3 The company's submission | Tolvaptan for treating autosomal dominant polycystic kidney disease | Guidance

Clinical effectiveness

3.1 The main source of evidence presented in the company's submission came from 1 phase‑III trial, TEMPO 3:4. This trial was an international, multicentre, double‑blind, placebo‑controlled, parallel‑arm, randomised controlled trial in which 1445 patients were randomised in a 2:1 ratio to either the tolvaptan (n=961) or the placebo (n=484) arm. Patients aged 18–50 years with rapidly progressing autosomal dominant polycystic kidney disease (ADPKD) and an estimated glomerular filtration rate (eGFR) of 60 ml per minute or more and with a total kidney volume (TKV) of 750 ml or more (as measured by MRI) were included in the trial. Tolvaptan and placebo were administered orally, twice daily. The dose was titrated at weekly intervals over a 3‑week period, initially administered at doses of 45 mg and 15 mg, in the morning and afternoon respectively, and titrated to 60 mg and 30 mg, and then to 90 mg and 30 mg, according to patient‑reported tolerability. Following the titration period, patients had the maximum tolerated dose for the remainder of the 36‑month treatment period. Patients were monitored every 4 months during the treatment period. Two additional follow‑up visits were also conducted 7 to 21 days after month 36 and 7 to 21 days after the first follow‑up visit. The baseline demographics were balanced in terms of age, sex, family origin and factors influencing ADPKD progression. The mean age of patients in the trial was 38.7 years. The mean TKV was 1705 ml in the tolvaptan group and 1668 ml in the placebo group. Patients having tolvaptan and placebo were evenly distributed at baseline across the Kidney Disease Outcomes Quality Initiative chronic kidney disease (CKD) stages 1 (34.5% and 35.9%), 2 (48.5% and 46.5%) and 3 (17.0% and 17.4%), respectively. Seventy‑three patients were from the UK.

3.2 The primary end point of the TEMPO 3:4 trial was the rate of TKV change from baseline for tolvaptan relative to placebo, as measured by MRI. Data on the rate of decline of renal function (listed in the final scope as an outcome measure) were also available. The results of TEMPO 3:4 showed that tolvaptan had a statistically significant relative reduction of 49.2% on TKV growth over 3 years when compared with placebo (absolute reduction of −2.71% per year; 95% confidence interval [CI] −3.27% to −2.15%; p<0.0001). Post‑hoc analyses of TKV in subgroups of patients at each CKD stage (1, 2, or 3 at baseline) showed a consistent and significant effect favouring tolvaptan across all stages; results were designated as academic in confidence by the company and therefore cannot be reported here.

3.3 In TEMPO 3:4 the composite secondary end point was time to onset of multiple ADPKD outcomes (worsening renal function, new onset hypertension, worsening hypertension, renal pain and worsening albuminuria). The results showed that tolvaptan treatment was associated with a 61% relative reduction in the risk of worsening renal function over 3 years compared with placebo (absolute reduction: 3 events per 100 person‑years; hazard ratio 0.39; 95% CI 0.26 to 0.57; p<0.001).

3.4 Rate of change in renal function was also included as a secondary end point in the trial, and was assessed by the reciprocal of the serum creatinine level as a measure of change in glomerular filtration rate (GFR). Subsequent analyses used other methods to estimate GFR, including the Chronic Kidney Disease Epidemiology Collaboration (CKD‑EPI) equation. Using the first measure, tolvaptan was associated with a statistically significant 31.6% relative reduction in the annual rate of renal function decline (absolute reduction of 1.20 mg/ml^‑1 serum creatinine; 95% CI 0.62 to 1.78; p<0.001), compared with placebo. When GFR was assessed using the CKD‑EPI equation, the relative reduction was 26.4% for tolvaptan compared with placebo (absolute reduction of 2.72 ml/min/1.73 m² per year over 3 years; 95% CI 0.60 1.36; p<0.001).

3.5 The company also presented evidence from an ongoing clinical study in which tolvaptan was the intervention of interest. TEMPO 4:4 is an open label, non‑randomised extension study of TEMPO 3:4. The study aims to determine whether tolvaptan modifies the progression of ADPKD and if the effects of tolvaptan are sustained over time. The available interim results of this study indicate that the benefit of treatment persisted for patients who continued taking tolvaptan compared with those having placebo in TEMPO 3:4.

3.6 The company also reported the results of a real‑world study called OVERTURE, which is an ongoing, multicentre, prospective, observational cohort study aiming to identify factors that may predict rapid progression toward, or higher frequency of, clinically relevant morbidities in ADPKD.

3.7 The most commonly reported adverse reactions from TEMPO 3:4 were thirst, polyuria, nocturia and pollakiuria, occurring in approximately 55%, 38%, 29% and 23% of patients respectively. Furthermore, tolvaptan was associated with the adverse effects of elevations of serum alanine and aspartate aminotransferases (ALT and AST respectively), with infrequent cases of concomitant elevations in total bilirubin. During the TEMPO studies 3 people met the criteria for a Hy's law case (hepatocellular injury, serum ALT or AST more than 3 times the upper limit of normal, total bilirubin more than twice the upper limit of normal), indicating the potential risk for serious drug‑induced liver injury. In all cases the abnormalities resolved after stopping treatment with tolvaptan. The company reported that after the implementation of the TEMPO Steering Committee's recommendation to increase monitoring to monthly intervals, no further Hy's law cases had been identified. In addition, the company noted that liver biochemistry monitoring was relatively infrequent in the TEMPO studies, and that more frequent monitoring would be expected in real‑world use, which would further lower the risk of people developing serious drug‑induced liver injury. The percentage of patients who stopped treatment was 23% in the tolvaptan group and 14% in the placebo group. Health‑related quality of life was not assessed in the TEMPO 3:4 trial.

Cost effectiveness

3.8 The company submitted an economic analysis for the cost effectiveness of tolvaptan, which was a patient‑level simulation model. The model used a lifetime horizon of up to 80 years, and a cycle length of 1 year. A half‑cycle correction was applied. The model used the perspectives of the NHS and personal social services, and costs and benefits were discounted by 3.5% per year. The model encompassed the disease pathway through 2 distinct modules: the first module captured the period of ADPKD progression up to the onset of end‑stage renal disease (ESRD) and the second module captured the management of ESRD when tolvaptan is no longer given. The ADPKD module encompassed 5 health states (CKD stages 1 to 4, a significant pain health state), and an end state of death. The ESRD module contained 5 health states (CKD stage 5, conservative care, haemodialysis, peritoneal dialysis, transplant) and an end state of death.

3.9 At the start of a simulation, the model generated a patient cohort based on the average baseline characteristics of the TEMPO 3:4 trial. Each patient within the cohort progressed in annual time increments (1‑year cycles). Within each cycle, the movements between CKD stages, the incidence of renal failure (CKD stage 5) and the incidence of all‑cause mortality were tracked. In the case of a patient's simulated eGFR falling below 15 ml/min/1.73 m², the patient moved on to the ESRD module. Once this stage was reached, patients could have conservative care management, dialysis or kidney transplantation. At the end of each cycle the patient's disease state was assessed and costs and appropriate health‑utility decrements were applied. Assuming a patient did not die in a given cycle, the simulation continued until the model time horizon (80 years or maximum age of 101 years) was reached. In the case of a fatal event, all costs, life years and quality‑adjusted life years (QALYs) were accumulated and the simulation ended for that patient. Once the simulation ended, the process started for the next patient in the cohort.

3.10 In the tolvaptan arm of the model, a treatment effect was applied to the underlying disease progression. This influenced the incidence and timing of ESRD (CKD stage 5) in this group. Further differences between the treatment and placebo groups included the following variables:

Treatment‑related discontinuation: for the first 3 years the discontinuation rates from TEMPO 3:4 were used, and after that a 0.5% treatment discontinuation was assumed.
Clinically significant pain events: for CKD stages 1 to 4 an annual probability of 0.05 was assumed for the tolvaptan arm and 0.07 for the placebo arm.
The incidence of clinically significant kidney pain was modelled separately from disease progression (see section 3.8).
Treatment utility decrement: in the base case no treatment‑related utility decrement was applied, but in the sensitivity analysis a treatment‑related utility decrement of 0.0123 was explored; this value was based on Sullivan et al. 2011 and was originally used by the company in its sensitivity analysis.

3.11 The underlying risk of disease progression was modelled using regression equations to predict annual change in TKV and eGFR. Baseline characteristics from the placebo arm of TEMPO 3:4 were used in the regression equation for estimated TKV progression in the first year. Thereafter, the patient characteristics of the previous cycle were used for each new cycle. Annual change in TKV was used as an intermediate step to model change in eGFR, which was the primary outcome of the model (eGFR was dependent on TKV in the previous cycle). This was repeated until the lifetime trajectory of TKV and eGFR of each patient was predicted. To estimate the treatment effect of tolvaptan (reduction in annual rate of renal decline for tolvaptan compared with standard care), the absolute change in eGFR from TEMPO 3:4 from the period between post‑titration baseline to the end of the study (3 years) was applied to the underlying disease progression. After the first 3 years, the treatment effect was assumed to persist and remain constant at a level of 31.6% for as long as treatment was continued.

3.12 The annual rate of treatment discontinuation observed during the TEMPO 3:4 trial was used in the first 3 cycles of the model (15.3%, 6.5% and 2.9%, for years 1, 2 and 3 respectively). For the remaining modelled years, an annual rate of discontinuation of 0.5% was applied. If a patient stopped treatment with tolvaptan, the natural history of disease progression was assumed to apply to their disease course. After progressing to the ESRD module, tolvaptan therapy was stopped.

3.13 For identifying health‑state utility values, the company conducted a systematic review of the literature. It identified 23 studies, but none of them reported health‑state utility value estimates for patients with ADPKD. In the model, health‑state utility values estimates from Gorodetskaya et al. (2005) were chosen in the base case for CKD stages 1 to 4. Gorodetskaya et al. reported estimates for CKD stages 1 to 4 and 5, using time trade‑off methods in a US sample. For the ESRD module health states, the estimates reported by Lee et al. (2005) were used. Lee et al. published EQ‑5D data from a UK sample on CKD stage 5 pre‑dialysis haemodialysis, peritoneal dialysis and functional transplant. The base‑case analysis assumed no disutilities for tolvaptan treatment. The disutility associated with dialysis complications was based on the NICE guideline on chronic kidney disease. The disutility associated with significant pain events was estimated from a study by Dolan et al. (1997). The model used baseline age‑adjusted utilities (general population values [Centre for Health Economics]) with utility decrements applied for the various health states in the model.

3.14 Adverse events were not explicitly modelled, but were incorporated in the costs and utilities of CKD and ESRD health states. Only clinically significant pain was included. The probability of clinically significant pain was derived from the TEMPO 3:4 study, and applied to CKD stages 1 to 4. For patients who stopped tolvaptan, the probability of clinically significant pain reported in the control arm was applied.

3.15 All‑cause mortality was modelled using age‑ and sex‑specific life tables from England and Wales. Patients with end‑stage renal disease are subject to a specific mortality risk, based on age‑specific (18–64 and 65+ years) observed dialysis survival rates, using a Weibull model. Time‑dependent mortality after transplant was based on the NHS Blood and Transplant Organ Donation and Transplantation Activity Report 2012–13. The cost of tolvaptan used for the base‑case analysis included a patient access scheme discount. Treatment costs were only calculated for the ADPKD module health states, until the year of discontinuation.

3.16 Monitoring costs were applied including liver function tests (done monthly for 18 months and 3‑monthly thereafter), 2 additional consultant visits in the first year of treatment and 1 additional consultant visit in the second year of treatment. Additional consultant time was added for reviewing liver function test results and issuing prescriptions. The resource use was based on the NICE guideline on chronic kidney disease and values were based on clinical expert opinion. The costs were calculated on the basis of Unit Costs of Health and Social Care (Curtis, 2014) and NHS Reference Costs 2012–13. Costs for CKD stages 1 and 2 health states included 1 consultant visit, 1 specialist nurse visit, 1 biochemistry test, 1 haematology test and 1 phlebotomy, based on Unit Costs of Health and Social Care and NHS Reference Costs. The resource use was based on clinical expert opinion. Costs for CKD stages 3 to 5 health states were sourced from the NICE guideline on chronic kidney disease. The costs for CKD stage 3 differed from CKD stage 4, based on a ratio published by Chamberlain et al. (2014). The cost of a significant pain event was based on NHS Reference Costs and HRG code. Costs for the ESRD module were based on HRG codes, NHS Reference Costs 2012–13, and the NICE guideline on peritoneal dialysis. Annual costs for dialysis were sourced from a study published by Baboolal et al. (2008). In the case of transplantation, the maintenance cost of the transplant was sourced from Kerr et al. (2012) and NICE technology appraisal guidance on immunosuppressive therapy for renal transplantation in adults. Costs associated with organ donation and transplantation activities were sourced from the NHS Blood and Transplant Organ Donation and Transplantation Activity Report 2013–14.

3.17 The model resulted in patients in the tolvaptan group spending less time (approximately 2 years) in ESRD and more time in CKD stages 2 to 4. The company's base‑case analysis after correcting a model code error, which was identified by the ERG during clarification, resulted in a probabilistic mean estimate of the incremental cost‑effectiveness ratio (ICER) for tolvaptan compared with standard care of £34,733 (with the patient access scheme) per QALY gained, representing a gain of 0.92 QALYs at a cost of £31,838 (with the patient access scheme).

3.18 Due to the nature of the model, which performs individual patient simulation and probabilistic simulations in a single analysis, all base‑case and sensitivity results were given as probabilistic mean values. Therefore the company did not do individual deterministic sensitivity analyses using alternative fixed estimates of model parameters; however, structural sensitivity analyses and scenario analyses were carried out. The 3 most influential scenario analyses were those that incorporated:

A treatment effect based on CKD‑EPI (ICER with the patient access scheme: £47,510 per QALY gained).
Using 'minimum' utility decrements for ESRD (exact utility decrements not specified, ICER with the patient access scheme: £40,615 per QALY gained).
Using a disutility of 0.0123 for being on tolvaptan treatment (ICER with the patient access scheme: £40,401 per QALY gained).

Evidence Review Group critique

3.19 The ERG provided a critique of the evidence provided by the company for the clinical and cost effectiveness of tolvaptan. It considered the company's approach to the decision problem was appropriate. However, regarding the comparator in the appraisal, it concluded that because standard care was not defined, there is some uncertainty on what measure comprised standard care and how this could have influenced the overall findings.

3.20 The ERG noted that the generalisability of the results was limited, because only 73 (5%) of the patients included in the trial were from the UK. It also noted that the trial only included patients aged 18–50, therefore patients over 50 years (when ESRD onset usually occurs) were not included in the trial. It also stated that the number of patients included in the CKD stage 3 subgroup was relatively low (17%) and evidence for this subgroup is limited. In addition, the ERG considered that, from the people considered eligible to have treatment, a high number had not been included, either because they had not met the inclusion criteria (TKV 750 ml or more, and estimated creatinine clearance 60 ml/min or more, n=530) or because they had declined to participate, or had other reasons for not participating (n=147).

3.21 Regarding the outcomes of interest, the ERG highlighted that there seems to be some uncertainty about how GFR should be estimated. In the base case it was estimated by measuring the reciprocal of serum creatinine level and as a secondary measurement it was also assessed by the CKD‑EPI equation.

3.22 The ERG noted that the primary outcome of the TEMPO 3:4 trial was outside the final scope and because the trial was powered for this primary outcome it is possible that the relevant outcomes defined in the final scope are underpowered. In its report, the ERG considered that TKV as a surrogate end point for annual eGFR decline had limited value, but that it was a good measure of extent of disease because it predicts future decline of renal function.

3.23 Regarding the adverse events, the ERG emphasised that 2 or more Hy's law cases, which were found in the clinical trial, is an important safety concern. Other adverse events (such as thirst or polyuria) may affect the ability of people to tolerate effective doses of tolvaptan. It was also reported that more people stopped treatment because of adverse events in the tolvaptan arm than in the placebo arm of the trial (15.4% compared with 5.0% respectively).

3.24 The ERG found that the regression analyses for disease progression based on TKV and eGFR were not described in detail (it was unclear to the ERG which covariates were initially examined, why only age and sex were included in the final models, why sex was included to predict TKV progression despite not being statistically significant and whether alternative models for the data were tested). The ERG highlighted that these analyses assumed that the rates of eGFR decline and TKV growth were constant. This was not tested and because eGFR is predicted from TKV, and TKV is dependent on age, the ERG considered that it was probable that eGFR would not be fully constant over time. The measurement of eGFR might result in uncertainty, but the ERG thought that the approach is justifiable.

3.25 The ERG noted that there is little evidence to substantiate the company's claim that the correlation between eGFR and TKV as observed in patients having no treatment may misrepresent the relationship in tolvaptan patients. The ERG also questioned the assumption that treatment effect persists for the duration of treatment. The ERG cautioned about accepting this assumption because the only longer‑term data was from an interim analysis from the 5‑year TEMPO 4:4 trial, which is an open‑label, non‑randomised extension study. Data from longer follow‑up periods, in which late onset adverse effects may arise, are not available. Therefore the ERG argued that there is little evidence to conclude whether the treatment effect would persist or decline.

3.26 The ERG stated that there is little evidence available to support the use of a 0.5% annual treatment discontinuation rate after year 3. It also considered the company's sensitivity analyses (0%, 2%) had too small a range and conducted exploratory analysis using a 6.5% discontinuation rate, which was equal to that observed in the second year in TEMPO 3:4. This resulted in an ICER for tolvaptan compared with standard care of £42,893 per QALY gained (with patient access scheme).

3.27 In its report, the ERG noted that the way the utilities were included in the model was subject to possible errors and double counting. The model calculated the total utility for a patient at any given time by subtracting the sum of the disutilities for health state, kidney pain and (in a scenario analysis only) treatment‑related disutility from the age‑adjusted baseline utility for each patient. The ERG also considered that including a disutility only for kidney pain and for no other adverse events (and therefore potentially favouring the tolvaptan arm) was not a conservative approach. Therefore, in its exploratory analysis the probability of a kidney pain event was set to equal for both arms. The ERG also noted that the disutility value (0.06) applied for haemodialysis and peritoneal dialysis complications is exaggerated and favours the tolvaptan arm. The conservative approach would be to set the value of this disutility to 0.02.

3.28 The ERG noted that modelling only clinically significant pain may have introduced a downward bias to the ICER, because of the assumption that the difference in kidney pain as observed in TEMPO 3:4 is independent from the effect of tolvaptan on disease progression. During clarification the ERG pointed out that the results of TEMPO 3:4 show different serious adverse events in the tolvaptan and in the placebo arms. The ERG asked the company to investigate treatment‑dependent adverse events in a scenario analysis. In its response the company argued that the differences are not sufficient to justify more detailed modelling and concluded that more detailed consideration of these events in the economic model would add little value and would not greatly impact the overall results. In its report the ERG stated that a conservative approach would be to apply a 0.0123 utility decrement for tolvaptan treatment; this value was investigated in one of the scenario analyses in the company submission.

3.29 The ERG noted that hepatotoxicity from tolvaptan treatment or other drug‑induced liver injury was not included in the model, despite 3 Hy's law cases being reported in TEMPO 3:4 and TEMPO 4:4 collectively. Finding 2 or more Hy's law cases is considered highly predictive that there is a risk of causing severe drug‑induced liver injury. Therefore, the ERG did an exploratory analysis, incorporating consequences of tolvaptan‑induced hepatotoxicity.

3.30 In its report the ERG noted that using non‑ADPKD‑specific mortality for CKD stages 1 to 4 can underestimate the mortality risk and this assumption may favour tolvaptan. To account for this the ERG explored a higher mortality (hazard ratio 2.0) in CKD stages 1 to 4 in an exploratory analysis.

3.31 The ERG critiqued the use of additional monitoring costs, noting that although liver function tests and additional consultant visits had been included in the model, the costs of further monitoring for patients who experience an abnormal test result had not been included. Therefore, the ERG explored the impact of higher costs in its exploratory analyses. In addition, it did not agree with the cost applied in the model for the CKD stage 3 health state, which was based on the calculated cost for CKD stage 4, adjusted using a reference from a study that the manufacturer acknowledged may not have been fully representative of the population in the UK. The ERG therefore used equal costs for CKD stages 3 and 4 in its exploratory analyses.

3.32 The ERG considered that the lack of 1‑way sensitivity analyses was a serious shortcoming and that the justification for excluding sensitivity analyses was not convincing.

3.33 The ERG noted that the company did not explore scenarios considering the extrapolation of the treatment effect, which it considered would probably be one of the most influential uncertainties.

ERG's exploratory analyses

3.34 The ERG criticised the company's assumption that the potential for drug‑induced liver injury does not lead to any costs or health losses and did an exploratory analysis incorporating the consequences of it. For this exploratory analysis, a worst‑case scenario was adopted assuming that all Hy's law cases would need a liver transplant at the end of year 1 and would die immediately after. In this exploratory analysis, the ICER increased from the company's base‑case ICER of £34,733 to £35,751 per QALY gained (with the patient access scheme).

3.35 The ERG also explored higher mortality values in CKD stages 1 to 4, because the company's base case used general population results, which might underestimate mortality for ADPKD. The ICER in this scenario increased the company's base‑case ICER to £34,754 per QALY gained.

3.36 The ERG conducted an analysis in which it assumed a treatment discontinuation of 6.5% after 3 years. This increased the company's base‑case ICER to £42,893 per QALY gained.

3.37 The ERG explored the effects of additional monitoring costs using 2 assumptions. It assumed that patients with serum alanine aminotransferase higher than 3 (4.4%) will need more monitoring, therefore in the exploratory analysis the monitoring was doubled for these patients. It was also assumed that after the second year, patients would need an extra consultation visit because of possible adverse events. The results showed that the company's base‑case ICER increased to £36,167 per QALY gained.

3.38 The ERG considered that the maintenance costs after kidney transplants are likely to be overestimated, therefore it subtracted the background management costs from the maintenance costs for all years, which resulted in an ICER of £39,264 per QALY gained.

3.39 After correcting a model code error, the ERG also implemented some changes to the model and calculated a base‑case ICER with its preferred assumptions. The assumptions were:

equal probability of kidney pain for both arms
equal CKD stage costs for CKD stage 3 and CKD stage 4
applying a disutility of 0.0123 for tolvaptan treatment
applying a disutility of 0.02 for haemodialysis and peritoneal dialysis complications.

This resulted in an ICER for tolvaptan compared with standard care of £43,280 per QALY gained.

3.40 The ERG conducted further analyses to explore the impact on the ICER of using the CKD‑EPI equation as an approximation for eGFR for both modelling the disease progression and treatment effect. This change to the model increased the company's base‑case ICER to £50,524 per QALY gained. The ERG also re‑calculated its preferred base case using the same assumptions as in the exploratory analyses (described in section 3.39) and using the CKD‑EPI equation as an approximation of eGFR. This increased the ICER to £64,515 per QALY gained (ERG's preferred base‑case ICER using the CKD‑EPI equation).

3.41 The ERG also did an exploratory analysis using the worst‑case scenarios as described in sections 3.34–3.38. This analysis contained only these scenarios, and did not include any of the ERG's preferred assumptions from the ERG's preferred base‑case ICER (section 3.39). When using the CKD‑EPI equation as an approximation for eGFR, the combination of these worst‑case scenarios resulted in an ICER of £72,705 per QALY gained (ERG's worst‑case scenario ICER using CKD‑EPI).

3.42 The results of the company's base‑case analysis, the ERG's preferred base case and the further analyses using the CKD‑EPI equation for assessing eGFR for each of the exploratory analyses described in sections 3.17, 3.39, 3.40 and 3.41 are presented in Table 1.

Table 1 ERG exploratory analyses (with the patient access scheme)

Technologies		Incremental cost	Incremental QALY	ICER
Company's base case (after correcting a model code error)
Standard care
Tolvaptan		£31,838	0.92	£34,733
Company's base case using the CKD‑EPI equation as an approximation for eGFR (after correcting a model code error)
Standard care
Tolvaptan	£36,411		0.72	£50,524
ERG's preferred base case
Standard care
Tolvaptan		£33,015	0.76	£43,280
ERG's preferred base case using the CKD‑EPI equation as an approximation for eGFR
Standard care
Tolvaptan		£37,956	0.59	£64,515
ERG's worst‑case scenario exploratory analyses using the CKD‑EPI equation as an approximation for eGFR
Standard care
Tolvaptan		£32,095	0.44	£72,705
CKD‑EPI, Chronic Kidney Disease Epidemiology Collaboration; eGFR, estimated glomerular filtration rate; ERG, Evidence Review Group; HD, haemodialysis; ICER, incremental cost‑effectiveness ratio; PD, peritoneal dialysis; QALYs, quality‑adjusted life years.

Company's submission of additional evidence

3.43 The company submitted additional evidence in response to consultation. In this additional evidence, the company presented a revised base‑case analysis for a subgroup with only CKD stages 2 and 3 at the start of treatment and who had evidence of rapidly progressing disease. The updated analysis using the subgroup will be referred to from this point as the company's 'revised base case'.

3.44 In the company's revised base case, the following adjustments were applied to the original model:

A revised patient access scheme discount on the tolvaptan NHS list price.
Baseline patient profile adjusted to represent ADPKD patients with CKD stages 2 to 3.
Subgroup‑specific relative reduction in renal function decline using the CKD‑EPI equation as an approximation for eGFR.
Annual change in renal function and percentage TKV change estimated using regression equations instead of data observed for the first 3 years over the trial period.
Corrected model code.
Equal costs for CKD stages 3 and 4.
Utility decrement of 0.02 for haemodialysis and peritoneal dialysis complications.
Utility decrement of 0.0025 for tolvaptan treatment.
Treatment discontinuation of 2.9% after year 3 for the remainder for the modelled time horizon.
Decreased post‑kidney transplant costs.
Increased mortality (hazard ratio: 1.6).
Increased monitoring costs for the first 2 years while assuming the same monitoring costs as in its original base case for subsequent years.

3.45 From the revisions listed above, the company's revised base case matched the preferred assumptions described by the ERG in its exploratory analyses for the following: corrected model code (see section 3.39); equal costs for CKD stages 3 and 4 (see section 3.31); utility decrement of 0.02 for haemodialysis and peritoneal dialysis complications; decreased costs after kidney transplant (see section 3.38).

3.46 In the company's revised base case, the underlying risk of disease progression was modelled using regression equations to predict the annual change in TKV and eGFR. Baseline characteristics were adjusted from those of the placebo arm of TEMPO 3:4 to reflect a population with CKD stages 2 and 3, which was 48.4% female with a mean age of 44 years, mean eGFR of 60 ml/min/1.73 m², and a mean TKV of 2300 ml. These baseline characteristics were used in the regression equation for estimated TKV progression in the first year. Thereafter, the patient characteristics of the previous cycle were used and updated for each new cycle. Annual change in TKV was used as an intermediate step to model change in eGFR, which was the primary outcome of the model (eGFR was dependent on TKV). This was repeated until the lifetime trajectory of TKV and eGFR of each patient was predicted.

3.47 To estimate the treatment effect of tolvaptan for the subgroup (note that only patients with CKD stages 2 or 3a at baseline from TEMPO 3:4 were included in the analysis), the company took the relative reduction in the slope of renal function decline, which was 29.7% for tolvaptan compared with placebo, as measured by the CKD‑EPI equation. The company then applied this treatment effect to the underlying disease progression. As with the company's original submission, after the first 3 years the treatment effect was assumed to persist and remain constant for as long as treatment was continued.

3.48 The company argued that the utility decrement of 0.0123 assumed by the ERG in its original base case (see section 3.39) was overestimated. The company's revised base case assumed a treatment‑related utility decrement of 0.0025, which it estimated assuming that the number of QALYs gained because of pain reduction was equal to the number of QALYs lost because of the negative effect of tolvaptan treatment. Regarding serious kidney pain events, the company continued to apply probabilities for people having tolvaptan and placebo of 0.05 and 0.07 respectively. For treatment‑related discontinuation, the company continued to use the TEMPO 3:4 trial data for the first 3 years, after which the company then used a treatment discontinuation of 2.9% for the remainder of the modelled time horizon; this was the rate of discontinuation in the third and final year of TEMPO 3:4. All‑cause mortality was modelled according to figures specific to polycystic kidney disease from a publication by Florijn et al. (1995) using a hazard ratio of 1.6.

3.49 The company's revised base case assumed that monitoring and administration costs in year 1 and 2 were set at 1.044 times the annual monitoring cost, accounting for doubling the intensity of monitoring for the estimated 4.4% of people with raised serum alanine aminotransferase (ALT) levels. The company stated that the risk across study arms of elevated serum ALT was equal after the first 18 months and therefore it did not assume increased monitoring frequency after year 2.

3.50 In the company's revised base case in the subgroup with CKD stages 2 and 3, the probabilistic mean estimate of the ICER for tolvaptan compared with standard care was £23,503 (with the patient access scheme) per QALY gained.

ERG's critique of the company's submission of additional evidence

3.51 The ERG commented that the company had not adjusted the annual change in renal function and percentage TKV change to reflect the subgroup of people with CKD stages 2 and 3. The ERG noted that it was likely the reduction in renal function would be higher and the percentage TKV change would be lower for the subgroup, which would both have the effect of decreasing the ICER compared with the analyses the company had presented using values for the whole population. Hence, the ERG commented that the company's estimate was conservative. The ERG also commented that it would have preferred the annual change in renal function and percentage TKV change to be informed by the actual trial data for the first 3 years, rather than using regression equations, which it stated the company had not justified and which was not a conservative assumption.

3.52 The ERG noted that it had applied a utility decrement of 0.0025 in its revised base case. The ERG stated that it was unclear why the utility decrement as a result of tolvaptan treatment should 'cancel out' the utility decrement as a result of pain reduction (as argued by the company). In addition, the ERG stated that it was not convinced this assumption was justified given the evidence presented, hence the ERG stated it would prefer the use of the conservative utility decrement of 0.0123 for being on tolvaptan treatment.

3.53 The ERG noted that in the company's additional evidence, it had assumed a mortality hazard ratio of 1.6 , and agreed that this was a more plausible estimate than the value of 2.0 the ERG had previously assumed in its exploratory analysis (see section 3.35). The ERG further noted that the company had agreed with the increased monitoring costs applied except for the additional monitoring costs after the second year. The ERG commented that given the serum ALT elevation after 18 months for patients having tolvaptan compared with placebo (as noted by the company, see section 3.49), it was likely that the scenario proposed by the company was more plausible than the increased monitoring costs scenario proposed by the ERG in its exploratory analysis (see section 3.37).

3.54 In response to consultation, the company also submitted additional evidence regarding the treatment of missing data in TEMPO 3:4. In this response, the company presented 'jump to placebo' and 'tipping point' analyses. For the 'jump to placebo' analysis it was assumed that for all patients who withdrew from the TEMPO 3:4 study, there was 100% loss in the efficacy of tolvaptan. In this analysis, the company showed there was still a statistically significant improvement (p<0.0001) in the decline in eGFR for tolvaptan compared with placebo. The 'tipping point' analysis estimated that a 267% drop in the efficacy of tolvaptan was needed before the improvement in eGFR decline was no longer statistically significant. The company argued that based on these analyses, the impact of missing data had not negatively affected the overall conclusions about the efficacy of tolvaptan in ADPKD made from the TEMPO 3:4 study. The company also submitted additional evidence relating to whether TEMPO 3:4 was adequately powered to detect a statistically significant difference in eGFR using the CKD‑EPI equation. Using post‑hoc power calculations, the company estimated that for a statistical significance of 95%, the TEMPO 3:4 study provided over 99% power to detect a difference of 0.977 in the eGFR slope.

ERG's additional analyses

3.55 After critiquing the company's additional evidence submission, the ERG presented additional analyses for both the intention‑to‑treat population and the subgroup with CKD stages 2 and 3; these analyses included the Committee's preferred assumptions as described in the appraisal consultation document.

3.56 The ERG's additional analyses agreed with those of the company's revised base case with the exception of the following:

The ERG assumed equal kidney pain probability for both tolvaptan and placebo (company assumed probabilities of 0.05 and 0.07 respectively; see section 3.48).
The ERG assumed a utility decrement of 0.0123 for tolvaptan treatment (company assumed a utility decrement of 0.0025; see section 3.48).
The ERG used trial data to inform the annual change in renal function and percentage TKV change for the first 3 years (whereas the company used regression equations).

The ERG's additional analyses resulted in a probabilistic mean estimate of the ICER for tolvaptan compared with standard care of £43,514 per QALY gained in the intention to treat population, and £30,025 per QALY gained in the subgroup with CKD stages 2 and 3.

3.57 Full details of all the evidence are available.