A Systematic Review of the Clinical Efficacy and Safety of CFTR Modulators in Cystic Fibrosis

Several placebo-controlled trials have been recently published evaluating novel therapies targeting the defective CFTR protein. This systematic review examines the clinical efficacy and safety of CFTR modulators in individuals with cystic fibrosis (CF) with specific genetic mutations. Online sources were searched for placebo-controlled, parallel-design clinical trials investigating CFTR modulators from January 1, 2005 to March 31, 2018. The primary outcome of interest was FEV1% predicted (ppFEV1). Fourteen RCTs met our eligibility criteria. The largest improvement in ppFEV1 favouring treatment was observed for ivacaftor (IVA) in G551D individuals (≥6 years old). Both tezacaftor-ivacaftor (TEZ-IVA) and lumacaftor-ivacaftor (LUM-IVA) also improved ppFEV1 in F508del homozygous individuals but there was increased reporting of respiratory adverse events with LUM-IVA compared to placebo. IVA also significantly improved ppFEV1 in a sub-group of individuals ≥18 years old with an R117H mutation. No significant improvements in ppFEV1 were observed for IVA, LUM, or TEZ in F508del homozygous individuals, LUM or LUM-IVA in F508del heterozygous individuals, or ataluren in individuals with a nonsense mutation. Significant improvements in ppFEV1 and other clinical outcomes were observed for IVA in G551D individuals, TEV-IVA and LUM-IVA in F508del homozygous individuals, and IVA in adults with a R117H mutation.

Cystic fibrosis (CF) is a genetic condition caused by dysfunction of the cystic fibrosis transmembrane conductance regulator (CFTR) protein. CFTR is located at the apical surface of epithelial cells and the absence of CFTR activity leads to loss of chloride secretion and deficient fluid transport 1 . This results in thick and sticky secretions involving a range of epithelial tissues such as the airways and pancreatic ducts, eventually culminating in end-organ damage and failure. Since the discovery of the CFTR gene in 1989 2 , significant progress has been made in the understanding of how CFTR gene mutations alter protein structure and function leading to reduced CFTR activity 3 .
Although over 2000 variants in the CFTR gene have been identified to date, F508del accounts for most CFTR alleles in patients with CF. This particular mutation leads to abnormal CFTR folding and trafficking causing reduced delivery of CFTR to the cell surface 4 . Another class of CFTR mutations, referred to as "nonsense" mutations, leads to a premature termination codon and reduced synthesis and hence delivery of CFTR to the cell surface 5 . In contrast, "gating" mutations are missense mutations that lead to CFTR proteins that are sufficiently synthesized, processed and trafficked to the cell surface but once they arrive they have defective channel opening leading to diminished chloride secretion 6 .
With advances in our understanding of CFTR biology, a new class of small molecule therapies, referred to as CFTR modulators, have been identified using high-throughput small molecule screening; these drugs are unique as they directly target molecular defects in the CFTR protein to increase CFTR activity [7][8][9][10][11] . For example, CFTR "potentiators" are small molecules capable of increasing the amount of time the CFTR channel is spent in the open position and thus targets CFTR mutations with defective "gating" 10 . CFTR "correctors" are small molecules that can target mutations such as F508del as they can improve CFTR trafficking or transport to the cell surface by stabilizing the 3D conformation of the protein, even if misfolded 11 . Other CFTR modulators, including CFTR "amplifiers" and "translational read-through" agents increase the amount of CFTR protein produced, the latter being specific to mutations leading to a premature termination codon 12,13 .
In recent years, several placebo-controlled clinical trials have been conducted investigating the efficacy and safety of CFTR modulators but the results have varied depending on the specific CF genotype and therapy under investigation 8 . The primary objective of this systematic review was to evaluate the impact of CFTR modulators on lung function and other clinically important outcomes including pulmonary exacerbations, hospitalizations, respiratory symptoms, nutritional status, and adverse events in individuals with CF.

Methods
Search strategy. Our search strategy was developed in accordance with PRISMA guidelines 14 . A systematic search of online databases using key phrases was conducted to identify randomized, placebo-controlled trials published from January 1, 2005 to March 31, 2018. Online databases searched included: MEDLINE, EMBASE, ACP Journal Club, Cochrane Central Register for Controlled Trials (CENTRAL), Cochrane Database of Systematic Reviews (CDSR), Cochrane Methodology Register (CMR), Database of Abstracts of Reviews of Effects (DARE), Health Technology Assessment (HTA), and NHS Economic Evaluation Database (NHSEED). For comprehensiveness, clinical trial registries such as the European Medicines Agency, U.S. National Institute of Health, and the World Health Organization records were accessed and screened. We used the following key phrases which were designed to maximize sensitivity for detecting therapeutic trials in CF: ("cystic fibrosis" OR "CFTR") AND ("drug therapy" OR "clinical trial").
Selection criteria. The literature search and abstracts were reviewed for eligibility independently by two investigators (A.R.H and M.K.). Randomized controlled trials (RCTs) with a parallel design comparing CFTR modulators (e.g. potentiators, correctors, translational read-through agents) to placebo in patients with CF were included. Study inclusion/exclusion were summarized in a PRISMA flow diagram 14 . The level of agreement in the articles selected for full text review and then for inclusion in the review by the two investigators were reported and discrepancies were resolved by the principal investigator (B.S.Q).
Data extraction. The review protocol used in this study is available in the Appendix and was developed in accordance with the PRISMA statement 14 . Two reviewers (A.R.H. and M.K.) independently extracted data. The level of agreement in the data extracted broken down by study characteristics, risk of bias, and effects of the intervention by the two investigators were reported and discrepancies were resolved by the principal investigator (B.S.Q).
Risk of bias assessment. Risk of bias was assessed using the Cochrane Risk of Bias tool 15 . A detailed review of the randomization process, blinding, and allocation sequence concealment was performed.
Outcomes. Change in percent-predicted forced expiratory volume in one second (ppFEV 1 ) was our primary outcome. Secondary efficacy outcomes included protocol-defined pulmonary exacerbations (PEx), hospitalization due to PEx, respiratory symptoms (i.e., Cystic Fibrosis Questionnaire-Revised (CFQ-R) Respiratory domain), and nutritional status (i.e., body mass index and weight). Adverse events with a prevalence of >10% (and involving >2 subjects) from either experimental or control groups, serious adverse events (including deaths) leading to treatment discontinuation, and the prevalence of elevated liver function tests (LFTs) were evaluated. statistical analysis. The statistical analysis was performed using ReviewManager (RevMan 5.3, Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014) in accordance with the Cochrane Handbook 16 . For each clinical outcome, the results were stratified by genotype and type/dose of CFTR modulator. If two or more studies evaluated the same drug at the same dose in the same genotype, the data was pooled using a fixed-effect meta-analysis (Appendix). For the primary outcome, ppFEV 1 , sub-group analyses were planned based on age and baseline ppFEV 1 .

Results
Study selection. The search yielded a total of 789 potentially relevant articles and abstracts. Following fulltext review, thirteen articles (14 placebo-controlled, parallel-group studies) met the inclusion and exclusion criteria ( Fig. 1).

Characteristics of included studies.
A total of eight phase 3 and six phase 2 studies from thirteen original articles were identified. The article by Wainwright et al. included two phase 3 studies accounting for the discordance between the number of articles and studies 17 . The proposed class/mechanism of action for each CFTR modulator along with the number of studies evaluating the therapy is described in Table 1. Characteristics of the included studies and its participants are detailed in Table 2 and Appendix Table 1. The a priori outcomes of interest for the included studies are summarized in Appendix Table 2. Fig. 1.

Risk of bias of included studies. Risk of bias for each included article is summarized in Appendix
Most studies were considered 'low risk' for selection, performance, and attrition bias (Fig. 2) 17-20 . Effects of the intervention. Primary outcome. ppFEV 1 : Of all the CFTR modulators examined to date, individuals with a G551D mutation treated with IVA experienced the largest improvement in ppFEV 1 compared to placebo (n = 2 studies; n = 213; weighted absolute mean difference 10.8, 95% CI: 9.0-12.7) (Fig. 3A) with no heterogeneity (I 2 = 0%) in results between studies (Fig. 3B) 20,21 .
For F508del heterozygous individuals, there was no significant improvement in ppFEV 1 on LUM or LUM-IVA ( Fig. 3A) 24,26 . In a small study involving individuals with F508del/G551D, TEZ-IVA did not lead to a significant improvement in ppFEV 1 compared to IVA alone 22 .
For individuals with the R117H mutation on at least one allele, IVA did not lead to an overall improvement in ppFEV 1 compared to placebo, but there was a significant improvement in a pre-defined subgroup analysis restricted to adults (n = 50; absolute mean difference 5.0, 95% CI 1.2-8.8) 27 . For individuals with a nonsense mutation on at least one allele, ataluren did not result in a significant relative improvement in ppFEV 1 compared to placebo 18 .   Table 3. Of all the CFTR modulators examined, individuals (≥12 years old) with a G551D mutation receiving IVA derived the greatest reduction in PEx risk compared to placebo (n = 1 study; n = 161; OR 0.39, 95% CI: 0.21-0.74) (Appendix Fig. 2A) 20 . LUM-IVA and TEZ-IVA also significantly reduced the risk of PEx compared to placebo in F508del homozygous individuals (≥12 years old) but the risk reduction was less than that observed with IVA in G551D (Appendix Fig. 2A,B) 17,23 . In comparison to placebo, no significant reduction in PEx risk was observed for F508del homozygous individuals or individuals with the R117H mutation on at least one allele receiving IVA, nor for individuals with a nonsense mutation receiving ataluren (Appendix Fig. 2A) 18,25,27 .
Pulmonary exacerbations (PEx) requiring hospitalization: LUM-IVA reduced the risk of PEx requiring hospitalization in F508del homozygous individuals (Appendix Fig. 3A,B) 17 . TEZ-IVA also significantly reduced the rate of PEx leading to hospitalization compared to placebo (n = 1 study; n = 504; rate ratio 0.53, 95% CI 0.34-0.82) but a risk ratio could not be calculated 23 . Individuals with the G551D mutation on at least one allele treated with IVA also experienced a reduction in the risk of PEx requiring hospitalization but this was not statistically significant (Appendix Fig. 3A) 20 .
There was worsening of the CFQ-R Respiratory domain score for F508del homozygous and heterozygous individuals (≥18 years old) on LUM alone (Appendix Fig. 4A) 24 . In a small phase 2 study involving individuals with F508del/G551D, TEZ-IVA did not lead to significant improvement in the CFQ-R Respiratory domain compared to IVA alone 22 . For individuals with a nonsense mutation on at least one allele, ataluren did not modify CFQ-R Respiratory domain score compared to placebo 18 .
Nutritional outcomes (BMI and weight): For individuals with at least one G551D mutation (≥6 years old), significant improvements in weight were observed on IVA compared to placebo (n = 2 studies; n = 213; weighted absolute mean difference: 2.8 kg, 95% CI: 1.8-3.8) (Appendix Fig. 5A,B) 20,21 . For F508del homozygous individuals (≥12 years old), a clinically modest but statistically significant increase in BMI was observed for both doses of LUM-IVA compared to placebo (Appendix Fig. 6A,B) 17 ; however, no significant treatment effect was seen in individuals 6-11 years on LUM-IVA (Appendix Fig. 6A) 19 . TEZ-IVA did not lead to improvement in BMI compared to placebo in individuals 12 years and older (Appendix Fig. 6A) 23 . For F508del heterozygous individuals (≥18 years old), LUM-IVA did not result in significant improvement in weight or BMI compared to placebo 26 . There were no significant improvements in BMI compared to placebo among IVA treated individuals with an R117H mutation (Appendix Fig. 6A) or ataluren treated individuals with a nonsense mutation (data not shown) 18,27 .
Adverse event reporting: CFTR modulators were generally well tolerated compared to placebo ( Appendix  Figs 7-30). For studies involving F508del homozygous and heterozygous individuals, those assigned to LUM had increased dyspnea and "abnormal respiration" compared to placebo (Appendix Figs 11 and 13). F508del homozygous and heterozygous subjects assigned to LUM and LUM-IVA also had more respiratory-related adverse events leading treatment discontinuation compared to placebo (Appendix Table 4) 17,24 . For the one study involving individuals with a nonsense mutation, subjects receiving ataluren had increased incidence of acute kidney injury compared to placebo (15% vs. <1%) resulting in higher rates of treatment discontinuation 18 .
The prevalence of LFT abnormalities was generally similar between treatment and placebo, however there were a few exceptions. A greater proportion of G551D patients had severe ALT elevations (>8x ULN) on IVA compared to placebo (3.6% vs 0%) (Appendix Table 5 T a b le 2. Characteristics of Phase 2 and 3 Clinical Trials Included in the Systematic Review. * Note: 2 nd allele had a mutation predicted to result in the lack of CFTR production or otherwise expected to be unresponsive to ivacaftor (based on in vitro testing). Abbreviations: bid = twice daily; BMI = body mass index; CFQ-R = Cystic Fibrosis Questionnaire Revised; CFTR = cystic fibrosis transmembrane conductance regulator; FVC = forced vital capacity; IV = intravenous; LCI 2.5 = lung clearance index or N 2 washout until 2.5% of the starting N 2 endtidal concentration; N/A = not available; ppFEV 1 = forced expiratory volume in 1 second (FEV 1 )% predicted; OD = once daily; PEx = pulmonary exacerbation; sd = standard deviation.

Discussion
This study represents the most comprehensive systematic review of the efficacy and safety of CFTR modulators performed to date. While evidence-based recommendations for the use of CFTR modulators were recently published and provides a valuable resource for practicing clinicians, this review provides a more concise and up-to-date synthesis of all the placebo-controlled clinical trial data 29 . No prior systematic review has compared all investigational CFTR modulators from phase 2 and 3 RCTs in specific CF genotypes [30][31][32] . As this review highlights, patients with gating mutations such as G551D benefit the most from current CFTR modulators and those that are F508 homozygous have moderate benefit in comparison. Based on published parallel design trials, CFTR modulators have not been effective in F508 heterozygotes or those with nonsense mutations. However, in a recent phase 3 cross-over study evaluating IVA and TEZ-IVA in individuals ≥12 years old with F508del and a residual CFTR function mutation, improvements in ppFEV1 of 4.7% and 6.8%, respectively, were observed compared to placebo 33 . Furthermore, unpublished phase 2 data evaluating TEZ-IVA in combination with "next-generation" corrector molecules have demonstrated significant improvements in ppFEV1 in subjects with F508del and a minimal CFTR function mutation, some of whom have nonsense mutations.
When comparing the efficacy of CFTR modulators across all genotypes for ppFEV 1 , CF individuals (≥6 years old) with the G551D mutation on at least one allele receiving IVA experienced the largest benefit 20,21 . F508del homozygous subjects receiving TEZ-IVA (≥12 years old) and LUM-IVA (≥6 years old) also had improvements in ppFEV 1 compared to placebo but the effect sizes were modest compared to IVA in G551D 17,19,24 . Individuals (≥18 years old) with the R117H mutation on at least one allele treated with IVA experienced similar improvement in ppFEV 1 to F508del homozygous subjects treated with TEZ-IVA and LUM-IVA.
Similar to ppFEV 1 , the effect of CFTR modulators on PEx risk and respiratory symptoms were most pronounced with IVA in G551D adolescents and adults (≥12 years old), with a 60% reduction in PEx risk and a 7-point improvement in the CFQ-R Resp domain 20,21 . F508del homozygous adolescents and adults also had a 40-45% reduction in PEx risk on TEZ-IVA and LUM-IVA. While F508del homozygous subjects experienced improvements in the CFQ-R Resp domain on both TEZ-IVA and LUM-IVA, this was not clinically significant for LUM-IVA. Individuals with a R117H mutation also experienced improvements in the CFQ-R Resp domain on IVA, with a magnitude of change in the adults comparable to that observed with IVA in G551D. The effect of CFTR modulators on weight were most significant with IVA in G551D individuals (≥6 years old). While F508del homozygous individuals (≥12 years old) had improvement in BMI with LUM-IVA, the effect size was modest.
Most of the CFTR modulator therapies examined in this review were well tolerated with the exception of increased reporting of respiratory adverse events (e.g. dyspnea) leading to higher rates of treatment discontinuation in patients randomized to LUM and LUM-IVA. The molecular mechanism responsible for the adverse respiratory effects (e.g. dyspnea, abnormal respiration) for patients on LUM remain unclear but appears to be an off-target effect specific to LUM, as opposed to being related to F508del CFTR correction per se, as similar adverse effects have not been observed with F508del CFTR correction with TEZ-IVA 22,23,34 . There was also increased reporting of acute kidney injury for nonsense mutation patients assigned to ataluren compared to placebo. The  18 as the study authors did not report in their full text publication all outcomes listed in their study protocol including antibiotic use and hospitalization due to CF-related symptoms, disruption to school or work due to CF-related symptoms, and pharmacokinetics. Similarly, Ramsey et al. 20 did not report on all CFQ-R domain items or tertiary outcomes pre-defined in their clinical trial protocol including EQ-5D, oxygen saturation, and outpatient sick visits to the clinic or hospital for CF-related complications. Ratjen et al. 19 did not report data on exacerbations (time to first, number) and the Treatment Satisfaction Questionnaire despite these being listed as secondary endpoints in the publication. Wainwright et al. 17 did not report data on the EQ-5D or Treatment Satisfaction Questionnaire despite it being listed in their trial protocol.
www.nature.com/scientificreports www.nature.com/scientificreports/ long-term safety of CFTR modulator therapies beyond one year could not be assessed in this review and therefore the detection of infrequent or long-term side effects will require ongoing post-marketing surveillance 35,36 .
There are several potential limitations of this review. We excluded cross-over, open-label, and observational studies to avoid carryover effects and to ensure we incorporated the highest level of evidence. We also limited our inclusion to full-text studies which could have resulted in publication bias. We focused on pre-defined clinically important outcomes but did not include multiple-breath washout measurement (e.g. LCI 2.5 ) given the lack of clinical trials utilizing this outcome measure 19 .
There remain several gaps in the placebo-controlled evidence base for CFTR modulators. RCTs to date have excluded young children (<6 years old) and therefore the earliest age of safe use of CFTR modulators remains uncertain. However, small open-label 24-week studies have demonstrated a similar safety profile of IVA in children 1-5 years old with CFTR gating mutations compared to older age groups studied 37,38 . Most RCTs have also excluded CF individuals with severe lung disease (ppFEV 1 < 40%), individuals colonized/infected with bacteria associated with rapid lung function decline (e.g. Burkholderia cenocepacia, Mycobacterium abscessus), and individuals with very frequent pulmonary exacerbations requiring continuous or near continuous systemic antibiotics by virtue of requiring clinical stability and no systemic antibiotics 4 weeks prior to randomization and therefore the efficacy and safety of CFTR modulators in these sub-groups remain unclear. For example, based on observational data, F508del homozygous individuals with advanced lung disease started on LUM-IVA have increased respiratory-related adverse events leading to treatment discontinuation; therefore, closer monitoring following treatment initiation is recommended 39,40 .
Most placebo-controlled RCTs to date have been limited to a maximum duration of 48 weeks and therefore the long-term placebo-controlled effects of these therapies remain unclear. However, an open-label extension trial evaluating the long-term effects of ivacaftor up to 144 weeks has demonstrated sustained clinical benefits of ivacaftor on lung function, weight, patient-reported respiratory symptoms and PEx risk reduction with no new safety concerns 35 . Furthermore, based on combined data from an open-label extension trial and U.S. CF patient registry data, the rate of lung function decline over 3 years was lower in G551D patients treated with ivacaftor compared to propensity-matched controls from the CF registry, suggestive of a disease-modifying effect over the longer term.
In conclusion, based on randomized placebo-controlled parallel design trials, CFTR potentiation with IVA in individuals with a G551D mutation is safe, and results in robust clinical benefits compared to placebo and to date is superior to the effects observed with CFTR modulators in other CF genotypes. The effects of TEZ-IVA and LUM-IVA in F508del homozygous individuals are comparable with respect to the magnitude of change in ppFEV 1 and PEx risk reduction but TEZ-IVA is safer and leads to greater improvement in respiratory symptoms.