Immunosuppressive therapy for kidney transplantation in children and adolescents: systematic review and economic evaluation

Aetiology

In children, ESRD is usually due to innate structural abnormalities or genetic causes or is acquired in childhood through glomerulonephritis. 11 Figure 1 displays the causative diagnoses for children and adolescents (< 16 years old) with primary renal disease in 2013.

FIGURE 1.

Causative diagnoses for children and adolescents; primary renal disease percentage in incident and prevalent children and adolescents with established renal failure patients < 16 years old in 2013. Reproduced with permission from UK Renal Registry 17th Annual Report (figure 4.3, p. 99). 4 The data reported here have been supplied by the UK Renal Registry of the Renal Association. The interpretation and reporting of these data are the responsibility of the authors and in no way should be seen as an official policy or interpretation of the UK Renal Registry or the Renal Association.

Pathology

Table 1 displays the distribution of the UK primary renal diagnosis for end-stage renal failure over time, reported from 1999 to 2003, 2004 to 2008 and 2008 to 2013 in children and adolescents aged < 16 years. Renal dysplasia, which is abnormal tissue development in the kidney, is the primary renal disease diagnosis in approximately one-third of all children and adolescents with ESRD.

When chronic renal failure occurs, children and adolescents may experience malaise, nausea, loss of appetite, change in mental alertness, bone pain, headaches, stunted growth, change in urine outputs, urinary incontinence, pale skin, bad breath, poor muscle tone, tissue swelling and hearing deficit. Treatment of chronic renal failure depends on the degree of kidney function that remains and the age of the child/adolescent. Treatment may include dialysis, kidney transplantation, diet restrictions, diuretic therapy and medications (to help with growth and prevent bone density losses). 12

Graft function

Glomerular filtration rate (GFR) describes the flow rate of filtered fluid through the kidney. GFR is expressed in terms of volume filtered per unit time [sometimes this is also expressed per average surface area (1.73 m²)]. There are various methods used to calculate GFR [estimated glomerular filtration rate (eGFR)] from serum creatinine levels, age, sex and ethnic group (e.g. Modification of Diet in Renal Disease, Cockcroft–Gault, and Nankivell). Different methods are used for children and adolescents (e.g. Schwartz and Counahan–Barratt equations). Levels of eGFR represent the level of kidney function and Table 2 presents the NICE cut-off values for classification of CKD (NICE guidelines CG182). 16 These values apply to children aged > 2 years and up to (and including) adulthood. 17

TABLE 2 - Glomerular filtration rate categories

Notes

The eGFR and level of serum creatinine following a transplant can guide postoperative care as indicators of AR, recurrence of original kidney disease or development of de novo kidney disease.

Reproduced from NICE guidelines CG182. 16 © NICE 2014. All rights reserved. NICE copyright material can be downloaded for private research and study, and may be reproduced for educational and not-for-profit purposes. No reproduction by or for commercial organisations, or for commercial purposes, is allowed without the written permission of NICE.

GFR category	GFR (ml/minute/1.73 m2)	Terms
1	> 90	Normal or high
2	60–89	Mildly decreased
3a	45–59	Mildly to moderately decreased
3b	30–44	Moderately to severely decreased
4	15–29	Severely decreased
5	< 15	Kidney failure

Some children and adolescents may experience delayed graft function (DGF) after transplantation and Figure 2 shows a hypothetical graph to explain the relationship between normally functioning grafts, DGF and primary non-functioning (PNF) grafts. At 7 days post transplant, some of the children and adolescents who need dialysis and whose grafts are therefore classified as DGF will have grafts that never function. When this has been established, these grafts are classified as PNF.

FIGURE 2.

Hypothetical graph to explain graft function, DGF and primary non-functioning graft. Reproduced with permission from Bond et al. 18 Contains information licensed under the Non-Commercial Government Licence v1.0.

Prognosis

Data collected for survival rates of children and adolescents < 16 years of age starting RRT between 1999 and 2012 were collected from UK paediatric centres. 4 The median follow-up time was 3.5 years (ranging from 1 day to 15 years). There were a total of 99 deaths reported. Table 3 shows the survival hazard ratios (HRs) (following adjustment for age at start of RRT, sex and RRT modality) and highlights that children starting RRT at < 2 years of age, compared with 12- to 16-year-olds starting RRT, had a worse survival outcome with a HR of 5.0.

Various factors may influence survival following a kidney transplant. A study of 1189 child/adolescent kidney transplants in England between April 2001 and March 2012 found that 33 children and adolescents did not survive. 27 The most common causes of these 33 deaths were renal (n = 8; classified as ESRD, renal dysplasia and disorder of kidney/ureter), infections (n = 6) and malignancy (n = 5). 27 The age of the recipient was not found to significantly impact patient survival: age 0–1 years (100% survival), age 2–5 years (96% survival), age 6–12 years (97.5% survival) and age 13–18 years (97.4% survival). 27

Significance for patients

Living with ESRD may substantially challenge the well-being of children and adolescents. Not only will the disease impact physical health, but mental and social health may also be affected owing to increased hospital visits and the child or adolescent’s inability to take part in the same activities as their peers. 35 However, having a kidney transplant will improve the symptoms associated with ESRD and dialysis and reduce the time spent in hospital. 36 The median wait time for a child/adolescent requiring a kidney transplant in the UK is 342 days. 32

Kidney transplantation requires a lifelong regimen of immunosuppressive medication. Immunosuppressants may produce unpleasant side effects (including possible skin cancer, crumbling bones, fatigue, body hair growth, swollen gums and weight gain). 37 Nevertheless, favourable social and professional outcomes have been observed from a long-term follow-up (15.6 years ± 3 years) of people who had a kidney transplant as a child (aged 10 years ± 5 years). 38 Adherence to post-transplant immunosuppressive regimens is important for favourable clinical outcomes in children and adolescents39 and has been suggested as a core strategy to improve clinical outcomes. 40 In addition, failing to follow treatment may result in an increase in medical costs. 41

Acute rejection is common in the first year after kidney transplantation and treatment of AR involves a more intensive drug treatment than standard maintenance regimens, which in turn increases the possibility of adverse events (AEs). Should a graft be lost, the child/adolescent will face another wait for transplantation (if appropriate) and will need to undergo dialysis while waiting for transplantation (although a pre-emptive transplantation may be available), or need to undergo dialysis for life where transplantation is not possible.

The impact on a child/adolescent returning to or starting dialysis (of the psychological burden of graft failure and going back to a previous treatment) is little researched, but necessarily includes the impact of being on dialysis per se: dialysis is time-consuming and may affect education and normal family life and require changes in diet and fluid intake. Common side effects of dialysis (either haemodialysis or peritoneal dialysis) include fatigue, low blood pressure, invasive staphylococcal infections, muscle cramps, itchy skin, peritonitis, hernia and weight gain. 42

Finally, growth retardation in children and adolescents with ESRD is thought to be a combination of inadequate nutritional intake, acidosis, renal osteodystrophy and alterations to the growth hormone insulin-like growth factor. 8 Ensuring optimal growth or optimisation of final height is a major concern for children and adolescents with ESRD, as short stature may have an impact on social development, self-esteem and quality of life and is associated with an increase in the number of hospitalisations and behavioural and cognitive disorders, and a decrease in the level of education and employment in adulthood. 20,43–45

Unfortunately, data relating specifically to quality of life are currently available only in the adult population, among whom there are clear quality-of-life improvements from having a functioning kidney transplant compared with being on dialysis. 46–52

Significance for the NHS

Treatment for ESRD is considered resource-intensive for the NHS because current costs have been estimated to use 1–2% of the total NHS budget to treat 0.05% of the population (both adult and child/adolescent). 53 Based on data from the Department of Health, it is estimated that in 2008/9, the total expenditure on ‘renal problems’ in England was £1.3B, representing 1.4% of the NHS expenditure. 53 An economic evaluation of treatments for ESRD by de Wit et al. 54 showed that transplantation is the most cost-effective form of RRT with increased quality of life and independence for an individual.

There are no apparent reasons why RRT demand may dramatically increase in children and adolescents. However, it is projected that an increasingly overweight population will increase the demand for RRT, with a consequent increase in pressure on services from renal units and other health-care providers dealing with comorbidities. Increased resources may be needed for dialysis, surgery, pathology, immunology, tissue typing, histopathology, radiology, pharmacy and hospital beds. Demand is likely to be particularly significant in areas where there are large South Asian, African and African Caribbean communities and in areas of social deprivation, where people are more susceptible to kidney disease. 4

Short term

• Immediate graft function: the graft works immediately following transplantation, removing the need for further dialysis.

• Delayed graft function: the graft does not work immediately and dialysis is required during the first week post transplant. Dialysis has to continue until graft function recovers sufficiently to make it unnecessary. This period may last up to 12 weeks in some cases.

• Primary non-function: the graft never works after transplantation.

Long term

• Rejection rates: the percentage of grafts that are rejected by the recipients’ bodies; rejection can be acute or chronic.

• Graft survival: the length of time that a graft functions in the recipient.

• Graft function: a measure of the efficiency of the graft by various markers, for example GFR and serum creatinine levels.

• Patient survival: how long the recipient survives.

• Quality of life: how a person’s well-being is affected by the transplant.

Management of kidney transplants

If transplantation is the chosen method for RRT for a child/adolescent with ESRD then there are three main service provision steps required for the management of the transplant.

The first of these is organ procurement, which includes the identification and management of potential donors and assessment of donor suitability. HLAs are carried on cells within the body, enabling the body to distinguish between its ‘self’ or to recognise ‘non-self’ that should be attacked. The closer the HLA matching, the less vigorously the body will attack the foreign transplant and, consequently, the chances of graft survival are improved. HLA mismatch refers to the number of mismatches between the donor and the recipient at the A, B and DR loci, with a maximum of two mismatches at each locus. 32 Therefore, a match would have a score of zero and a complete mismatch would have a score of six. However, it should be noted that with the improvements in immunosuppressants, the significance of HLA matching has diminished. 55

The second step is the provision of immunosuppressive therapy. Immunosuppressants are the drugs taken around the time of, and following, an organ transplant. They are aimed at reducing the body’s ability to reject the transplant and thus at increasing patient and graft survival and preventing acute and/or chronic rejection (while minimising associated toxicity, infection and malignancy). Immunosuppressants are required in some form for all kidney transplant recipients (KTRs), except potentially when the donor is an identical twin.

The final service provision step is short- and long-term follow-up following transplantation. This step involves looking for indications of any kidney graft dysfunction and other complications. Complications fall into four categories.

1. Medical follow-ups to include rejections, nephrotoxicity of calcineurin inhibitors (CNIs) and recurrence of the native kidney diseases.

2. Anatomic complications of surgery to include renal artery thrombosis, renal artery stenosis, urine leaks from disruption of the anastomosis, ureteral stenosis and obstruction and lymphocele.

3. Other complications include infection, malignancy, new-onset of diabetes mellitus, liver disease, hypertension, CVD.

4. Ensuring growth is not impeded and maximal ‘catch-up’ growth is achieved. The 2010 NAPRTCS report suggests that the average final adult height of a renal transplant recipient has increased significantly from –1.93 SDS between 1987 and 1991 to –0.94 SDS between 2002 and 2010. 25

If the kidney loses its function, many of the physiological changes that occur mimic those seen with progressive renal diseases from other causes. Therefore, these symptoms should be managed in a similar way to the non-transplant population. However, it should be noted that the loss of a kidney transplant carries increased susceptibility to bruising and infection compared with pre-transplant kidney failure. 56

Once the kidney is confirmed to have been lost, the graft may or may not need to be surgically removed. The decision of whether or not the graft is removed is often made on a case-by-case basis taking into consideration all perceived benefits and risks. The immunosuppression regimen can then be tapered and withdrawn while the patient returns to dialysis and waits for a new kidney to become available.

Induction therapy

Basiliximab or daclizumab (DAC), used as part of a ciclosporin (CSA)-based immunosuppressive regimen, is recommended as an option for induction therapy in the prophylaxis of acute organ rejection in children and adolescents undergoing renal transplantation, irrespective of immunological risk. The induction therapy (BAS or DAC) with the lowest acquisition cost should be used, unless it is contraindicated. 1 The marketing authorisation for DAC has been withdrawn at the request of the manufacturer.

Maintenance therapy

Tacrolimus (TAC) is recommended as an alternative option to CSA when a CNI is indicated as part of an initial or a maintenance immunosuppressive regimen for renal transplantation in children and adolescents. The initial choice of TAC or CSA should be based on the relative importance of their side effect profiles for the individual patient. 1

Mycophenolate mofetil is recommended as an option as part of an immunosuppressive regimen for child and adolescent renal transplant recipients only when:

• there is proven intolerance to CNIs, particularly nephrotoxicity which could lead to risk of chronic allograft dysfunction or

• there is a very high risk of nephrotoxicity necessitating the minimisation or avoidance of a CNI until the period of high risk has passed. 1

The use of MMF in CCS reduction or withdrawal strategies for child and adolescent renal transplant recipients is recommended only within the context of randomised clinical trials. 1

Mycophenolate sodium is currently not recommended for use as part of an immunosuppressive regimen in child or adolescent renal transplant recipients. 1

Sirolimus is not recommended for children or adolescents undergoing renal transplantation except when proven intolerance to CNIs (including nephrotoxicity) necessitates the complete withdrawal of these treatments. 1

As a consequence of following this guidance, some medicines may be prescribed outside the terms of their UK marketing authorisation. Health-care professionals prescribing these medicines should ensure that children and adolescents receiving renal transplants and/or their legal guardians are aware of this and that they consent to the use of these medicines in these circumstances. 1

Induction therapy

Basiliximab (Simulect,^® Novartis Pharmaceuticals) is a monoclonal antibody which acts as an interleukin 2 receptor antagonist. It has a UK marketing authorisation for prophylaxis of AR in allogeneic renal transplantation in children (aged 1–17 years). The summary of product characteristics (SPC) states it is to be used concomitantly with CSA for microemulsion- and CCS-based immunosuppression, in patients with panel reactive antibodies < 80%, or in a triple maintenance immunosuppressive regimen containing CSA for microemulsion, CCSs and either azathioprine (AZA) or MMF. 7

Rabbit antihuman thymocyte immunoglobulin is a gamma immunoglobulin. It has a UK marketing authorisation for the prevention of graft rejection in renal transplantation. The SPC states it is usually used in combination with other immunosuppressive drugs. It is administered intravenously. The UK marketing authorisation is not restricted to adults only. 7

Maintenance therapy

Tacrolimus is a CNI that is available in an immediate-release formulation (Adoport,^® Sandoz; Capexion,^® Mylan; Modigraf,^® Astellas Pharma; Perixis,^® Accord Healthcare; Prograf,^® Astellas Pharma; Tacni,^® Teva; Vivadex,^® Dexcel Pharma). All of these formulations of TAC have UK marketing authorisations for prophylaxis of transplant rejection in kidney allograft recipients. The marketing authorisations include adults and children. 7 TAC (Modigraf^®, Astellas Pharma) is available in a granule form which can be suspended in liquid and may be more suitable for those who struggle swallowing pills.

Tacrolimus is also available in a prolonged-release formulation (Advagraf,^® Astellas Pharma). It has a UK marketing authorisation for prophylaxis of transplant rejection in kidney allograft recipients. The marketing authorisation is restricted to adults. The Commission on Human Medicines advises that all oral TAC (including both TAC-IR and TAC-PR) medicines in the UK should be prescribed and dispensed by brand name only. 7

Belatacept is designed to selectively inhibit CD28-mediated co-stimulation of T-cells. BEL has a UK marketing authorisation for prophylaxis of graft rejection in adults receiving a renal transplant, in combination with CCSs and a mycophenolic acid (MPA; Myfortic,^® Novartis Pharmaceuticals). The SPC recommends that an interleukin 2 receptor antagonist for induction therapy is added to this BEL-based regimen. The SPC states that the safety and efficacy of BEL in children and adolescents aged 0–18 years have not yet been established. This formulation does not have a UK marketing authorisation for the prophylaxis of transplant rejection in renal transplantation in children and adolescents. 7

Mycophenolate mofetil is a prodrug of MPA which acts as an antiproliferative agent (Arzip,^® Zentiva; CellCept,^® Roche Products; Myfenax,^® Teva); generic MMF is manufactured by Accord Healthcare, Actavis, Arrow Pharmaceuticals, Dr Reddy’s Laboratories, Mylan, Sandoz and Wockhardt). It has a UK marketing authorisation for use in combination with CSA and CCSs for the prophylaxis of acute transplant rejection in people undergoing kidney transplantation. The UK marketing authorisation is not restricted to adults (dosage recommendations for children aged 2–18 years are included in the SPC). 7

Mycophenolate sodium is an enteric-coated formulation of MPA. This formulation has the same UK marketing authorisation as MMF; however, this is restricted to adults. This formulation does not have a UK marketing authorisation for the prophylaxis of transplant rejection in renal transplantation in children and adolescents. 7

Sirolimus (Rapamune,^® Pfizer) is an antiproliferative with a non-calcineurin-inhibiting action. It has a UK marketing authorisation for the prophylaxis of organ rejection in adult patients at low to moderate immunological risk receiving a renal transplant. It is recommended to be used initially in combination with CSA and CCSs for 2–3 months. It may be continued as maintenance therapy with CCSs only if CSA can be progressively discontinued. This formulation does not have a UK marketing authorisation for the prophylaxis of transplant rejection in renal transplantation in children and adolescents. 7

Everolimus (Certican,^® Novartis Pharmaceuticals) is a proliferation signal inhibitor and is an analogue of SRL. EVL does not currently have a UK marketing authorisation for immunosuppressive treatment in kidney transplantation in children and adolescents. 7

Renal societies (UK)

• British Renal Society (www.britishrenal.org/).

• Renal Association (www.renal.org/).

• UK Renal Registry (www.renalreg.com/).

• Kidney Research UK (www.kidneyresearchuk.org/).

• British Kidney Patient Association (www.britishkidney-pa.co.uk/).

• National Kidney Federation (www.kidney.org.uk/).

Renal societies (international)

• American Society of Nephrology (www.asn-online.org/).

• American Association of Kidney Patients (www.aakp.org/).

• National Kidney Foundation (US; www.kidney.org/).

• Canadian Society of Nephrology (www.csnscn.ca/).

• Kidney Foundation of Canada (www.kidney.ca/).

• Australian and New Zealand Society of Nephrology (www.nephrology.edu.au/).

• Kidney Health Australia (www.kidney.org.au/).

• Kidney Society Auckland (www.kidneysociety.co.nz/).

Previous Health Technology Assessment review

Studies included in the previous HTA review (Yao et al. 2) were screened using the inclusion criteria for the Peninsula Technology Assessment Group (PenTAG) review (Inclusion and exclusion criteria).

Reference lists

Reference lists of included guidelines, systematic reviews, company submissions and clinical trials were scrutinised in order to identify additional studies.

Ongoing trials

Searches for ongoing trials were also undertaken. Terms for the intervention and condition of interest were used to search the following trial registers for ongoing trials: ClinicalTrials.gov and Controlled Trials (ISRCTN). Trials that did not relate to immunosuppressive therapies for kidney transplantation in children and adolescents were removed by hand-sorting. All searches for ongoing trials were carried out in January 2015. The search strategies can be found in Appendix 1.

Adult randomised controlled trial evidence

In addition, as specified in the review protocol, all child/adolescents RCT and non-RCT evidence included in this review was compared with adult evidence identified from parallel HTA 09/46/01 appraisal. 68 This parallel HTA was conducted by PenTAG to inform the ongoing technology appraisal of immunosuppressive therapy for kidney transplantation in adults (review of technology appraisal guidance 85; NICE appraisal ID 456). The NIHR Evaluation, Trials and Studies Coordinating Centre reference for the adult report is 09/46/01 (www.nice.org.uk/guidance/indevelopment/gid-tag348/documents).

Study design

The clinical effectiveness review included:

• eligible studies – RCTs in children and adolescents (≤ 18 years), RCTs of adults and children/adolescents in which a subgroup analysis of children and adolescents is reported, and non-randomised controlled studies (comparative quasi-experimental and observational studies were considered)

• search strategy – databases were searched to identify RCTs, systematic reviews of RCTs and systematic reviews of non-randomised controlled studies. Individual non-randomised controlled studies were identified via the bibliographies of systematic reviews (i.e. individual non-randomised controlled studies were not searched for directly).

For the purpose of this review, a systematic review was defined as one that has:

• a focused research question

• explicit search criteria that are available to review, either in the document or on application

• explicit inclusion/exclusion criteria, defining the population(s), intervention(s), comparator(s), and outcome(s) of interest

• a critical appraisal of included studies, including consideration of internal and external validity of the research

• a synthesis of the included evidence, whether narrative or quantitative.

Interventions

Studies evaluating the use of the following immunosuppressive therapies for renal transplantation were included.

Comparator

Studies using the following comparators were included.

Population

The population is children and adolescents aged ≤ 18 years undergoing kidney transplantation. The kidney donor may be living related, living unrelated or deceased. Patients receiving multiorgan transplants and those who have received transplants and immunosuppression previously were excluded.

Outcomes

The outcome measures to be considered are:

• patient survival

• graft survival

• graft function

• time to and incidence of AR

• severity of AR

• growth

• adverse effects (AE) of treatment

• HRQoL.

Randomised control trials

Four reviewers (LC, MHa, HC and TJ-H) independently assessed the quality of all studies included in the clinical effectiveness review. The internal and external validity of RCTs was assessed according to criteria based on Centre for Reviews and Dissemination (CRD) guidance69 (Table 8).

Non-randomised control trials

There is no agreed recommended appraisal tool for the assessment of non-randomised studies. 70 The CRD handbook suggests considering the study design, risk of bias, other issues related to study quality, choice of outcome measure, statistical issues, quality of reporting, quality of the intervention and generalisability. 69 Therefore, the internal and external validity of non-RCTs was assessed according to criteria based on CRD guidance69 (Table 9).

Meta-analyses

When data permitted, the results of individual studies comparing the same regimens were pooled using the methods described below.

A random-effects model was assumed for all meta-analyses. For binary data, an odds ratio (OR) was used as a measure of treatment effect and the DerSimonian–Laird method was used for pooling. 71 For continuous data (e.g. graft function), mean differences were calculated if the outcome was measured on the same scale in all trials. If applicable, publication bias was assessed using funnel plots, the Harbord test was used for binary outcomes [OR, log-standard error (SE)] and the Egger test for continuous data. All analyses were performed in Stata 13 (StataCorp LP, College Station, TX, USA).

For studies with more than one intervention arm (that were separately compared with the same control arm), the number of events and the total sample size in the control arm were divided equally across the comparisons, and when pooling mean differences the total sample size in the control arm was adjusted and divided equally across the comparisons. However, if only one experimental arm was eligible for the analysis, all participants and events assigned to the control arm were included. If the number of events was zero in one of the studies arms, a value of 0.5 was added to all study arms to allow for statistical analyses.

Randomised control trials

Our searches returned 5079 unique titles and abstracts, with 784 papers retrieved for detailed consideration. To ensure the inclusion of trials with mixed child/adolescent and adult populations that reported separate results for children and adolescents, the searches and title and abstract screening were not limited to children and adolescents. Update searches conducted on 7 January 2015 returned 416 unique titles and abstracts. Forty papers were retrieved for detailed consideration.

Of the 824 full-text papers retrieved, 793 were excluded (a list of these records with reasons for their exclusion can be found in Appendix 2, Table 135). Although RCTs in mixed populations were identified, none included subgroup analysis by age – providing separate results for children/adolescents and adults – and were therefore excluded from the review (a list of these records can be found in Appendix 2, Table 136). Three RCTs (published in one abstract72 and seven papers73–79) met the inclusion criteria.

Only one abstract72 was included in the review. This abstract included new data related to Offner et al. 73 and sufficient methodological information to inform the quality appraisal. In addition, there were 23 articles that were systematic reviews and all eligible systematic reviews were tabulated (see Appendix 3, Table 137).

The process is illustrated in detail in Figure 8.

FIGURE 8.

Clinical effectiveness review: Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram. SR, systematic review.

In summary, three RCTs (published in seven papers73–79 and one abstract72) were found eligible and are included in this review (Table 10).

Non-randomised trials

The systematic reviews were used to identify non-RCTs. We screened the titles and abstracts of 226 unique references identified by the PenTAG systematic review searches (including 43 records from update searches) and retrieved 38 papers for detailed consideration. All eligible systematic reviews were tabulated (see Appendix 3, Table 137).

In total, four non-RCTs met the inclusion criteria and were considered eligible for inclusion (Table 11). All of these were included in the previous HTA by Yao et al. 2 so no new non-RCTs were identified. However, in 2007 one of the four non-RCT studies83 published 5-year follow-up data85 that were not included in the previous HTA.

Randomised control trials

Non-randomised studies

An overview of the nine non-randomised studies included in Yao et al. 2 with reasons for inclusion/exclusion in the current review is provided in Table 13. Five studies were excluded from the PenTAG review (see Table 13):

• Duzova et al. 91 (compared BAS with no induction) administered triple therapy of (CSA or TAC) + (AZA or MMF) + CCS; however, a breakdown of the numbers (and results) in each combination was not reported and the mean recipient age was 14.9 ± 3.6 years (range 7–21 years).

• Pape et al. 92 recruited a child with a combined kidney–liver transplantation.

• Swiatecka-Urban et al. 93 included children, adolescents and adults (inclusion criteria aged < 21 years).

• Neu et al. 94 included children, adolescents and adults (inclusion criteria aged > 2 years and < 21 years) and the use of induction therapy varied in the study.

• Steffen et al. 95 was published as an abstract only and did not include enough information to allow critical appraisal.

In summary, four non-randomised studies were included in the PenTAG review and all were also included in the previous HTA review by Yao et al. 2 No new non-randomised studies were identified in PenTAG systematic review searches.

Adults

The previous TA99 included evidence from 25 adult RCTs. In comparison, the updated HTA 09/46/01 appraisal ‘Immunosuppressive therapy for kidney transplantation in adults’ included 86 trials: 11 induction studies, 73 maintenance studies and two studies of both induction and maintenance treatment. An overview of the 25 adult RCTs included in Yao et al. 2 with reasons for inclusion/exclusion in the parallel HTA review68 is provided in Appendix 5 (see Table 139).

If relevant, the adult evidence from the HTA 09/46/01 appraisal was summarised and compared with child/adolescent evidence included in the PenTAG review.

Randomised controlled trials

In total, three RCTs were assessed: two induction studies and one maintenance study. 73,75,77

Randomised controlled studies

Baseline characteristics of the three included RCTs73,75,77 are summarised in Table 16. All three studies were conducted over multicentres in Europe. Only Offner et al. 73 reported the countries involved (Germany, France and Switzerland). 73 Mean age across the studies’ arms ranged from 10.1 years to 11.5 years. The proportion of adolescents (with 12 or 13 years old being the cut-off point for adolescence in the three studies; see Table 16 for details) is 36.6% to 54.4% across the study arms. Boys represented 56.0–67.4% of participants. Two studies had a high proportion of white participants (87–95%),73,77 with one trial not reporting ethnicity. 75 The proportion of living donors across the study arms ranges from 15.5% to 35.8%. The proportion of first transplants is high, ranging from 85% to 96% across the arms. Finally, HLA antigen mismatch ranges from 2.3 to 2.7 across the three trials. A close antigen match is no longer considered critical owing to the more effective immunosuppressive therapy, but a better HLA match may lead to longer graft survival.

Non-randomised studies

Similarly, baseline characteristics of the four included non-randomised studies [Antoniadis et al. 81 (non-RCT), Benfield et al. 82 (historically controlled study), Garcia et al. 80 (retrospective cohort study) and Staskewitz et al. 200183 (historically controlled study)] are summarised in Table 17. 80–83 The Antoniadis et al. 81 study was conducted in one Greek centre, the Benfield et al. 82 study was conducted in two centres in the USA and the Staskewitz et al. 83 study was conducted in 12 German centres. Garcia et al. 80 did not report where or within how many centres their study was performed, but the authors are all based in Brazil and, therefore, it is likely that this study was completed in Brazil. Not surprisingly, the baseline characteristics of the non-RCTs vary not only across the studies, but also within the studies. Mean age across the study arms ranges from 9.0 years to 11.5 years; however, none of the non-RCTs reports the proportion of adolescents included. Boys represented 50.0–66.7% of participants. Two studies had a high proportion of white participants (75–100%),80,83 one study reported that between 19% and 25% of participants were black (dependent on treatment group),82 while one study did not report ethnicity. 81 Most studies included a high proportion of living donors (75 –100%). However, one study reported only 6% living donors in one treatment group and 9% in the other treatment group. 83 This was the only study reporting mean HLA mismatches (2.69–2.89). 83

Mortality

Both RCTs73,75 provided data on mortality for BAS versus no induction or PBO (Table 18). Grenda et al. 75 reported the longest follow-up data at 2 years post transplant. No evidence of a statistically significant difference in overall survival between BAS and comparator arms was reported at any time point.

Graft loss

Both RCTs73,75 provided data on graft loss for BAS versus no induction or PBO (Table 19). Grenda et al. 75 reported the longest follow-up data of 2 years. No evidence of a significant difference between the BAS and control arms was reported for any data point.

The pooled results at 6-month follow-up did not find any significant difference between BAS and control arms for graft loss [OR = 93 favours BAS; 95% confidence interval (CI) 0.29 to 2.97, I² = 0%, τ² = 0; Figure 9].

FIGURE 9.

Graft loss: randomised control trials. Control, no induction/PBO control arms. τ² = 0. Studies included were Offner et al. 73 and Grenda et al. 75

Graft function

Both RCTs73,75 reported graft function estimated using the Schwartz equation (ml/minute/1.73 m²; Table 20). There were no statistically significant differences between BAS and control arms at any data point (between 6 months and 2 years). Both RCTs reported 6-month and 2-year follow-ups, no standard deviation (SD) was reported at 2 years by Offner et al. 73 and no SD was reported at 6 months or 2 years by Grenda et al. 75

To allow for combining the results at 6-month and 2-year follow-ups, a SD of 26 ml/minute/1.73 m² was used (‘average’ SD calculated from SD available at 6-month and 2-year follow-ups; Figure 10). The pooled results do not suggest any difference for eGFR between BAS and control arms: weighted mean difference (WMD) = –4.20 (favours controls; 95% CI –9.60 to 1.20, I² = 0%) at 6 months and WMD = –1.38 (favours controls; 95% CI –7.20 to 4.44, I² = 0%) at 2 years. Grenda et al. 75 also reported incidences of DGF (defined as requiring dialysis for more than 1 day during the first study week). The rate of DGF was not statistically significantly different between the two arms: 11 out of 99 participants (11%) and 5 out of 93 participants (5%) in BAS and no induction arms, respectively (p-value was not reported). 75

FIGURE 10.

Graft function (eGFR): randomised control trials. Control, no induction/PBO control arms. For data points with no SD reported, a SD of 26 was used. Graft function was estimated using the Schwartz equation (ml/minute/1.73 m²). Studies included were Offner et al. 73 and Grenda et al. 75

Acute rejection

Both RCTs73,75 provided data on biopsy-proven acute rejection (BPAR) for BAS versus no induction or PBO (Table 21). Grenda et al. 75 reported the longest follow-up data of 2 years. No evidence of a statistically significant difference between the BAS and the comparators arms was reported for any data point. The pooled results at 6 months did not find any difference between BAS and control arms for BPAR: OR = 0.71 (favours BAS; 95% CI 0.40 to 1.27, I² = 15.7%, τ² = 0.03; Figure 11).

FIGURE 11.

Biopsy-proven acute rejection: RCTs. Control, no induction/PBO control arms. τ² = 0.03. Studies included were Offner et al. 73 and Grenda et al. 75

In addition, Grenda et al. 75 also reported BPAR separately for younger and older age groups (< 12 years and ≥ 12 years, respectively). The incidence of BPAR was lower in the patients < 12 years in the no induction arm (4/42, 10%) than the same age group with BAS (6/46, 13%), although this difference was not statistically significant (Wilcoxon–Gehan test, p-value was not reported). Conversely, incidences of BPAR were higher for the patients ≥ 12 years with no induction (15/51, 29%) than the same age group with BAS (13/53, 25%); however, again, this difference was not statistically significant (Wilcoxon–Gehan test, p-value was not reported).

Finally, the data from Offner et al. 75 of 79 BAS and 65 PBO on study participants (reported in an abstract by Jungraithmayr et al. 72) found a cumulative AR rate of 33% versus 35% in the BAS and PBO arms, respectively, at 2 years and a cumulative AR rate of 41% versus 45% in the BAS and PBO arms, respectively, at 5 years. Results were not statistically significant at either data point. 72

Time to BPAR (Table 22) was only reported by Grenda et al. 75 The median time to BPAR appears to be similar between the two arms (p-values were not reported in the study). 75 Time to first BPAR episode or treatment failure within the first 6 months post transplant was the primary efficacy end point in Offner et al. 73 The proportion of children and adolescents (Kaplan–Meier estimates) achieving this efficacy point was 16.7% in the BAS arm and 21.7% in the PBO arm. The difference was not statistically significant; HR of 0.72 (favours BAS; 95% CI 0.42 to 1.26). 73

Severity of BPAR was reported by Offner et al. 73 and Grenda et al. 75 (Table 23). All BPAR episodes in BAS treated patients were mild (grade IA or IB), whereas 8 out of 18 episodes in the PBO group were moderate (grade IIA) in Offner et al. 73 Similarly, there seemed to be more moderate BPAR episodes (Banff 2) in the no induction group than the BAS group in Grenda et al. 75 However, Offner et al. 73 also performed biopsies in children who had not recently experienced clinical signs of rejection or undergone biopsy (at 6 months, n = 64 and n = 60 in BAS and PBO groups, respectively) to identify subclinical rejections. The rate (p = 0.055) and severity (p-value not reported) of subclinical rejections was higher in the BAS group (25.0%) than in the PBO group (11.7%). 73

Adverse events

Two RCTs73,75 provided data on AEs for BAS versus no induction or PBO. Offner et al. 73 reported AEs that occurred in at least 10% of the safety population. Grenda et al. 75 reported AEs that occurred in at least 10% in either treatment arm. The AEs reported in these trials are summarised in Table 24.

In one trial,73 more infections were found with BAS than with PBO (OR = 2.23; favours PBO; 95% CI 1.03 to 4.68). 73 In Grenda et al. 75 toxic nephropathy was higher in the BAS arm than in the no induction arm (14.1% vs. 4.3%, respectively; p = 0.03). Similarly, abdominal pain was higher in the BAS arm than no induction (11.1% vs. 2.2%, respectively; p = 0.02). 75

Grenda et al. 75 also reported changes in glucose metabolism disorders. None of the children and adolescents had a glucose metabolism disorder {described as glucose tolerance decreased, hyperglycaemia or diabetes mellitus using the modified coding symbols for a thesaurus of adverse reaction terms [The Coding Symbols for a Thesaurus of Adverse Reaction Terms (COSTART) dictionary]} at baseline. However, during the study, 13 patients (13.1%) in the BAS arm and 10 patients (10.8%) in the no induction arm developed a glucose metabolism disorder within the first 6 months. One new case of impaired glucose metabolism was noted at 1 year; this new case resolved at 2 years.

Mortality

Graft loss

Graft function

Acute rejection

Non-randomised controlled trials

Four non-RCTs80–83 provided data on BPAR (Table 32)80–83 and three studies compared MMF with AZA. 81–83 The remaining study80 compared TAC + AZA with CSA + MMF. No statistically significant difference in BPAR was found between the MMF arm and AZA arms, and between TAC + AZA and CSA + MMF.

TABLE 32 - Biopsy-proven acute rejection: non-randomised studies

N/A, not applicable; NR, not reported.

Reported in text as 4 out of 17, 23%.

All ORs were calculated by PenTAG.

Study	Treatment	3 months		6 months
		n events/N participants, %	OR (95% CI)	n events/N participants, %	OR (95% CI)
Garcia et al. 200280	BAS + TAC + AZA + CCS	1/12, 8	0.45 (0.04 to 5.78)	NR	N/A
	BAS + CSA + MMF + CCS	2/12, 17		NR
Antoniadis et al. 199881	CSA + MMF + CCS	NR	N/A	0/7, 0	0.08 (0.003 to 1.94)
	CSA + AZA + CCS	NR		3/7, 43
Staskewitz et al. 200183	CSA + MMF + CCS	NR	N/A	10/65, 15	0.52 (0.21 to 1.29)
	CSA + AZA + CCS	NR		14/54, 26
Benfield et al. 199982	CSA + MMF + CCS	NR	N/A	4/17, 24a	0.56 (0.13 to 2.47)
	CSA + AZA + CCS	NR		6/17, 35

The pooled results at a 6-month follow-up suggested borderline statistically non-significantly lower BPAR in MMF compared with AZA (OR = 0.48, 95% CI 0.23 to 1.02, I² = 0%, τ² = 0; Figure 12).

FIGURE 12.

Biopsy-proven acute rejection: non-randomised studies. τ² = 0. Studies included were Antoniadis et al. ,81 Staskewitz et al. 83 and Benfield et al. 82

In addition, Garcia et al. 80 reported the severity of AR (Table 33); one Banff 3 episode was reported in TAC + AZA and two Banff 1 episodes were reported in CSA + MMF. No study reported time to BPAR.

TABLE 33 - Severity of AR: non-randomised studies

Note

No Banff 2 AR was reported, assumed 0 and 0 events of Banff 2 in each arm.

Study	Treatment	3 month n events/N participants, %
		Banff 1	Banff 2	Banff 3
Garcia et al. 200280	BAS + TAC + AZA + CCS	0/12, 0	0/12, 0	1/12, 8
	BAS + CSA + MMF + CCS	2/12, 17	0/12, 0	0/12, 0

Adverse events

Mortality

Graft loss

Graft function

Acute rejection

Adverse events

Mortality

Graft loss

Graft function

Acute rejection

Adverse events

Comparison with the previous Health Technology Assessment and the parallel Health Technology Assessment in adults

The results of the current review are similar to the previous HTA. 2

In addition, the child RCT evidence is similar to the conclusions of the parallel HTA in adults. However, the adult evidence found less BPAR in BAS than PBO or no induction (OR = 0.53; favours BAS; 95% CI 0.40 to 0.70, I² = 0.0%, τ² = 0.0; pooled results at 1-year follow-up with five studies).

The comparison of the child/adolescent RCT evidence with the previous HTA and the parallel HTA in adults is summarised in Table 41.

Randomised controlled trial evidence

One RCT of maintenance therapy (reported in three publications) evaluating TAC (compared with CSA) in children and adolescents was identified. 77 No RCTs were identified that evaluated TAC-PR, MMF, MPA, SRL, EVL or BEL in children and adolescents.

From the RCTs, we found no significant difference in survival or graft loss between TAC and CSA. 77 However, a significantly higher graft function (mean eGFR of 71.5 ml/minute/1.73 m², SD 22.9 ml/minute/1.73 m², in TAC vs. mean eGFR of 53.0 ml/minute/1.73 m², SD 21.6 ml/minute/1.73 m², in CSA; t-test = 4.03; p < 0.01 at 4-year follow-up), and less BPAR (OR = 0.29, favours TAC, 95% CI 0.15 to 0.57 at 6-month follow-up) was found in TAC compared with CSA. 77

Non-randomised controlled trial evidence

Three non-RCTs evaluating MMF (compared with AZA) in children and adolescents were identified. 81–83 One non-RCT compared TAC + AZA with CSA + MMF. 80 No non-RCTs were identified that evaluated TAC-PR, MPA, SRL, EVL or BEL in children and adolescents.

Induction

More infections were found in children treated with BAS than in those treated with PBO (OR = 2.23, favours PBO; 95% CI 1.03 to 4.68). 73 In addition, Grenda et al. 75 found that toxic nephropathy and abdominal pain was higher in the BAS arm than no induction (p = 0.03 and p = 0.02, respectively). 75 The previous HTA reported only post-transplant diabetes mellitus90 and the rest of the data were confidential and were omitted from the report. 2

In addition, the child RCT evidence is largely similar to the conclusions of the parallel HTA in adults. The adult 1-year follow-up RCT evidence identified in the parallel HTA did not find any significant differences in NODAT, PTLD, malignancy, infections and CMV infections between BAS and PBO or no induction. 98,100,101,103,104

Maintenance therapy

There were no statistically significant differences between TAC and CSA for a range of AEs (any infections, UTIs, bacterial infections, viral infections, PTLD, solid tumour, hypertension, any AEs and NODAT). 77 This is similar to the conclusions of the previous HTA. 2 However, Trompeter et al. 77 also reported that a deficiency of magnesium in the blood and diarrhoea were more common with TAC than with CSA, while excessive hair growth, flu syndrome and swollen gums were less common with TAC than with CSA. 77 In addition, there were no statistically significant differences between MMF and AZA for UTI, CMV infections, respiratory infections, herpes simplex, oral thrush and diarrhoea identified in the non-randomised evidence. 81 Similarly, no statistically significant differences between TAC + AZA and CSA + MMF in CMV infections and NODAT were identified in the non-randomised evidence. 80

However, the parallel HTA in adults found more cases of NODAT in TAC than CSA (OR = 2.22; favours CSA; 95% CI 1.42 to 3.46, I² = 0%; pooled results of eight studies at 1-year follow-up). 106–109,113,125–127 In addition, no difference in CMV infections114,117 and infection116 were found between MMF and AZA regimens in the adult evidence at 1-year follow-up.

Searches

Bibliographic literature searching was conducted on 8 April 2014. The searches took the following form: (terms for kidney or renal transplant or kidney or renal graft) AND (terms for the interventions under review) AND (a costs or economic literature search filter). The search was date limited 2002–current in line with the previous assessment and the searches were updated on 15 January 2015. The search was not limited by language and it was not limited to human-only studies.

The following databases were searched: MEDLINE and MEDLINE In-Process (via Ovid), EMBASE (via Ovid), NHS EED (via Wiley Online Library), Web of Science (ISI – including conference proceedings), Health Economic Evaluations Database (HEED) (via Wiley) and EconLit (via EBSCOhost). The search strategies are recorded in Appendix 1.

Screening

Inclusion and exclusion criteria were the same as for the clinical effectiveness systematic review (see Chapter 3, Inclusion and exclusion criteria), with the following exceptions (as specified in the appraisal protocol):

• Non-randomised studies were included (e.g. decision model-based analyses, or analyses of patient-level cost and effectiveness data alongside observational studies).

• Full cost-effectiveness analyses, cost–utility analyses and cost–benefit analyses were included (economic evaluations that report only average cost-effectiveness ratios were included only if the incremental ratios can be easily calculated from the published data).

• Studies that measure only costs but not health benefits were excluded except for stand-alone cost analyses from the perspective of the UK NHS.

• Only economic evaluations from the UK, the USA, Canada, Australia and western Europe were included as these settings may include data generalisable to the UK.

All records were dual screened. Titles and abstracts were screened for relevance by two reviewers (RMM and LC), with disagreements resolved by discussion. Full texts were retrieved for references judged to be relevant and were screened for eligibility by the same reviewers, with disagreements resolved by discussion.

The bibliographies of review articles not judged eligible for inclusion were examined by one reviewer (LC) to identify other potentially relevant references. These references were retrieved and checked for eligibility in the same way as full texts from database searches.

Quality assessment

Studies meeting the criteria for inclusion were assessed by one reviewer (RMM) using the checklist developed by Evers et al. 130 When studies were based on decision models they were also quality assessed using the checklist developed by Philips et al. 131,132

Synthesis

Economic studies were summarised and synthesised using tabulated data and narrative synthesis.

Identified studies

The electronic database search for cost-effectiveness evidence, including update searches conducted on 18 November 2014, identified 2090 records. After deduplication 1378 records remained, all of which were screened by title and abstract. Of these, 86 full texts were assessed for eligibility. Thirty-eight full texts were deemed to meet the eligibility criteria for the review. The process is illustrated in detail in Figure 13.

FIGURE 13.

Cost-effectiveness review: PRISMA flow diagram. CEA, cost-effectiveness analyses; CUA, cost–utility analyses. a, The previous HTA review by Yao et al. ;2 includes child, adolescent and adult evidence.

Only one study2 was identified that met the inclusion criteria. This was the HTA report of the previous NICE appraisal on the topic in children or adolescent patients. The rest of the subsection is devoted to reviewing this study.

Yao et al. 2 reports the methods and results of economic analyses submitted to the previous NICE appraisal on the topic by three sponsoring companies. All of these analyses used an equation estimated from regression analysis (meta-model) of child/adolescent simulation outcomes of immunosuppressive regimens derived from a model originally developed by one company (Novartis) for informing its submission to the corresponding NICE review on adult patients. The adult metamodel was developed by the Technology Assessment Group at Birmingham and the individual companies adapted it to children and adolescents. After critically appraising the evidence submitted by the companies, the group at Birmingham then produced its own analysis by adapting the metamodel to children and adolescents.

Briefly, the Birmingham model was a Markov model spanning a 10-year horizon after the initial transplant. It consisted of three states: functioning graft, graft failed (dialysis) and death. In common with models in this clinical area, surrogate outcomes were used to extrapolate beyond the end of follow-up in the RCT evaluating the relative effects of immunosuppressive regimens in terms of BPAR. The model used a HR of 1.41 for graft failure up to 7 years post transplant for children and adolescents (≤ 18 years) treated for an AR before discharge versus those not treated. The Birmingham group then used this surrogate relationship to translate 12-month differences in BPAR rates between immunosuppressive regimens from RCT studies in children and adolescents for therapies other than MMF and DAC, for which adult RCT data were used, into 10-year graft survival differences. The study also adjusted the resource use and costs for age–weight immunosuppressive doses in children and adolescents.

Table 43 presents the characteristics of the analysis by Yao et al. 2 Results were presented for two pairwise comparisons of induction regimens and two pairwise comparisons of initial and maintenance immunosuppressive regimens. In the comparisons of induction therapy regimens, BAS was found to result in lower total costs and higher quality-adjusted life-years (QALYs) than no induction in patients managed with either TAC or CSA in a CNI-containing triple immunosuppressive therapy including AZA and steroids. In terms of the initial and maintenance immunosuppressive regimens, TAC was found to have an incremental cost per QALY gained of £145,540 relative to CSA, while the corresponding figure for MMF relative to AZA was £194,559 when these therapies were combined with CSA and steroids. It is worth noting that the latter comparison was based on efficacy data from studies on MMF use in adults. Table 44 summarises the base-case results. However, altering the hazard (risk) ratio of graft loss with AR from 1.41 (which was based on a single observational study in children and adolescents) to a HR of 1.96 (derived from a pooled analysis of adult observational studies) and arbitrarily increasing the cost of dialysis from the base-case value of £21,000 (which was estimated from data on adults) to £50,000, as a way of accounting for the higher staff-to-patient ratios in children and adolescents, resulted in a cost per QALY gained of £34,000. 2

The technology assessment review team at Birmingham developed these analyses after considering evidence submitted by three companies using the Birmingham original model, which related to adult patients. The companies had found their sponsored drugs to result in lower total costs and higher QALYs, when compared with the triple therapy of CSA, AZA and steroids (CSA + AZA + CCS). Although the independent assessment by the Birmingham group confirmed the companies’ finding that BAS induction was expected to reduce total costs and increase QALYs, its results for initial and maintenance immunosuppression were contrary to those obtained by the companies, as TAC, AZA and steroids had an incremental cost-effectiveness ratio (ICER) above £30,000 relative to CAS and the same was found for CSA with MMF and steroids. Moreover, the technology assessment team at Birmingham found these results robust to uncertainty in the hazard rate used to extrapolate differences in AR rates to long-term estimates of health benefit.

These analyses represent the only available evidence about the costs and benefits of immunosuppressive regimens in recipients of kidney transplants aged ≤ 18. However, this evidence is based on regimens that may no longer represent routine practice in terms of therapies used (MMF has become part of standard immunosuppressive therapy) and dosages (lower doses of TAC are being used as they are perceived to have a better efficacy and safety profile).

As for the methodology behind this evidence, the assessment was based on a meta-analysis of the evidence on AR rates, although for MMF this included studies in adult patients. The study did not account for costs and HRQoL effects of changes in graft function and omitted the effect of differences between regimens in terms of the graft function on longer-term prognosis. Recent evidence from studies in adults suggest that quality of life133 and costs60 do vary significantly with renal function and this casts some doubt on the conclusion by the Birmingham group that small QALY differences are generally found between regimens. It is also questionable whether or not the surrogate relationship between AR and graft survival was validly implemented, because the estimated HR used to predict graft survival was estimated from AR rates occurring before discharge post transplantation, while the efficacy data used to model treatment differences were based on 12-month outcomes post transplantation. In addition, lack of data prevented the analysis from accounting for side effect differences between regimens, to which results were found to be sensitive. The quality assessment of these analyses are summarised in Table 45.

Review of economic models and their results in the submission

The submission provides an overview of model structures and conclusions of previous cost-effectiveness analyses of renal transplantation immunosuppressive regimes. From searches of electronic databases (NHS EED, The Cochrane Library, MEDLINE and other sources not specified) it identified and included 12 studies in its review (although the Astellas submission states that 11 studies were included). No details were provided about the inclusion criteria for the review of economic studies but all of the reviewed studies were conducted in adults.

One of the included studies compared TAC-IR with TAC-PR (US study140); four studies compared TAC with CSA (two in continental Europe,141,142 one in the UK143 and the remaining study was from the USA and measured only costs of medication113); seven studied SRL in CNI avoidance or minimisation strategies versus TAC (one from the USA,144 another from the UK,145 two more from Germany146,147 and three studies from Colombia, Mexico and Poland148–150).

The submission briefly described the main results of these studies without critically assessing their validity and applicability to a UK setting, although it mentions the limited transferability of results from non-UK studies (10 out of the 12). It concludes that the evidence supports the view that TAC is cost-effective when compared with CSA, but that it is ambiguous in relation to the comparison against SRL in a CNI avoidance or minimisation strategy. The submission also includes a section where three published models are described. 144,148,149 No assessment of their strengths and weakness was presented. These models are all of adult patient populations and are therefore not included in the review of cost-effectiveness studies of this monograph. 144,148,149

Economic evaluation by the company

The cost-effectiveness analysis submitted by Astellas is an adaptation of a published Markov model-based assessment of the cost-effectiveness of TAC, in either its extended release formulation, Advagraf, or the current standard therapy of immediate-release (Prograf151) in adult KTRs. The model describes the annual transitions between four health states starting from kidney-only transplantation: functioning graft without history of AR, functioning graft having experienced AR, graft failure (dialysis) and death (Table 46). Owing to the lack of child/adolescent data, the Astellas submission is based on a review of short-term safety and efficacy outcomes of immunosuppression in adults, reported by RCTs published study until June 2014. These were then extrapolated using registry data on child/adolescent graft and patient survival. The base-case analyses submitted by the company discount costs and QALY outcomes at an annual rate of 3.5%.

Critical appraisal

The analysis presented by Astellas (Table 53 shows the quality checklist) covers a number of appropriate comparators, including new regimens BEL and regimens with modes of action different from that of CNIs (i.e. EVL and SRL), as well alternative TAC formulations that are believed by the company to be used in routine practice (i.e. Modigraf and specials). However, it omits one relevant comparator: CSA. There is no justification in the submission as to why this drug regimen was not considered. This suggests that the results presented may be misleading owing to the exclusion of a relevant comparator. In addition, all of the regimens analysed by Astellas were evaluated in combination with MMF. This seems to contradict the assertion in the company’s submission that ‘Most children in the UK receive triple immunosuppression therapy with a CNI (CSA or TAC), a DNA proliferation inhibitor (usually azathioprine), and a CCS following kidney transplantation’ (Astellas’ submission, page 1). Astellas also reported the results of sensitivity analyses that varied the mean starting age of patients in the cohort modelled, but as the analysis was censored/stopped at age 18 years, it is difficult to assign any meaningful interpretation to their findings that the results were sensitive to such variation.

There are two logical concerns with the Astellas model-based analysis. First, by accounting for the advantages in adherence of Advagraf in its comparison with Prograf, it makes the comparison of outcomes of Advagraf with those of other immunosuppressive regimens in the model invalid, as no allowance was made for any effects of adherence on graft survival for the other regimens analysed in the model. Indeed this undermines the fundamental assumption in the model that all significant differences in any drug regimen comparison may be accounted for by the effect through the surrogate, in this case the rate of AR. 169 Thus, regardless of the validity of the comparative analysis of Advagraf and Prograf, indirect comparisons of model results between advagraf and SRL, and EVL and BEL are invalid. Second, although the model was adjusted to include the effect of adherence on graft survival in the Advagraf versus Prograf comparison, the patient survival curves (for the functioning and failed graft states) were left unchanged, thus the same set of patient survival curves was applied to all immunosuppressive options analysed. This implies the empirically questionable assumption that improvements in graft survival, such as those obtained with Advagraf relative to Prograf (and indeed relative to all other model arms), do not translate in direct patient survival benefits. This inconsistent logic in turn leads to underestimating the benefits of Advagraf and overestimating its costs.

Inspection of the Microsoft Excel^® 2010 (Microsoft Corporation, Redmond, WA, USA) model spreadsheets revealed that the TAC drug regimen options (Advagraf and Prograf) and EVL were the only treatment arms populated by actual data on immunosuppressive drug use from the RCT sample that served as the source for the respective efficacy data; drug consumption values for BEL and SRL regimens were based on treatment guidelines (BNF or SPC). Adult dosages (per kg body weight) of these treatments were used to estimate costs in the model. The only therapies for which child-specific doses were used in calculating resource utilisation in the analysis were MMF and EVL. There are important distinctions with adults that are likely to cast doubt on these drug dosage values. In particular, as acknowledged by the authors in relation to TAC pharmacokinetic studies, children and adolescents appear to eliminate the drug more rapidly than older adults. Further, in relation to steroids, there are concerns about the effects of the medication on growth, which are likely to lead to its more limited use in children and adolescents than in adults.

There is inadequate use of the registry data used to extrapolate short-term efficacy outcomes from RCT in the model. The model used the data from the NHSBT report 2013–1432 on patient survival rates for kidney-only transplant recipients in the UK (table 28, p. 35, Astellas’ submission) to populate the patient survival parameters of patients with a functioning graft, ignoring the fact that such data on survival rates were likely to include deaths from both patients with a functioning and a failed graft. Instead, the probability of death in the graft functioning state should have been calculated as the remainder of the annual probability of death from the NHSBT patient survival data minus the product of probability of mortality in the graft failure state and the proportion of patients with a failed graft. In other words, the Astellas model is likely to overestimate mortality in the functioning graft states, which, in turn, underestimates the benefits of gains in efficacy (i.e. reductions in AR in the model) that any regimen may have over another, for example TAC over the comparators. Thus, the results reported by Astellas in the submission may be treated as conservative estimates of the costs and benefits of its TAC regimes. In relation to the evidence presented in support of Advagraf, its quality is limited by the omission of CSA as a comparator therapy and the fact that the Advagraf versus Prograf comparison is based on what is, in effect, a different model of the outcomes of renal transplantation from that used to compare Prograf with all the other regimens. In fact, the model used for comparing Advagraf with Prograf contradicts the fundamental premise of the model used to compare Prograf with all regimens other than Advagraf: that AR captures all important drivers of clinically meaningful outcomes.

One other issue relates to the way the model was structured. Although the model allowed repeat transplantation to occur for a given individual, only for the first transplantation were the costs and HRQoL of subsequent dialysis accounted for. Although the proportion of patients with more than one retransplantation may be small, this assumption could have been important to the conclusions derived from the comparison with CSA, had such comparator been included.

In addition, Astellas chose to use values of time to retransplantation for patients on dialysis that were obtained from adult studies, whose mean wait for a retransplant was 3 years. 155 This was in contradiction with the company’s submission, which stated that ‘Children tend to be prioritised in deceased donor organ allocation systems: the median wait for a kidney in the UK during 2003–2006 for patients aged < 18 years was 277 days’. 170

There is also an anomaly with regards to the timing of transplantation. Markov models typically imply that transitions occur at the end of the period represented by each cycle. In the present case, the cycle length was 1 year and the authors of the Astellas model correctly decided on using half-cycle corrections to reduce the inaccuracy in expected costs and QALY calculations arising from more frequent average state transitions. However, the model assumed that a proportion of patients undergo retransplantation in the very first cycle and that these made a transition from the failed graft state to a functioning graft post retransplantation state as if the retransplant had occurred at the start of the period so that they spent the whole cycle length (6 months, owing to the half-cycle correction) with a functioning graft after retransplantation in the first cycle. However, this is incorrect as in a cohort of de novo kidney transplant patients, the discrete Markov process transition from a functioning first graft to a functioning retransplant requires two sequential intervening events to occur (i.e. graft failure and retransplantation) and a minimum of two cycles, one for each event.

In terms of the values used to populate the model, the costs of dialysis – one of the most influential parameters in the analysis – was derived from a microcosting study of the treatment pathway of a typical (i.e. adult) patient at six hospital units. This study166 sought to inform the introduction of Payment by Results in the NHS. 171 It did not include the costs of access surgery, managing dialysis complications and capital building costs. Reference costs for dialysis are now available that may reflect more representative data. 58 On this basis of this feature and the observation that children and adolescents tend to require higher staff-to-patient ratios than adults,2 it is expected that the costs of dialysis have been underestimated by the Astellas analysis.

The analysis does not account for discounts in price paid by hospitals for TAC-IR (Prograf), MMF, steroids and CSA (in the SRL CNI minimisation regimen), which were found to be one-third, one-tenth, one-tenth and a half of the list prices, respectively (see Tables 49 and 97). The implications of these differences are further explored in the next section (see Chapter 5).

Target population and subgroups

The target population was children and adolescents undergoing kidney-only transplantation (i.e. people receiving multiorgan transplants are not included). The upper age limit for the population ‘children and adolescents’ is not always clear as young people aged 16–18 years may receive their treatment in child/adolescent or adult centres. 172 Although some data sets include only young people aged < 16 years, the population for the economic assessment is children and adolescents aged < 18 years. The vast majority of transplant kidneys for this population come from DBD and living related donors (UK Transplant Registry standard data set, see Appendix 10 for further details) (Cathy Hopkinson, NHSBT, 15 October 2014, personal communication).

The UK Transplant Registry standard data set contains data on all solid organ transplants in the UK between 1995 and 2012. It allows linkage of multiple transplants for a single recipient and includes graft and patient survival (measured in days). A total of 34,803 records refer to kidney-only transplants, of which 29,759 recorded both graft and patient survival, 4937 recorded graft survival only (although it may be inferred that the patient survived at least as long as the graft), 24 recorded patient survival only, and 83 recorded neither graft nor patient survival.

The population modelled is incident KTRs and did not include prevalent KTRs (i.e. people who received a kidney transplant in the past) or those suffering from AR (although a number of the interventions separately have marketing authorisation for the treatment of AR).

To explore the impact of age at time of transplantation on cost-effectiveness, subgroups were identified by age (Table 55). In addition to this, the average cost-effectiveness of interventions was calculated by determining weighted average total discounted costs and QALYs for each year of age. It was assumed that the same number of transplants would be conducted in 16- and 17-year-olds as for 15-year-olds in order to estimate the cost-effectiveness for those under 18 years. No other subgroups were analysed as there was no evidence from child/adolescent RCTs identified in the systematic review of clinical effectiveness to support economic evaluation of these subgroups.

The weight and body surface area of child/adolescent KTRs are important for dosing and are highly dependent on age. It was assumed that the weight of child/adolescent KTRs would follow the median weight of UK children and adolescents160,161 (Figure 14). In scenario analyses it was assumed instead that the weight of child/adolescent KTRs would follow the ninth centile weight of UK children and adolescents to reflect the possibility that child/adolescent KTRs may have had their growth impaired by renal failure.

FIGURE 14.

Median weight of UK children and adolescents according to age.

Body surface area was then calculated from weight based on the table for body surface area estimation in the BNF for Children,173,174 as shown in Table 56.

Setting and location

The NHS in England (although some data sources have been UK-wide, particularly the UK Renal Registry and the UK Transplant Registry standard data set).

Study perspective

In line with the NICE reference case,165 the perspective adopted on outcomes was all direct health effects for patients (and, when relevant, carers) and the perspective adopted on costs was that of the NHS and Personal Social Services.

Interventions and comparators

As the immunosuppressive agents are used in combination and in sequence, we used treatment regimens as interventions and comparators rather than individual agents, although the cost-effectiveness of an individual agent versus another individual agent can then be evaluated by considering the cost-effectiveness of regimens which are identical but for the use of the intervention agent or the comparator.

Regimens were included as interventions or comparators if they were in current use in the NHS or if they would plausibly be used in the NHS and there was sufficient clinical evidence to estimate the costs and outcomes for KTRs receiving those regimens. It was necessary to include regimens that are not in current clinical practice to allow all the interventions being appraised to have their cost-effectiveness appraised. The only regimen which is a pure ‘comparator regimen’ (in that it contains no agents listed as interventions in the scope) is CSA + AZA.

Two regimens were included which were not included in the economic assessment for adults: BAS + TAC + AZA and r-ATG + TAC + AZA. The first was added as it is in common use in the NHS and the second was added to allow comparison of BAS and r-ATG in combination with TAC-IR and AZA.

Table 57 presents the regimens considered in this analysis as well as an indication of whether or not the Assessment Group believes the regimen to be a licensed combination for children and adolescents (however, no warranty or representation is given as to the correctness of the information presented in this regard, which reflects the Assessment Group’s understanding of the marketing authorisation as stated in the summaries of product characteristics; this understanding has not been confirmed by a clinician or pharmacist and, therefore, its accuracy cannot be guaranteed, particularly as regards drug combinations).

In its submission, Astellas also included the following regimens, which we have not modelled:

• SRL and CSA (with BAS induction) – note that we have modelled SRL and TAC without BAS induction (although the SPC for SRL specifies it is to be used in combination with CSA, we found significantly more RCT evidence in the adult population for which it was used in combination with TAC)

• EVL and CSA (with BAS induction) – note that we have modelled this without BAS induction because there were slightly more patients in adult RCTs receiving this regimen without induction

• TAC-IR (‘specials’ for first 3 years followed by Prograf for remaining life of graft) and MMF (with BAS induction)

• TAC-IR (Modigraf for first 3 years followed by Prograf for remaining life of graft) and MMF (with BAS induction).

The last two regimens are for children and adolescents who are unable to swallow Prograf capsules (although, inconsistently, they are assumed to be able to swallow MMF capsules and prednisolone tablets) and able to swallow Modigraf suspension (our expert advisory group has suggested some children cannot swallow Modigraf suspension and require fully liquid formulations, which can be purchased from specialist manufacturers rather than being prepared as specials by pharmacists or carers).

Time horizon

The time horizon was 50 years for consistency with the parallel HTA in adults and to ensure that all important differences in costs or outcomes between the technologies are included.

Discount rate

In line with the NICE reference case, the discount rate for costs and health effects was 3.5% per annum. 165

Choice of health outcomes

The primary health outcome of the independent economic assessment was QALYs for each comparator regimen, in line with the NICE reference case. 165

Secondary outcomes included:

• undiscounted life-years (life expectancy)

• undiscounted life-years with a functioning graft

• undiscounted life-years on dialysis

• likelihood of experiencing at least one episode of AR

• likelihood of developing NODAT

• likelihood of receiving a second, third or fourth transplant.

Decision tree

For each of the three RCTs in children and adolescents, a decision tree was created which calculated the following outcomes for each arm:

• costs (discounted and undiscounted) of immunosuppression, AR and AEs during the trial duration

• life-years up to the trial duration with functioning graft and with dialysis

• QALYs (discounted and undiscounted) during the trial duration

• for extrapolation using the Markov model

  • proportion of KTRs alive with functioning graft at the end of the trial duration

  • proportion of KTRs who are dialysis-dependent at the end of the trial duration

  • probability of AR within 12 months

  • probability of NODAT within 12 months

  • graft function (mean eGFR) at 12 months.

The discounted costs and QALYs from the decision tree and from the Markov model extrapolation were then combined. Cost-effectiveness results were presented both with ICERs and with incremental net health benefit figures (calculated at £20,000 and £30,000 per QALY). Cost-effectiveness results were also calculated by restricting the time horizon to the trial duration, that is, without extrapolating using the Markov model.

For simplicity, it was assumed that no KTRs losing their graft would be retransplanted within the trial duration. For Offner et al. ,73 with follow-up of only 1 year, this is likely to be a very reasonable assumption. For Grenda et al. 75 and Trompeter et al. ,77 with follow-up of 2 and 5 years, respectively, this may result in a bias against the arm with greater graft loss.