Analysis of body mass index, weight loss and progression of idiopathic pulmonary fibrosis

The INPULSIS trials (NCT01335464 and NCT01335477) were two randomised, double-blind, placebo-controlled trials of nintedanib in patients with IPF, performed at 205 sites in 24 countries. Eligibility criteria for the INPULSIS trials have been described [15]. Briefly, the participants were aged ≥40 years, with a diagnosis of IPF, FVC ≥50% predicted and diffusing capacity of the lungs for carbon monoxide (DLco) 30–79% predicted.

Patients were randomised 3:2 to receive nintedanib 150 mg twice daily or placebo for 52 weeks, with a follow-up visit 4 weeks after treatment discontinuation. Treatment interruption and dose reduction to 100 mg twice daily were allowed to manage adverse events. After an adverse event had resolved, the dose could be increased back to 150 mg twice daily. Patients who discontinued study drug prematurely were asked to attend all scheduled visits and undergo all examinations as originally planned. FVC was measured at baseline and at weeks 2, 4, 6, 12, 24, 36, and 52 using sponsor-provided spirometers. Weight was measured at baseline and at weeks 2, 4, 6, 12, 24, 36, and 52.

The pre-specified subgroup analyses of the INPULSIS trials have been reported [15, 16]. Here, we describe post-hoc analyses of efficacy and safety outcomes in subgroups by BMI at baseline and by weight loss over 52 weeks. Efficacy outcomes assessed were the annual rate of decline in FVC (mL/year); absolute change from baseline in FVC (mL); absolute change from baseline in FVC % predicted; time to first investigator-reported acute exacerbation; absolute change from baseline in St. George’s Respiratory Questionnaire (SGRQ) total score; time to absolute decline in FVC ≥10% predicted or death; and time to death, all over 52 weeks. An acute exacerbation was defined based on worsening or development of dyspnoea and the appearance of new abnormalities on high-resolution computed tomography (HRCT), with the exclusion of known causes of acute worsening in respiratory function [15]. The SGRQ is a self-administered questionnaire, comprising three domains (symptoms, activity, impact), which assesses health-related quality of life in patients with respiratory disease on a scale of 0 to 100 [17]. Safety was assessed based on adverse events reported by the investigators (irrespective of causality) with onset after the first dose and up to 4 weeks after the last dose of study drug, which were coded according to preferred terms in the Medical Dictionary for Regulatory Activities (MedDRA) version 20.1. Reductions in weight reported by the investigators as adverse events (based on perceived clinical relevance rather than a defined degree of weight loss) were coded under the preferred term “weight decreased”.

Analyses were conducted using data from patients who received ≥1 dose of nintedanib or placebo. In the primary analysis, the annual rate of decline in FVC (mL/year) was analysed using a random coefficient regression model (with random slopes and intercepts) with fixed effects for trial, treatment, sex, age and height and random effect of patient-specific intercept and time. To assess the effect of BMI at baseline or weight loss on the treatment effect of nintedanib, we repeated the original analyses with the addition of covariates for race (White; Asian; Black/African-American) and either BMI at baseline (< 25; ≥25 to < 30; ≥30 kg/m²) or weight loss over 52 weeks (weight gain or ≤ 5% weight loss; > 5% weight loss) to the model. In addition, we assessed outcomes in subgroups by BMI at baseline (< 25 kg/m²; ≥25 to < 30 kg/m²; ≥30 kg/m²) and by weight loss over 52 weeks (weight gain or ≤ 5% weight loss; > 5% weight loss). Weight loss ≤5% or > 5% over 52 weeks was assessed based on the annual rate of decline in weight analysed using a random coefficient regression model with fixed effects for trial, treatment, sex, age and height and random effect of patient specific-intercept and time. We also analysed the annual rate of decline in FVC (mL/year) in subgroups based on BMI below and at least the median at baseline, and in subgroups by weight loss over 52 weeks (weight gain/no weight loss; > 0 to ≤5% weight loss; > 5 to ≤10% weight loss; > 10% weight loss) based on the change from baseline in weight at week 52 (or, if no week 52 measurement was available, the last measurement after baseline but before week 52). In analyses of the annual rate of decline in FVC, the term subgroup and the interaction terms treatment-by-subgroup, time-by-subgroup and treatment-by-time-by-subgroup were included in the model. The model allowed for missing data, assuming that they were missing at random; missing data were not imputed. The interaction p-values are an indicator of the difference in the effect of nintedanib versus placebo among the subgroups. The other statistical analyses are described in Supplemental Appendix 1. Analyses were not adjusted for multiplicity. Adverse events are presented descriptively.

A total of 638 patients with IPF were treated with nintedanib and 423 with placebo. Baseline characteristics were balanced between the nintedanib and placebo groups. Most (79.3%) of the participants were male. Mean (SD) age at baseline was 66.8 (8.0) years, weight was 79.0 (16.6) kg, BMI was 27.9 (4.6) kg/m², and FVC % predicted was 79.6 (17.8).

At baseline, 307 (28.9%), 453 (42.7%) and 301 (28.4%) patients had BMI < 25 kg/m², ≥25 to < 30 kg/m² and ≥ 30 kg/m², respectively. Compared with patients with BMI ≥30 kg/m², greater proportions of patients with BMI < 25 or ≥ 25 to < 30 kg/m² were of Asian race, and greater proportions with BMI < 25 kg/m² had never smoked (Supplemental Table 1). Low correlation was observed between BMI at baseline and FVC (mL) at baseline (Supplemental Figure 1).

Low correlation was observed between BMI at baseline and the rate of decline in FVC over 52 weeks (Supplemental Figure 1). However, in the placebo group, the mean rate of decline in FVC (mL/year) over 52 weeks was numerically greater in patients with BMI < 25 kg/m² (− 283.3 [SE 22.4]) than BMI ≥25 to < 30 kg/m² (− 207.9 [20.9]) or BMI ≥30 kg/m² (− 175.7 [27.0]) (Figs. 1 and 2; Table 1). In the nintedanib group, the mean rate of decline in FVC over 52 weeks was similar across subgroups with baseline BMI < 25 kg/m², ≥25 to < 30 kg/m² and ≥ 30 kg/m² (mL/year) (− 142.9 [SE 20.8], − 104.0 [16.0] and − 104.5 [21.4] mL/year, respectively) (Figs. 1 and 2; Table 1). Nintedanib reduced the annual rate of decline in FVC versus placebo in all the subgroups by baseline BMI with a treatment effect that was numerically greater in patients with lower BMI, but the p-value for treatment-by-time-by-subgroup interaction did not indicate heterogenous treatment effects across the subgroups (p = 0.31). Similar results were observed when the rate of decline in FVC was analysed in subgroups by baseline BMI above and below the median (27 kg/m²) (Supplemental Table 2). When BMI at baseline (< 25; ≥25 to < 30; ≥30 kg/m²) and race were included in the model analysing the rate of decline in FVC (mL/year) over 52 weeks, the difference between treatment groups was consistent with the original analysis (Supplemental Table 3).

The mean increase (worsening) in SGRQ total score at week 52 in placebo-treated patients was numerically greater in patients with baseline BMI < 25 kg/m² than ≥25 to < 30 kg/m² and ≥ 30 kg/m² (7.0 versus 4.5 and 4.2, respectively). The proportions of placebo-treated patients who had an investigator-reported acute exacerbation was similar across these subgroups (7.9, 7.7 and 7.0%, respectively), while the proportions of patients who died was numerically greater in patients with baseline BMI < 25 kg/m² than ≥25 to < 30 kg/m² or ≥ 30 kg/m² (10.7% versus 6.0 and 7.0%, respectively).

Differences between the nintedanib and placebo groups in absolute changes from baseline in FVC (mL and % predicted) and change in SGRQ total score were numerically greater in patients with BMI < 25 kg/m² than ≥25 to < 30 kg/m² or ≥ 30 kg/m², but the p-values for treatment-by-subgroup interaction did not indicate heterogenous treatment effects across the subgroups (Table 1). Differences between the nintedanib and placebo groups in the proportion of patients with an absolute decline in FVC ≥10% predicted or death, and the proportion of patients who died were consistent across subgroups by BMI at baseline (Table 1).

The mean (SE) annual rate of decline in weight over 52 weeks was − 3.3 (0.2) kg/year and − 1.5 (0.2) kg/year in the nintedanib and placebo groups, respectively. Based on the annual rate of decline in weight, in the nintedanib group, 397 (62.2%) patients had weight gain or ≤ 5% weight loss and 241 (37.8%) had > 5% weight loss over 52 weeks, and in the placebo group, 338 (79.9%) had weight gain or ≤ 5% weight loss and 85 (20.1%) had > 5% weight loss over 52 weeks. Compared with patients with ≤5% weight loss, patients with > 5% weight loss over 52 weeks had a greater proportion of females, lower mean FVC and DLco % predicted at baseline, and higher (worse) mean SGRQ total score at baseline (Supplemental Table 4).

In the placebo group, the rate of decline in FVC (mL/year) was numerically greater in patients with > 5% than ≤5% weight loss over 52 weeks (− 312.7 [SE 32.2] versus − 199.5 [SE 14.4] mL/year) (Figs. 3 and 4; Table 2), as was the mean increase (worsening) in SGRQ total score at week 52 (13.6 versus 3.0). In the placebo group, the proportion of patients who had an acute exacerbation was numerically greater (11.8% versus 6.5%), but the proportion of patients who died was numerically lower (5.9% versus 8.3%), in patients with > 5% than ≤5% weight loss.

When weight loss over 52 weeks and race were included in the model analysing the annual rate of decline in FVC (mL/year), the difference between nintedanib and placebo was consistent with the original analysis (Supplemental Table 3). Nintedanib reduced the annual rate of decline in FVC versus placebo both in patients with ≤5 and > 5% weight loss over 52 weeks, with a greater effect in patients with > 5% than ≤5% weight loss (p = 0.0008 for treatment-by-time-by-subgroup interaction) (Figs. 3 and 4; Table 2). Nintedanib also had a greater effect on the annual rate of decline in FVC in patients with greater weight loss based on the change from baseline in weight at week 52 (no weight loss, > 0 to ≤5% weight loss, > 5 to ≤10% weight loss and > 10% weight loss) (p = 0.0017 for treatment-by-time-by-subgroup interaction) (Supplemental Figure 2). Differences between the nintedanib and placebo groups in change in SGRQ total score and the proportion of patients with an absolute decline in FVC ≥10% predicted or death were numerically greater in patients with > 5% than ≤5% weight loss, but the p-values for treatment-by-subgroup interaction did not indicate heterogenous treatment effects between the subgroups (Table 2).

In both the nintedanib and placebo groups, the proportions of patients with adverse events of decreased appetite, weight decrease, and progression of IPF were greater in patients with lower baseline BMI and in patients with > 5% than ≤5% weight loss over 52 weeks (Tables 3 and 4). The proportion of patients with diarrhoea adverse events was greater in those with > 5% than ≤5% weight loss (Table 4). The adverse event profile of nintedanib was similar across subgroups by baseline BMI and weight loss over 52 weeks, with gastrointestinal adverse events reported more frequently in patients treated with nintedanib than placebo (Tables 3 and 4).