Trastuzumab for the treatment of HER2-positive metastatic gastric cancer

The manufacturer identified 1 phase 3 randomised controlled trial (ToGA) evaluating the efficacy of trastuzumab plus cisplatin and a fluoropyrimidine (that is, capecitabine or 5-fluorouracil) in people with inoperable locally advanced or recurrent and/or metastatic adenocarcinoma of the stomach or gastro-oesophageal junction. People in the trial had tumours with high levels of HER2 protein (see section 3.4), and had received no prior treatment for their advanced or metastatic disease.

The trial randomised 594 people to receive either chemotherapy (n=296) or trastuzumab plus chemotherapy (n=298). The chemotherapy group in the ToGA trial received intravenous cisplatin (80 mg/m²) on day 1 of each cycle with either oral capecitabine (1,000 mg/m²) twice daily for 14 days, or an intravenous infusion of 5-fluorouracil (800 mg/m²) on days 1 to 5 of each of the six 3-weekly cycles. In addition to six 3-weekly cycles of chemotherapy, the treatment group received trastuzumab (8 mg/kg loading dose on day 1, followed by 6 mg/kg intravenous infusion every 3 weeks) until disease progression. In both groups, the choice of fluoropyrimidine was at the discretion of the investigator; 87% received capecitabine and 13% received 5-fluorouracil.

People whose tumours were classed as HER2 positive were included in the ToGA trial. IHC and FISH tests were done at the same time (parallel testing) according to the trial protocol. At the time of randomisation, tumours that were IHC3 positive, or those that tested FISH positive were defined as HER2 positive. Changes in the understanding of HER2 testing during the ToGA trial resulted in HER2 positive being defined as tumours that were IHC2 positive and FISH positive, or IHC3 positive. From the full population of 594 in the ToGA trial, 446 people (75%) had tumours that met this narrower definition. The European marketing authorisation was granted for this population (referred to as the EMA subgroup). Of this subgroup, 218 people received treatment with chemotherapy alone and 228 people received treatment with trastuzumab plus chemotherapy.

At baseline, characteristics in the ToGA trial were balanced between the treatment groups. These characteristics included sex, age, weight, region, and type of tumour (that is, intestinal, diffuse or mixed tumour types). A high proportion of people in the ToGA trial were male (76%) and 55% were from Asian countries. Almost all people had metastatic gastric cancer (97%), and accordingly the marketing authorisation was granted for the metastatic disease only.

The primary outcome in the ToGA trial was overall survival. The trial was terminated early in accordance with a revised stopping rule recommended by the Independent Data Monitoring Committee. At the time of the clinical cut-off, the median duration of survival follow-up was 18.6 months in the trastuzumab plus chemotherapy group and 17.1 months in the chemotherapy alone group. The hazard ratio for overall survival in the EMA subgroup was 0.65 (95% confidence interval [CI] 0.51 to 0.83) corresponding to a median survival for the trastuzumab plus chemotherapy group of 16 months compared with 11.8 months for the chemotherapy alone group (4.2-month improvement in survival). A median survival of 13.8 months was reported for the total trial population receiving trastuzumab plus chemotherapy compared with 11.1 months in the chemotherapy alone group (2.7-month improvement in survival).

The manufacturer reported the results for secondary outcomes including progression-free survival and overall response rate. For the EMA subgroup the hazard ratio for progression-free survival was 0.64 (95% CI 0.51 to 0.79), corresponding to a median progression-free survival for the trastuzumab plus chemotherapy group of 7.6 months compared with 5.5 months for the chemotherapy alone group (a 2.1-month improvement in progression-free survival). Results for overall response rate were reported for the total trial population only, and demonstrated a statistically significantly higher overall response rate in the trastuzumab plus chemotherapy group (47.3%) compared with the chemotherapy alone group (34.5%), odds ratio 1.70 (95% CI 1.22 to 2.38, p=0.002).

Quality of life was assessed in the ToGA trial as a secondary objective using the European Organisation for Research and Treatment of Cancer (EORTC) QLQ-C30 (global health status, functioning and symptoms) and QLQ-STO22 (containing 22 items associated with dysphagia, pain, reflux, eating restrictions, anxiety, dry mouth, body image and hair loss). Both treatment groups in the trial showed improvements in quality of life over the course of treatment. A statistical analysis of differences in quality of life between the treatment groups was not presented. Additionally, EQ-5D data were collected at baseline and every 3 weeks until disease progression.

In both groups, 68% of people had grade 3 or 4 adverse events. The most frequent were disorders of the blood and lymphatic system, gastrointestinal disorders, and metabolism and nutritional disorders. More people treated with trastuzumab plus chemotherapy had asymptomatic reductions in left ventricular ejection fraction; however, the difference in symptomatic cardiac events was not statistically significant.

The manufacturer developed a Markov economic model to assess the cost effectiveness of trastuzumab plus chemotherapy to treat people with HER2-positive metastatic gastric cancer. The model had 3 distinct health states: progression-free survival, disease progression and death. The model had a cycle length of 1 month and an 8-year time horizon (considered to be a lifetime horizon). Both costs and benefits were discounted at a rate of 3.5%. One-way sensitivity analyses were undertaken on utility values, survival analysis, unit costs and various resource use assumptions. Probabilistic sensitivity analysis was also undertaken to explore parameter uncertainty in the model.

The treatment regimens included in the manufacturer's economic evaluation were:

• trastuzumab plus cisplatin and capecitabine

• trastuzumab plus cisplatin and 5-fluorouracil

• epirubicin plus cisplatin and capecitabine

• epirubicin plus cisplatin and 5-fluorouracil

• epirubicin plus oxaliplatin and capecitabine.

The clinical estimates (transition probabilities) in the model were derived from the overall survival and progression-free survival estimates from the trastuzumab plus chemotherapy group in the ToGA trial (EMA subgroup).

The manufacturer's literature search for comparator data did not find trials in which triple regimens including epirubicin were directly compared with triple regimens including trastuzumab. However, 4 studies were identified that evaluated 1 of the regimens of interest:

• Tobe (1992) and Kim (2001), which compared epirubicin plus cisplatin and 5-fluorouracil with cisplatin and 5-fluorouracil

• Yun (2010), which compared epirubicin plus cisplatin and capecitabine with cisplatin and capecitabine

• REAL-2 (2008), which compared the non-inferiority of capecitabine with 5-fluorouracil in triple chemotherapy regimens including epirubicin.
  
  Additionally, a meta-analysis was identified (Wagner 2006) evaluating the efficacy of a triple regimen including an anthracycline (epirubicin) compared with a regimen that did not include an anthracycline. However, the results of this analysis were not used in the economic model because the largest trial assessed a comparison between epirubicin, cisplatin and 5-fluorouracil and cisplatin and 5-fluorouracil plus mitomycin (Ross 2002), and the other trials were presented in abstract form (Kim) or only included a small population with more severe disease (Tobe).

The manufacturer explored the possibility of conducting a network meta-analysis between the 3 comparator triple regimens used in UK clinical practice (that is, epirubicin plus cisplatin and capecitabine, epirubicin plus cisplatin and 5-fluorouracil, and epirubicin plus oxaliplatin and capecitabine) and the chemotherapy comparator group from the ToGA trial, to obtain the comparator effectiveness data for the model. However, the manufacturer concluded that the results of a network meta-analysis would not produce reliable or meaningful results because the only study to evaluate epirubicin plus cisplatin and capecitabine compared with cisplatin and capecitabine (Yun) did not report the primary outcome of the ToGA trial (overall survival). Additionally, the studies that compared cisplatin and 5-fluorouracil with epirubicin plus cisplatin and 5-fluorouracil (Tobe and Kim) did not have comparable patient populations. Therefore, the manufacturer presented a narrative summary of the results from the Yun, Tobe, Kim and REAL-2 trials.

For epirubicin plus cisplatin and capecitabine, the manufacturer concluded that the estimates of overall survival and progression-free survival could be assumed to be equivalent to those from the chemotherapy comparator group in the ToGA trial. This was because the Yun study showed no evidence of a significant difference (hazard ratio of progression-free survival for epirubicin plus cisplatin and capecitabine compared with cisplatin and capecitabine 0.96, 95% CI 0.58 to 1.57). The dose of cisplatin in the ToGA trial was also considered to be higher than it would be in UK clinical practice when added to a triple regimen including epirubicin, and that the 2 regimens could therefore be regarded as equivalent.

For epirubicin, cisplatin and 5-fluorouracil the manufacturer concluded that the estimates of progression-free survival could be assumed to be equivalent to those from the chemotherapy comparator group in the ToGA trial. The evidence to support this conclusion came from the study by Tobe (hazard ratio of overall survival for epirubicin plus cisplatin and 5-fluorouracil compared with cisplatin and 5-fluorouracil 0.57, 95% CI 0.27 to 1.2) and the study by Kim (hazard ratio of overall survival for epirubicin plus cisplatin and 5-fluorouracil compared with cisplatin and 5-fluorouracil 0.83, 95% CI 0.42 to 1.61). Additionally, the manufacturer used an overall survival benefit of capecitabine over 5-fluorouracil (hazard ratio 1.15) from a meta-analysis (Okines 2009).

For the third comparator regimen (epirubicin plus oxaliplatin and capecitabine), the manufacturer concluded that it could be considered equivalent to epirubicin plus cisplatin and capecitabine in effectiveness. This was based on the REAL-2 trial. The manufacturer stated that the results of this study indicated that oxaliplatin was as effective as cisplatin (hazard ratio of overall survival for oxaliplatin arms compared with cisplatin arms 0.92, 95% CI 0.80 to 1.10). Overall survival and progression-free survival estimates for epirubicin plus oxaliplatin and capecitabine were therefore considered equivalent to those for epirubicin plus cisplatin and capecitabine.

Health-related quality of life was estimated for progression-free survival and progressive disease health states. A utility value for the progression-free survival health state was calculated using results from the EQ-5D data collected at baseline and then every 3 weeks until progression in the ToGA trial. The manufacturer estimated a baseline utility value of 0.7292, which increased daily by 0.000142 during progression-free survival. For the progressive disease health state, an estimate from the literature was used because EQ-5D data were not collected after disease progression in the ToGA trial. The utility value of 0.577 for progressive disease was taken from NICE's technology appraisal guidance on sunitinib for the treatment of gastrointestinal stromal tumours. Utility values associated with adverse events were not included in the model.

The model included costs for HER2 testing, drug acquisition, drug administration, monitoring during progression-free survival, treating adverse events, care costs in progression-free survival after chemotherapy treatment was stopped, and supportive care costs after progression of disease. The cost of HER2 testing was based on a sequential testing strategy in which only people who tested IHC2 positive received a FISH test. Total drug costs included an amount for wastage based on an assumption that 80% of centres using trastuzumab to treat gastric cancer would also use it to treat breast cancer and would share vials, thereby implying no wastage. Cardiac monitoring was assumed to be done using a multiple-gated acquisition scan or an echocardiogram and to take place once every cycle for people treated with epirubicin and once every 3 months for people treated with trastuzumab, in accordance with the summaries of product characteristics. The costs of grade 3 or 4 adverse events with an incidence of at least 5% in any of the treatment groups were included.

The model suggested that trastuzumab plus cisplatin and capecitabine compared with epirubicin plus cisplatin and capecitabine, and compared with epirubicin plus oxaliplatin and capecitabine, produced a mean gain of 4.8 months of life for both comparisons. Trastuzumab plus cisplatin and 5-fluorouracil compared with epirubicin plus cisplatin and 5-fluorouracil produced a mean gain of 4.3 months of life. In the manufacturer's incremental cost-effectiveness analysis of all 5 regimens, epirubicin plus cisplatin and 5-fluorouracil, and epirubicin plus oxaliplatin and capecitabine, were both less effective and more expensive than the other regimens (that is, they were dominated). Additionally, although the total cost of the trastuzumab plus cisplatin and 5-fluorouracil regimen was lower than that of trastuzumab plus cisplatin and capecitabine, it was also less effective and had a higher incremental cost-effectiveness ratio (ICER) than trastuzumab, cisplatin and capecitabine (that is, it was extendedly dominated). For the 2 remaining regimens, trastuzumab plus cisplatin and capecitabine had an additional cost of £13,064 and produced an additional 0.25 QALYs over epirubicin plus cisplatin and capecitabine. The ICER was £51,927 per QALY gained. The manufacturer examined other scenarios in a probabilistic sensitivity analysis. Distributions were applied to utility values, unit costs, monthly supportive care costs, adverse event probabilities, survival curves, parametric parameters, and progression-free survival monthly Kaplan–Maier estimates. This resulted in ICERs that ranged from £37,180 to £95,238 per QALY gained. The probability that trastuzumab plus cisplatin and capecitabine was cost effective at £30,000 was 0%.

The manufacturer also presented the results of pair-wise comparisons between trastuzumab plus cisplatin and 5-fluorouracil and epirubicin plus cisplatin and 5-fluorouracil, and between trastuzumab plus cisplatin and capecitabine and epirubicin plus oxaliplatin and capecitabine. The ICERs for these comparisons were £50,838 and £40,711 per QALY gained respectively.

The ERG stated that the manufacturer identified the only trial evaluating trastuzumab in the treatment of HER2-positive metastatic gastric cancer (the ToGA trial). This trial was a well-conducted phase 3 randomised controlled trial with appropriate validity assessment. The ERG considered that the ToGA trial demonstrated improved overall survival of people treated with trastuzumab when added to cisplatin plus either capecitabine or 5-fluorouracil. The economic model was considered appropriate for the decision problem and the general approach employed by the manufacturer to estimate lifetime cost effectiveness met the requirements of the NICE reference case.

The ERG identified some small errors in the manufacturer's model related to inconsistent assumptions about adverse events, distributions of survival data and resource use. Correcting these errors reduced the ICER from £51,927 per QALY gained to £49,005 per QALY gained. This ICER represented an additional cost of £12,332 and an additional 0.251 QALYs from treatment with trastuzumab plus cisplatin and capecitabine over epirubicin plus cisplatin and capecitabine.

The ERG highlighted the following main areas of concern in the manufacturer's submission:

• The relevance of the trial data to a UK population.

• The comparator data considered for network meta-analysis.

• The relative effectiveness estimates of comparators in the model.

• The utility values applied during progression-free survival.

• The frequency of cardiac monitoring with trastuzumab and epirubicin.

• The use of parallel and sequential HER2 testing strategies.

The ERG stated that the efficacy of standard triple regimens in people with HER2-positive gastric cancer was unknown. It was assumed that the HER2-positive population is equivalent to a mixed-HER2 population containing an unknown proportion of HER2-positive people. The ERG noted that the trial population in the ToGA trial was substantially younger than the UK population of people with gastric cancer, of whom only 17% die before the age of 65. The population of the ToGA trial also differed substantially from the UK clinical population, in that over 50% of people in the trial were from Asian countries.

The ERG considered that the most relevant comparators in the context of current clinical practice were epirubicin plus cisplatin and capecitabine, epirubicin plus cisplatin and 5-fluorouracil, epirubicin plus oxaliplatin and capecitabine, and epirubicin plus oxaliplatin and 5-fluorouracil. However, it stated that epirubicin plus oxaliplatin and capecitabine was likely to be used more often in routine clinical practice in England and Wales than the 6% suggested in the manufacturer's submission. Following clarification by the manufacturer, the ERG concluded that all relevant trials had been identified for inclusion in a possible network meta-analysis to establish a link between the triple regimens with the chemotherapy comparator group of the ToGA trial. The ERG considered that the manufacturer's decision not to attempt a network meta-analysis using these trials was justified, mainly because of differences in the trial populations.

The ERG stated that the manufacturer's assumption of no difference in effect between the chemotherapy comparator regimen in the ToGA trial and the comparator regimens in the economic evaluation largely rested on the critique of the meta-analysis by Wagner. This study found a benefit for triple regimens including an anthracycline over regimens that did not include an anthracycline. The ERG stated that the manufacturer's decision to exclude the meta-analysis on the basis that it contained a trial that had only been published in abstract form (Kim) and a trial with a small population of people with more severe disease (Tobe) was inconsistent with the decision to include these studies individually. As a result of dismissing the meta-analysis, the small phase 2 Kim, Tobe and Yun trials were used as evidence of no effect between double and triple regimens, and the ERG stated that these were underpowered to detect a statistically significant benefit of treatment. The ERG therefore considered that the manufacturer's assumption of no difference in effectiveness between the chemotherapy comparator group in the ToGA trial and triple regimens including epirubicin was not justified. The ERG noted that the balance of evidence suggested that there may be an advantage in adding epirubicin to a double regimen.

The ERG explored the effect on the ICER using the hazard ratio from the Yun study (0.96), which indicated a small benefit of adding epirubicin to a double regimen. When applied to progression-free survival only, the ICER increased from the revised base case of £49,005 per QALY gained to £49,754 per QALY gained. When applied to progression-free survival and overall survival, the ICER increased from £49,005 per QALY gained to £52,709 per QALY gained.

The ERG also explored the effect of the manufacturer's assumption of no difference between triple regimens including oxaliplatin and triple regimens including cisplatin. From the REAL-2 trial, a hazard ratio for overall survival and progression-free survival (0.87) was derived that suggested a benefit of oxaliplatin regimens over cisplatin regimens. As a result, the cisplatin regimen no longer dominated the oxaliplatin regimen in the incremental analysis, which increased the ICER for trastuzumab plus cisplatin and capecitabine. When applied to overall survival alone, the ICER increased from £49,005 per QALY gained to £50,745 per QALY gained. When applied to overall survival and progression-free survival, the ICER increased from £49,005 per QALY gained to £54,114 per QALY gained.

The ERG stated that the manufacturer's base-case approach was relatively optimistic because utility values were assumed to increase by 0.000142 for each day in progression-free survival. The ERG suggested that it would be more appropriate to apply a small decrease in utility values over time to reflect the change in utility over time for an equivalent group of people from UK general population norms for EQ-5D. The ERG calculated this utility decrement to be 0.003502 per year. The ERG explored the impact of applying this assumption to the corrected manufacturer's base case and reported an increase from £49,005 per QALY gained to £51,309 per QALY gained.

The ERG highlighted the manufacturer's base-case assumption that cardiac monitoring took place once every cycle with epirubicin and once every 3 months with trastuzumab. Although the summary of product characteristics for epirubicin recommends an echocardiogram before and after each treatment cycle, the ERG's clinical advisers considered that, in the UK, cardiac monitoring is not done this often. The ERG carried out an exploratory analysis assuming that cardiac monitoring was carried out at the same frequency for epirubicin as for trastuzumab. This resulted in the ICER increasing from the corrected base-case estimate of £49,005 per QALY gained to £50,816 per QALY gained.

The ERG raised a number of issues relating to HER2 testing in the manufacturer's base case. These included the potential costs of repeat tests needed because of test failures and that the effectiveness estimates in the model were derived from the ToGA trial, which used a parallel testing strategy rather than a sequential testing strategy. The ERG also noted the estimate that 17.8% of people with metastatic gastric cancer whose tumours are HER2-positive came from the full trial population; however, in the UK subgroup of people in the ToGA trial, this was 26%. To explore the impact of variations in the HER2-positive rate, the ERG undertook an exploratory analysis in which the HER2 rate was varied between 5% and 30%, indicating the lower and upper limits which it considered reasonable. The resulting ICERs from this analysis ranged from £52,866 per QALY gained for the lower assumption of 5% to £48,395 per QALY gained for the higher assumption of 30%.

To consider the combined potential impact of some of the uncertainties raised, the ERG undertook 2 alternative exploratory analyses in which the following key assumptions were considered to be equally plausible:

• The hazard ratio of epirubicin plus oxaliplatin and capecitabine compared with epirubicin plus cisplatin and capecitabine was changed to 0.87.

• The hazard ratio of epirubicin plus cisplatin and capecitabine compared with cisplatin and capecitabine was changed to 0.96 for progression-free survival and overall survival (indicating a survival benefit for epirubicin plus cisplatin and capecitabine compared with cisplatin and capecitabine alone).

• Utility values during progression-free survival were changed to incorporate an expected decrease in utility in line with the general population over time.

• The frequency of cardiac monitoring for epirubicin was changed to be the same as trastuzumab.
  
  The ERG presented the results of these exploratory analyses separately for sequential and parallel HER2 testing strategies. In both analyses, epirubicin plus cisplatin and capecitabine was no longer the dominant comparator regimen and the relevant comparator for trastuzumab plus cisplatin and capecitabine was epirubicin plus oxaliplatin and capecitabine. When trastuzumab plus cisplatin and capecitabine was compared with epirubicin plus oxaliplatin and capecitabine, the QALY gain was 0.149 at an incremental cost of £9,987, giving an ICER of £66,982 per QALY gained using a sequential testing strategy. When a parallel testing strategy was assumed, the QALY gain was the same, however, the incremental costs increased to £10,681, giving an ICER of £71,637 per QALY gained.

Finally, the ERG undertook a separate exploratory analysis in which the hazard ratio was 0.77 (indicating a survival benefit) for epirubicin plus cisplatin and capecitabine, and epirubicin plus oxaliplatin and capecitabine, compared with cisplatin and capecitabine alone. This estimate of effectiveness was inferred from the Wagner meta-analysis in which a triple regimen including an anthracycline therapy had been compared with a regimen without an anthracycline. When applied to overall survival only, the QALY gain was approximately 0.13 at an incremental cost of £10,993, giving an ICER of £84,373 per QALY gained. When applied to overall survival and progression-free survival, the QALY gain was approximately 0.13 at an incremental cost of £11,226, giving an ICER of £99,797 per QALY gained.

The manufacturer's response to the appraisal consultation document (ACD) included an alternative base-case analysis and a new economic analysis for a subgroup of people from the ToGA trial who were IHC3 positive.

The manufacturer's alternative base case incorporated 3 of the 4 parameter changes in the ERG's exploratory analysis (see section 3.33). These were:

• a hazard ratio of 0.87 for epirubicin plus oxaliplatin and capecitabine compared with epirubicin plus cisplatin and capecitabine

• a hazard ratio of 0.96 for overall survival for epirubicin plus cisplatin and capecitabine compared with cisplatin plus capecitabine

• the same frequency of cardiac monitoring for epirubicin as trastuzumab.
  
  The decrement in utility values during progression-free survival in the ERG's exploratory analysis was not incorporated. The manufacturer included the estimate of increasing utility values during progression-free survival used in its original base-case analysis. The ICER for trastuzumab plus cisplatin and capecitabine in comparison with epirubicin plus oxaliplatin and capecitabine was £62,829 per QALY gained. The other treatments in the analysis were either dominated or extendedly dominated. A probabilistic sensitivity analysis for the same comparison produced an ICER with a mean value of £67,786 per QALY gained.

A new economic analysis based on a subgroup of people who tested IHC3 positive (that is, people with very high levels of HER2) was provided by the manufacturer. The analysis included 279 people, of whom 144 received treatment with trastuzumab plus cisplatin and either capecitabine or 5-fluorouracil, and 135 received treatment with cisplatin plus either capecitabine or 5-fluorouracil.

The clinical evidence submitted for the IHC3-positive subgroup included an analysis of overall survival. Analyses of progression-free survival and time to progression were provided as commercial-in-confidence. The hazard ratio of overall survival for the trastuzumab plus chemotherapy group compared with the chemotherapy alone group was 0.57 (95% CI 0.41 to 0.79), corresponding to a median survival for the trastuzumab plus chemotherapy group of 18 months compared with 12.4 months for the chemotherapy alone group (5.6-month improvement in survival). The manufacturer also provided an adjusted hazard ratio to account for an imbalance in baseline characteristics between the treatment groups in the IHC3-positive subgroup. The adjusted hazard ratio of overall survival for the trastuzumab plus chemotherapy group compared with the chemotherapy alone group was 0.51 (95% CI 0.36 to 0.72).

The manufacturer presented the results of an economic analysis for the IHC3-positive subgroup using the adjusted hazard ratio. The modelled mean overall survival for trastuzumab plus cisplatin and capecitabine was 7.4 months when compared with epirubicin plus cisplatin and capecitabine, and 6.2 months when compared with epirubicin plus oxaliplatin and capecitabine. The incremental analysis of all interventions resulted in an ICER of £42,969 per QALY gained for trastuzumab plus cisplatin and capecitabine compared with epirubicin plus cisplatin and capecitabine. This represented incremental costs of £16,654 and 0.388 incremental QALYs gained. All other treatment regimens were either dominated or extendedly dominated. A probabilistic sensitivity analysis produced a mean ICER of £43,970 per QALY gained.

The ERG commented that the manufacturer's alternative base-case analysis allowed the quality of life during progression-free survival to rise above that of the general population. It considered that this was not a plausible assumption. Therefore, the ERG re-ran the analyses presented by the manufacturer using a ceiling utility value equal to general population utility value estimates. The results were presented separately for deterministic and probabilistic analyses. The incremental analysis of all interventions resulted in an ICER of £63,081 per QALY (deterministic) and £71,463 per QALY gained (probabilistic) for trastuzumab plus cisplatin and capecitabine compared with epirubicin plus oxaliplatin and capecitabine. All other regimens were either dominated or extendedly dominated.

The ICERs calculated by the ERG for the IHC3-positive subgroup were presented separately using the manufacturer's adjusted hazard ratio (stratified analysis, see section 3.39) and the unadjusted hazard ratio (unstratified analysis). The incremental analysis of all interventions using the stratified hazard ratio resulted in an ICER of £43,206 per QALY gained (deterministic) and £44,490 per QALY gained (probabilistic) for trastuzumab plus cisplatin and capecitabine compared with epirubicin plus cisplatin and capecitabine. When the unstratified hazard ratio was used, the optimal comparator treatment in the incremental analysis changed depending on whether a deterministic or probabilistic analysis was used. In the deterministic analysis, the ICER was £49,970 per QALY gained for trastuzumab plus cisplatin and capecitabine compared with epirubicin plus cisplatin and capecitabine. For the probabilistic analysis, the ICER was £51,934 per QALY gained for trastuzumab plus cisplatin and capecitabine compared with epirubicin plus oxaliplatin and capecitabine.

Full details of all the evidence are in the manufacturer's submission and the ERG report.