The Pompe Registry (NCT00231400) was initiated in 2004 and is a long-term, multi-national, observational program in patients with Pompe disease that is sponsored and managed by Sanofi (Cambridge, MA, USA). Physician investigators may enroll patients into the Registry irrespective of patients’ age, clinical presentations, treatment, or clinical trial participation status provided there is a confirmed Pompe diagnosis. Registry sites are encouraged to enter all patient data; however, these data do vary among patients both in the types of data collected and in their frequency of collection according to patient needs, local practices, resources, and available skills.

Each participating center obtained written informed consent from patients or, for minors, their guardians, to submit their data to the Pompe Registry and to use and disclose this information anonymously in aggregate analyses. Local sites’ Institutional Review Boards or Independent Ethics Committees approved the Pompe Registry protocol, informed consent form, and local authorization documents for data submissions.

The study included data downloaded from the Pompe Registry as of September 2021 for patients with LOPD aged ≥ 5 years at treatment initiation with alglucosidase alfa. The ≥ 5-year-old inclusion criterion was chosen to be consistent with a previous analysis and because children younger than 5 years may have difficulty performing an accurate FVC maneuver [5]. LOPD was defined as symptom onset at ≤ 12 months of age without cardiac enlargement/cardiomyopathy or symptom onset at > 12 months of age [8].

For patients’ data to be included in the analysis, patients had to have a baseline upright FVC value measured between 6 months before and 1 month after treatment initiation, along with ≥ 2 additional FVC records in the period from 1 month to 13 years after first treatment (giving a total of ≥ 3 FVC values per patient). Only upright FVC was included in the analysis, as there were fewer supine FVC data available, and supine measurements can be prone to bias if positional dyspnea interferes with patients’ ability to perform the test optimally. At least 6 months were required between the first and last FVC measurement. Measurements beyond 13 years from first treatment were not included as only < 5% of patients had a record beyond this time limit.

Exclusion criteria, which paralleled those of a prior Pompe Registry analysis [5], were invasive ventilation at alglucosidase alfa initiation or an implausible annual change in a patient’s FVC, as defined by a change in slope, estimated by linear regression, across all their measurements from baseline through follow-up on alglucosidase alfa of > +10% or < −10%/year. Patients whose first treatment was not alglucosidase alfa were excluded. Patients who switched from alglucosidase alfa to a different treatment were included in the analysis until the date of the switch.

Demographic and clinical information including FVC measurements over time was collected in case report forms. Date of symptom onset was derived from the earliest of (1) diagnosis date reported for ophthalmic, respiratory, gastrointestinal/hepatic, renal, or musculoskeletal symptoms related to Pompe disease; (2) date of respiratory support use, date of ambulatory symptoms (ambulatory with difficulty, ambulation lost, wheelchair use), or date of cardiac enlargement/cardiomyopathy from echocardiogram or chest X-ray (onset at > 12 months of age for this study cohort’s definition of individuals with LOPD); or (3) directly reported age of onset of Pompe disease symptoms. Diagnosis date was derived from the earliest of confirmatory enzyme assay date, genotype assay date, or physician-reported diagnosis date (directly reported by the Registry site).

Descriptive analyses were generated for demographic and clinical variables overall and according to categories of age at baseline (≥ 5 to < 18 years or ≥ 18 years) and FVC % predicted at baseline (< 55%, 55 to 80%, or > 80%). The baseline FVC categories chosen were based on prior Pompe disease studies [2, 5, 9] and published guidelines for spirometry testing [10, 11]. Medians, 25th and 75th percentiles, and ranges are presented for continuous variables.

Upright FVC was analyzed using % predicted values. Change in FVC over time was assessed using linear mixed effects models to account for repeated measures within patients over time. To assess how change in FVC over time might vary based on time from treatment initiation, we used piecewise linear regression to estimate the annual change in FVC over three periods: 0 to 6 months, > 6 months to 5 years, and > 5 to 13 years from treatment initiation. The periods were chosen a priori based on results from clinical trials that suggest an initial increase in FVC after treatment initiation in alglucosidase alfa-naïve patients [2] and previous observational studies (including from the Pompe Registry) that found a period of stability in the first 5 years of treatment [1, 5]. We fitted separate slopes for the > 6 months to 5 years and > 5 to 13 years periods to assess whether the change in FVC in the later period was different from that seen in the first 5 years after treatment initiation.

The final models included fixed effects for: time from treatment initiation (providing an estimate of the slope for the initial 0 to 6 months of treatment), two linear spline terms to allow the slopes to vary for the > 6 months to 5 years and > 5 to 13 years periods, baseline age, sex, and use of non-invasive respiratory support at baseline, and random effects for: the intercept (FVC at baseline), time, and the two linear spline terms for time. The slopes from the three time-related variables give the estimated annual change in FVC in absolute percentage points/year (%/year). An unstructured covariance matrix was used to accommodate the varying number of measurements and timing of measurements across patients, and models were estimated using restricted maximum likelihood (REML). Fixed effects for age, sex, and baseline non-invasive ventilation were included, as all three were significantly associated with baseline FVC and their inclusion improved model fit as assessed by the Akaike Information Criterion.

In addition to estimating FVC slopes in the full study population, we assessed whether slopes in the three periods varied according to age at treatment initiation (≥ 5 to < 18 years or ≥ 18 years), baseline FVC (< 55%, 55 to 80%, or > 80%), or time from diagnosis to first treatment (below or at/above the median time for adult patients, < 1.5 or ≥ 1.5 years, respectively) using interaction terms between the three time-related variables and indicator variables for the subgroups. The median time from diagnosis to first treatment was selected as a cut-off because, in a prior shorter-term Pompe Registry analysis [5], a shorter time from diagnosis to first treatment was associated with a higher FVC.

For the models by baseline FVC category, the first period was extended to 1 year instead of 6 months after treatment initiation due to limited data within each baseline FVC subgroup. Thus, results for this analysis present estimated changes in FVC for the periods of 0 to 1 year, > 1 to 5 years, and > 5 to 13 years from treatment initiation. To test for overall significant differences in FVC slopes between subgroups, we used likelihood ratio tests comparing models with and without interaction terms between the patient subgroup and the time variables. Likelihood ratio tests were based on the maximum likelihood estimation of the linear mixed models, as REML methods cannot be used to test fixed effects.

We assessed the robustness of the results with sensitivity analyses excluding patients with very low FVC at baseline (< 30%), excluding patients diagnosed before 2006 (when alglucosidase alfa became widely available—prior to 2006, alglucosidase alfa was available only in clinical trials or for compassionate use), and excluding follow-up measurements from > 10 to 13 years after treatment initiation (as less than one-quarter of individuals, 106/485 (21.9%), had an assessment beyond 10 years).

To assess the variation in patients’ FVC responses in each period, we calculated individual predicted slopes for each patient using both the fixed and random effects estimates from the main model. The results are presented as box plots.

All statistical analyses were performed using SAS (version 9.4; SAS Institute Inc., Cary, NC, USA), and P values < 0.05 were considered statistically significant.