Erythropoiesis‑stimulating agents (epoetin and darbepoetin) for treating anaemia in people with cancer having chemotherapy

Outcomes related to anaemia

4.1.7 The Assessment Group's random‑effects analysis of mean haemoglobin change included 16 trials (n=3170) and showed a statistically significant weighted mean difference (WMD) between patients treated with an ESA and patients treated without an ESA of 15.9 g/litre from the start to the end of treatment (95% confidence interval [CI] 1.33 to 1.84). There was considerable heterogeneity between the trials (I²=75.9%, p<0.001), although in all trials ESAs increased haemoglobin concentration. The fixed‑effects analysis also showed a statistically significant difference in haemoglobin change in favour of the ESA group, and also showed considerable heterogeneity (WMD 14.9 g/litre, 95% CI 1.37 to 1.60, I²=75.9%, p<0.001). Although ESAs increased haemoglobin concentration across all subgroups, there were statistically significant (p<0.05) differences between the types of ESA and between chemotherapy treatments. The analysis showed that epoetin treatment offered greater benefits than darbepoetin treatment, and that the ESAs were more effective in the trials with mixed chemotherapy than in the trials with platinum‑based chemotherapy only, trials with non‑platinum‑based chemotherapy only or trials in which the cancer treatment was not reported. The Assessment Group emphasised that the subgroup results were uncertain because of the small number of studies, and because it had not adjusted the subgroup analyses for multiple testing.

4.1.8 Using a random‑effects model and the results of a meta‑analysis of 10 trials (n=2228), the Assessment Group reported a statistically significant difference in haematological response in favour of ESA treatment compared with treatment without an ESA (risk ratio [RR] 3.29, 95% CI 2.84 to 3.81). Using a fixed‑effects model, the risk ratio was 3.41 (95% CI 2.96 to 3.92). All the individual trials showed a beneficial effect of ESA treatment with little heterogeneity (I²=6.4%, p=0.383).

4.1.9 Fewer patients randomised to ESA treatment than patients randomised to treatment without an ESA (554 of 2480 compared with 835 of 2299 patients respectively) needed blood transfusions in the 22 trials that assessed transfusion needs. The risk ratio was 0.63 (95% CI 0.57 to 0.69) for the random‑effects analysis and 0.62 (95% CI 0.51 to 0.67) for the fixed‑effects analysis, indicating a statistically significant difference between treatment arms. The Assessment Group found little heterogeneity (I²=10.5%, p=0.315), and all but 1 study showed a beneficial effect of ESA treatment.

4.1.10 In addition to evaluating whether patients needed blood transfusions, the Assessment Group evaluated whether there was a difference in how many units of blood a patient having transfusions needed. The Assessment Group reported that, based on 10 studies evaluating 1920 patients, patients randomised to an ESA compared with patients not randomised to an ESA needed fewer units of blood transfused (WMD −0.87, 95% CI −1.28 to −0.46 using the random‑effects model; and WMD −0.64, 95% CI −0.79 to −0.48 using the fixed‑effects model). The Assessment Group found moderate heterogeneity between trials (I²=59.3%, p=0.006), and all but 1 study showed that patients treated with an ESA needed fewer units of blood transfused than did patients treated without an ESA. The effect of ESA treatment in reducing the number of units of blood transfused was consistent across all subgroups, except for the subgroup characterised by having taken part in studies with treatment lengths of 17–20 weeks (WMD 0.10, 95% CI −0.59 to 0.79).

Outcomes related to cancer

4.1.11 Whether or not a patient's cancer responded to treatment was measured as 'complete tumour response' in 7 studies comprising 1909 patients. Randomisation to ESA treatment was associated with a pooled risk ratio of 1.10 (95% CI 0.86 to 1.41) for complete tumour response compared with randomisation to treatment without an ESA. The Assessment Group did not find significant heterogeneity between the trials; however, the direction of effect did vary between trials. The fixed‑effects analysis showed similar results of no difference between patients treated with and without an ESA (RR 1.20, 95% CI 0.85 to 1.71). The Assessment Group highlighted that the review by Wilson et al. (2007) showed that randomisation to ESAs compared with randomisation to treatment without an ESA worsened tumour response (RR 1.31, 95% CI 1.08 to 1.60), whereas the review by Tonia et al. (2012) did not find any difference (RR 1.02, 95% CI 0.98 to 1.06).

4.1.12 To assess whether ESAs prolonged or shortened overall survival, the Assessment Group extracted summary data from the Cochrane review by Tonia et al. (2012) which had used individual patient data. The Assessment Group's meta‑analysis included 18 trials comprising 4454 patients, in which 818 out of 2317 patients in the ESA group and 744 out of 2137 patients treated without an ESA had died. The pooled hazard ratio for the association of treatment with an ESA and death was 0.97 (95% CI 0.83 to 1.13), showing no difference in survival; there was moderate heterogeneity between the trials (I²=42.4%, p=0.03). The fixed‑effects analysis showed a similar result (hazard ratio [HR] 0.98, 95% CI 0.89 to 1.08) as did the review by Wilson et al. (2007) (HR 1.03, 95% CI 0.92 to 1.16). This differed from the findings of the Cochrane review by Tonia et al., which reported that ESAs increased the risk of death (HR 1.05, 95% CI 1.00 to 1.11). The Assessment Group emphasised that its analysis included only studies complying with the licensed ESA starting dose, whereas the Cochrane review did not restrict trials based on dose.

4.1.13 The Assessment Group's meta‑analysis of mortality during the study period included 14 studies comprising 2967 patients. The Assessment Group reported no difference in the risk of death (HR 0.86, 95% CI 0.67 to 1.11) and no heterogeneity between the trials (I²=16.4%, p=0.274). The fixed‑effects analysis also showed no difference in the risk of death (HR 0.87, 95% CI 0.70 to 1.09), whereas the Cochrane review by Tonia et al. (2012) showed that ESA treatment increased the risk of death (HR 1.17, 95% CI 1.03 to 1.29).

Adverse events

4.1.14 The Assessment Group conducted meta‑analyses (using data from the Cochrane review by Tonia et al. 2012) to address whether, and to what degree, ESA treatment was associated with the following adverse events: thromboembolic events (14 trials, n=4013); hypertension (9 trials, n=2032); thrombocytopenia and haemorrhage (7 trials, n=1715); seizures (1 trial, n=289); and pruritus (pruritus, rash and irritation; 6 trials, n=869). The random‑effects analysis showed that ESA treatment increased the risk of thromboembolic events (RR 1.46, 95% CI 1.07 to 1.99), hypertension (RR 1.80, 95% CI 1.14 to 2.85) and pruritus (RR 2.04, 95% CI 1.11 to 3.75) compared with treatment without an ESA. ESA treatment was not associated with seizures (RR 1.19, 95% CI 0.33 to 4.38), or thrombocytopenia and haemorrhage (RR 0.93, 95% CI 0.65 to 1.34). The Assessment Group reported similar results for the fixed‑effects analyses.

Subgroups

4.1.15 The Assessment Group presented subgroup analyses exploring key elements of the recommendations in NICE technology appraisal guidance 142. It presented results for subgroups according to chemotherapy, cancer type and other anaemia treatments as follows: patients with any cancer having platinum‑based chemotherapy (5 trials, n=1119); patients with ovarian cancer having platinum‑based chemotherapy (1 trial, n=122); patients having iron supplementation (16 trials); and patients unable to have blood transfusions. The Assessment Group noted that the results were uncertain given the small number of studies supplying data for each subgroup, and it had not adjusted the results for multiple testing. However, it noted that, in patients having platinum‑based chemotherapy, the response was generally better than in patients who received other chemotherapy. The Assessment Group did not identify any trials that evaluated the use of ESAs in patients unable or unwilling to have blood transfusions. It commented that it had trouble interpreting the results of the trials that used ESAs with iron because there are many types of iron supplements, and because few publications reported these results.

4.1.16 The Assessment Group conducted an analysis of patients with a haemoglobin concentration of 110 g/litre or less at the start of treatment (14 trials) and with target haemoglobin values of 130 g/litre or less (2 trials); the Assessment Group considered that these patients more closely reflected the marketing authorisations for ESAs. For anaemia‑related outcomes, when using this subset of trials, the Assessment Group found estimates of the effectiveness of ESAs similar to those from meta‑analyses from all of the trials included in the review. The analysis showed that ESAs do not increase or decrease the risk of death (inclusion concentration of 110 g/litre or less; HR 0.91, 95% CI 0.70 to 1.20). The analysis also showed that the risks of thromboembolic events (RR 1.29, 95% CI 0.66 to 2.54) and hypertension (RR 1.68, 95% CI 1.03 to 2.74) were slightly lower in this subgroup than in the overall population. When assessing the 2 trials in which investigators also limited the target haemoglobin concentration to 130 g/litre or less, ESA treatment did not increase or decrease the risk of death (HR 0.50, 95% CI 0.20 to 1.23).

Health‑related quality of life

4.1.17 In its review of health‑related quality of life, the Assessment Group included 13 randomised controlled trials that measured quality of life. The Assessment Group reported that treatment with an ESA compared with treatment without an ESA improved quality of life more, and reported a difference in scores of Functional Assessment of Cancer Therapy‑Fatigue (FACT‑F) (WMD 2.54, 95% CI 1.42 to 3.65), with low heterogeneity between the trials (I²=14.9%, p=0.32). The Assessment Group reported similar results for its fixed‑effects analysis. The Assessment Group stated that a clinically important difference in quality of life is considered to be a value of greater than 3.0 (Cella et al. 2002). For the FACT‑General (G) and FACT‑Anaemia (An) outcomes, the Assessment Group included 3 studies that showed no difference between the treatment arms (FACT‑G: WMD 2.98, 95% CI −0.83 to 6.78; FACT‑An: WMD 2.60, 95% CI −0.52 to 5.72). However, the Assessment Group noted that the quality‑of‑life data were limited by substantial missing data.

Assessment Group's cost‑effectiveness analysis

4.2.4 The Assessment Group developed a simple empirical model to assess the cost effectiveness of ESA treatment. The model had 2 arms (treatment with or without an ESA) and 2 components: a short‑term component (during treatment and during the time over which the haemoglobin returns to normal concentrations) and a long‑term component. The Assessment Group modelled short‑term QALYs as changes in haemoglobin concentrations over time seen in the clinical trials, and long‑term QALYs by estimating overall survival in each arm and applying a long‑term utility common to both arms. The Assessment Group based the analyses on a lifetime time horizon from an NHS and personal social services perspective. Costs and benefits were discounted at an annual rate of 3.5%. The mean age of modelled patients was 59.1 years and the mean weight was 66.6 kg, which was taken from the Assessment Group's systematic review of clinical effectiveness.

4.2.5 To estimate the magnitude of effectiveness of ESAs, the Assessment Group used trials reporting intention‑to‑treat analyses. The Assessment Group took parameters, including difference in haemoglobin change from baseline, difference in number of red blood cell units transfused, overall survival hazard ratio, relative risk and probability of adverse events (thromboembolic events, hypertension and thrombocytopenia), directly from its random‑effects meta‑analyses. For other parameters such as change in haemoglobin from baseline in patients treated without an ESA and mean number of red blood cell units transfused in patients treated without an ESA, the Assessment Group calculated the weighted averages from the control arms of the studies included in its meta‑analyses. The Assessment Group estimated patients' baseline haemoglobin concentration as 103.8 g/litre based on the weighted average of the studies included in its review. In its base case, the Assessment Group assumed that all ESAs are equally effective.

4.2.6 The Assessment Group assumed a 'normalisation period' in the model, when patients' haemoglobin concentrations increase at a constant rate until they reach normal concentrations. Based on the opinions of clinical experts, the Assessment Group assumed the same rate (2 g/litre per week) for both treatment arms, that is, patients treated with or without an ESA during chemotherapy. This value was consistent with previous cost–utility studies.

4.2.7 To extrapolate overall survival in patients treated with or without ESAs, the Assessment Group first modelled survival in the control arm by taking a weighted geometric average of the overall survival rate seen in the control arms of all the included trials. It chose an exponential distribution. It then derived a hazard ratio from its meta‑analysis, and used this to estimate survival in the ESA arm. The Assessment Group estimated a mean overall survival of 2.76 years for patients treated with an ESA and 2.67 years for patients treated without an ESA. In its base case, the Assessment Group assumed that patients treated with ESAs died later than those not treated with ESAs, and used a hazard ratio of 0.97 in its base case. It also explored alternative scenarios, notably that treatment with ESAs does not prolong survival (HR=1.0).

4.2.8 To estimate the utility contributing to the short‑term QALY gains in the model, the Assessment Group did not use the FACT‑F scores measured in some of the trials. Instead, it modelled utility as a function of haemoglobin concentration. It used utility values from the literature, specifically from a study by Harrow et al. (2011) in which SF‑36 utility values were measured in 13,433 women in the USA with cancer and valued by the UK general public using the standard gamble technique to transform them to SF‑6D values. The Assessment Group highlighted that the patient population in the study included only patients who were not treated with ESAs and who may or may not have been having chemotherapy. The Assessment Group then expressed the SF‑6D values as EQ‑5D values using regression analyses from Brazier et al. (2004). The SF‑6D utility increase of 0.009 per unit increase in haemoglobin concentrations translated to an EQ‑5D utility increase of 0.028 per increase of 10 g/litre in haemoglobin concentration. The Assessment Group applied an increase in utility of 0.028 per each 10 g/litre rise in haemoglobin until a patient's haemoglobin concentration reached 120 g/litre. The Assessment Group adjusted for mean difference in haemoglobin concentrations between the treatment arms in the model.

4.2.9 To estimate utility in the long‑term component of the model, that is, after a patient's haemoglobin had reached 120 g/litre, the Assessment Group applied age‑related utility calculations from Ara and Brazier (2010) to the utilities reported by Tengs and Wallace (2000), resulting in a constant utility value of 0.76 for both treatment arms. The Assessment Group stated that, because of sparse data, the estimated utility was uncertain and could affect the overall QALYs accrued. The Assessment Group did not include disutilities associated with adverse reactions when calculating QALYs because it considered that the trials did not clearly report adverse events. However, the Assessment Group stated that including disutilities associated with adverse reactions would increase the ICERs because patients treated with ESAs experienced more adverse reactions than patients not treated with ESAs (see section 4.1.14).

4.2.10 To cost the ESAs, the Assessment Group used the list price per 1000 units from the British national formulary (BNF, March 2014) for Eprex (£5.53), Binocrit (£5.09), NeoRecormon (£7.01), Eporatio (£5.99) and Retacrit (£5.66), and per microgram for Aranesp (£1.47). To calculate a mean weekly dose, the Assessment Group combined into a single parameter the rates of withdrawing from ESA treatment, increasing the dose, and decreasing the dose reductions estimated from the trials included in its review of clinical effectiveness. The Assessment Group used an average body weight of 66.6 kg to convert from weight‑based doses to fixed doses. The Assessment Group estimated a fixed dose of 24,729 units per week for Eprex, Binocrit and Retacrit, and fixed doses of 31,021 units for NeoRecormon, 22,859 units for Eporatio and 141.1 micrograms for Aranesp. The Assessment Group assumed that the duration of ESA treatment was 12 weeks based on its review of clinical effectiveness.

4.2.11 Based on the opinions of clinical experts, the average cost per administration of an ESA used in the model was £8.16. This was estimated from the personal social services research unit (PSSRU) and weighted by the probability of being administered by a district nurse (21.6%), a GP nurse (21.6%) or a hospital staff nurse (16.3%), or being self‑administered (40.6%). In its base‑case analysis, the Assessment Group assumed that patients would have ESAs once a week (based on the marketing authorisations and the included trials) for 12 weeks. The Assessment Group did not model ESA treatment after a patient achieved haemoglobin of 120 g/litre, although it noted that in clinical practice some patients continue ESA treatment up to 4 weeks after chemotherapy.

4.2.12 The unit cost for the supply of red blood cells was taken from NHS Blood and Transplant and inflated to 2014/15 prices. In the absence of more recent data, the Assessment Group derived the cost of an appointment for a transfusion from the study by Varney and Guest (2003). The Assessment Group assumed that patients who do or do not have treatment with ESAs are equally likely to need iron supplements; therefore, it did not include the cost of iron supplements in the analysis.

4.2.13 The Assessment Group assumed that a patient would have blood tests regularly during chemotherapy, whether or not they were treated for anaemia, and that patients treated with an ESA would have 4 additional blood tests post‑chemotherapy. The Assessment Group estimated the cost for the additional blood tests from PSSRU and NHS reference costs. The Assessment Group obtained costs of treating adverse reactions (thromboembolic events, hypertension and thrombocytopenia) by pooling the results of studies included in its review, and from NHS reference costs. The Assessment Group assumed that patients in the model would experience, at most, 1 adverse reaction of each type. It assumed that the dosing schedule, administration cost, cost of red blood cell transfusion, additional blood tests and adverse reactions were similar for all the ESAs.

4.2.14 The base‑case analysis that used parameters from all studies resulted in ICERs ranging from £19,429 per QALY gained for Binocrit to £35,018 per QALY gained for NeoRecormon compared with no ESA treatment. The incremental costs of the ESAs compared with no ESA treatment ranged from £1371 for Binocrit to £2472 for NeoRecormon, whereas the incremental QALY gain was 0.0706 for all ESAs compared with no ESA treatment.

4.2.15 The Assessment Group noted that more than three‑quarters of the total QALY gain associated with ESA use (0.0706) were accrued in the long‑term component of the model (0.0582). The Assessment Group found that the estimated short‑term QALY gain of 0.0124 was lower than those reported in other analyses of cost effectiveness, such as in the review by Wilson et al. (2007), which reported a short‑term QALY gain of 0.030.

4.2.16 In the Assessment Group's probabilistic analysis, the ICERs ranged from £14,724 per QALY gained for Binocrit to £27,226 per QALY gained for NeoRecormon. The 95% credible intervals covered a range of £2322 per QALY gained to dominated (that is, the ESAs had higher costs and lower QALYs than treatments not including ESAs). At a maximum acceptable ICER of £20,000 per QALY gained, Binocrit had less than 25% chance of being cost effective, whereas the other ESAs had less than 20% chance of being cost effective.

Scenario analyses

4.2.17 The Assessment Group explored a scenario in which patients using ESAs do not live longer than patients not using ESAs. This resulted in a long‑term QALY gain of 0 and an overall QALY gain of 0.0124 (reflecting the short‑term QALY gain only). This analysis resulted in an ICER of more than £110,000 per QALY gained for patients using ESAs compared with patients not using ESAs. The probabilistic sensitivity analysis resulted in ICERs ranging from £96,754 per QALY gained for Binocrit to £174,193 per QALY gained for NeoRecormon.

4.2.18 In a second scenario analysis, the Assessment Group applied the best contract prices available for the ESAs to its base‑case analysis, rather than using BNF prices. The contract prices reflect the actual prices paid by the NHS for ESAs based on a 'price‑volume' agreement with the companies. The contract prices used in this scenario represented the latest tenders to London hospitals provided to NICE for this appraisal by the Commercial Medicines Unit and the South East England Specialist Pharmacy Services, with the companies' consent. These prices are designated as commercial in confidence. The ICERs are also commercial in confidence because they allow the contract prices to be calculated. Using these prices, the ICERs were considerably lower. Retacrit generated the lowest ICER and Aranesp the highest ICER; however, the Assessment Group stated that the probabilistic analysis of incremental net health benefits suggests that the cost effectiveness of the ESAs were similar.

4.2.19 When the Assessment Group combined these 2 scenario analyses, applying contract prices and assuming that people using ESAs do not live any longer than people not using ESAs, the ICERs were lower than the base‑case estimates (these ICERs are designated commercial in confidence). The probability that the ESA with the lowest contract price would be cost effective at a threshold of £20,000 per QALY gained was above 50%. The Assessment Group noted that the most important drivers of the cost‑effectiveness results included the price at which the NHS buys ESAs, and the assumption that ESAs prolong survival.

4.2.20 In another scenario, using the base‑case drug prices and assuming that ESA treatment improves survival (as estimated from the base case) but for the first 3 years only (after which the death rate is equal for both treatment arms), the Assessment Group estimated an ICER range of £42,584 per QALY gained for Binocrit to £76,751 per QALY gained for NeoRecormon. The Assessment Group highlighted that the results suggest that 66% of the long‑term QALY gain and 54% of the total QALY gain in the base case accrues over the first 3 years after ESA treatment.

4.2.21 To estimate ICERs more closely reflecting the marketing authorisation, the Assessment Group performed a scenario analysis using only trials in which the haemoglobin concentration of patients was 110 g/litre or less when starting treatment. The baseline haemoglobin concentration estimated using this subgroup of trials was 94 g/litre compared with 103.8 g/litre estimated in the base case. The Assessment Group used most cost and utility input parameters from the base case. The resulting deterministic ICERs ranged from £12,593 per QALY gained for Binocrit to £23,013 per QALY gained for NeoRecormon. The probabilistic analysis resulted in ICERs ranging from £10,363 to £19,157 per QALY gained, with the upper limit of the 95% credible intervals indicating that treatment without ESAs dominated treatment with any ESA.

Univariate sensitivity analyses

4.2.22 The Assessment Group performed various univariate (one‑way) sensitivity analyses around the base‑case ICERs. To assess the effect of the utility associated with increasing haemoglobin concentrations on the ICERs, the Assessment Group assumed alternate values of 0.009 (SF‑6D value from Harrow et al. 2011) and 0.016 (EQ‑5D value from Tajima et al. 2010) for anaemia related to chronic kidney disease. Using the values of 0.016 and 0.009 slightly increased the base‑case ICERs. However, applying a higher utility value of 0.06 (Wilson et al. 2007) decreased the base‑case ICERs to below £30,000 per QALY gained for NeoRecormon and Aranesp, and to below £20,000 per QALY gained for all the other ESAs compared with treatment without an ESA. When the Assessment Group included in the model long‑term costs of £20,000 per year associated with ongoing cancer treatment (such as costs of maintenance therapy, subsequent chemotherapy cycles or relapse), the ICERs of all the ESAs increased to levels above £30,000 per QALY gained. Applying alternative dosing schedules within the marketing authorisations of the ESAs (see section 3) generally increased the ICERs slightly. All other scenarios had little effect on the base‑case ICERs, including using the ESA administration schedule for chronic kidney disease‑related anaemia (that is, nurse administration [25%] and self‑administration [75%]), and using higher and lower values for the cost of a blood transfusion appointment and the cost of treating adverse reactions.

4.2.23 The Assessment Group highlighted the large difference between the lowest base‑case ICER reported in the current review (£19,429 per QALY gained) and the base‑case ICER reported in NICE technology appraisal guidance 142 by Wilson et al. (2007) (£150,342 per QALY gained). In exploring these differences, the Assessment Group adjusted its model to incorporate some parameters used in NICE technology appraisal guidance 142. The adjustments increased the base‑case ICER for the most cost‑effective ESA from £19,429 to £109,055 per QALY gained. The Assessment Group noted that the parameters from the current appraisal that affected the results were:

Clinical effectiveness

4.3.2 The Committee considered the clinical effectiveness of ESAs. It heard from the Assessment Group that none of the studies that evaluated ESAs was in line with the current UK marketing authorisations. It was also aware that most of the trials were conducted before the European Medicines Agency revised the marketing authorisations of the ESAs to stipulate a haemoglobin concentration of 100 g/litre or lower at the start of treatment. The Committee was aware that the Assessment Group analysed a subset of studies in which patients were treated with ESAs if their haemoglobin concentration was 110 g/litre or lower in an attempt to match the ESAs' marketing authorisations more closely, while also maintaining a large enough group of studies to generate a reliable estimate. The Committee concluded that the Assessment Group's analysis reflecting the population closer to the marketing authorisations was relevant to UK clinical practice.

4.3.3 The Committee examined the results of the Assessment Group's systematic review. It noted that the meta‑analyses suggested that ESAs increase haemoglobin concentrations, improve haematological responses and reduce the need for a blood transfusion compared with treatment without an ESA. The Committee also considered the results of the subgroup analyses and noted that most of the subgroups included a small number of studies, which limited the interpretation of the results. Having heard from the Assessment Group that the trials were 'flawed', the Committee was concerned about the quality of the included studies and the effect this had on interpreting the results. However, it heard from the Assessment Group that the flaws related mostly to inadequate reporting rather than poor design. The Committee also heard from the clinical expert that the results for the anaemia‑related outcomes were consistent with what clinicians see in practice. The Committee concluded that ESAs were effective in increasing haemoglobin concentrations, improving haematological responses and reducing the need for blood transfusions.

4.3.4 The Committee discussed the overall survival results. It noted that the point estimate for the hazard ratio suggested that ESAs prolong life but that the difference between the treatment arms was not statistically significant at a 0.05 significance level (see sections 4.1.12 and 4.1.16). It heard from the Assessment Group that the trials were not designed to address overall survival, and that the follow‑up periods in the trials varied. The Committee understood that the Cochrane review by Tonia et al. (2012) suggested that ESAs increased the risk of death, but noted that this may have been influenced by the fact that the authors did not restrict the review based on the ESA dose used in the trials (see section 4.1.12). The Committee also understood that the Assessment Group's analysis, which included only studies complying with the licensed ESA starting dose, showed that ESAs have no effect on survival. The Committee heard from the clinical expert that the main aim of treatment with ESAs is to make people feel better, and not necessarily to extend life. The Committee considered various explanations for the variable survival results associated with ESAs from the trials, including using unlicensed doses of ESAs, promoting tumour growth by improving oxygen supply to the cancer, using ESAs at high starting haemoglobin concentrations, or achieving haemoglobin concentrations that would now be considered too high in light of the revised marketing authorisations. The Committee considered that a survival benefit from ESA treatment could reflect that blood transfusions lower survival, having heard from the clinical expert that some evidence supports an association between blood transfusions and increased mortality. Based on the balance of the evidence presented, the Committee concluded that it could not assume that ESA treatment either prolonged or shortened survival compared with treatments that did not include ESAs.

4.3.5 The Committee considered the health‑related quality‑of‑life results, which showed a statistically significant difference in FACT‑F scores between patients treated with an ESA and patients treated without an ESA, as well as the Assessment Group's comments that there were several methodological concerns that may lead to bias. However, the Committee accepted the comments from the clinical expert and the patient expert that ESA treatment improves people's wellbeing and enables them to perform everyday tasks. It concluded that the available evidence suggests that ESA treatment improves health‑related quality of life compared with treatment without ESAs.

4.3.6 The Committee considered the adverse reactions associated with ESAs. It noted from the Assessment Group's meta‑analyses that ESAs increased the risks of thromboembolic events, hypertension and pruritus compared with treatment without an ESA. The Committee heard from the clinical expert that thromboembolic events were the most common serious adverse reactions associated with ESAs, and that these were mostly venous thrombosis and pulmonary embolism. It also heard from the Assessment Group that these adverse reactions occurred rarely in the trials. The Committee considered that the risk of these adverse reactions in the trials might be associated with the high starting and target haemoglobin concentrations in the trials; this was because the Assessment Group's meta‑analyses showed that the risks of thromboembolic events and hypertension were slightly lower in the subgroup with haemoglobin concentrations of 110 g/litre or less when starting treatment than in the overall population. It noted that the safety concerns led the European Medicine Agency to revise the marketing authorisation. The Committee concluded that the current evidence suggests that the risks of adverse reactions are lower when ESAs are used in line with their current marketing authorisations.

4.3.7 The Committee considered the relative effectiveness of the different ESAs. It understood that the Assessment Group's subgroup analysis suggested that epoetins increase haemoglobin concentration more than darbepoetin alfa does. However, the Committee recognised that the analysis of darbepoetin contained few studies, that the confidence intervals of the estimates were wide, and that the Assessment Group did not adjust the analyses for multiple testing. The Committee noted that Binocrit and Retacrit are biosimilar medicines, that is, new biological products that are similar to the biological reference medicine (Eprex). The Committee understood that, unlike conventional pharmaceuticals, which can be copied by chemical synthesis, biopharmaceuticals are complex molecules and are difficult to replicate fully. It also understood that biosimilar products may have a different safety profile from the biological reference medicine biopharmaceutical product. The Committee noted that biosimilar products are regulated by the European Medicines Agency through a centralised procedure, and that the European Medicines Agency's legislation on biosimilars defines the studies needed to demonstrate safety and efficacy to the biological reference medicine. The Committee was aware that making specific recommendations about the safety of a drug falls outside the remit of NICE, and that current advice for prescribing recommends that biopharmaceutical products should be prescribed by brand name. The Committee considered that there was no evidence to suggest differences between the biosimilars and the biological reference medicine, and noted the limitations in the Assessment Group's subgroup analyses comparing different ESAs. The Committee also heard from the clinical expert that ESAs did not appear to differ in their clinical effectiveness, and that the choice of ESA in clinical practice usually depends on price, and occasionally on difference in dosing frequency. The Committee therefore concluded that it was likely that the ESAs did not differ in clinical effectiveness.

Cost effectiveness

4.3.8 The Committee considered the Assessment Group's economic model and whether its assumptions were appropriate. It noted that the Assessment Group assumed in the model that all ESAs had the same effectiveness. In light of its conclusion that the ESAs did not differ clinically (see section 4.3.7), the Committee concluded that this assumption was reasonable.

4.3.9 The Committee considered whether the modelled treatment duration of 12 weeks was reasonable, noting that the marketing authorisations allow ESAs to be used up to 4 weeks after chemotherapy ends, and that the treatment duration in the trials varied from 12 to 28 weeks. The Committee considered that this could affect the costs and benefits of ESAs. However, it heard from the clinical expert that it was common clinical practice to use ESAs for 12 weeks only, that is, during chemotherapy. The Committee noted that the model assumed that haemoglobin returned to normal at concentrations of 120 g/litre, which is in line with the target haemoglobin concentration of 100 to 120 g/litre stated in the marketing authorisations for ESAs. It heard from the clinical expert that this assumption was reasonable, although some clinicians may prefer to stop treating at the lower end of the target range. The Committee concluded that the treatment duration and haemoglobin concentrations assumed in the model were appropriate.

4.3.10 The Committee considered the utility values applied in the economic model. It accepted the Assessment Group's choice to use the study by Harrow et al. (2011) to estimate the short‑term utility values. The Committee noted that the sample size was large enough and that although the SF‑6D data were collected in women with cancer in the USA, they were valued by the UK general population. It accepted the Assessment Group's mapping of the SF‑6D utility to EQ‑5D values, in the absence of directly derived EQ‑5D data. The Committee was concerned that the Assessment Group did not include disutilities associated with adverse reactions in the QALY calculation given that most adverse reactions occurred more frequently in the ESA arms. However, it recognised that there would be minimal effect on the ICERs given that the adverse reactions in the studies were rare. The Committee heard from the patient expert that it is possible for people to self‑administer ESAs at home, which is more convenient for the person and costs the NHS less than hospital attendance for a blood transfusion. The Committee noted that the benefits from reducing the need for hospital visits were not captured in the QALY calculation. It also considered that there were potential relative health benefits of ESAs associated with avoiding blood transfusions given that any risks from transfusion were not included in the model. The Committee was generally satisfied with the Assessment Group's approach to estimating the utility values but concluded that the QALY gain from ESAs may have been underestimated.

4.3.11 The Committee considered the costs used in the model. It noted that the prices of ESAs used in the base case were based on BNF list prices, but that the NHS procures ESAs on a 'price‑volume' agreement on a confidential basis with the companies. The Committee noted that NICE's Guide to the methods of technology appraisal 2013 indicates a preference for using nationally available price reductions in the reference‑case analysis to reflect the price relevant to the NHS. The Committee concluded that the contract prices were the most relevant prices to the NHS and therefore the appropriate prices on which to base its decision.

4.3.12 The Committee considered the scenario assessing the subgroup of people with haemoglobin concentrations of 110 g/litre or lower at the start of ESA treatment. It noted that the ICERs for this scenario were approximately a third lower than the base case (see sections 4.2.14 and 4.2.21). The Committee was aware that this was mostly because the overall survival hazard ratio estimated from the meta‑analysis for this subgroup was 0.91 compared with 0.97 used in the base case. However, the Committee agreed that other model parameters specific to this subgroup, such as the lower risks of adverse events and lower haemoglobin concentrations at the start of treatment, contributed to the lower ICERs in this subgroup. The Committee concluded that using ESAs only at haemoglobin concentrations that reflect the marketing authorisations would slightly reduce the base‑case ICERs.

4.3.13 The Committee considered whether ESAs were a cost‑effective use of NHS resources and which assumptions it should use to derive the most plausible ICER. The Committee noted that the prices of the drugs and the assumption that ESAs prolong survival most strongly influenced the cost‑effectiveness results (see sections 4.2.17 to 4.2.20). The Committee had concluded that there was not enough evidence to suggest a survival gain with ESAs and therefore agreed that the model should incorporate a hazard ratio of 1 instead of 0.97. It also agreed that it was appropriate to use contract prices because these are what the NHS pays. Therefore, the Committee concluded that the scenario assuming equal survival and using contract prices was the most plausible. The Committee noted that the probabilistic ICERs for this scenario were all below £30,000 per QALY gained, although the credible intervals indicated a degree of uncertainty. The Committee considered that including disutilities associated with adverse reactions could increase the ICERs slightly. However, it concluded that the benefits of ESA treatment associated with avoiding blood transfusions (see section 4.3.10) and starting ESA treatment only at haemoglobin concentrations in line with the marketing authorisations (see section 4.3.12) would likely reduce the ICERs. The Committee agreed that the most plausible ICER was below £20,000 per QALY gained, and that ESAs could be considered a cost‑effective use of NHS resources and should be recommended as an option for treating anaemia in people with cancer having chemotherapy. The Committee noted that, because it assumed that the ESAs were equally effective, using the ESA with the lowest acquisition cost for a course of treatment would best employ scarce NHS resources. It understood from comments received during consultation that the current tendering process in the NHS for ESAs takes into account other factors related to the drugs, such as safety, efficacy and dosing frequency. However, the Committee noted that it had already considered these factors in its deliberations. The Committee therefore also recommended that if different ESAs are equally suitable, the product with the lowest acquisition cost for the course of treatment should be used.