Overall, 332 patients were included in this analysis (Fig. 1), with 1609 person-years of observation time and 88 deaths (26.5%). Patient characteristics of the full study population are described in Table 1. Median age at diagnosis of Pompe disease was 2.7 months (range 0.0–11.2). Median age at first treatment was 3.6 months (range 0.1–11.6 months). Overall, 51 patients (15.4%) were diagnosed by newborn screening (NBS). Among the 332 patients, 213 (64.2%) patients were CRIM-positive, 70 (21.1%) were CRIM-negative, and CRIM status was unknown for 49 (14.8%) patients. Immune tolerance induction (ITI) use was reported for 29 of 70 CRIM-negative patients (41.4%) and for 32 of 213 CRIM-positive patients (15.0%). At baseline, 11.3% patients used non-invasive ventilation only, 2.8% used both non-invasive and invasive ventilation, and 1.4% used invasive ventilation only. By the end of follow-up, 23.4% of patients had reported some use of non-invasive ventilation either at baseline or during follow-up, 9.6% had reported use of both non-invasive and invasive ventilation over time, and 12.8% had reported use of invasive ventilation only. Median age at death (n = 88) was 23.5 months (range 5.1–187.8). Additional characteristics of the study population are presented in Additional file 2: Table S1.

Derivation of study population from the Pompe Registry

The study population is described according to vital status in Additional file 2: Table S2. Deceased patients tended to be older at diagnosis and at first treatment and had longer times from first symptom to diagnosis and from diagnosis to first treatment, compared to patients alive at the end of follow-up. Deceased patients were more likely to be CRIM-negative, to use non-invasive or invasive ventilation, to have been diagnosed/treated in earlier time periods, and to have started treatment at the label dose, and were less likely to have ever received ITI or higher doses of ALGLU.

The initial treatment dose of ALGLU was at/about label dose for 81.3% of patients, while 18% of patients starting treatment at a higher dose: 11.4% started at/around 20 mg/kg/week, 3.6% at 40 mg/kg EOW, and 3% at 40 mg/kg/week (Table 1). Doses varied in patients over time, with about half of patients (51.8%) receiving a higher than label dose at some point during follow-up. The rest (48.2%) remained at/about label dose throughout follow-up.

Additional file 2: Table S3 describes baseline and most recent dose by categories of average relative dose. Almost all patients in the average dose category of ‘Label dose’ at the end of follow-up (n = 153) received at/around label dose at both baseline and most recent follow-up; 99.1% of the person-time in this group was in the label dose category. Patients with average dose in the category ‘Between label and double dose’ at the end of follow-up (n = 70) tended to start at label dose (88.6%), but were likely to have moved to higher doses, with only 28.5% receiving label dose or lower at their most recent record. This group spent 62.3% of total person-time in the label dose category. Patients in the average dose category of ‘Double dose’ at end of follow-up (n = 54) or ‘Above double to quadruple dose’ (n = 45) were more likely to start and end on higher than label doses. Around half of the patients in both groups were on a higher dose at baseline, and nearly all were on a higher dose at the most recent follow-up. At most recent follow-up, 80% of those in the ‘Above double to quadruple dose’ group were receiving 40 mg/kg/week, and none were at the label dose; overall, this group spent half of its person-time (49%) in the 40 mg/kg/week dose category.

Additional file 2: Table S4 presents detailed information on the study population by year of first treatment.

Higher dose treatments have become more common over time (Fig. 2, Additional file 2: Table S4). Between 85 and 93% of patients began treatment at the label dose for the periods 2003–2009, and 2010–2013, compared with 67.7% during 2014–2016 and 74.2% in 2017 or later.

Changing trends in clinical practice: alglucosidase alfa dose over time by year of first treatment

Additional file 2: Table S4 further describes dosing changes over time. Among higher dose regimens, 20 mg/kg/week was the most used; however, 40 mg/kg EOW and 40 mg/kg/week have become more common recently. In the 2014–2016 period, 24.6% of patients received 20 mg/kg/week initially, with 1 (1.5%) patient starting at 40 mg/kg EOW and 3 (4.6%) at 40 mg/kg/week. From 2017, the percentage starting at 20 mg/kg/week fell to 12.4%, while those receiving 40 mg/kg EOW (4.5%), and 40 mg/kg/week (7.9%) increased. Median age at first higher dose was lower for patients in the most recent periods, ranging from 24 to 43 months for patients in the first three time-periods, falling to 6.6 months for 2014–2016 and 5.4 months for 2017 and later.

Table 2 summarizes the number of events and total observation time in the study population overall and by dose at the time of event. Among 332 patients, there were 88 deaths over 1609.4 person-years (PY) of follow-up. Approximately three-quarters of patients who died were in the label dose category at the time of death, with three deaths in the 40 mg/kg EOW group and 13 deaths in the 20 mg/kg/week group. There were no deaths among patients on 40 mg/kg/week, although this dose category accounted for 9.9% of total person-time observed.

Patients receiving higher than label doses at the time of death tended to be older at death (median ages at death: 21.7 months [25th, 75th percentile: 11.7, 41.5] for the label dose group, 108.3 months [25th, 75th percentile: 20.6, 187.8] for the 40 mg/kg EOW group, and 43.5 months [25th, 75th percentile: 26.1, 63.3] for the 20 mg/kg/week group); although this is based on a small number of patients in the higher dose groups.

The overall crude (unadjusted) incidence rate of death was 5.47 per 100 PY. For patients not currently on treatment, the crude incidence rate was 121.87 per 100 PY; there were 7 deaths in this group and only 5.7 PY of observation time. This suggests that these patients discontinued treatment shortly before death. The crude incidence rate of death was lower in the three higher dose groups than in the label dose group (1.34 per 100 PY for 40 mg/kg EOW, 4.95 per 100 PY for 20 mg/kg/week, 0 for 40 mg/kg/week, vs. 6.85 per 100 PY for label dose). Patterns for the distribution of person-time and events across categories of average dose received over time were similar to those observed for current dose category at the time of event.

There were 270 patients (1236.8 PY of observation time) in the analysis of invasive ventilation-free survival, with 97 events and a crude incidence rate of invasive ventilation or death of 7.84 per 100 PY. The distribution of person-time and events across dose categories was similar to that of the main survival analysis, with approximately three-quarters of events in patients on the label dose. There were seven events (7.2%) in the 40 mg/kg EOW group and 13 (13.4%) in the 20 mg/kg/week group. There was one event in the 40 mg/kg/week category; representing 1.0% of events, while this dose category contributed 10.9% of total person-time observed. There were no composite events among patients with an average dose over time in the above double to quadruple dose category.

Table 3 presents the survival analysis for risk of death and risk of the composite event of death or invasive ventilation according to average dose received over time. Higher average dose was significantly associated with improved survival and invasive ventilation-free survival. The adjusted hazard ratio (HR) per 1-unit increase in average relative dose was 0.40 (95% confidence intervals [CI] 0.22–0.73; p = 0.0030). Patients with average dose above double to quadruple dose had the greatest reduction in risk of death (HR = 0.10; 95% CI 0.01–0.82, p = 0.03) compared to those with average dose around the label dose. Higher doses were also associated with a lower relative risk of the composite outcome, death or invasive ventilation, with an HR per 1-unit increase in average relative dose of 0.48 (95% CI 0.28–0.84; p = 0.010). The full model results including HR estimates for covariables are presented in Additional file 2: Table S5.

The association between average relative dose and risk of death was similar across sensitivity analyses (Additional file 2: Table S6). The HR for a 1-unit increase in average relative dose ranged from 0.24 to 0.47 (versus 0.40 for the main model) and remained statistically significant in models adjusted for: baseline dose category, year of first treatment, invasive ventilation use (updated over time), or time from diagnosis to first treatment, and in models restricted to patients first treated in 2006 or later or to patients treated before 6 months of age.

The model results for current dose category and 3-month and 6-month lagged dose category are presented in Additional file 2: Table S7. Few deaths in the higher dose categories resulted in wide CIs or inestimable HRs. Compared to patients on the label dose, patients receiving 40 mg/kg EOW were at significantly lower risk of death (HR 0.08, 95% CI 0.01–0.70, p = 0.02; n = 3 deaths). Risk was not significantly lower for those receiving 20 mg/kg/week (HR 0.85, 95% CI 0.40–1.80, p = 0.68; n = 13 deaths). There were no deaths in the 40 mg/kg/week group.

HRs for the 40 mg/kg EOW group were similar, and remained statistically significant, when exposure was lagged by 3 or 6 months, suggesting the current dose results were not due to reverse causation. HRs for the 20 mg/kg/week group were lower (though still not significant) after applying a 3- or 6-month lag. Thus, the current dose results for 20 mg/kg/week may be somewhat attenuated due to reverse causation, that is, patients increasing to a “rescue dose” of 20 mg/kg/week shortly before death.

Risk of death was significantly higher for patients in the none/very low dose category (HR 7.67, 95% CI 1.90–30.9, p = 0.004). The HR for this category was substantially lower with a 3- or 6-month lag between dose exposure and death, suggesting that the increased risk for patients currently on no/very low dose is partly due to reverse causation, that is, patients discontinuing treatment immediately prior to death.

Table 4 presents sub-group analyses by age at first treatment and by CRIM status. Higher average dose over time was associated with improved survival irrespective of age at first treatment. In patients first treated at < 3 months, the HR was 0.29 per 1-unit increase in average relative dose (95% CI 0.09–0.90, p = 0.03) and in patients first treated at ≥ 3 months, the HR was 0.43 per 1-unit increase in average relative dose (95% CI 0.21–0.87, p = 0.02; p-interaction = 0.81).

Higher average dose was associated with significantly improved survival in CRIM-positive patients (HR 0.44, 95% CI 0.20–0.98, p = 0.04). The HR for CRIM-negative patients suggested a lower risk of death with higher doses, though not statistically significant (HR 0.24, 95% CI 0.04–1.37, p = 0.11). The CRIM-negative group was relatively small, with 70 patients and 29 deaths. The p-value for interaction between average dose and CRIM status was not significant (p = 0.71), suggesting that the association of higher average dose with improved survival was not limited to one CRIM subgroup versus the other.