Predicting weight gain in patients with Cystic Fibrosis on triple combination modulator

Abstract

Background

Cystic fibrosis (CF) is caused by CF transmembrane conductance regulator (CFTR) gene mutations producing dysfunctional CFTR proteins leading to progressive clinical disease. Elexacaftor-tezacaftor-ivacaftor (ETI) remarkably improves lung disease but is associated with substantial weight gain.

Study Design and Methods

We performed a single-center longitudinal study predicting 6-month weight gain after ETI initiation. We used linear mixed effects modeling (LME) to determine association of ETI treatment with changing body mass index (BMI). Using linear regression, we examined BMI prediction models with distinct combinations of main effects to identify a model useful for patient counseling. We used up to eight commonly observed clinical characteristics as input variables (age, sex, percent predicted FEV1 [FEV1%], F508del homozygous state, pancreatic sufficiency, HgbA1c, prior modulator use and prior year number of pulmonary exacerbations).

Results

We evaluated 154 patients (19–73 years old, 54% female, FEV₁%=19–121, 0–6 prior year pulmonary exacerbations). LME demonstrated an association between ETI use and weight increases. Exhaustive testing suggested a parsimonious linear regression model well-fitted to data that is potentially useful for counseling. The two variable model shows that on average, BMI decreases by 0.045 (95% Confidence Interval [CI]=−0.069 to −0.021, p<0.001) for every year of age and increases by 0.32 (CI=0.14 to 0.50, p=0.001) for each additional prior year exacerbation at the time of ETI initiation.

Interpretation

Young patients with many prior year pulmonary exacerbations likely have the largest 6 month weight gain after starting ETI.

Introduction and Background

Cystic fibrosis (CF) transmembrane conductance regulator (CFTR) gene mutations result in dysfunctional CFTR proteins and progressive clinical disease. CF involves multiple organs, including the lungs, pancreas, GI tract and liver. CF was first described in 1938, and the observed median lifespan for a person born with CF was 6–7 months.¹ Novel FDA-licensed drugs targeting defective CFTR beginning in 2011 accelerated intervening gains in expected lifespan for people with CF: according to the 2022 Cystic Fibrosis Foundation Annual Data Report, people born with CF in 2021 are expected to live to a median age of approximately 65.6 years.²

Malnutrition and progressive lung disease were the two most important factors originally precipitating death in CF.¹ Higher weight continues to predict better survival.^3,4 Past studies demonstrated an association between malnutrition status and lung function, and BMI goals were established as part of optimal nutrition care in CF: 22–25 for women and 23–25 for men.^5,6

Modulator drugs targeting defective CFTR protein synthesis altered the clinical landscape of CF, including aspects related to malnutrition and body weight. Currently there are four drugs or drug combinations approved in the US to treat specific CFTR mutations: ivacaftor, lumacaftor-ivacaftor, tezacaftor-ivacaftor and elexacaftor-tezcaftor-ivacaftor (ETI). All the drugs are associated with some weight gain. However, ivacaftor and ETI are associated with the most pronounced weight gain.^7–9 On average, clinical trial participants randomized to ETI experienced a weight gain of 1.13 BMI points compared to 0.09 for participants randomized to placebo during 24 weeks of observation.⁷ Other recent studies evaluated weight gain on ETI over longer time periods. In the PROMISE post approval study (n=487),¹⁰ average increase in BMI at the six month mark was 1.24 kg/m². In a single-center retrospective study (n=134) average BMI increase was 1.47 kg/m² after an average of 12.2 months on ETI.¹¹ In a multi-center retrospective study assessing effectiveness of ETI in advanced lung disease (n=26), average BMI increase after 48 weeks on ETI was 2.08 kg/m².¹²

Weight gain in the CF population is recently contentious. Initially increased weight following initiation of modulator drugs was celebrated as a reversal of malnutrition. However, as many patient weights continued to increase, concern mounted regarding how much weight is too much. Both patients and clinicians share this concern. From a clinical perspective, increased weight is associated with improved lung function, even when weight is above BMI cut points of 25 (overweight) and 30 (obese).¹³ However, weight gain is a difficult experience for many patients: culturally, weight gain is often considered a moral failing, and patients may express shame regarding their weight gain.¹⁴ This study originated from a common patient question before starting ETI, “Will I be someone who gains a lot of weight?” To address urgent patient concerns at the time of a decision to prescribe ETI, we examined commonly available clinical data in order to create a predictive model of weight gain associated with ETI treatment to enhance clinicians assessment individual risk for excessive weight gain and facilitate discussion of healthy eating and movement habits at the start of ETI.

Methods

After initial evaluation (IRB 00144047), the Institutional Review Board of the University of Utah exempted our study from further review and did not require informed consent due to the minimal risk nature of the study. We reviewed the charts of patients followed in our clinic for at least a year prior to taking ETI with clinical measurements of weight within two months before and up to twelve months after starting the modulator. We included all patients who started and continued medication for at least 3 months following drug licensing in November, 2019 through September, 2020. We followed patients from initiation of treatment through January 11, 2021. We excluded patients who started ETI less than three months prior to chart review, were pregnant or had undergone solid organ transplant.

We recorded weights at 3, 6 and 9–12 months after starting the modulator, mutation type, sex, age, previous modulator use, pancreatic sufficiency status, diabetes status, hemoglobin A1c (HgbA1c), percent predicted forced expiratory volume in one second (FEV₁%) as estimated using the National Health and Nutrition Examination Survey III equations because these were the equations in use for clinical decision making during the study period,¹⁵ and number of hospitalizations due to pulmonary exacerbations of CF in the year prior to starting the modulator. We noted any non-standard dosing of the modulator and documented dietary interventions for weight loss after starting the modulator, which might have introduced bias into the analysis.

Statistical Analysis

We performed all analyses in the R statistical environment.¹⁶ We performed summary statistics of age, BMI on or within 2 months before the date of modulator initiation and at each of one or two follow up appointments. Because we had repeated weight measurements following start of ETI, we performed linear mixed effects modeling^17,18 to establish that there is a relationship between ETI use and BMI change. We used BMI as the output variable and time to measurement as the input variable adjusted by baseline BMI, age, sex, number of prior exacerbations and other recorded variables as fixed effects and the individuals as random effects.

Because the output of linear mixed effects modeling is complex to explain and impractical to use for making predictions with individual patients in a clinical setting, we performed univariable linear regressions^19,20 of weight gain within 6 months after starting ETI as the dependent variable with pre-ETI BMI, age, sex, FEV₁%, mutation type, pancreatic sufficiency, HgbA1c, use of previous modulators and number of hospitalizations for lung exacerbations in the year prior to starting ETI and number of hospitalizations in the year after starting ETI as independent.

Because a single multivariable model prediction might be easier to present and to understand for a patient and caretaker, respectively, than a series of predictions based on up to eight separate univariable models, we proceeded to evaluate multivariable models as more suitable to counsel patients about potential weight gain when starting ETI. We performed exhaustive linear regression modeling with change in BMI within 6 months as the dependent variable and all possible combinations with 1 to all 8 of the clinical input variables without interaction terms. We considered the adjusted R-squared, F-statistic, second order Akaike Information Criterion,¹⁹ Bayesian Information Criterion¹⁹ and log likelihood^20,21 of each of the 255 possible combinations of predictors. Because we selected from 255 possible models, we performed false discovery rate analysis of model results.²²

To exclude the possibility of results derived from extreme outlying data, we performed sensitivity analyses of gender, prior CFTR modulator use and extreme numbers of prior-year pulmonary exacerbations. Recognizing that our study may be limited in generalizability by being a single center study, we performed power calculations by boot-strapping our results to understand whether a future multi-center study would be feasible.

Results

Among 446 adult patients seen at our center during the study period, 154 started ETI. The age range overall was 19 to 73 years. Of these, 45 patients had 3 month follow up data, 42 had 6 month data, and 67 patients had 3 and 6 month data and met inclusion criteria (Figure 1). Patients were similar in the three groups (Table 1), except those with only 3 month data had lower FEV₁% and more pulmonary exacerbations in the prior year. Only one patient had a weight measured 9–12 months after initiation of ETI because the timing coincided with the summer peak of the COVID19 pandemic in 2020, when quarantine and isolation public health measures were most stringent. Nearly all patients were seen via telehealth without recorded weights during that period.

Figure 1:

CONSORT²³ flow diagram depicting inclusion and exclusion of study participants.

Table 1.

Baseline Patient Characteristics (n=154)

Characteristic*	Times BMI Data Available After starting EFI
	Only 3 Month n = 45	Only 6 Month n = 42	Both 3 and 6 Months n = 67
Mean Age in Years (SD, Range)	33 (11.1, 19–73)	35.3 (12.1, 19–62)	31.2 (11.8, 19–73)
Fraction Female	0.644	0.476	0.478
Fraction F508del Homozygous	0.711	0.548	0.537
Fraction Pancreatic Insufficient	0.933	0.881	0.91
Mean Percent Predicted FEV1 (SD)	66 (27.5)	78.9 (22.7)	71 (24.7)
Exacerbations in Year Prior to Starting ETI (Range)§	1.31 (0–6)	0.81 (0–4)	1 (0–6)
Pre-ETI BMI (SD)	21.6 (2.97)	24.1 (4.5)	23.1 (4.34)

^*Groups compared using linear regression for each variable except exacerbation numbers in year prior to ETI. Patients with data at both 3 and 6 month follow up times (third column) were the reference group for comparisons

^†p = 0.02 compared to patients with data at both 3 and 6 months.

^§Comparisons used quasipoisson regression

^¶p = 0.004 by quasi-poisson regression compared to patients with only 6 month data.

Abbreviations: ETI: elexacaftor-tezacaftor-ivacaftor; BMI: Body Mass Index (kg/m²)

Linear mixed effects modeling including all potential predictive variables as fixed effects (Supplemental Table 1) found an average increase in BMI of 0.39 kg/m² at the visit 6 months after starting ETI. FEV₁%, sex, pancreatic sufficiency status, HgbA1C levels, F508del homozygous state and diabetes status had no indication of relationships with the change in BMI. Higher starting BMI, younger age and more frequent prior pulmonary exacerbations had the strongest relationships to 6-month increases in BMI. In a parsimonious model restricted to these variables, younger age and more frequent pulmonary exacerbations predicted larger gains in BMI after starting ETI (Supplemental Table 2).

Univariable linear regressions with change in BMI after six months of therapy for the purpose of easing complexity of clinical counseling found associations with BMI at the time of ETI initiation for age and number of prior exacerbations (Table 2 and Figure 2). We found no univariable associations between BMI change at six months with pre-ETI BMI, sex, FEV₁%, pancreatic sufficiency status, HgbA1c (continuous), diabetes status or F508del homozygous state.

Table 2.

Univariable Linear Regression Models of BMI Change in 6 Months

Univariable	Estimate (95% CI)*	p
Pre-ETI BMI (kg/m2)	−0.11 (−0.18 to −0.036)	0.004
Age (Years)	−0.042 (−0.066 to −0.017)	0.001
Number of Prior Exacerbations	0.31 (0.13 to 0.49)	0.001
FEV1%	−0.007 (−0.019 to 0.004)	0.214
Sex (0=Male, 1=Female)	0.14 (−0.44 to 0.72)	0.629
Pancreatic Sufficiency (0=insufficient, 1=sufficient)	−0.46 (−1.6 to 0.68)	0.431
Hgb A1C (mg/dL)	−0.03 (−0.38 to 0.32)	0.872
F508del Homozygous Status (0=No, 1=Yes)	0.04 (−0.55 to 0.64)	0.883
Diabetes Status (0=No, 1=Yes)	−0.38 (−1.9 to 1.1)	0.614

^*95% Confidence Interval; Intercept terms are omitted.

Figure 2. Associations of BMI Change at 6 months with Age and Number of Pulmonary Exacerbations in the Year Prior to Starting ETI.

A. Patients starting with high BMI have smaller changes in BMI than patients with normal or low BMI. B. Older patients have a smaller change in BMI than younger patients. C. Patients with higher numbers of exacerbations in the year prior to starting ETI, up to 5 or more, have a greater increase in BMI after 6 months of therapy. The effects remain significant if patients with 5 or more exacerbations are excluded or if those patients are considered in a single 5 or more category.

To derive a single multivariable model for use with counseling, we performed exhaustive linear regression modeling of all possible combinations of the eight clinical input variables without interaction terms (255 combinations). In the model with best-fit to the data, the BMI increase six months after starting ETI was 2.46 kg/m² on average which was reduced by 0.045 for every year of age and increased by 0.32 for each pulmonary exacerbation in the year prior to starting ETI (Table 3).

Table 3.

Multivariable Model of BMI Change in 6 Months^*

Variables	Estimate (95% CI)†	p
(Intercept)	2.46 (1.63 to 3.29)	<0.001
Age (Years)	−0.045 (−0.069 to −0.022)	<0.001
Number of Prior Exacerbations	0.32 (0.15 to 0.50)	0.001

^*Example Applications:

⁽¹⁾For a 20 year old with 5 prior exacerbations, the change in BMI would be predicted to be 2.46−0.045 × 20 + 0.32 × 5 = 2.46−0.9 + 1.6 = 3.16 BMI units.

⁽²⁾A 40 year old with 5 prior exacerbations would have 2.46−0.045 × 40 + 0.32 × 5 = 2.26 BMI units.

⁽³⁾A 20 year old with no prior exacerbations would have 2.46 – 0.045 × 20 + 0.32 × 0 = 1.56 BMI units.

⁽⁴⁾A 60 year old with no prior exacerbations would have 2.46 – 0.045 × 60 + 0.32 × 0 = −0.24 BMI units.

^†95% Confidence Interval

Four other models found by exhaustive testing had comparable fits to the data but all included one or two additional terms from among pre-ETI BMI, F508del homozygous state, CF-related diabetes status and FEV₁% (Supplemental Table 3). While each of these models had similar fit to the model presented in Table 3, none of the terms, pre-ETI BMI, F508del homozygous state, CF-related diabetes status or FEV₁% were significant within their respective models, making these models more difficult to use for patient counseling (Supplemental Figure 1) than the model selected after exhaustive testing (Table 3).

Because we fitted 255 models, we explored the model false discovery rates based on the F-statistics for fit to the underlying data. Our best model and the other 4 with similar fits (Table 3 and Supplemental Tables 3–4) had false discovery rates < 1 × 10^-5.

In sensitivity analyses, we found no effect of substitution of self-determined gender for biological sex at birth, and no effect of history of use of previous CFTR modulators in any model. Because a few patients had extreme numbers of exacerbations in the year prior to starting therapy, we repeated our analysis after exclusion of patients with more than four prior exacerbations and repeated analysis again including those patients but reassigning to a single group with five or more in the year prior to enrollment. We found similar results with both methods providing reassurance that our results were not driven by data from individuals with extreme disease.

Using the variables from the multivariable model of BMI change after six months of ETI therapy with the most useful variables (Table 3) and generally best fit to data (Supplemental Table 4), we explored the feasibility of a future prospective study. Bootstrapped power simulations indicate that roughly 300 patients similar to those in our study with data at the 6 month follow up time (109 patients, Table 1) would provide 90% power at an alpha level of 0.01, and 1000 similar patients would provide 98% power at an alpha level of 0.001 to produce results similar to the current study.

None of the 154 study patients were lost to 6 month follow up after initiating ETI. However, only 112 patients at three months and 109 patients at six months (with 67 patients in both groups) had weight data to allow analysis of weight change. None of the patients with weight data at the 3 or 6 month time points had missingness of other data.

Discussion

We performed a retrospective study of BMI change after 3, 6 and 9–12 months of therapy with the highly effective CFTR modulator combination treatment of ETI among adults with CF. From the model that best fit the underlying data (Table 3), we found that younger age and larger numbers of exacerbations prior to starting therapy were strongly associated with increased BMI in the period following ETI initiation. These results suggest that we can advise young patients with many exacerbations in the year prior to starting ETI that they may gain more than the average weight gain on ETI, and this gain will likely occur over a multi-month period.

FEV₁% is the most important single predictor of survival in CF.^3,4 However, it is independent of weight-for-age z-score for the prediction of survival. In our study, it has no significant univariable association with increasing BMI after treatment with CFTR modulators. In an alternative model with good fit to the data discovered through exhaustive testing of all possible models without interactions, FEV₁% was included (Supplemental Table 3, Model 4), nevertheless, it was not statistically significant even if inclusion improved model fit, and the lack of significant relationship (Supplemental Figure 1), makes the encoded information and this specific model difficult to use for patient counseling.

History of greater numbers of prior pulmonary exacerbations strongly predicts increased BMI after six months of therapy in the models found in linear mixed effects modeling and exhaustive multivariable modeling of all input variables. Carnovale et al reported a mean BMI increase of 2.08 kg/m² 48 weeks after ETI in patients with advanced lung disease.¹² Our results are similar considering that we studied a younger and healthier patient population for about 6 months resulting in a lower observed mean BMI increase with wide range.

Gramegna analyzed weight gain on ETI in a small cohort of 92 patients in Italy and found pre-ETI BMI values and CFTR residual function mutations to explain much of the heterogeneity in weight gain.²⁴ In contrast, we found that pre-ETI BMI improved the fit of predictive models but was not significant. Previous studies with ivacaftor found mutation type to be predictive of weight gain.²⁵ We did not directly test the impact of residual function mutations, however, F508del homozygosity had no significance in models with good fit to the data.

Hypotheses to explain relationships between age, frequent exacerbation history and BMI gains after starting ETI might invoke several etiologies that remain to be fully demonstrated. Mice studies suggest that ETI might lower gut inflammation, however evidence in humans is very limited.^5,26 A reduction in inflammation may alter appetite, food intake and energy costs associated with more frequent or prolonged pulmonary illness episodes. Despite many efforts,^27–29 the interpretation of measurements of inflammation for clinical relevance in CF are not established, although some advances have been made ³⁰. The relationship of inflammation with CFTR modulator use is incompletely explored. Ivacaftor alone has been found to decrease resting energy expenditure⁵ thus it is plausible ETI would also decrease resting energy expenditure, leading to weight gain. Additionally, Gelfond demonstrated ivacaftor improved proximal small intestinal pH profile³¹, which was associated with weight gain. ETI may improve olfaction sensation ³² which could positively impact appetite.

By determining which patients may be most at risk for dramatic weight gain on ETI, this study can help guide clinicians on which patients to spend extra time with counseling diet changes.

Limitations

Our study was limited by inclusion of patients from only one center which serves only adults with CF. Our study was not able to include body composition data, which has been suggested as a more useful tool than BMI ³³ however most clinics do not have access to body composition data and demonstration that this data improves clinical care remains to be ascertained. Our practices may differ from practices elsewhere that affect weight, potentially limiting generalizability of our results. However, the results provide a starting point for studies with recruitment from a wider range of settings, and our power calculations are encouraging that such studies, whether retrospective or prospective are feasible with regards to recruitment numbers.

Our results centered on prediction models that incorporate clinical characteristic variables provide no direct information on specific mechanisms related to weight gain that may be associated with use of CFTR modulators, however, the data provide some guidance that mechanisms underlying pulmonary exacerbations may be important to explore further and provide information that may be useful in counseling of patients prior to starting CFTR modulator therapy.

Hypotheses suggesting that weight gain after initiation of CFTR modulator therapy are linked to improving FEV₁%, perhaps via generalized decreased work of breathing, were not tested by our study, because work of breathing improvements are future events relative to the initiation of ETI and thus cannot be used to predict weight change outcomes and provide counseling at the point in time when decisions related to ETI therapy are made. The impact of ETI on work of breathing should be investigated independently.

Our efforts are directed towards prediction and patient counseling, thus we presented our best fitting parsimonious multivariable model (Table 3), but that model may not be the most useful, depending on the clinical situation. We performed exhaustive testing of all possible models for the input variables that we measured. Among the alternative models that also fit the data well (Supplemental Tables 3 and 4), there may be models that are better suited to specific individuals in different clinical situations.

In summary, age and the number of exacerbations prior to starting ETI most strongly predicted weight gain over the following six months. In light of previous research and the history of clinical CF, weight continues to be an important consideration to patients and clinicians alike.^3,4,14 Clinicians should focus on improving or maintaining good dietary and physical activity habits especially with young patients starting ETI with many prior exacerbations.

Abbreviation List

AIC

Akaike Information Criterion

AICc

second order Akaike Information Criterion

BIC

Bayesian Information Criterion

BMI

Body Mass Index (kg/m²)

CF

cystic fibrosis

CFTR

cystic fibrosis transmembrane conductance regulator

ETI

elexacaftor-tezacaftor-ivacaftor

FEV₁%

percent predicted forced expiratory volume in one second

HgbA1c

hemoglobin A1c