A randomised controlled trial and economic evaluation of intraoperative cell salvage during caesarean section in women at risk of haemorrhage: the SALVO (cell SALVage in Obstetrics) trial

Article history

The research reported in this issue of the journal was funded by the HTA programme as project number 10/57/32. The contractual start date was in October 2012. The draft report began editorial review in December 2016 and was accepted for publication in March 2017. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ report and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this report.

Declared competing interests of authors

none

Copyright statement

© Queen’s Printer and Controller of HMSO 2018. This work was produced by Khan et al. under the terms of a commissioning contract issued by the Secretary of State for Health. This issue may be freely reproduced for the purposes of private research and study and extracts (or indeed, the full report) may be included in professional journals provided that suitable acknowledgement is made and the reproduction is not associated with any form of advertising. Applications for commercial reproduction should be addressed to: NIHR Journals Library, National Institute for Health Research, Evaluation, Trials and Studies Coordinating Centre, Alpha House, University of Southampton Science Park, Southampton SO16 7NS, UK.

Background and rationale

Objectives

The primary objective of the SALVO trial was to determine if the routine use of cell salvage during both elective and emergency caesarean section, in women at risk of haemorrhage, reduced the need for donor blood transfusion in comparison with current practice, for which salvage is not routinely used. In addition, we sought to assess the consistency of the effect across subgroups defined by indication for caesarean, and to determine the effect on secondary outcomes including the number of units of donor blood transfused, fall in haemoglobin level, maternal morbidity resulting from postoperative anaemia (time to first mobilisation, duration of hospital stay and postnatal fatigue), maternal exposure to fetal blood, and its cost-effectiveness in comparison with current practice.

Trial design

The SALVO trial was a multicentre individually randomised controlled trial with cost-effectiveness analysis.

Setting

The trial was conducted in 26 hospitals with large obstetric units, in 23 NHS Trusts in England, Wales and Scotland (see Appendix 1 for a list of sites). These units each cared for between 3800 and 8000 births annually and performed between 900 and 2000 caesarean sections per year.

Participants

Screening and consent procedures

Screening and antenatal information

In addition to participant information sheets and informed consent forms, we provided sites with short patient pamphlets (which were used to provide information about the study during the antenatal period) as well as posters. All patient recruitment materials were approved by the Research Ethics Committee (REC) prior to use.

Information about the study was distributed to as many women as possible who were ‘booked’ to deliver at participating centres during their pregnancy and again on admission to delivery suite, whether they were intending a normal (vaginal) delivery or an elective caesarean section. This process was individualised at each participating centre depending on their routine practice to ensure that the maximum number of women were offered information well in advance of delivery. In some centres, women were provided with information about the trial at their routine anomaly scan appointment at 18–22 weeks’ gestation. The provision of study information was documented in the woman’s medical record or handheld notes and a sticker applied to indicate whether or not she was interested in taking part in the study. It was also documented at this point whether or not they would still be interested in taking part in the study in an emergency situation. Written informed consent was obtained by a trained health professional (obstetrician, anaesthetist or midwife) with delegated authority from the principal investigator. All women were assessed to ensure that they had the capacity to provide consent. The process and timing for obtaining written consent varied according to clinical urgency (see Figure 1).

FIGURE 1.

Consent procedure flow chart. CS, caesarean section; IOCS, intraoperative cell salvage.

Recruiting women undergoing elective caesarean section

Eligible women requiring elective caesarean section were provided with further information and the opportunity to ask questions at the time the operation was booked. They were then approached for written consent at the preoperative assessment clinic or on the day of surgery. Randomisation took place on the day of surgery.

Recruiting women undergoing emergency caesarean section

Women who were booked for delivery received information regarding the trial during their pregnancy so that there was sufficient time to consider participation in the trial, should an emergency caesarean section be required. On admission to delivery suite, women’s notes were checked to ensure this information had been supplied and the opportunity for further discussion provided.

After recruitment of half of the required target sample, a substantial protocol amendment was submitted and approved in order to facilitate recruitment of women undergoing emergency caesarean section. This allowed women to be approached for the first time on the delivery suite if they were found to be in the latent stage of labour (i.e. not yet in established first stage of labour according to NICE guidelines59,60) or were comfortable with epidural analgesia, provided that all of the following criteria were fulfilled.

• They were willing to receive the trial information and were subsequently willing to discuss the participant information sheet and have any questions answered if desired.

• They had either 0- to 3-cm cervical dilatation and were not contracting regularly (i.e. a maximum of one contraction in 10 minutes, with contraction lasting < 30 seconds) or they were comfortable with effective epidural analgesia in place.

• They were given at least 1 hour to decide whether or not they would be interested in taking part, should they require a caesarean section. If their situation changed (i.e. labour became established during that hour or they were no longer comfortable under epidural analgesia, or required a caesarean section before the hour elapsed), they were not approached for inclusion. After 1 hour, the women were approached for further discussion and the opportunity for questions about the study. If the women had a contraction during the discussion, the health professional involved would pause and wait for the contraction to finish. Permission to continue with the discussion was then sought.

Women in established labour (i.e. 4-cm cervical dilatation and regular painful contractions), or who were not comfortable with epidural analgesia, were not approached for the first time on delivery suite. Women who were distressed and not in a position to absorb the information on the patient leaflet were not approached for the first time on delivery suite.

Consent was obtained if a decision for caesarean section was made.

Documenting written informed consent

Consent comprised a dated signature from the woman and the dated signature of the person who obtained informed consent. It was clearly stated that the participant was free to withdraw from the trial at any time, for any reason, without prejudice to future care and with no obligation to give the reason for withdrawal. A copy of the signed informed consent document was given to the woman. One copy was retained in the woman’s medical notes and another by the principal investigator in the investigator site file.

Verbal consent and timing of written informed consent

All participants undergoing elective caesarean section gave written informed consent before the intervention. Likewise, the majority of emergency caesarean sections in the absence of acute fetal distress were conducted in a controlled manner with ample time for regional anaesthesia to be established, and written consent was obtained at this stage once the decision for caesarean section had been made.

In some emergency situations, the urgency meant that there would be insufficient time for written consent to be obtained prior to the emergency caesarean section. Under these circumstances, if the woman had the capacity to consent and had previously indicated an interest in taking part in the trial, verbal consent was obtained by an authorised health professional as described above, and documented on the randomisation checklist. Written consent was then sought once the urgency of the situation was over and the caesarean section complete.

Intervention

All staff were sufficiently trained and familiar in the use of the cell salvage machine, in accordance with local procedures and requirements. The majority of sites used conventional cell-saver machines with separate set-up for collection and processing of shed blood, whereas some sites used continuous transfusion systems. To confirm eligibility for randomisation, investigators needed to verify that women met the inclusion/exclusion criteria for the trial as well as gaining informed consent. An eligibility checklist was completed prior to randomisation.

Women were randomly allocated to either:

1. caesarean section with cell salvage (intervention group), set up routinely with collection of shed blood from the outset of surgery and return of any processed blood obtained

2. caesarean section without cell salvage (control group), with transfusion of donor blood according to standard local guidelines.

The intervention group was treated as follows: blood was aspirated from the surgical field, the red cell component isolated by centrifugation and retransfused after washing and filtration. The ability to return salvaged blood is dependent on sufficient volume being collected and processed. Blood was uniformly returned to women in the cell salvage group if this volume threshold was reached and it was a protocol requirement that cell-saver machines were fully set up for both collection and processing upfront at commencement of surgery and that all available processed blood was retransfused regardless of volume. The use of a LDF for transfusion of salvaged blood was not mandated as part of the study intervention protocol, but left up to local guidance. Any reports of severe, unanticipated hypotension and the potential association with the presence of LDFs were monitored. Likewise, the use of one versus two suction devices, the latter having one dedicated to amniotic fluid only at uterotomy, as well as salvage machine ‘bowl size’, was at the discretion of the participating site. Swab washing was encouraged, as it was thought to increase the volume of blood available for processing and, thus, for retransfusion,61 but was ultimately left to the local investigator’s discretion.

The control group was treated as follows: participants received standard current practice (without cell salvage) with allogeneic donor blood transfusion as the standard treatment, if required. In life-threatening acute haemorrhage, women were managed at the discretion of attending clinicians in line with the standard of care for such an emergency,1,40 potentially including the use of cell salvage in the control group.

Follow-up

Participants were followed up until discharge or transfer from the participating hospital only. Postnatal investigations included assessment of postoperative haemoglobin levels, collection of multidimensional fatigue inventory (MFI)62 questionnaires completed by patients (with any missed MFI questionnaires followed up for completion up to 2 weeks after discharge), documentation of AEs, mobilisation and discharge times, and, for RhD-negative women with RhD-positive babies, assessment of exposure to fetal blood by Kleihauer tests and anti-D given.

We took the opportunity to undertake an observational study of practice around anti-D prophylaxis in RhD-negative women who gave birth to a RhD-positive baby. There are UK guidelines stating that all RhD-negative women giving birth to a RhD-positive baby should receive a minimum of 500 IU anti-D Ig as a standard dose following delivery to minimise the risk of RhD alloimmunisation. 34 These guidelines, published in 2014, also recommend that after cell salvage the minimum standard dose should be higher, at 1500 IU anti-D. The maternal sample should be tested after delivery to assess the level of FMH to guide if additional anti-D doses are required following the standard dose. In the majority of centres, the Kleihauer test is undertaken as an initial screening test but, as this is a manual test with a high coefficient of variation, the guidelines also make further recommendations. Given the crudeness of Kleihauer test results, these guidelines recommend flow cytometry tests to be performed for Kleihauer test results of ≥ 2 ml, and repeat administrations and repeat testing after 72 hours for any Kleihauer test results of > 4 ml. 63 All centres participating in the SALVO trial would have been expected to have local guidelines on anti-D prophylaxis. We aimed to collect data around anti-D prophylaxis and FMH testing in all RhD-negative women recruited to this study to assess current practice. We did not attempt to collect follow-up data on the development of red cell sensitisation either to the RhD or indeed other red cell antigens in either group, as this was outside the scope of this particular study.

Outcomes

The primary outcome was the use of donor blood transfusion. Reducing the proportion of women with this outcome should lead to fewer transfusion-related complications.

Safety considerations

Adverse events were defined as any untoward medical occurrence in a participant receiving trial intervention, including occurrences that were not necessarily caused by, or related to, that intervention. An AE was therefore defined as any unfavourable and unintended sign (including an abnormal laboratory finding), symptom or disease temporarily associated with study activities.

A SAE was defined as an AE that fulfilled at least one of the following criteria: fatal, life-threatening, required prolongation of hospitalisation beyond 7 nights after caesarean section for maternal reasons, resulted in persistent or significant disability, was a congenital anomaly or birth defect, or was otherwise considered medically significant by the investigator.

The AEs and SAEs were documented if they occurred between randomisation and discharge. They were only reported if they related to the mother, except for SAEs that fulfilled the criteria of congenital anomaly above. The local principal investigator responsible for the care of the participant, or in his or her absence an authorised medic within the research team, was responsible for assessing the severity, causality and expectedness of an AE, and for assessing whether or not the event was serious according to the definitions given above.

If an AE was not defined as serious, the AE was documented in the participants’ medical notes (when appropriate) and on the case report form (CRF). All reported AEs were subject to a central medical review and coded and grouped by a clinician member of the trial team.

All SAEs occurring during the trial observed by the investigator or reported by the participant, whether or not attributed to the trial, were documented in the participants’ medical notes (when appropriate) and reported to the trials office within 24 hours of the site becoming aware of the event. All SAEs were followed up until resolution or the event being considered stable. The chief investigator, or a delegated clinical coapplicant, reviewed all SAE reports within 24 hours and raised any queries to be addressed to the sites. Locally, all serious incidents (such as maternal deaths) occurring at a UK NHS site were subject to root cause analyses. 64

Any SAEs considered both related to the intervention and unexpected were reported to the sponsor and the pragmatic clinical trials unit (PCTU) quality assurance manager within 24 hours, and to the main REC within 15 days. Although there were some known or theoretical potential risks associated with the trial (including maternal exposure to fetal blood, AFE, severe hypotension and transfusion reaction), none was considered to fulfil the criteria of being ‘expected’. Therefore, any SAEs that were at least possibly related to the trial intervention were reported as unexpected SAEs.

If applicable, it was the chief investigator’s responsibility to take any urgent safety measures to ensure the safety and protection of the clinical trial participants from any immediate hazard to their health and safety, in which case the REC was informed immediately by telephone and in writing within 3 days.

Annual progress reports to the REC included a listing of all related and unexpected SAEs. All SAEs were reported to the Data Monitoring Committee (DMC) and Trial Steering Committee (TSC) on the occasion of their meetings (i.e. every 6–12 months). The DMC viewed data with knowledge of treatment. In the event of a participant dying as a result of the study protocol or study interventions, any postmortem findings were to be provided to the chief investigator, who would report the findings to the DMC for continuous safety review.

Data collection and quality assurance

The SALVO study met the requirements of the Data Protection Act 1998,65 NHS Caldicott principles,66 the Research Governance Framework for Health and Social Care67 and REC approval. Identifiable information collected from participants, including name, date of birth, hospital number and contact details, was considered confidential and collected and stored only at the local NHS site. Study data were collected using paper CRFs, with all data being pseudonymised using a unique participant number and transmitted to the trials office by secure NHS e-mail transmission or post.

The following data were collected through CRFs.

• Before surgery: eligibility, obstetric history, indication for caesarean section, prognostic factors for haemorrhage, demographics, due dates and labour data, preoperative haemoglobin and platelet count.

• During surgery: time of delivery, time into and out of theatre, transfusion of donor blood products, set-up of cell salvage machine (if applicable), including consumables used and volume of blood returned, reasons for no return of salvaged blood, documentation of any technical failure of cell salvage, additional staff required in theatre owing to cell salvage.

• Between surgery and discharge: transfusion of donor blood products, postoperative haemoglobin, FMH measured by Kleihauer test, anti-D administration, flow cytometry for Kleihauer results of > 2 ml, repeat Kleihauer and anti-D administration for initial Kleihauer results of > 4 ml, time of mobilisation and discharge, MFI, AEs including admission to higher level of care (HLC).

The CRF data were verified by the trials office and queries raised with individual sites for discrepancies identified. Data were input at the trials office by delegated staff into a bespoke Oracle Database [11g Release 2 (Oracle Corporation, Redwood Shores, CA, USA)] with a Java [SE 7 (Oracle Corporation, Redwood Shores, CA, USA)] user interface, set up and managed by the PCTU. Data quality was monitored through source data verification on samples of patient records during on-site monitoring and during remote self-monitoring activities, and through central statistical monitoring with discrepancies raised from database extracts, highlighting outliers and discrepancies. On-site and self-monitoring activities also included verification of eligibility, informed consent and completeness of local trial documents according to a predefined trial monitoring plan.

On 69 occasions, stratification factors were found to be entered incorrectly during the randomisation procedure, but these were corrected to the true values in the analysis, which adjusted for stratification factors.

Additional quality control measures undertaken included a cross-check of primary outcome data against local transfusion laboratory records on recommendation of the TSC. Manual data entry was also subject to quality control procedures according to predefined procedures, with 100% of primary outcome data being checked and 10% of all other data being checked with an allowable error threshold of 2% for non-primary outcome data.

Sample size

Establishing a baseline rate for the primary outcome was not straightforward because estimates in the published literature for blood transfusion in caesarean section varied widely (1.8–23.5%). 44,68 Factors influencing this figure include country of origin, indication for caesarean section (emergency or elective) and local transfusion policy. Our audits in two centres conducted at the time of study planning put transfusion rates for an unselected caesarean section population at around 5%. A detailed audit of donor blood use at the Royal Hallamshire Hospital Sheffield during the period 2009–10, without cell salvage in routine use, was performed by cross-reference of perioperative records, blood bank data and electronic records stored in cell salvage machines. It reported that in a recent series of 1647 caesarean sections over 10 months, 89 women were transfused with donor blood, giving a rate of 5.4% (Ian Wrench, Royal Hallamshire Hospital Sheffield, 2011, personal communication). A similar audit at Birmingham Women’s Hospital of all caesarean sections carried out in 2006 showed that, of 1674 women, 83 (5.0%) received a transfusion. 43 Both auditing units delivered approximately 7300 women per year with a comparable caesarean section rate and could thus be considered representative of UK tertiary obstetric unit practice. Our pilot sample55 was too small to assist in providing reliable information on sample size calculations. In the light of reported contemporary observations and audited data on transfusion rates, the assumption of a 5% event rate was used to base the main sample size calculation on.

The expected effect estimate was informed by the literature. Our systematic review43 and its most recent update36 showed only one small trial, published in 1998,44 which randomised a total of 68 participants to either cell salvage or standard care. The transfusion rate was 23.5% in the control group and 2.9% in the cell salvage group. The control event rate was considerably higher than that observed in current UK practice and inconsistent with literature from other sources. 21,22 This was probably due to a sample at an exceptionally high risk of haemorrhage. Weaknesses that raise the risk of bias (e.g. inadequate concealment of randomisation) precluded reliance on it alone to inform our calculations. Non-obstetric literature evaluating cell salvage in interventions with a moderate to high risk of transfusion had two high-quality systematic reviews: a HTA report citing a RR of exposure to allogeneic blood of 0.59 (95% CI 0.48 to 0.73) with salvage22 and a Cochrane review reporting a RR of 0.62 (95% CI 0.50 to 0.70) for transfusion with salvage compared with normal practice. 21 Detecting a smaller effect size would have been possible but the larger sample size required had to be balanced against the cost and practicability of undertaking such a trial. Based on the HTA report and Cochrane review cited,21,22 we assumed a relative risk for the intervention effect of 0.6 (at a control event rate of 5%, the intervention group would have a transfusion rate of 3%).

Therefore, the planned sample size was a total of 3050 women (1525 per group), to detect an absolute difference in the transfusion rate of 2% and given a power of 80% for a two-sided test, a type I error rate of 5% and event rate of 5% in the control group. Our sample size allowed for primary outcome data and follow-up loss of 1% of randomised cases.

The planned trial sample was also to represent an even split between elective and emergency caesarean sections, the rationale for which was as follows: the primary event rate in the control group was based on data representing caesarean section across ‘all-comers’ in obstetric units, including both emergencies and elective cases. It included all indications at increased risk of haemorrhage. Ideally, this distribution would be faithfully and proportionally represented in the trial population, but there was good reason to suspect that clinicians would find it much more difficult or be more reluctant to recruit those patients at higher risk of haemorrhage, such as emergency indications or in cases of placental abnormality. Equally, a decision to limit recruitment to these high-risk groups alone, while desirable to maximise the primary outcome event rate and reduce sample size, was likely to result in reluctance to take part in the study at all. Adoption of such narrow eligibility criteria may have restricted sites from ever gaining a sufficient rate of recruitment to become confident in the trial processes and rendered the conduct of the trial unviable. In addition, at the time the study was designed, there was an increasing trend for obstetric units to have started utilising cell salvage in the routine, uncomplicated elective caesarean section population to facilitate the generation of an effective skill base among clinical staff to support the deployment of the technology when deemed necessary, even though the majority of these would not suffer significant blood loss. Therefore, a pragmatic compromise to these conflicting requirements was to exclude those elective cases with the very lowest risk of haemorrhage (elective first caesarean section for breech or maternal request) while at the same time prespecifying a desired equal distribution across elective and emergency cases.

Between June 2013 and March 2014, the majority of the elective patient population was recruited relatively rapidly, exceeding our projected target accrual. The emergency patient population was recruited more slowly, along with high-risk elective cases (see Appendix 2, Figure 11). Although sites adapted to the more challenging recruitment of these participants, particularly once the changes to the consenting procedures that were introduced through a substantial protocol amendment had started to take effect, an extension of the projected recruitment duration by 11 months was necessary to allow completion of the target sample size.

Randomisation

Randomisation to the allocated intervention (allocation ratio 1 : 1) was done using a bespoke web-based randomisation system hosted by the University of Bristol. Randomisation of participants was done on the delivery ward by local study staff. The randomisation used random permuted blocks of variable sizes to ensure that trial staff conducting randomisation could not reliably predict the next allocation. Randomisation was stratified by four criteria: centre, type of caesarean section (emergency vs. elective), presence of abnormal placentation versus normal placentation and multiple pregnancy (twins or more) versus singleton pregnancy.

Blinding

Allocation concealment with third-party randomisation helped minimise selection bias. However, given the nature of the intervention, it was not possible to blind local treatment staff and data entry staff to the allocation. Performance bias may cause transfusion rates to vary. This risk was minimised by ensuring that each centre had an intraoperative transfusion protocol for use in theatre and recovery to standardise operative transfusion triggers across both study groups in each centre. Some centres adopted an agreed haemoglobin threshold for transfusion, which was to be applied equally to both groups.

Sites were encouraged to blind postnatal carers to group allocation after caesarean section. The allocation was not recorded in routine case notes, but this did not represent formal blinding as theatre notes were available. The carers on postnatal wards were a different group of staff to the carers on labour wards and operating theatres, and it was on the postnatal wards that the decisions for postoperative donor blood transfusions were made, based on the postoperative haemoglobin level and maternal symptoms. In the event of the need for a donor blood transfusion, serum haemoglobin was measured by blood sample, and pre- and post-transfusion results recorded. This allowed monitoring of numeric transfusion thresholds between units and groups. In the event that between-group variations in haemoglobin transfusion triggers were indeed evident, consideration was given for adjusting for such differences in the final analysis.

The study statistician remained blinded until completion of data collection and sign off of the statistical analysis plan so as not to bias the analysis, and the chief investigator remained blinded until completion of the analysis. For interim reporting purposes to the DMC during the running of the trial, an independent statistician employed by the PCTU produced summaries of unblinded data for a closed report to the DMC.

Statistical methods

Governance and oversight

The SALVO trial was undertaken following clinical trials database registration (registry number ISRCTN66118656) and the required regulatory approvals and local NHS permissions (UK REC North West – Haydock, reference number 12/NW/0513). The study was funded by the National Institute for Health Research (NIHR) as part of the HTA programme (reference number 10/57/32).

A trial management group (TMG) was responsible for the day-to-day running of the trial, with support from the PCTU at Queen Mary University of London. The TMG reported to the TSC, which was composed of an obstetrician, an independent statistician and a consumer representative, and which convened every 6–12 months and provided overall supervision of the trial. This included giving advice on trial protocol and changes thereof, resolving problems brought to it by the TMG, monitoring the progress of the trial, protocol adherence and patient safety, considering new information and recommendations of the DMC and other authorities, and approving reports and papers for publication.

The DMC consisted of an independent statistician, obstetrician and anaesthetist. The DMC met approximately every 12 months during the running of the trial and reviewed accruing trial data in order to assess whether or not there were any ethics or safety issues as to why the trial should not continue. Interim reports were supplied to the DMC in strict confidence and included unblinded data provided by a PCTU statistician independent of the trial. The DMC formulated recommendations for the attention of the TSC. Both committees also monitored the pooled primary outcome event rate (i.e. across both arms) and formulated recommendations to encourage recruitment of the full spectrum of patients likely to benefit from the intervention.

Patient and public involvement

Working with organised consumer groups capable of identifying research priorities and introducing their ideas into research programmes was a crucial part of our activities leading to the trial. The National Childbirth Trust (NCT) significantly strengthened the project, being well placed to reflect on their experience in relation to avoiding the need for donor blood transfusion and to encourage participation. A volunteer for the NCT collaborated with the project from its inception, advised on the pilot protocol and agreed to provide representation on the TSC. An additional patient and public representative was identified through ‘Katie’s Team’, the Queen Mary University of London women’s health research advisory group, and included in the project at a later stage. This representative participated in TSC and clinical investigator group meetings, reviewed the plain English summary for this report, and advised on dissemination strategies.

In preparation for the trial, a survey was conducted among women who received cell salvage, showing that they perceived the intervention as reassuring, safe and preferable to donor blood transfusion (our primary outcome).

Summary of changes to the project protocol

No changes were made to the objectives, outcomes, eligibility criteria, sample size or statistical parameters during the course of the trial. Three substantial and four minor amendments to the protocol were implemented during the trial, which concerned changes to recruitment materials and strategies, clarifications and administrative changes to the protocol, and an extension of the overall recruitment period.

Participants

Between June 2013 and April 2016, 3054 participants requiring caesarean section from 26 participating hospitals were recruited (Figure 2). Of these, 26 participants had to be excluded owing to issues surrounding consent or eligibility, as follows.

• Nine participants gave verbal consent as per protocol (PP), but written consent could not be obtained postoperatively.

• Four participants gave written informed consent, but the consent form was destroyed or missing and consent could not be reobtained.

• Four participants had not given consent and were randomised in error, but were not exposed to any trial intervention.

• One participant was found to have given invalid consent owing to language issues.

• One participant gave verbal consent but withdrew consent after surgery.

• Seven participants were found not to have met the eligibility criteria.

FIGURE 2.

Participant enrolment and follow-up. Reproduced from © 2017 Khan et al. 72 This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Therefore, 1517 participants were assigned to cell salvage (intervention) and 1511 to usual care (control). Pregnancies resulting in vaginal delivery after assignment (n = 17 in the cell salvage group and n = 18 in the control group) were excluded from the analysis, as were patients who were transferred to a different hospital prior to delivery (n = 2 in the cell salvage group and n = 1 in the control group) and who were therefore lost to follow-up. However, the baseline characteristics for these 38 patients are included in the tables of baseline characteristics. For the analysis, this left 2990 participants (n = 1498 in the cell salvage and n = 1492 in the control group, respectively).

Baseline data

The main characteristics of participants were similar at baseline (Table 1). The distribution of participants across the different sites is summarised in Table 2.

Implementation of cell salvage

In the intervention group, 1432 (95.6%) participants received their allocated treatment with the cell salvage machine set up. There were 24 cases (1.6%) for which the cell salvage machine was unavailable or out of order and 42 cases (2.8%) for which the machine was simply not set up in deviation of the protocol. In the group receiving salvage, 50.8% had salvaged blood returned averaging 259.9 ml (Table 3). In the control group, 1434 (96.1%) participants received their assigned intervention without the cell salvage machine set up. In 15 cases (1.0%) the cell salvage machine was used in an emergency and in 43 cases (2.9%) it was set up from the start of the operation, in deviation from the protocol.

Primary outcome

Overall, the donor blood transfusion rate was 2.5% in the group assigned to cell salvage and 3.5% in the control group, but this result did not reach statistical significance [adjusted OR 0.65, 95% CI 0.42 to 1.01; p = 0.056 (Table 4)]. In a subgroup analysis exploring the consistency of treatment effects in procedures undertaken at different levels of urgency, the transfusion rate was 3.0% in women assigned to salvage and 4.6% in the control group among emergency caesareans (adjusted OR 0.58, 95% CI 0.34 to 0.99), whereas it was 1.8% in the intervention group and 2.2% in the control group among elective caesareans (adjusted OR 0.83, 95% CI 0.38 to 1.83), but the interaction was not statistically significant (p = 0.46, see Table 4).

There was no significant difference in the clinical effectiveness of the intervention between centres (p-value for random slope = 0.091). In a prespecified sensitivity analysis assuming that a donor blood transfusion would have been required, had salvaged blood not been returned in the control group (n = 31), the effect of cell salvage was significant [5.6% vs. 2.5%, adjusted OR 0.39, 95% CI 0.26 to 0.59; p < 0.001 (see Table 4)]. When excluding cases of placental abruption from the primary analysis, little difference in the results was seen, with one transfusion excluded in the control group as a result [3.4% vs. 2.5%, adjusted OR 0.67, 95% CI 0.43 to 1.03; p = 0.071 (see Table 4)].

We also reviewed primary outcome events against available transfusion guidelines in order to determine whether or not the lack of blinding introduced bias. When specific haemoglobin thresholds were defined in local guidelines, we compared these with participants’ reported postoperative haemoglobin values. We found 25 instances (cell salvage group n = 14, control group n = 11) in which donor blood was administered postoperatively without locally defined haemoglobin thresholds having been reached. The TMG did not deem the difference between intervention groups a cause for concern, also accounting for the fact that other less quantifiable factors may also be taken into account when deciding on donor blood transfusion. Therefore, we did not adjust our analysis accordingly.

Secondary outcomes

Allocation to cell salvage did not have an effect on the units of donor blood transfused [adjusted mean difference (MD) –0.12, 95% CI –0.80 to 0.57; p = 0.74 (Table 5)].

A small difference was detected between cell salvage and control groups for time to mobilisation [median 0.74 days vs. 0.72 days, adjusted hazard ratio (HR) 1.11 days, 95% CI 1.03 to 1.19 days; p = 0.006 (see Table 5)]. This represented a shorter absolute median time to mobilisation of 0.02 days, that is, approximately half an hour. A small difference was also observed in length of hospital stay [median 2.131 days vs. 2.126 days, adjusted HR 1.08 days, 95% CI 1.00 to 1.16 days; p = 0.050 (see Table 5)]. This represented a shorter absolute median hospital stay by 0.005 days, that is, approximately 10 minutes.

Analysis of postoperative haemoglobin levels showed no difference between cell salvage and control groups [adjusted MD 0.63, 95% CI –0.09 to 1.35; p = 0.085 (see Table 5)] and this was also the case for fall in haemoglobin level from baseline [adjusted MD –0.68, 95% CI –1.40 to 0.04; p = 0.066 (see Table 5)].

Among RhD-negative women giving birth to RhD-positive babies, allocation to cell salvage was associated with greater FMH, defined as a Kleihauer testing result of ≥ 2 ml [10.5% vs. 25.6%, adjusted OR 5.63, 95% CI 1.43 to 22.14; p = 0.013 (see Table 5)]. It should be noted that for 67 patients, Kleihauer testing was done but results could not be classified as they were reported as ‘< 4 ml’ or similar estimations, according to local guidelines; these results are therefore not available for analysis. Anti-D was routinely administered in the vast majority of RhD-negative mothers with RhD-positive babies (99.2% cases in control group, 98.6% cases in cell salvage group) (Table 6), although a total of three women across both groups did not seem to have received a minimum standard dose of anti-D following delivery, with a risk of RhD alloimmunisation. The dose of anti-D that was administered is summarised in Table 6, with further detail on management of large FMH detailed in Table 7. Out of the 140 RhD-negative mothers in the cell salvage group, only 40.6% received a dose of 1500 IU, with 56.5% receiving 500 IU. It is worth nothing that the updated guidance34 recommending a higher standard anti-D dose of 1500 IU after cell salvage was published in 2014; therefore, practice in some centres with doses of 500 IU may predate publication of these guidelines.

Kleihauer tests were undertaken on the majority of RhD-negative participants in both groups (92.2% cases in the control group and 95.0% cases in the cell salvage group, see Table 6). A total of 5% of women in cell salvage group and 7.8% women in the control group did not have Kleihauer tests undertaken following delivery. It should be noted that a dose of 500 IU of anti-D covers a FMH of 4 ml and, although a 1500-IU anti-D dose covers a FMH of ≈12 ml, a small proportion of women may have had an even higher fetomaternal bleed. A Kleihauer test is therefore recommended to determine if additional doses of anti-D are needed in addition to the standard dose.

Repeat Kleihauer tests were undertaken 72 hours post anti-D administration for the majority of applicable participants and further anti-D was administered for two participants (see Table 6). Overall, eight women (n = 2 in the control group and n = 6 in the intervention group) had larger instances of FMH with a Kleihauer test result of > 4 ml. Of the two in the control group, one woman had the largest observed FMH on the study, with 37 ml. She was managed as per guidelines with flow cytometry for confirmation of the FMH volume, additional anti-D dose administered and a repeat Kleihauer test undertaken. The other had an initial Kleihauer test result of > 4 ml and so was administered 1500 IU of anti-D but had no flow cytometry undertaken. Neither of these two women in the control group received cell salvage. Out of the six women in the intervention group, five had received cell-salvaged blood back. All six women had confirmatory flow cytometry done and repeat Kleihauer tests undertaken.

Breakdowns of FMH by sucker use and return of salvaged blood were also summarised (Table 8). On a descriptive level, sucker use appeared to have little effect on the proportion of participants experiencing FMH (28.3% when one sucker was used vs. 25.0% when two suckers were used). Return of salvaged blood appeared to increase FMH (13.0% in cases of no salvaged blood returned vs. 48.4% in cases for which salvaged blood was returned).

Administration of other donor products (namely fresh-frozen plasma, platelets and cryoprecipitate) is summarised in Table 9.

The MFI questionnaire was completed for 2408 (80.5%) participants. Analysis of MFI statement scores showed no significant differences between allocation groups for the fatigue categories of ‘general fatigue’, ‘physical fatigue’, ‘reduced motivation’ or ‘reduced activity’. There was a modest difference between the cell salvage and control groups for ‘mental fatigue’ [adjusted MD –0.30, 95% CI –0.59 to –0.01; p = 0.043 (Table 10)].

As for process outcomes, the cell salvage machine was set up for 1490 participants across both groups and salvaged blood was processed and returned for a total of 761 participants [35 out of 58 (60.3%) in the control group and 726 out of 1432 (50.8%) in the cell salvage group]. Volume of salvaged blood was recorded and summarised and reasoning behind no return of salvaged blood was also denoted, with the most common reason being that no blood was produced by the cell salvage machine. Further measures detailing implementation of the intervention, such as machine settings and equipment use, are summarised in Table 11.

Adverse events

There were 453 AEs reported in total across both groups, with 191 participants experiencing 220 AEs in total in the control group and 199 participants experiencing a total of 233 AEs in the cell salvage group. There was no significant difference between allocation groups for experiencing an AE (adjusted OR 1.02, 95% CI 0.81 to 1.29; p = 0.84, see Table 5). Table 12 shows details and descriptions of AEs. One case of transfusion reaction associated with allogeneic donor blood was observed in the control group. There was no case of AFE observed, with or without use of a LDF.

In 18 cases, AEs were considered to be related to cell salvage, with 15 events possibly related, two events probably related and one event definitely related to the intervention. Table 13 shows details of related AEs. Of the 18 AEs classed as related to cell salvage, the majority (n = 16) were also in the context of the use of a LDF. These included transient episodes of hypotension that might have been related to the return of cell-salvaged blood, as well as haemorrhagic and infective complications. As cell salvage removes clotting factors and platelets, it can theoretically lead to coagulopathy unless coagulation products are simultaneously given, and there is also a potential risk of returning infective agents such as bacteria in the salvaged blood. The ultimate judgement on whether or not these complications might have been caused by cell salvage lay with the local principal investigator.

There were 36 SAEs reported during the SALVO trial. Of these, 32 are included in the AE table (Table 14), with one SAE having three AEs pertaining to it. There were four additional SAEs concerning the offspring (e.g. congenital anomalies), which did not require reporting for the main AE analysis. One fatal event was observed among trial participants. It was considered unrelated to the intervention. This maternal death occurred in a patient who died on the sixth day following her delivery.

Two serious adverse reactions were reported in this trial (i.e. SAEs that were considered related to the intervention). The first was reported as a reaction to salvaged blood. The patient became tachycardic, flushed and had difficulty breathing, starting shortly after the start of the retransfusion and resolving completely once the transfusion was stopped. The event was classed by the local investigator as life-threatening and as most likely due to the use of a LDF. The second event was a sudden onset of hypotension, following retransfusion of 600 ml of cell-salvaged blood. The patient recovered fully. The event was also reported as life-threatening and as most likely secondary to the use of a LDF.

Further exploratory analyses

There was no significant difference in the clinical effectiveness of cell salvage on secondary outcomes between elective and emergency caesarean section (Table 15) or in the effect of cell salvage on reducing donor blood transfusion between participants with normal and abnormal placentation (p-value for interaction term = 0.28, Table 16).

In a sensitivity analysis assuming that a donor blood transfusion would have been required had salvaged blood not been returned in the control when the cell salvage machine was set up in an emergency (n = 8), the effect of cell salvage on donor blood transfusion was significant [4.0% vs. 2.5%, adjusted OR 0.56, 95% CI 0.36 to 0.86; p = 0.008 (Table 17)].

We observed that swab washing greatly increased the proportion of participants who received salvaged blood (16.0% when swabs were not washed vs. 81.3% when swabs were washed) and that the volume of blood returned was higher when swabs were washed [mean (SD) = 32.8 (100.5) when swabs were not washed vs. 219.3 (169.8) when swabs were washed] (Table 18). In a comparison between participants who did and did not have swabs washed within those who had the cell salvage machine set up, no significant difference in the transfusion rates was observed [adjusted OR 0.79, 95% CI 0.39 to 1.57; p = 0.50 (Table 19)].

Introduction

This chapter reports the economic evaluation carried out alongside the SALVO trial. The primary objective of the study was to determine whether or not the routine use of cell salvage during caesarean section, in women at risk of haemorrhage, reduced the need for donor blood transfusion compared with standard care.

Methods

To compare the costs and outcomes of cell salvage and standard care in the SALVO trial, a decision-analytic model was deemed the most suitable method of presenting the alternative pathways and collating the data for analysis and sensitivity analysis. In a decision-analytic model, consequences are expressed as probabilities, weighted against costs and outcomes to derive an expected value for each alternative option. 73 The economic evaluation took the form of a cost-effectiveness analysis from the perspective of the health-care provider based on the principal clinical outcome of the trial. The main comparison is the use of cell salvage versus standard care. The results are reported in terms of the additional cost per donor blood transfusion avoided by using cell salvage compared with standard care. Standard care is defined for the purposes of the trial as ‘transfusion of donor blood according to standard local guidelines’. Costs were calculated in 2014–15 Great British pounds. Given the objectives of the trial and the duration of follow-up, only a within-trial economic analysis was carried out and outcomes beyond this point were not considered relevant.

Model structure

A decision tree model was developed in TreeAge Pro 2016 (TreeAge Software, Inc., Williamstown, MA, USA). The structure was informed by the objectives of the study and the pathways indicated by the clinical data. The model pathways (Figure 3) represent that of the trial in which patients undergoing a caesarean section were randomised to receive either cell salvage or standard care. Square boxes represent decision nodes, for which there is a choice to be made between strategies. Circles represent chance nodes, for which there are a number of subsequent events that could happen and each event is assigned a probability that it will occur. Triangles represent terminal nodes, signifying the last stage in the model.

FIGURE 3.

Decision tree structure.

Figure 3 shows that the model starts with the choice of transfusion strategies considered in the SALVO trial:

• cell salvage

• standard care.

Women allocated to either transfusion strategy have a possibility of receiving the treatment to which they were allocated or not. In both pathways, if the cell salvage machine was switched on, women have a possibility of receiving cell salvage, either on its own or in combination with donor blood transfusion. There is also a possibility that the woman may not require a transfusion.

The pathways of the model represent, as far as possible, the clinical procedures carried out in the study. The model combines the probability of a woman following a particular path and the associated costs. Probabilities, detailed in Table 20, were obtained from the trial and attached to each pathway. The cost and outcome measures that were incorporated into the model were collected prospectively during the SALVO trial using forms filled out at the pre-, intra- and postoperative phase and at the time of discharge from hospital. Intraoperative resource use and costs were estimated as the mean cost per caesarean section procedure conducted for each treatment pathway in the model and postoperative resource use and costs were estimated as the mean cost per patient in both treatment strategies represented in the model.

TABLE 20 - Probabilities used in the model

Treatment pathway	Trial data (n/N)	Probability	Distribution
Cell salvage intended
Cell salvage intended → allocated treatment received (machine was on)	1432/1498	0.96	Beta
Allocated treatment received → salvaged blood returned	726/1432	0.51	Beta
Allocated treatment received → salvaged blood not returned	703/1432	0.49	Beta
Salvaged blood returned → donor blood transfusion given	22/726	0.03	Beta
Salvaged blood returned → donor blood transfusion not given	704/726	0.97	Beta
Salvaged blood not returned → donor blood transfusion given	9/703	0.01	Beta
Salvaged blood not returned → donor blood transfusion not given	697/703	0.99	Beta
Cell salvage intended → allocated treatment not received (machine was off)	66/1498	0.04	Beta
Allocated treatment not received → donor blood transfusion given	6/66	0.09	Beta
Allocated treatment not received → donor blood transfusion not given	60/66	0.91	Beta
Standard care intended
Standard care intended → allocated treatment received (machine was off)	1434/1492	0.96	Beta
Allocated treatment received → donor blood transfusion given	47/1434	0.03	Beta
Allocated treatment received → donor blood transfusion not given	1387/1434	0.97	Beta
Standard care intended → allocated treatment not received (machine was on)	58/1492	0.04	Beta
Allocated treatment not received → salvaged blood returned	35/58	0.60	Beta
Allocated treatment not received → salvaged blood not returned	23/58	0.40	Beta
Salvaged blood returned → donor blood transfusion given	4/35	0.11	Beta
Salvaged blood returned → donor blood transfusion not given	31/35	0.89	Beta
Salvaged blood not returned → donor blood transfusion given	1/23	0.04	Beta
Salvaged blood not returned → donor blood transfusion not given	22/23	0.96	Beta

Analysis

Given the objectives of the trial and the duration of follow-up, a within-trial economic analysis was carried out. The analysis took the perspective of the NHS following current recommendations from NICE. 82 The main economic analysis was a cost-effectiveness analysis with results expressed as cost per donor transfusion avoided.

We carried out three main analyses on the trial data. In analysis 1, the base case was based on the ITT principle. In this method, patients are compared within the treatment groups to which they were originally randomised irrespective of the treatment received. 83 This method of analysis allows the estimates to follow real-life scenarios in which patients may not always receive the planned treatment. Not using ITT analysis can often exaggerate the benefits of a given intervention. 83

Analysis 2 was based on the treatment received by patients irrespective of randomisation (the PP analysis). Within the SALVO trial, equal numbers of patients were randomised to either the cell salvage or control group. However, because some clinicians managing women in the control group had access to a cell salvage machine, it was possible that women in the control group could receive cell salvage in place of a donor blood transfusion. A PP analysis was carried out to look at the effect of treatment received on the outcome estimates. Therefore, in analysis 2, all patients who received cell salvage were compared with those who received standard care, irrespective of the treatment to which they were randomised.

Analysis 3 considered only patients who underwent an emergency caesarean section. This analysis was considered necessary as the SALVO trial found that numerically, there was a greater reduction in rate of transfusion within the emergency patient group than the elective patient group. This analysis followed the same methodology as analyses 1 and 2. Probabilities were obtained from the trial and attached to each pathway in the existing model. For the analysis, intra- and postoperative resource use data were obtained from the SALVO trial. Intraoperative costs were estimated for each item to arrive at a mean cost per caesarean section procedure for each treatment pathway in the model. Postoperative costs were estimated for each item based on their occurrence in each branch of the model to arrive at a mean cost per patient for each branch (Tables 35–37).

Sensitivity analysis

Deterministic sensitivity analyses and probabilistic sensitivity analyses (PSAs) were carried out for each analysis to explore the effects of the inherent uncertainty in parameter estimates on model results. Deterministic sensitivity analysis involves varying one or more parameters while keeping the others at their baseline value. Although deterministic sensitivity analyses can be helpful to identify which model inputs are important in driving a decision or to identify threshold values, comprehensive representation can be obtained by undertaking a probabilistic sensitivity analysis (PSA), in which the uncertainty around a parameter is represented with a probability distribution. Monte Carlo simulation was used to sample from these distributions to allow the effect of parameter uncertainty to be evaluated. This involved 1000 repeated random draws from the distributions to indicate how variation in the model parameters would affect the results and hence illustrate the decision uncertainty. Beta distributions were used for probability data and Gamma distributions for costs. 84,85

The results of the analyses are presented in terms of incremental cost-effectiveness ratios (ICERs), which reflect the additional cost per donor blood transfusion avoided of cell salvage compared with standard care. The results of the cost-effectiveness analysis are presented using scatterplots and cost-effectiveness acceptability curves (CEACs) to reflect sampling variation and uncertainties in the appropriate threshold cost-effectiveness value.

Results

Analysis 1: intention to treat

The results of the analysis are shown in Table 38. The strategy in which standard care was intended was the least costly, with the average cost per patient estimated at £1244. However, the cell salvage intended group was only slightly more expensive, with the average cost per patient estimated at £1327. The cell-salvage-intended strategy was the most effective at avoiding a transfusion. The estimated incremental cost-effectiveness ratio (ICER) for the cell-salvage-intended strategy compared with standard care was £8110 per donor blood transfusion avoided. This means that it would cost an additional £8110 to avoid a donor blood transfusion through cell salvage compared with standard care.

TABLE 38 - Results for the base-case analysis

Transfusion strategy	Average cost per patient (£)	Effectiveness: donor blood transfusion avoided	ICER (£)
Standard care intended	1244	0.965
Cell salvage intended	1327	0.975	8110

The scatterplot (Figure 4) shows the modelled uncertainty in the cost and effectiveness of the cell salvage intended strategy compared with the standard care intended strategy from 1000 Monte Carlo simulations. In this, the ICER of each simulation is plotted on the cost-effectiveness plane providing information about the joint density of the differences in cost and effectiveness between the two strategies. From Figure 4, it is evident that although cell salvage is a more effective transfusion strategy, it is uncertain whether it is more or less costly than standard care. The CEAC (Figure 5) shows that the probability that cell salvage is cost-effective increases as the willingness to pay for a donor blood transfusion avoided increases. Given that there is not a prespecified threshold of willingness to pay for a blood transfusion avoided, as in the case of QALYs of £20,000–30,000 are the recommended cut-off points by NICE, the identification of the probability of cell salvage being cost-effective is less straightforward. However, the CEAC shows that if the maximum willingness to pay for a donor blood transfusion avoided was, for example, £50,000, the probability of cell salvage being cost-effective would be 62%.

FIGURE 4.

Incremental cost-effectiveness scatterplot for donor blood transfusion avoided (ITT).

FIGURE 5.

Cost-effectiveness acceptability curve for donor blood transfusion avoided (ITT).

Analysis 2: per protocol

The results of the analysis are shown in Table 39. In terms of cost, standard care is, again, the least costly strategy with a mean cost per patient estimated at £1238. The cell salvage strategy was the most effective at avoiding a transfusion. The estimated ICER for the cell salvage strategy compared with standard care was £8252 per donor blood transfusion avoided.

TABLE 39 - Results for the PP analysis

Transfusion strategy	Average cost per patient (£)	Effectiveness donor blood transfusion avoided	ICER (£)
Standard care	1238	0.967
Cell salvage	1330	0.978	8252

The scatterplot (Figure 6) shows that although cell salvage is a more effective transfusion strategy, it is again uncertain whether it is more or less costly than standard care. This uncertainty has been graphed in the CEAC (Figure 7). The graph shows that if the maximum willingness to pay for a donor blood transfusion avoided was £50,000, the probability that cell salvage was cost-effective would be 63%.

FIGURE 6.

Incremental cost-effectiveness scatterplot for donor blood transfusion avoided (PP).

FIGURE 7.

Cost-effectiveness acceptability curve for donor blood transfusion avoided (PP).

Analysis 3: emergency caesarean

The results of the analysis are shown in Table 40. The strategy in which standard care was followed remains the least costly, but higher in comparison with the previous two analyses, with the average cost per patient estimated at £1352. The cell salvage group had an average cost per patient estimated at £1407, which was also slightly higher than the previous analyses. The cell salvage strategy continues to be the most effective at avoiding a transfusion. The estimated ICER for the cell salvage strategy compared with standard care was £13,713 per donor blood transfusion avoided.

TABLE 40 - Results for the emergency-caesarean analysis

Transfusion strategy	Average cost per patient (£)	Effectiveness donor blood transfusion avoided	ICER (£)
Standard care	1352	0.986
Cell salvage	1407	0.990	13,713

The scatterplot (Figure 8) shows again that although cell salvage is a more effective transfusion strategy, it remains uncertain whether it is less or more costly than standard care. The CEAC (Figure 9) shows that the probability that cell salvage is cost-effective remains between 47% and 55% as the willingness to pay for a donor blood transfusion avoided increases.

FIGURE 8.

Incremental cost-effectiveness scatterplot for donor blood transfusion avoided (emergency only).

FIGURE 9.

Cost-effectiveness acceptability curve for donor blood transfusion avoided (emergency only).

Discussion

Aim and overview

The provision of a reliable donor blood transfusion service has critical implications for maternal health. In health-care systems in which this is available, maternal mortality due to haemorrhage is almost one thousand fold less than in those in which it is not. 86,87 Donor blood transfusion is a safe intervention with remarkably few associated AEs, although these may be serious and even, although rarely, fatal. In the face of such a proven clinical intervention, any new technique seeking to further reduce mortality would have to be extremely effective and require an unfeasibly large trial in order to demonstrate it. Despite these limitations to evaluation of a new technology, for many clinicians this would be intuitively considered the aim of introducing cell salvage into obstetric clinical practice. More realistically, cell salvage might reduce reliance on donor blood, the production and delivery of which remains relatively expensive (£120 per unit78) and while representing a ‘pooled’ national resource, may suffer from local hospital shortages when consumption is increased by a case of unexpected severe haemorrhage. However, cell salvage is a technology with its own costs and, therefore, it was clear from the outset of the SALVO trial that the health economic evaluation would be a crucial aspect of the trial, in order to show a worthwhile reduction in spend on donor blood units that was greater than the cost of the intervention. With an increasing reluctance to transfuse donor blood to even quite substantially anaemic women, owing to the possible adverse effects, cell salvage offered the possibility of safely increasing postoperative haemoglobin levels, leading to additional savings in patient care and hospital stay. These again, required rigorous health economic evaluation to provide meaningful conclusions, particularly as other clinical interventions88 for optimising preoperative haemoglobin levels (e.g. iron therapy) or reducing intraoperative blood loss (e.g. tranexamic acid or interventional radiology) may be cheaper and more efficacious.

A number of end points were considered as candidates for a primary outcome. The selection of transfusion rate was ultimately a pragmatic one, based on objectiveness, ease of collection and the existence of pre-existing data. In addition, a patient questionnaire given to women who had undergone the procedure highlighted the reassuring nature of receiving one’s own blood instead of donor blood [see original SALVO protocol (URL: www.journalslibrary.nihr.ac.uk/programmes/hta/105732#/ (accessed August 2017)].

Main findings

This large, multicentre RCT demonstrated modest evidence that routine, prophylactic use of cell salvage during caesarean section reduced the need for donor blood transfusion. It was associated with increased maternal exposure to fetal blood among Rh-negative mothers. Small differences were observed between groups for time to mobilisation and length of hospital stay but not in other secondary outcomes. Although numerically there appeared to be a greater effect in the emergency group compared with the elective group, the difference in effect between subgroups was not statistically significant. Exploratory analysis did not suggest a subgroup benefit in women with abnormal placentation. Although it appeared to increase the volume of salvaged blood returned, the use of ‘swab washing’ when conducting cell salvage did not appear to affect the need for donor blood transfusion.

Health economic analysis demonstrated that it would cost £8110 to avoid a blood transfusion with the use of cell salvage as used in the study. This cost could potentially be reduced by varying both the indication for cell salvage in caesarean section and by changing the technique used. Set-up for a continuous cell-saver machine, with ‘collection’ only, until sufficient volume was obtained for processing, could save the cost of the processing consumables. Swab washing could be relinquished as a technique, as it did not appear to have any effect on donor blood transfusion rate.

Strengths and limitations of the trial

To our knowledge, the SALVO trial is the largest multicentre evaluation of cell salvage in caesarean section to date. The RCT was prospectively registered, robustly conducted, independently monitored, rigorously analysed and transparently reported (see Figure 2). This should provide for confidence in validity and reliability of the findings. The study sample was diverse, spread across more than 20 UK centres. Only two indications for caesarean section were considered exclusions: first elective caesarean section for either breech or maternal request. The very low probability that these cases might require donor blood meant that excluding them left only women with a recognisable increased risk of haemorrhage and, thus, potential need for transfusion, increasing the power of the study to detect a difference. This broad base for inclusion adds to the generalisability of the findings.

The trial recruited to target and had comparability at baseline and compliance with assignment, minimal loss on follow-up and primary outcome. A substantial challenge to the conduct of an individually randomised trial is obtaining consent from women in labour who require emergency caesarean. 89,90 Therefore, meeting the target for recruitment in a challenging clinical context was a major achievement. We considered this group to be the one most likely to derive specific benefit from this health technology. Another challenge was the promotion of equipoise among participating clinicians, who were keen to adopt the technology, without robust evidence in cases for which anxiety for life-threatening haemorrhage was high, particularly for cases of abnormal placentation.

Owing to the nature of the intervention and because mothers are usually awake for caesarean section, it was considered impractical to formally blind either clinicians or patients to the group allocation. In view of this and the local variation in transfusion practices, the trial collected transfusion policies from each unit and then reviewed transfusion decisions in light of these. Although clinicians were found to commonly give donor blood in deviation to local policy, no difference was found in the rate with which this non-adherence to local policy occurred between intervention and control groups.

A criticism of sample size and power with presumption of likelihood of type II error could risk erroneous conclusions. We highlight that a p-value that is in the region of a 0.05, regardless of the side of the threshold in which it lies, deserves careful consideration with respect to use of the evidence for guiding practice. The failure to achieve statistical significance cannot be attributed to insufficient data when the study is completed to the planned size with independent monitoring. We would like to propose the following considerations in interpretation of our main finding: (1) addition of new data does not guarantee that the p-value threshold for significance will be reached91 and (2) the point estimate is the most plausible estimate of the true effect. This being the case, we believe that our main finding meets the criteria for accurate decision-making.

From the outset, we were aware that we were investigating the effect of an intervention in a heterogeneous population in terms of baseline risk. This heterogeneity is principally derived from the indication for caesarean section. This situation guided the use of covariate-adjusted and subgroup analyses as an integral part of trial planning, analysis and inference. The credibility of findings in subgroup analysis depends on a number of factors. We planned these a priori and limited the number of subgroup analyses to the bare minimum in order to limit the risk of spurious significance associated with multiple hypothesis testing. Caution must be exercised in the conduct and interpretation of evidence derived from subgroup analysis; however, not investigating or ignoring results of subgroup analyses could also lead to incorrect inferences. Although interaction proved statistically insignificant, suggesting no evidence for an inconsistent effect of cell salvage between elective and emergency cases, taking into account the observed point estimates and CIs from subgroup analyses for the primary outcome measure, we believe that our findings concerning the effect in emergency caesareans merit consideration.

The role of our sensitivity analysis was to evaluate the integrity of the primary analysis conducted based on the ITT principle. The design and analysis was predicated on adherence to assignment, whether control or cell salvage. We sought for consistency between the results of primary analysis and the results of sensitivity analysis to examine the credibility of the main finding. We planned a priori to assess if the erroneous return of cell-salvaged blood in the control group could potentially avert the use of donor transfusion. In the throes of a developing surgical emergency, it may be thought of as a useful intervention to deal with ongoing haemorrhage by clinicians handling cases in the control group. We could not prevent such a clinical intervention in an ethically consistent trial policy. Therefore, we proposed to reclassify such cases as having experienced the primary outcome. Study structures and team members strove to promote adherence to assignment but protocol deviations are common in pragmatic trials and several cases assigned to the control group did indeed receive salvage, even in the absence of such an acute emergency situation. One proposal to handle this problem could include alternative trial designs, such as cluster randomisation, but this approach does not necessarily guarantee avoidance of performance bias and generally reduces statistical power. Our approach to sensitivity analysis maintained the ITT principle, avoiding PP and as-treated analyses that have a tendency to produce spurious significance. Our sensitivity analyses confirmed the main finding for the primary outcome.

The SALVO trial studied both emergency and elective (planned) caesarean sections, despite the fact that owing to a higher incidence of haemorrhage in the former, a significant result might be more likely to be detected if the trial had excluded electives. There were two reasons for this. First, some centres were known to have already commenced using cell salvage routinely for elective cases in the absence of any evidence, so the question of evaluating effectiveness in this group remained pertinent. Second, emergencies represented a population much more challenging to recruit. The elective sample gave centres the opportunity to deploy and prove the trial processes in a much more straightforward population. Despite this, two groups of electives were known to have extremely low rates of haemorrhage and were therefore excluded (first caesareans for either breech or maternal request). The effect of cell salvage in the emergency group, while non-significant with regards to interaction with the elective group, will be interesting to clinicians. On the other hand, the finding of no effect in the placental abnormalities subgroup is less relevant for policy as guidance already exists for use of salvage in these high-risk cases. 40

At the time that the SALVO trial was conceived and designed, cell salvage techniques used in obstetrics were markedly heterogeneous and dependent on local attitudes and expertise. It was clear from the outset that a robust trial would require close to optimal use of the intervention in order for the results to be accepted by the clinical community. Optimal use in this context maximises the volume of salvaged blood returned to the mother. Therefore, a number of technical aspects of the machine use were made mandatory for trial patients to achieve this aim. Some other aspects that might increase blood return were left to local preference, as it was felt that it would be difficult to commission enough recruiting centres to complete the study if these technical elements of cell salvage management were made mandatory. Subgroup analysis of swab washing failed to show any effect on the primary outcome. We conclude that there are no other identifiable mechanisms, utilising current cell salvage technology, which might significantly increase the return of salvaged blood to the patient compared with trial practices capable of casting any doubt on the validity of our results.

External validity and generalisability

We consider the external validity of the SALVO trial to be very robust. It was conducted in a broad group of obstetric units, ranging from large, tertiary teaching hospitals to small district general hospitals. It recruited from a diverse range of indications for caesarean section, with few exclusion criteria. The nature of the intervention was maintained as pragmatic as possible, consistent with efficacy and adequate recruitment, yet delivered in a predictable manner that is easy to emulate outside the context of a trial. The adherence to protocol was higher than expected for an intervention that had already begun to enter routine practice with a consequent potential loss of equipoise when the trial commenced.

Red cell immunisation

The UK SHOT haemovigilance scheme has repeatedly highlighted suboptimal practice in relation to the management of anti-D prophylaxis in cases of caesarean section with rhesus incompatibility. 92 They have stressed the need for improved awareness of national guidelines, supported by education and training among all health-care professionals involved. 92 We had the novel opportunity to observe practice around anti-D prophylaxis, in relation to cell salvage in particular.

Although total omissions of anti-D prophylaxis after delivery occurred in only a small number of cases in either group (total n = 3), there are many other opportunities for improved adherence to national guidelines, in particular regarding the recommended minimum anti-D dose of 1500 IU following cell salvage34 and the need for FMH testing to assess if further anti-D is needed beyond the standard dose. This highlights a need for close communication between clinicians and laboratory teams to ensure that relevant testing is undertaken followed by subsequent appropriate management to minimise the risk of RhD sensitisation.

Although secondary analysis indicated a significantly higher risk of FMH of ≥ 2 ml based on Kleihauer testing in the cell salvage group, the clinical implications of this result are unclear. We have incomplete data on flow cytometry results to confirm FMH volumes. Further analysis of available results indicate that the majority of women with bleeds > 4 ml did receive appropriate doses of anti-D with further follow-up testing to check for fetal red cell clearance as per guidelines, but there were some omissions. We are unable to comment on the overall efficacy of anti-D prophylaxis and the subsequent risk of RhD sensitisation. The SHOT scheme is currently collecting data on all pregnant women who have produced immune anti-D detected for the first time, to better understand the reasons underlying RhD sensitisation.

Although we are unable to comment on the risk of alloimmunisation to other red cell antibodies following cell salvage as opposed to standard care, our data support the assertion that the use of cell salvage significantly increases the risk of FMH. When just the women receiving cell-salvaged blood are considered, the rate of FMH, based on the definition used in the study, is more than four times that of the control group. The implications of this may go beyond the increased cost of a larger anti-D dose required in women receiving cell-salvaged blood. Although we did not specifically study other red cell antigens, it is plausible and likely that this increased FMH would cause increased rates of introduction of these antigens into the maternal circulation. As no anti-D equivalent is available for these antigens, maternal antibody formation will not be avoided and these antibodies (e.g. anti-Kell) may make cross matching blood for these women significantly more difficult in the future, incurring additional health-care costs. As the rate and severity of these potential complications is completely unknown, we were unable to incorporate it into our health economic analysis.

Adverse and serious adverse events

The AEs and SAEs were spread evenly across the two groups. An increase in the rate of AFE has long been considered a potential AE of returning cell-salvaged blood at caesarean section, even though research indicates it is removed by the cell saver. It is reassuring that we did not observe any cases of AFE in either group, but particularly not in the subgroup who actually received salvaged blood back, whether or not a LDF was used. It is notable that all the AEs definitely or probably related to cell salvage occurred when a LDF was in use, consisting of acute haemodynamic and respiratory reactions to the return of salvaged blood. These reactions have been well reported in the literature and are thought to be due to an effect the filter induces on the salvaged blood, rather than due to the blood per se. 18,23–26 We did not observe any definitely or probably cell-salvage-related AEs when a filter was not used. Out of the 15 possibly related AEs, a filter had been used in 13. These included a range of AEs, including infective and haemorrhagic complications that could easily have been unrelated to the use of cell salvage. Ultimately, it was for the local principal investigator to make this judgement on relatedness. There was one case of reaction to donor blood in the control group, from which the mother made a full recovery. This is in keeping with the known rates of reaction. 18 One maternal death occurred in a trial participant, who had been allocated to the cell salvage intervention group. A local case review was carried out by the trust involved and did not find any link to the use of cell salvage.

Adverse events from donor blood transfusion are potentially serious but are also very rare, with a rate of 1 in 16,000. 18 Mortality associated with donor blood transfusion is even more uncommon, with a rate of 1 in 100,000. 18 We have demonstrated that the cost to avoid a transfusion event when routinely using cell salvage in caesarean section is £8110. When considered with the observed increase in FMH and potential long-term effects of this, which currently remain uncharacterised, it remains unclear whether or not cell salvage is beneficial in this patient group. The exception to this is in cases such as for Jehovah’s Witnesses, for whom donor blood cannot be used and cell salvage represents the only therapeutic option. We are also unable to comment on the specific benefit in the group at a high risk of torrential haemorrhage such as placenta accreta, as we had insufficient recruits in this group to come to any meaningful conclusions.

Conclusions

Acknowledgements for contributions to the project

The team would like to thank the following people:

Rebecca Harmston and Fleur Parker for lay advice.

Harry Gee (retired, Clinical Academic Obstetrician), Pollyanna Hardy (National Perinatal Epidemiology Unit, University of Oxford), Richard Gray (Medical Research Council Population Health Research Unit, University of Oxford), Mike Grocott (NIHR Respiratory Biomedical Research Unit, University of Southampton) and Andrew Shennan (Department of Women’s Health, King’s College London) for their advice and support as part of the TSC and DMC.

Sarah Weist and Felipe Castro Cardona at Barts Health NHS Trust for lead research midwife support. Aoife Harvey for trial management support.

Anna Placzek, Malika Barakat and James Heighway at the Queen Mary University of London Women’s Health Research Unit for data management and administrative support.

Anitha Manivannan, Natasha Stevens, Jeanette Hansen, Amy Hoon, Irene Simmonds, and Zuhur Balayah at PCTU for monitoring, quality assurance and early trial coordination support. Nadine Marlin, Lauren Bell, Lauren Greenberg and Miland Joshi at PCTU for statistical support. Mike Waring, Tom Power and Sandy Smith at PCTU for database and IT support. Glenn Poon at PCTU for logo design.

Angela Devine and Mhairi Harkness for input into the original trial application.

All the women who agreed to participate in the SALVO trial.

Contributions of authors

Khalid S Khan (Professor, Women’s Health, Queen Mary University of London) was the chief investigator. He contributed to the study design and protocol writing, study management, oversight of study conduct and the initial writing and final editing of the report.

Philip Moore (Consultant Anaesthetist) was a coapplicant, contributed to study design and writing the protocol, provided day-to-day clinical advice, oversight as a member of the TMG, medical review of clinical study data, contributed to study management and the writing and final editing of the report.

Matthew Wilson (Consultant Anaesthetist) was a coapplicant, contributed to study design and writing the protocol, contributed to study management as a member of the TMG and to the writing and final editing of the report.

Richard Hooper (Senior Statistician) contributed to the design of the study, provided statistical supervision and advice as a member of the TMG, and contributed to the writing and final editing of the report.

Shubha Allard (Consultant Haematologist) contributed to the design of the study, to study management as a member of the TMG, provided transfusion-related supervision and advice, and contributed to the final report.

Ian Wrench (Consultant Anaesthetist) contributed to the design of the study, study management as a member of the TMG, provided clinical advice and contributed to the final report.

Tracy Roberts (Professor of Health Economics) contributed to the design and planning of the health economic analyses, provided health economic supervision and advice throughout the study, and contributed to the final report.

Carol McLoughlin (Health Economist) contributed to the health economic analysis plan, conducted the health economic analysis and contributed to the final report.

Lee Beresford (Statistician) wrote the statistical analysis plan, provided statistical support, performed the statistical analysis of the study and contributed to the writing and final editing of the report.

James Geoghegan (Consultant anaesthetist) contributed to the study design, led on the pilot study, provided clinical advice and contributed to the final report.

Jane Daniels (Senior Research Fellow) contributed to the study design and the final report.

Sue Catling (Consultant Anaesthetist) contributed to the study design, provided clinical advice and contributed to the final report.

Vicki A Clark (Consultant Anaesthetist) contributed to the study design, provided clinical advice and contributed to the final report.

Paul Ayuk (Consultant Obstetrician) contributed to the study design, provided clinical advice and contributed to the final report.

Stephen Robson (Professor of Fetal Medicine) contributed to the study design, provided clinical advice and contributed to the final report.

Fang Gao-Smith (Professor in Anaesthesia) contributed to the study design, provided clinical advice and contributed to the final report.

Matthew Hogg (Consultant in Obstetrics and Gynaecology) contributed to study conduct and oversight as a member of the TMG and contributed to the final report.

Louise Jackson (Health Economist) contributed to the health economic analysis plan, provided health economic analysis support and reviewed the final report.

Doris Lanz (Senior Clinical Trial Coordinator) supervised the general management and day-to-day conduct of the trial, supervised data collection and contributed to the writing of the final report.

Julie Dodds (Senior Research Manager) supervised the running of the trial, provided direction and support as a member of the TMG and contributed to the final report.

Publication

Khan KS, Moore PAS, Wilson MJ, Hooper R, Allard S, Wrench I, et al. Cell salvage and donor blood transfusion during cesarean section: a pragmatic, multicentre randomised controlled trial (SALVO). PLOS Med 2017;14:e1002471.

Data sharing statement

We shall make data available to the scientific community with as few restrictions as feasible, while retaining exclusive use until the publication of major outputs. Requests for anonymised data should be sent to the corresponding author.

Disclaimers

This report presents independent research funded by the National Institute for Health Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, NETSCC, the HTA programme or the Department of Health.