3 Evidence

The appraisal committee (section 5) considered evidence submitted by Vertex Pharmaceuticals and a review of this submission by the evidence review group (ERG). See the committee papers for full details of the evidence.

Clinical effectiveness

3.1 The company did a systematic review of the literature to identify studies on the clinical effectiveness and safety of lumacaftor–ivacaftor for treating cystic fibrosis in people who are homozygous for the F508del mutation. It identified 2 phase III randomised controlled trials, TRAFFIC and TRANSPORT, and 1 ongoing extension study, PROGRESS.

3.2 TRAFFIC and TRANSPORT were international multicentre (including 5 UK centres) double-blind, phase III placebo-controlled trials in people 12 years and over with cystic fibrosis who are homozygous for the F508del mutation. People were randomised in a 1:1:1 ratio to:

lumacaftor 600 mg once daily plus ivacaftor 250 mg twice daily (TRAFFIC, n=183; TRANSPORT, n=185)
a fixed-dose combination of lumacaftor 400 mg–ivacaftor 250 mg twice daily (TRAFFIC, n=182; TRANSPORT, n=187) or
placebo (TRAFFIC, n=184; TRANSPORT, n=187).

People continued to have their usual cystic fibrosis management (standard of care) in all trial arms. In both TRAFFIC and TRANSPORT, people had treatment for 24 weeks and were then enrolled into the 96‑week PROGRESS extension study if they completed treatment. Patients stopped treatment if they could not tolerate the study drug. For lumacaftor–ivacaftor, only data relating to the licensed dosage (fixed-dose combination of lumacaftor 400 mg–ivacaftor 250 mg twice daily) were presented in the company's submission.

3.3 People were eligible for TRAFFIC and TRANSPORT if they had a confirmed diagnosis of cystic fibrosis (defined as a sweat chloride value of 60 mmol/litre or more, or 2 cystic fibrosis-causing mutations and either chronic sinopulmonary disease or gastrointestinal or nutritional abnormalities) and a forced expiratory volume in 1 second (FEV₁) of 40–90% of predicted normal. The company stated that the designs of the trials were almost identical, except that ambulatory electrocardiography screening was included in TRAFFIC and adolescent pharmacokinetic assessments were included in TRANSPORT. The company considered that the baseline characteristics in both trials were generally balanced across treatment arms. However, more people had inhaled antibiotics in the placebo arms (TRAFFIC, 66.3%; TRANSPORT, 72.7%) than in the lumacaftor–ivacaftor arms (TRAFFIC, 62.1%; TRANSPORT, 59.9%).

3.4 The primary outcome in TRAFFIC and TRANSPORT was the absolute change from baseline in percent predicted FEV₁ (ppFEV₁) at week 24, based on a mixed-effects model for repeated measures. The company noted that this was calculated by averaging the mean absolute change at weeks 16 and 24 to reduce variability. The analysis of efficacy outcomes was based on a 'full analysis set' population (that is, people who were randomised into the trials and had received at least 1 dose of the study treatment). All outcomes were assessed on day 1, day 15 and at weeks 4, 8, 16 and 24. The company noted that consistent and sustained improvements in ppFEV₁ were seen from as early as day 15 up until week 48 (that is, at week 24 of PROGRESS). People who had taken lumacaftor–ivacaftor plus standard of care for a total of 48 weeks had an absolute change from baseline in ppFEV₁ of 2.6%. The results for the primary outcome of TRAFFIC, TRANSPORT and a pre-specified pooled analysis are in table 1.

3.5 The company stated that the results (treatment effect) of its pre-specified subgroup analyses were consistent with the results for the overall population. It highlighted that 28 people having lumacaftor–ivacaftor plus standard of care had a ppFEV₁ value less than 40% at baseline, but the clinical benefit and safety profile seen in this group with severe lung dysfunction was comparable with the overall population.

Table 1 Mean absolute and relative change from baseline in ppFEV₁ at week 24

ppFEV ₁	TRAFFIC		TRANSPORT		Pooled analysis
ppFEV ₁	LUM–IVA (n=182)	PBO (n=184)	LUM–IVA (n=187)	PBO (n=187)	LUM–IVA (n=369)	PBO (n=371)
Primary outcome: Absolute change from baseline in ppFEV ₁ (%)
Within-group change (SE)	2.16 (0.53)	−0.44 (0.52)	2.85 (0.54)	−0.15 (0.54)	2.49 (0.38)	−0.32 (0.38)
Mean difference (95% CI)	2.6 (1.2 to 4.0)		3.0 (1.6 to 4.4)		2.8 (1.8 to 3.8)
Secondary outcome: Relative change from baseline in ppFEV ₁ (%)
Within-group change (SE)	3.99 (0.92)	−0.34 (0.91)	5.25 (0.96)	0.00 (0.96)	4.64 (0.67)	−0.17 (0.66)
Mean difference (95% CI)	4.3 (1.9 to 6.8)		5.2 ¹ (2.7 to 7.8)		4.8 (3.0 to 6.6)
Secondary outcome: Response (≥5% increase in average relative change from baseline in ppFEV ₁ )
Proportion of patients (%)	37	22	41	23	39	22
Odds ratio (95% CI)	2.1 (1.3 to 3.3) p=0.002		2.4 (1.5 to 3.7) p=0.001²		2.2 (1.6 to 3.1) p<0.001
Abbreviations: CI, confidence interval; LUM–IVA, lumacaftor–ivacaftor; PBO, placebo; ppFEV₁, percent predicted forced expiratory volume in 1 second; SE, standard error. The company did not report the mean baseline ppFEV₁ for each treatment arm. ¹ Taken from the company's response to clarification. Reported to be 5.3 in the company's original submission. ² p value ≤0.025; however, the company stated that it was not considered statistically significant within the framework of the testing hierarchy. Bold text indicates statistically significant result.

3.6 Secondary outcomes were the frequency and severity of pulmonary exacerbations, and changes in BMI. The company stated that lumacaftor–ivacaftor reduced the rate of pulmonary exacerbations and the need for hospitalisation and intravenous antibiotics compared with placebo (see table 2). It also noted that lumacaftor–ivacaftor improved a person's BMI compared with placebo (see table 3).

Table 2 Company's analysis of pulmonary exacerbations data

Pulmonary exacerbations ¹	TRAFFIC		TRANSPORT		Pooled analysis
Pulmonary exacerbations ¹	LUM–IVA (n=182)	PBO (n=184)	LUM–IVA (n=187)	PBO (n=187)	LUM–IVA (n=369)	PBO (n=371)
Total number of exacerbations at week 24 (event rate per 48 weeks)
Number (rate)	73 (0.71)	112 (1.07)	79 (0.67)	139 (1.18)	152 (0.70)	251 (1.14)
Rate ratio	0.66 (p=0.02)²		0.57 (p<0.001)²		0.61 (p<0.001)
Number of exacerbations needing hospitalisation at week 24 (event rate per year)
Number (rate)	17 (0.14)	46 (0.36)	23 (0.18)	59 (0.46)	40 (0.17)	105 (0.45)
Rate ratio	0.38 (p=0.0008)		0.39 (p=0.0002)		0.39 (p<0.0001)
Number of exacerbations needing IV antibiotics at week 24 (event rate per year)
Number (rate)	33 (³)	62 (³)	31 (0.23)	87 (0.64)	64 (0.25)	149 (0.58)
Rate ratio	(p=0.0050) ³		0.36 (p<0.0001)		0.44 (p<0.0001)
Mean duration in days of pulmonary exacerbations
Total	7.81	13.07	8.45	18.23	8.14	15.67
Total	p<0.0001		p<0.0001		p<0.0001
Hospitalisation	NR	NR	NR	NR	2.48	7.64
IV antibiotics	NR	NR	NR	NR	3.79	10.13
Abbreviations: IV, intravenous; LUM–IVA, lumacaftor–ivacaftor; NR, not recorded; PBO, placebo. ¹ Estimated using a negative binomial regression model that included treatment, study, sex, age group at baseline, and ppFEV₁ severity at screening. ² p value ≤0.025; however, the company stated that it was not considered statistically significant within the framework of the testing hierarchy. ³ The company stated that these rates could not be estimated because the negative binomial model did not converge. Bold text indicates statistically significant result.

Table 3 Absolute change from baseline in BMI at week 24

BMI	TRAFFIC		TRANSPORT		Pooled analysis
BMI	LUM–IVA (n=182)	PBO (n=184)	LUM–IVA (n=187)	PBO (n=187)	LUM–IVA (n=369)	PBO (n=371)
Baseline (SD)	21.68 (3.169)	21.03 (2.956)	21.32 (2.894)	21.02 (2.887)	21.50 (3.034)	21.02 (2.918)
Within-group change (SE)	0.32 (0.071)	0.19 (0.070)	0.43 (0.066)	0.07 (0.066)	0.37 (0.048)	0.13 (0.048)
Mean difference (95% CI)	0.13 (−0.07 to 0.32)		0.36 (0.17 to 0.54)		0.24 (0.11 to 0.37)
Abbreviations: CI, confidence interval; LUM–IVA, lumacaftor–ivacaftor; PBO, placebo; SD, standard deviation; SE, standard error. Bold text indicates statistically significant result.

3.7 Health-related quality of life was measured using the Cystic Fibrosis Questionnaire-Revised (CFQ‑R) and the EuroQol‑5 dimensions‑3 levels survey (EQ‑5D‑3L); see table 4. CFQ‑R is measured on a scale of 0–100, with higher scores representing better health. An absolute change of at least 4 points is considered a minimal clinically important difference for the CFQ‑R respiratory domain. The company stated that people in the trials had very high baseline EQ‑5D‑3L values because they are born with cystic fibrosis and perceive their quality of life to be 'normal' (that is, equivalent to people without cystic fibrosis). As a result, people with cystic fibrosis score their health-related quality of life as high, so statistically significant improvements in health-related quality of life are unlikely to be seen because of this ceiling effect. It noted that this is a challenge commonly reported in cystic fibrosis trials.

Table 4 Health-related quality-of-life data at week 24

Health-related quality of life	TRAFFIC		TRANSPORT		Pooled analysis
Health-related quality of life	LUM–IVA (n=182)	PBO (n=184)	LUM–IVA (n=187)	PBO (n=187)	LUM–IVA (n=369)	PBO (n=371)
Cystic Fibrosis Questionnaire-Revised: respiratory domain
Baseline (SD)	69.29 (17.4)	70.54 (16.03)	67.36 (18.5)	67.05 (18.4)	68.31 (18.0)	68.78 (17.3)
Within-group change (SE)	2.60 (1.192)	1.10 (1.161)	5.66 (1.169)	2.81 (1.153)	4.10 (0.834)	1.88 (0.818)
Mean difference (95% CI)	1.5 (−1.69 to 4.69)		2.9 (−0.27 to 5.98)		2.2 (−0.01 to 4.45)
EuroQol-5 dimensions-3 levels survey (EQ-5D-3L)
Baseline (SD)	0.9237 (0.104)	0.9217 (0.098)	0.9171 (0.10837)	0.9267 (0.10462)	Not reported by the company
Within-group change (SE)	0.0006 (0.0074)	0.01 (0.0076)	0.0117 (0.00673)	0.0108 (0.00683)
Mean difference (95% CI)	0.0095 (−0.0109, 0.0298)		−0.0009 (−0.0192, 0.0174)
Abbreviations: CI, confidence interval; LUM–IVA, lumacaftor–ivacaftor; PBO, placebo; SD, standard deviation; SE, standard error. Bold text indicates statistically significant result.

3.8 Adverse event data were available from the pooled analysis of TRAFFIC and TRANSPORT, and from PROGRESS (see table 5). The most common adverse events reported for lumacaftor–ivacaftor compared with placebo were cough (28.2% compared with 40.0%), diarrhoea (12.2% compared with 8.4%), dyspnoea (13.0% compared with 7.8%), haemoptysis (13.6% compared with 13.5%), headache (15.7% compared with 15.7%), increase in sputum production (14.6% compared with 18.9%), infective pulmonary exacerbation (35.8% compared with 49.2%), nasopharyngitis (13.0% compared with 10.8%), nausea (12.5% compared with 7.6%) and upper respiratory tract infection (10.0% compared with 5.4%). No deaths were reported in either TRAFFIC or TRANSPORT, and 1 death was reported in PROGRESS, which was considered unrelated to treatment.

Table 5 Summary of adverse event data

Number of people (%)	Pooled analysis (24 weeks)		PROGRESS (0–48 weeks): LUM–IVA (n=544)
Number of people (%)	LUM–IVA (n=369)	PBO (n=370)	PROGRESS (0–48 weeks): LUM–IVA (n=544)
Any AE	351 (95.1)	355 (95.9)	532 (97.8)
Any grade 3 or 4 AE	45 (12.2)	59 (15.9)	100 (18.4)
At least 1 serious AE	64 (17.3)	106 (28.6)	159 (29.2)
Stopping treatment because of AE	17 (4.6)	6 (1.6)	34 (6.3)
Abbreviations: AE, adverse event; LUM–IVA, lumacaftor–ivacaftor; PBO, placebo.

Cost effectiveness

3.9 The company submitted an individual patient-level microsimulation model that compared lumacaftor–ivacaftor plus standard of care with standard of care alone in people 12 years and older with cystic fibrosis who are homozygous for the F508del mutation. The company used a 4‑week cycle length for the first 2 years and yearly thereafter. It did the economic analysis from an NHS and personal social services perspective and chose a lifetime time horizon. Costs and health effects were discounted at an annual rate of 3.5% and a half‑cycle correction was applied.

3.10 Baseline characteristics (age, sex, weight-for-age z-score and ppFEV₁) were taken from 1,097 people in TRAFFIC and TRANSPORT who had ppFEV₁ data available at baseline. Statistical bootstrapping methods were used to randomly create a group of 1,000 people (see table 6). Baseline diabetes and infection status were taken from the UK Cystic Fibrosis Registry, and every person was assumed to have pancreatic insufficiency. Each person's data were run through the company's model twice (that is, once for lumacaftor–ivacaftor plus standard of care, and once for standard of care alone). The company ran its economic model for 6 replications on the group of 1,000 people and used different random numbers for each replication.

Table 6 Baseline characteristics

Characteristic	Mean of total trial population (n=1,097)	UK Cystic Fibrosis Registry
Age (years)	25.5	19.6
Male	50.6%	Not reported
BMI	21.2	Not reported
Percent predicted forced expiratory volume in 1 second (ppFEV₁)	60.6%	75%

3.11 Survival was estimated using a 2‑part calculation in the company's model:

Firstly, the age-specific background mortality was derived from UK Cystic Fibrosis Registry data (2013). The company fitted a series of parametric curves to a Kaplan–Meier analysis of 6,082 cystic fibrosis patients (all genotypes) divided into groups based on their year of birth (ranging from 1980 to 2008). The company simulated patient-level data based on digitised curves and the number of patients in each group using the exponential, generalised gamma, Gompertz, log-logistic, log-normal and Weibull functions. The company stated that the curves estimated from the generalised gamma, Gompertz and Weibull functions provided the best statistical fit. In the base-case analysis, the company used the Weibull function because it considered it provided the most valid long-term survival projections based on visual inspection and clinical expert opinion (that is, an estimated median survival of 40.8 years, with approximately 0% alive by 80 years).
Secondly, the age-specific mortality was adjusted to take into account 9 clinical and patient characteristics that the company considered as predictors of survival based on a Cox proportional hazards model published by Liou et al. (2001): ppFEV₁, pulmonary exacerbations, age, sex, weight-for-age z-score, pancreatic sufficiency, diabetes, infection with Burkholderia cepacia and Staphylococcus aureus. These clinical and patient characteristics were updated at the end of each cycle, and subsequently used to adjust the underlying survival function.

3.12 The company stated that the ppFEV₁ of people having lumacaftor–ivacaftor plus standard of care increased by 2.8% by week 16 and was maintained until week 24 in its economic model, to reflect the changes seen in TRAFFIC and TRANSPORT. However, the ppFEV₁ of people having standard of care alone was assumed to remain unchanged over the first 24 weeks of the company's economic model. After week 24 in the model, ppFEV₁ declined for people having standard of care alone and for people having lumacaftor–ivacaftor plus standard of care. The decline in ppFEV₁ was age dependent for standard of care alone based on a large US and Canadian observational study of 4,161 adults and 1,359 children. Decline in ppFEV₁ was not age dependent for lumacaftor–ivacaftor plus standard of care based on TRAFFIC, TRANSPORT and PROGRESS using a mixed-model analysis (see table 7). The company stated that it also included a lower bound ppFEV₁ of 15% to avoid unrealistically low values. The company's model also included pulmonary exacerbations needing intravenous antibiotics and hospitalisation, and modelled a person's BMI based on weight-for-age z-scores using data from TRAFFIC and TRANSPORT (see table 7). The company also assumed that 24.7% of people with a ppFEV₁ below 30% had a lung transplant. Post-lung transplant mortality was assumed to be 15.2% in the first year, and 6.1% for each subsequent year based on 6,766 adults with cystic fibrosis in the UK who had a lung transplant between 1990 and 2012.

Table 7 Summary of the company's ppFEV₁, exacerbation, and weight-for-age z-score inputs

Input		LUM–IVA plus SoC	SoC
ppFEV₁	From week 16–24	Baseline +2.8%	Baseline
ppFEV₁	Annual change after week 24	Age <18: −0.68% Age 18–24: −0.68% Age ≥25: −0.68%	Age <18: −2.34% Age 18–24: −1.92% Age ≥25: −1.45%
Annual rate of pulmonary exacerbations		Predicted, conditional on ppFEV₁ and age, multiplied by 0.442	Predicted, conditional on ppFEV₁ and age
Weight-for-age z-scores	First 24 weeks	Baseline +0.068	Baseline
Weight-for-age z-scores	After 24 weeks	Baseline +0.068	Baseline
Abbreviations: LUM–IVA, lumacaftor–ivacaftor; ppFEV₁, percent predicted forced expiratory volume in 1 second; SoC, standard of care.

3.13 The drug costs for lumacaftor–ivacaftor were based on the list price (£2,000 per week) and were assumed to reduce by 89% after 12 years because of patent expiry (see section 4.18). In the company's economic model, approximately 6.8% of people having lumacaftor–ivacaftor stopped treatment during the first 24 weeks to reflect TRAFFIC and TRANSPORT, and after 24 weeks their ppFEV₁ declined at the rate estimated for standard of care alone. The company assumed that after 24 weeks, no more people stopped treatment with lumacaftor–ivacaftor. It included an adherence rate of 90% for lumacaftor–ivacaftor, but noted that the adherence rate in the trials was 96.5%. The company's costs for managing cystic fibrosis were dependent on lung function and were based on a retrospective 24‑month study in 8 UK specialist centres of 200 people with cystic fibrosis who are homozygous for the F508del mutation. Hospitalisation costs for pulmonary exacerbations were assumed to reduce by 61% for people having lumacaftor–ivacaftor plus standard of care, based on the rate ratio of pulmonary exacerbations needing hospitalisation in TRAFFIC and TRANSPORT. The company included adverse reactions that were reported in more than 5% of people having lumacaftor–ivacaftor plus standard of care compared with standard of care alone, costed as a GP visit. It also included costs associated with lung transplant and monitoring (liver function tests).

3.14 To estimate the health-related quality of life in the economic model, the company used a multivariate mixed-model repeated measures regression analysis to model the relationship between EQ‑5D utility values, lung function (ppFEV₁) and pulmonary exacerbations reported in TRAFFIC and TRANSPORT. Therefore, the utility for a given patient varied throughout the time horizon of the company's economic model. The company did not apply any utility decrements for adverse events other than pulmonary exacerbations. Utility values for lung transplant were taken from Whiting et al. (2014) and the weighted-average utility for people post-transplant was estimated to be 0.81.

3.15 Table 8 presents a summary of the company's base-case and probabilistic cost-effectiveness results for lumacaftor–ivacaftor plus standard of care compared with standard of care alone. Table 9 presents a summary of the health outcomes predicted by the company's base-case analysis.

Table 8 Summary of company's base-case and probabilistic results

	LUM–IVA plus SoC	SoC	Increment
Base-case analysis
Life years	13.78	10.32	3.46
QALYs	12.38	8.92	3.45
Costs	£1,131,202	£377,632	£753,570
ICER (£/QALY)			£218,248
Probabilistic sensitivity analysis
Life years	13.82	10.34	3.48
QALYs	12.42	8.94	3.49
Costs	£1,125,946	£377,152	£748,794
ICER (£/QALY)			£214,838
Abbreviations: ICER, incremental cost-effectiveness ratio; LUM–IVA, lumacaftor–ivacaftor; QALY, quality-adjusted life year; SoC, standard of care.

Table 9 Summary of health outcomes predicted by company's base-case analysis

Outcome	LUM–IVA plus SoC	SoC	Increment
Projected median survival (years)	43.84	36.15	7.69
Undiscounted life years	24.52	15.05	9.47
Mean ppFEV₁ cumulative change	−13.51	−21.89	8.37
Mean years with ppFEV₁ ≥70%	4.08	1.14	2.94
Mean years with ppFEV₁ 40–70%	17.10	8.84	8.26
Mean years with ppFEV₁ 30–40%	2.58	2.66	−0.08
Mean years with ppFEV₁<30%	0.77	2.42	−1.65
Annual rate of pulmonary exacerbation	0.46	1.24	−0.78
Percent having lung transplant	1.82%	6.80%	−4.98%
Mean years until lung transplant	46.49	19.34	27.14
Abbreviations: ICER, incremental cost-effectiveness ratio; LUM–IVA, lumacaftor–ivacaftor; ppFEV₁, percent predicted forced expiratory volume in 1 second; SoC, standard of care.

3.16 The company presented the results of a univariate sensitivity analysis and several scenario analyses. The univariate sensitivity analysis suggested that the base-case incremental cost-effectiveness ratios (ICERs) were most sensitive to the rate of ppFEV₁ decline for lumacaftor–ivacaftor, the discount rate and costs of managing cystic fibrosis. The company presented the results of several scenario analyses (see table 10) and subgroup analyses (see table 11).

Table 10 Company's scenario analyses

Scenario	LUM–IVA plus SoC		SoC		ICER (£/QALY)
Scenario	Total cost	Total QALYs	Total cost	Total QALYs	ICER (£/QALY)
Base case	£1,131,202	12.38	£377,632	8.92	£218,248
Discount rate 1.5%	£1,381,148	16.56	£467,146	10.83	£159,678
Rate of ppFEV₁ decline (LUM–IVA): +20%	£1,121,358	12.04	£377,632	8.92	£238,795
Rate of ppFEV₁ decline (LUM–IVA): −20%	£1,140,078	12.76	£377,632	8.92	£199,003
Rate of ppFEV₁ decline (SoC): Canadian cystic fibrosis population	£1,131,202	12.38	£350,697	8.07	£181,366
PE rate: all events	£1,114,588	12.09	£377,632	8.92	£233,018
Utility values: TRAFFIC and TRANSPORT by ppFEV₁ strata	£1,131,202	12.52	£377,633	9.25	£230,769
Utility values: Tappenden et al.	£1,131,202	11.09	£377,632	7.97	£241,109
Utility values: Acaster et al.	£1,131,202	9.52	£377,632	6.86	£283,458
Stop treatment at rate of 1.9% post 24 weeks	£1,092,338	12.27	£377,633	8.92	£213,910
Survival curve: Gompertz	£939,058	10.00	£292,406	7.18	£228,830
Adherence: 96.5%	£1,185,593	12.38	£377,633	8.92	£234,000
Abbreviations: ICER, incremental cost-effectiveness ratio; LUM–IVA, lumacaftor–ivacaftor; PE, pulmonary exacerbation; ppFEV₁, percent predicted forced expiratory volume in 1 second; QALYs, quality-adjusted life years; SoC, standard of care.

Table 11 Company's subgroup analyses by baseline ppFEV₁

Subgroup	LUM–IVA plus SoC		SoC		ICER (£/QALY)
Subgroup	Total cost	Total QALYs	Total cost	Total QALYs	ICER (£/QALY)
Baseline ppFEV₁ >40%	£1,176,340	13.07	£393,337	9.40	£213,336
Baseline ppFEV₁ <40%	£745,575	5.76	£231,284	4.05	£300,688
Baseline ppFEV₁ >70%	£1,366,094	17.72	£493,464	13.34	£199,481
Baseline ppFEV₁ <70%	£1,053,685	10.48	£334,864	7.30	£225,907
Abbreviations: ICER, incremental cost-effectiveness ratio; LUM–IVA, lumacaftor–ivacaftor; ppFEV₁, percent predicted forced expiratory volume in 1 second; QALY, quality-adjusted life years; SoC, standard of care.

Evidence review group comments on clinical effectiveness

3.17 The ERG stated that the company's systematic literature review was of reasonable quality and all relevant randomised controlled trials were identified.

3.18 The ERG stated that TRAFFIC and TRANSPORT were generally of good quality. It was aware that the expert statements NICE received suggested they were the largest trials of a cystic fibrosis therapy to date. The ERG's clinical adviser also considered that the trial populations were generalisable to people in clinical practice in England.

3.19 The ERG stated that because both trials included people with mild to moderate cystic fibrosis (that is, ppFEV₁ of 40–90% at screening), the clinical evidence may not be generalisable to people with severe cystic fibrosis, or people with very mild cystic fibrosis.

3.20 The ERG stated that the company's method used to pool the results from TRAFFIC and TRANSPORT was likely to be appropriate, but insufficient details were provided by the company for the ERG to determine this.

3.21 The ERG's clinical adviser noted that estimating the mean absolute change from baseline ppFEV₁ at week 24 by averaging the mean absolute change at weeks 16 and 24 was common in cystic fibrosis trials and considered acceptable.

3.22 The ERG's clinical adviser stated that absolute changes in ppFEV₁ were more clinically relevant than relative changes, and that an absolute change in ppFEV₁ of 5% or more would be considered clinically important. The ERG concluded that although lumacaftor–ivacaftor plus standard of care had statistically significant effects on key outcomes compared with standard of care alone, it was unclear how clinically significant they were.

3.23 The ERG noted that because the company's trials were short, the long-term effects of lumacaftor–ivacaftor were uncertain.

Evidence review group comments on cost effectiveness

3.24 The ERG stated that the company's model appeared to capture the important features of cystic fibrosis.

3.25 The ERG stated that it was not possible to compare the baseline characteristics of the company's trial population with the subgroup of people included in the Cystic Fibrosis Registry who are homozygous for the F508del mutation and with a ppFEV₁ of 40–90%. As a result, it was unclear whether the differences in mean age and ppFEV₁ were because of different characteristics among the subtypes of cystic fibrosis or the result of differences between the trial population and the relevant UK cystic fibrosis population (see table 6). The ERG further highlighted that most of the natural history parameters in the company's model were informed by data for the whole UK cystic fibrosis population and not by data for the population with cystic fibrosis who are homozygous for the F508del mutation. Therefore, the ERG concluded that any differences between the modelled and real populations, and the impact this may have on efficacy and cost effectiveness, should be considered when interpreting the company's results.

3.26 The ERG acknowledged that the company had highlighted the challenges of estimating survival from the Cystic Fibrosis Registry:

There is selection bias with groups born earlier, because of a lack of available follow‑up data earlier in their lifetime, which may artificially inflate survival rates.
Observed survival in groups born more recently is relatively immature, making long-term extrapolation potentially unreliable.

3.27 The ERG highlighted that using the absolute difference in ppFEV₁ by averaging across the 16‑week and 24‑week measurements was more favourable for lumacaftor–ivacaftor than using the 24‑week measurement alone.

3.28 The ERG stated that short-term benefits were assumed to persist over much longer time horizons in the company's model because the long-term benefit of lumacaftor–ivacaftor on ppFEV₁ was based on 48‑week data. The ERG further considered that using different and non-randomised data sets for the long-term extrapolations may bias the estimates for each treatment group.

3.29 The ERG noted that the company had not provided any long-term evidence to support the assumptions around the benefits of lumacaftor–ivacaftor on pulmonary exacerbations (maintained for as long as people stayed on treatment) and weight-for-age z-score (maintained for the remainder of a person's life irrespective of whether they stopped treatment). Therefore these were associated with uncertainty.

3.30 The ERG highlighted that the company assumed the impact of lumacaftor–ivacaftor on pulmonary exacerbations was independent from, rather than partially caused by, its effect on ppFEV₁. The ERG was aware that the company's clinical experts verified this assumption, but the ERG noted that the company risked double counting the benefits of treatment.

3.31 The ERG considered that no robust rationale was provided by the company for the assumed price reduction after 12 years (see section 3.13). The ERG stated that the company's disease management costs were taken from a population that included people with a different mutation (G551D) and not only the F508del mutation as specified by the company.

3.32 The ERG considered that the company's assumption that pre-transplant health-related quality of life depended only on ppFEV₁ and pulmonary exacerbations may not be justified if other treatment-related factors affect health-related quality of life (for example, adverse events with lumacaftor–ivacaftor).

3.33 The ERG explored the impact of applying a conservative assumption in the company's economic model. The assumption was that after the time horizon of the trial, the effect of lumacaftor–ivacaftor on pulmonary exacerbations was based solely on any differences in ppFEV₁ (see section 3.30). This analysis explored by the ERG estimated incremental costs of £704,645 and an incremental quality-adjusted life year (QALY) gain of 2.59, with an estimated ICER of £272,265 per QALY gained for lumacaftor–ivacaftor plus standard of care compared with standard of care alone.

3.34 The ERG also presented an exploratory analysis that included the following changes (see table 12):

Setting the adherence rate to 96.5% rather than 90% so that the same adherence rate is used for both effectiveness and cost data (see section 3.13).
People could stop lumacaftor–ivacaftor treatment after 24 weeks. After this time, the rate for people stopping treatment was assumed to be 1.9% annually, in line with a rate used by the company in its scenario analysis.
The mean absolute change in ppFEV₁ from baseline was based on the 24-week data alone rather than the average of the 16‑week and 24‑week data (that is, replacing an absolute increase of 2.8% [see section 3.12] with an absolute increase of 2.45%). The absolute increase of 2.45% was estimated by the ERG from a graph in the company's submission showing the mean absolute change in ppFEV₁ from baseline at various time points of the trials.

Table 12 Summary of ERG's exploratory analysis

	LUM–IVA plus SoC	SoC	Increment
Life years	13.56	10.32	3.24
QALYs	12.14	8.92	3.22
Costs	£1,092,269	£377,632	£714,637
ICER (£/QALY)	–		£221,992
Abbreviations: ICER, incremental cost-effectiveness ratio; LUM–IVA, lumacaftor–ivacaftor; QALYs, quality-adjusted life years; SoC, standard of care.

3.35 The ERG presented a sensitivity analysis around the company's assumed price reduction using the exploratory analysis model (see table 13).

Table 13 Summary of the ICERs for the ERG's sensitivity analysis of generic pricing

		Percent price reduction for generic medicine
		89%	80%	70%¹	60%¹	50%¹
Time until generic alternative becomes available	10 years	£203,100	£215,971	£230,272	£244,573	£258,874
	12 years	£221,992²	£232,953	£245,132	£257,311	£269,490
	15 years	£244,675	£253,342	£262,972	£272,602	£282,232
	20 years	£271,764	£277,692	£284,279	£290,865	£297,452
	Never	£330,385³	£330,385³	£330,385³	£330,385³	£330,385³
¹ Costs were calculated by extrapolating costs from the 89% and 80% scenarios. ² ERG's exploratory analysis (see section 3.34, table 12). ³ The company's base-case incremental cost-effectiveness ratio increased from £218,248 to £349,337 per QALY gained when the price reduction for lumacaftor–ivacaftor was removed.

3.36 The ERG also presented a sensitivity analysis exploring the impact of applying the lower and upper bound of the 95% confidence interval for the annual ppFEV₁ decline estimated from weeks 4–48 in the company's trials for people having lumacaftor–ivacaftor (see table 14).

Table 14 Summary of ERG's sensitivity analysis around the annual decline in ppFEV₁ in people having lumacaftor–ivacaftor

	LUM–IVA plus SoC	SoC	Increment
Lower bound of 95% confidence interval (1.58% ppFEV ₁ decline per year after 24 weeks in people having lumacaftor–ivacaftor)
Life years	11.80	10.32	1.48
QALYs	10.41	8.92	1.49
Costs	£1,061,163	£377,632	£683,532
ICER (£/QALY)	–		£459,045
Upper bound of 95% confidence interval (−0.16% ¹ ppFEV ₁ decline per year after 24 weeks in people having lumacaftor–ivacaftor)
Life years	16.07	10.32	5.76
QALYs	14.73	8.92	5.81
Costs	£1,164,047	£377,632	£786,415
ICER (£/QALY)	–		£135,464
Abbreviations: ICER, incremental cost-effectiveness ratio; LUM–IVA, lumacaftor–ivacaftor; ppFEV₁, percent predicted forced expiratory volume in 1 second; QALY, quality-adjusted life years; SoC, standard of care. ¹ The upper bound of 95% confidence interval for annual ppFEV₁ decline indicated a slight improvement with lumacaftor–ivacaftor.

Lumacaftor–ivacaftor for treating cystic fibrosis homozygous for the F508del mutation