Iressa).mp. (13,204)

34. Vatalanib.mp. (2098)

35. Panitumumab.mp. (3458)

36. platinum compounds/ or platinum.mp. (40,273)

37. Taxoids/ or taxoid$.mp. (3083)

38. exp antineoplastic protocols/ (61,147)

39. Antineoplastic Agent/ (204,229)

40. Angiogenesis Inhibitor/ (12,049)

41. Antimetabolite/ (5559)

42. antineoplastic agents, alkylating/ (12,712)

43. antineoplastic agents, phytogenic/ (204,229)

44. or/18-43 (479,713)

45. Lung non Small Cell Cancer/ (45,891)

46. (non small cell or non-small-cell or nonsmall cell or nsclc).mp. (53,758)

47. (lung$ or pulmon$).mp. (1,175,884)

48. (cancer or tumor$ or tumour$ or carcino$ or blastom$ or squamous or neoplas$ or sarcom$ or lymphom$ or adenocarcinom$).mp. (3,339,510)

49. 46 and 47 and 48 (53,042)

50. 45 or 49 (53,042)

51. 17 and 44 and 50 (861)

52. limit 51 to yr=“2006 -Current” (586)

53. (2009$ or 201$).em. (4,309,768)

54. 52 and 53 (419)

55. remove duplicates from 54 (416)

56. limit 55 to embase (405)

Update of search strategy from appendix 10.4:

AstraZeneca UK Ltd. Single Technology Appraisal (STA) for Gefitinib for the first line treatment of locally advanced or metastatic non-small lung cancer (submission to National Institute for Health and Clinical Excellence) [Word document provided by AstraZeneca]. Luton, UK: AstraZeneca UK Ltd, 2010 (cited 18 July 2012). 233pp.

MEDLINE (OvidSP): 2009 to August 2012 week 4

Searched 30 August 2012:

1. Economics/ (26,369)

2. “costs and cost analysis”/ (40,051)

3. Cost allocation/ (1921)

4. Cost-benefit analysis/ (54,902)

5. Cost control/ (19,311)

6. Cost savings/ (7775)

7. Cost of illness/ (15,410)

8. Cost sharing/ (1769)

9. “deductibles and coinsurance”/ (1348)

10. Medical savings accounts/ (462)

11. Health care costs/ (23,671)

12. Direct service costs/ (974)

13. Drug costs/ (11,210)

14. Employer health costs/ (1044)

15. Hospital costs/ (6965)

16. Health expenditures/ (12,574)

17. Capital expenditures/ (1914)

18. Value of life/ (5232)

19. exp economics, hospital/ (18,137)

20. exp economics, medical/ (13,308)

21. Economics, nursing/ (3868)

22. Economics, pharmaceutical/ (2362)

23. exp “fees and charges”/ (26,011)

24. exp budgets/ (11,515)

25. (low adj cost).mp. (16,966)

26. (high adj cost).mp. (6653)

27. (health?care adj cost$).mp. (3110)

28. (fiscal or funding or financial or finance).tw. (66,638)

29. (cost adj estimate$).mp. (1193)

30. (cost adj variable).mp. (27)

31. (unit adj cost$).mp. (1276)

32. (economic$ or pharmacoeconomic$ or price$ or pricing).tw. (141,586)

33. or/1-32 (403,794)

34. Carboplatin/ or carboplatin.mp. or Paraplatin.mp. (11,002)

35. Cisplatin/ or Cisplatin.mp. or Platinol.mp. (47,742)

36. Paclitaxel/ or Paclitaxel.mp. or Taxol.mp. (22,534)

37. Topotecan/ or Topotecan.mp. or Hycamtin.mp. (2326)

38. (irinotecan or Campto).mp. (6288)

39. (docetax?l or Taxotere).mp. (7860)

40. (vinorelbine or Navelbine).mp. (2936)

41. (gemcitabine or Gemzar).mp. (8196)

42. (zactima or ZD6474).mp. (170)

43. (bevacizumab or Avastin).mp. (6308)

44. (pemetrexed or Alimta).mp. (1286)

45. (erlotinib or Tarceva).mp. (2611)

46. (bortezomib or Velcade).mp. (3482)

47. (vinflunine or Javlor).mp. (148)

48. (cetuximab or Erbitux).mp. (2956)

49. (gefitinib or Iressa).mp. (3609)

50. Vatalanib.mp. (249)

51. Panitumumab.mp. (586)

52. platinum compounds/ or platinum.mp. (22,109)

53. Taxoids/ or taxoid$.mp. (7866)

54. exp antineoplastic protocols/ (95,249)

55. Antineoplastic Agents/ (171,356)

56. Angiogenesis Inhibitors/ (13,184)

57. Antimetabolites/ (7082)

58. antineoplastic agents, alkylating/ (7002)

59. antineoplastic agents, phytogenic/ (22,154)

60. or/34-59 (336,727)

61. lung neoplasms/ (148,687)

62. carcinoma, non-small-cell lung/ (27,182)

63. (non small cell or non-small-cell or nonsmall cell or nsclc).mp. (32,947)

64. (lung$ or pulmon$).mp. (849,108)

65. (cancer or tumor$ or tumour$ or carcino$ or blastom$ or squamous or neoplas$ or sarcom$ or lymphom$ or adenocarcinom$).mp. (2,650,656)

66. 63 and 64 and 65 (32,779)

67. 61 or 62 or 66 (151,633)

68. 33 and 60 and 67 (367)

69. limit 68 to yr=“2006 -Current” (204)

70. (2009$ or 201$).ed. (2,869,749)

71. 69 and 70 (142)

72. remove duplicates from 71 (129)

Update of search strategy from appendix 10.4:

AstraZeneca UK Ltd. Single Technology Appraisal (STA) for Gefitinib for the first line treatment of locally advanced or metastatic non-small lung cancer (submission to National Institute for Health and Clinical Excellence) [Word document provided by AstraZeneca]. Luton, UK: AstraZeneca UK Ltd, 2010 (cited 18 July 2012). 233pp.

MEDLINE In-Process Citations (OvidSP): 2009 to 29 August 2012

MEDLINE Daily Update (OvidSP): 2009 to 29 August 2012

Searched 30 August 2012.

1. Economics/ (1)

2. “costs and cost analysis”/ (14)

3. Cost allocation/ (0)

4. Cost-benefit analysis/ (32)

5. Cost control/ (3)

6. Cost savings/ (7)

7. Cost of illness/ (11)

8. Cost sharing/ (2)

9. “deductibles and coinsurance”/ (0)

10. Medical savings accounts/ (1)

11. Health care costs/ (38)

12. Direct service costs/ (2)

13. Drug costs/ (17)

14. Employer health costs/ (0)

15. Hospital costs/ (3)

16. Health expenditures/ (11)

17. Capital expenditures/ (2)

18. Value of life/ (0)

19. exp economics, hospital/ (5)

20. exp economics, medical/ (2)

21. Economics, nursing/ (0)

22. Economics, pharmaceutical/ (0)

23. exp “fees and charges”/ (7)

24. exp budgets/ (4)

25. (low adj cost).mp. (2925)

26. (high adj cost).mp. (456)

27. (health?care adj cost$).mp. (296)

28. (fiscal or funding or financial or finance).tw. (4322)

29. (cost adj estimate$).mp. (64)

30. (cost adj variable).mp. (3)

31. (unit adj cost$).mp. (75)

32. (economic$ or pharmacoeconomic$ or price$ or pricing).tw. (10,972)

33. or/1-32 (18,244)

34. Carboplatin/ or carboplatin.mp. or Paraplatin.mp. (451)

35. Cisplatin/ or Cisplatin.mp. or Platinol.mp. (1599)

36. Paclitaxel/ or Paclitaxel.mp. or Taxol.mp. (927)

37. Topotecan/ or Topotecan.mp. or Hycamtin.mp. (71)

38. (irinotecan or Campto).mp. (297)

39. (docetax?l or Taxotere).mp. (501)

40. (vinorelbine or Navelbine).mp. (127)

41. (gemcitabine or Gemzar).mp. (524)

42. (zactima or ZD6474).mp. (10)

43. (bevacizumab or Avastin).mp. (695)

44. (pemetrexed or Alimta).mp. (96)

45. (erlotinib or Tarceva).mp. (257)

46. (bortezomib or Velcade).mp. (313)

47. (vinflunine or Javlor).mp. (12)

48. (cetuximab or Erbitux).mp. (245)

49. (gefitinib or Iressa).mp. (219)

50. Vatalanib.mp. (5)

51. Panitumumab.mp. (57)

52. platinum compounds/ or platinum.mp. (4025)

53. Taxoids/ or taxoid$.mp. (27)

54. exp antineoplastic protocols/ (47)

55. Antineoplastic Agents/ (185)

56. Angiogenesis Inhibitors/ (18)

57. Antimetabolites/ (2)

58. antineoplastic agents, alkylating/ (4)

59. antineoplastic agents, phytogenic/ (15)

60. or/34-59 (8676)

61. lung neoplasms/ (77)

62. carcinoma, non-small-cell lung/ (30)

63. (non small cell or non-small-cell or nonsmall cell or nsclc).mp. (2057)

64. (lung$ or pulmon$).mp. (24,317)

65. (cancer or tumor$ or tumour$ or carcino$ or blastom$ or squamous or neoplas$ or sarcom$ or lymphom$ or adenocarcinom$).mp. (86,846)

66. 63 and 64 and 65 (2012)

67. 61 or 62 or 66 (2060)

68. 33 and 60 and 67 (13)

69. limit 68 to yr=“2006 -Current” (13)

70. (2009$ or 201$).ed. (556,714)

71. 69 and 70 (4)

72. remove duplicates from 71 (4)

Update of search strategy from appendix 10.4:

AstraZeneca UK Ltd. Single Technology Appraisal (STA) for Gefitinib for the first line treatment of locally advanced or metastatic non-small lung cancer (submission to National Institute for Health and Clinical Excellence) [Word document provided by AstraZeneca]. Luton, UK: AstraZeneca UK Ltd, 2010 (cited 18 July 2012). 233pp.

NHS Economic Evaluation Database (NHS EED) (Internet): 1 January 2009 to 30 August 2012

www.york.ac.uk/inst/crd/index_databases.htm

Searched 23 August 2012.

1. (carboplatin or Paraplatin) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (10)

2. (Cisplatin or Platinol) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (19)

3. (Paclitaxel or Taxol) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (32)

4. (Topotecan or Hycamtin) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (2)

5. ((irinotecan or Campto)) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (8)

6. ((docetaxal or docetaxel or Taxotere)) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (37)

7. ((vinorelbine or Navelbine)) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (11)

8. ((gemcitabine or Gemzar)) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (8)

9. ((zactima or ZD6474)) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (0)

10. ((bevacizumab or Avastin)) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (18)

11. ((pemetrexed or Alimta)) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (14)

12. ((erlotinib or Tarceva)) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (9)

13. ((bortezomib or Velcade)) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (3)

14. ((vinflunine or Javlor)) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (0)

15. ((cetuximab or Erbitux)) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (8)

16. ((gefitinib or Iressa)) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (3)

17. (Vatalanib) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (0)

18. (Panitumumab) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (1)

19. (Advanced non-small cell lung cancer) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (10)

20. (NSCLC) IN NHSEED WHERE PD FROM 01/01/2009 TO 30/08/2012 (9)

21. #1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9 OR #10 OR #11 OR #12 OR #13 OR #14 OR #15 OR #16 OR #17 OR #18 (104)

22. #19 OR #20 (16)

23. #21 AND #22 (14)

Update of search strategy from appendix 10.4:

AstraZeneca UK Ltd. Single Technology Appraisal (STA) for Gefitinib for the first line treatment of locally advanced or metastatic non-small lung cancer (submission to National Institute for Health and Clinical Excellence) [Word document provided by AstraZeneca]. Luton, UK: AstraZeneca UK Ltd, 2010 (cited 18 July 2012). 233pp.

CINAHL (EBSCOhost): January 2009 to 24 August 2012

Searched 30 August 2012.

1. S1 (MH “Financial Management”) OR (MH “Financial Support”) OR (MH “Financing, Organized”) OR (MH “Business”) (17,359)

2. S2 (MH “Economics”) (5503)

3. S3 S2 not S1 (4967)

4. S4 (MH “Health Resource Allocation”) OR (MH “Health Resource Utilization”) (12,749)

5. S5 TX cost or costs or economic* or pharmacoeconomic* or price* or pricing* (164,114)

6. S6 S3 or S4 or S5 (172,311)

7. S7 PT ( Editorial or Letter or News ) OR MH Animal studies OR SO Cochrane library OR AU Anonymous (279,784)

8. S8 S6 not S7 (159,174)

9. S9 MH Carboplatin (607)

10. S10 MH Cisplatin (1427)

11. S11 MH Paclitaxel (1551)

12. S12 MH Antineoplastic Agents (12,767)

13. S13 MH Antimetabolites (106)

14. S14 MH Antimetabolites, antineoplastic (608)

15. S15 MH Angiogenesis Inhibitors (1370)

16. S16 S9 or S10 or S11 or S12 or S13 or S14 or S15 (16,663)

17. S17 TX carboplatin or paraplatin or cisplatin or Platinol or Paclitaxel or Taxol or Topotecan or Hycamtin or irinotecan or Campto or docetax?l or Taxotere or vinorelbine or Navelbine or gemcitabine or Gemzar or zactima or ZD6474 or bevacizumab or Avastin or pemetrexed or Alimta or erlotinib or Tarceva or bortezomib or Velcade or vinflunine or Javlor or cetuximab or Erbitux or gefitinib or Iressa or Vatalanib or Panitumumab or platinum or taxoid* (7964)

18. S18 TX carboplatin or paraplatin or cisplatin or Platinol or Paclitaxel or Taxol or Topotecan or Hycamtin or irinotecan or Campto or docetax?l or Taxotere or vinorelbine or Navelbine or gemcitabine or Gemzar or zactima or ZD6474 or bevacizumab or Avastin or pemetrexed or Alimta or erlotinib or Tarceva or bortezomib or Velcade or vinflunine or Javlor or cetuximab or Erbitux or gefitinib or Iressa or Vatalanib or Panitumumab or platinum or taxoid* (7964)

19. S19 MH lung neoplasms (8574)

20. S20 MH carcinoma, non-small-cell lung (2144)

21. S21 TX ( non small cell or non-small-cell or nonsmall cell or nsclc ) OR TX ( lung* or pulmon* ) OR TX ( cancer or tumor* or tumour* or carcino* or blastom* or squamous or neoplasm* or sarcom* or lymphoma* or adenocarcinom* ) (254,889)

22. S22 S18 or S19 or S20 (15,916)

23. S23 S16 or S17 (19,714)

24. S24 S22 and S21 and S8 Limiters - Published Date from: 20060101-20121231 (382)

25. Entry date limit (2009-2012) applied in EndNote. Final results = 241

Update of search strategy from appendix 10.4:

AstraZeneca UK Ltd. Single Technology Appraisal (STA) for Gefitinib for the first line treatment of locally advanced or metastatic non-small lung cancer (submission to National Institute for Health and Clinical Excellence) [Word document provided by AstraZeneca]. Luton, UK: AstraZeneca UK Ltd, 2010 (cited 18 July 2012). 233pp.