Human Genetics

Search results

After screening 4286 titles and abstracts, 237 full-texts were retrieved and screened. Thirty-eight studies (encompassing 106 unique records) were included in the review: three RCTs (LOTS[[3], PROPEL[[24]and COMET[[25]), three RCT extension studies[[26,27,28], seven Pompe disease registry studies[[29,30,31,32,33,34,35]and 25 single-group prospective studies[[2, 6, 36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58]. The process of identifying and selecting records is presented in the PRISMA flowchart (Fig. 2).

Study details and baseline characteristics

The LOTS RCT evaluated the safety and efficacy of alglucosidase alfa compared to placebo (plus BSC). The COMET and PROPEL RCTs assessed the safety and efficacy of avalglucosidase alfa and cipaglucosidase alfa plus miglustat, respectively, compared to alglucosidase alfa. Mean ages across the three trials ranged from 44 to 48 years (Table 2, Additional file 2). The proportion of participants using a walking aid at baseline was 43% in LOTS and 23% in PROPEL (data were not reported for COMET). LOTS and COMET recruited only ERT-naïve patients, whereas PROPEL recruited mostly ERT-experienced patients.

Of the seven registry studies, there were three studies from Sanofi’s international Pompe Registry cohort[[29,30,31], with sample sizes ranging from 396 to 1390 patients, three studies from the French Pompe disease registry[[32,33,34](range 29 to 177 patients) and one Spanish Pompe registry study (N = 113)[[35], (Table 1, Additional file 3). The largest study was reported only as a conference abstract[[31]. The mean ages when starting ERT ranged between 45 and 56 years, except for the Spanish registry study which was of a younger cohort (mean 29 years). Follow-up durations ranged from one to 10 years.

All other prospective studies (n = 25) included in the review were single-group studies: none compared an ERT with a specific type of best supportive care, although some studies did compare ERT patients with those not taking an ERT. Sample sizes ranged from 11 to 209 patients (Table 3, Additional file 3). Eleven studies included 30 or fewer patients and 6 studies included 100 or more patients; the number of patients included in individual analyses was often notably smaller than the number recruited e.g. some patients were not physically able to undertake an assessment of 6-minute walk distance. Two studies were reported only as conference abstracts[[41, 47]. All studies were of alglucosidase alfa, except for the NEO1/NEO-EXT study of avalglucosidase alfa[[57], and the ATB200-02 study of cipaglucosidase alfa plus miglustat[[58]. Most studies were conducted in Italy, Germany or the Netherlands. Five of the 25 studies[[39, 40, 45, 55, 59]reported results for a child cohort or subgroup (mean age ranged between 6 and 12 years), with the remaining studies being of adults (mean age range between 43 and 53 years). In one small study[[40]all patients were ERT-experienced at study entry, with patients typically having been on an ERT for around nine years. There was substantial variation across studies in the proportion of patients requiring wheelchair or respiratory support at baseline. Follow up durations ranged from 6 months to 15 years, with most patients being followed up for between 2 and 5 years.

IPD from eligible trials

IPD was sought from the three identified RCTs: PROPEL, LOTS and COMET. However, no IPD was provided by any of the sponsors. Amicus Therapeutics was contacted regarding sharing data from the PROPEL trial. In email correspondence, Amicus Therapeutics representatives indicated they were working on a process and platform to make these data available but failed to respond to further email contacts. Access to data from the LOTS and COMET trials was sought from the sponsor Sanofi via the data-sharing platform Vivli. Sanofi declined the request stating that they considered the proposed research not to be in the interest of patients or the patient community.

In the absence of IPD from the sponsors of the RCTs, we digitised the mean difference plots of the outcomes (FVC % predicted and 6MWD) using the PlotDigitzer website (https://plotdigitizer.com/) to obtain estimates of the mean differences at 12/13 weeks, 24/25 weeks, 36/38 weeks, and 49/52 weeks.

Risk of bias assessment

Risk of bias assessment results are presented in Additional file 2, Table 1. The COMET trial results were judged to have a low overall risk of bias but the LOTS and PROPEL trial results were judged to be at high risk of bias. Both LOTS and PROPEL had high risk judgements for the ‘bias in the selection of the reported result’ domain, since both trials failed to report results for all pre-specified analyses, as noted in the respective EMA reports[[60]. ^, [[61]

Another study quality issue identified was the reporting of results using only means in the published reports. Results data from regulatory documents showed skewing of the 6MWD data by outliers, with the means and medians differing substantially. The reporting of only means in the presence of outliers in a sample is not an accurate representation of the efficacy data. For example, the FDA reported that in COMET the mean change in 6MWD from baseline to week 49 for the alglucosidase arm was − 1.7 m, whereas the median was 16.0 m;[[20]the EMA reported that in LOTS the mean change in 6MWD from baseline to week 78 for the alglucosidase arm was 26.1 m, whereas the median was 15.0 m[[62].

NMAs of RCT evidence

Results from the primary analysis are reported on Table 1 and results of the sensitivity analyses and additional analyses are presented in Additional file 2.

In the primary analysis of FVC % predicted at 49/52 weeks, the results indicate that all three ERTs exhibit numerical superiority over placebo (which includes BSC). However, the estimated mean differences did not reach statistical significance for any of the ERTs (Table 2). In contrast, for 6MWD there were statistically significant improvements compared to placebo for both alglucosidase alfa (by around 25 m) and avalglucosidase alfa (by around 54 m). Cipaglucosidase alfa with miglustat, while numerically superior to placebo, did not show statistically significant differences. Credible intervals for this comparison were wide, reflecting the small number of ERT-naïve patients in the PROPEL trial.

Analysis of additional time points revealed a consistent pattern (Additional File 2 – Tables 8 and 9), with no statistically significant differences between any ERT and placebo for FVC % predicted at any time point. For each ERT versus placebo for 6MWD, statistically significant differences were observed favouring alglucosidase alfa beginning at week 12/13 and persisting through later time points. Similarly, statistically significant differences for avalglucosidase alfa were estimated from week 24/26, with sustained effects observed at subsequent time points. Cipaglucosidase alfa with miglustat showed numerical superiority across all remaining time points, but differences did not reach statistical significance.

Intra-ERT comparisons showed a numerical difference between avalglucosidase alfa and alglucosidase alfa at 49/52 weeks for both FVC% predicted and a meaningful numerical difference in 6MWD. Sensitivity analyses conducted at other time points, including an analysis at 49/52 weeks using imputed values for the COMET trial, similarly did not demonstrate statistically significant differences for either outcome. Results of the sensitivity analysis (Additional File 2, Table 7) also show a smaller numerical difference between avalglucosidase alfa and alglucosidase alfa; mean difference 12.43 m, (95%CrI: -13.17 to 38.07) vs. 28.87 m (95% CrI: 1.74 vs. 55.66). Cipaglucosidase alfa with miglustat exhibited numerical inferiority compared to avalglucosidase alfa for both outcomes across all time points and sensitivity analyses (Additional File 2, Tables 6, 7, 8 and 9); however, the differences were not statistically significant, and credible intervals were wide. Compared to alglucosidase alfa, cipaglucosidase alfa with miglustat demonstrated numerical inferiority for both outcomes in the primary and sensitivity analysis, but the differences were small and credible intervals were wide. This pattern of numerical inferiority was consistent across all time points (except week 24/26) for 6MWD. Differences for FVC% predicted were, however, inconsistent, with weeks 12/13 and 24/26 favouring alglucosidase alfa, and the week 37/38 analysis favouring cipaglucosidase alfa with miglustat. At all-time points and for both outcomes, differences between alglucosidase alfa and cipaglucosidase alfa with miglustat were not statistically significant.

The combined treatment effectiveness of all ERTs demonstrated significant numerical superiority over placebo for 6MWD at all time points (Additional File 2, Table 16). For FVC% predicted, ERTs also showed numerical superiority; however, the difference was not statistically significant using the random-effects model.

Evidence in the ERT-experienced population was limited to a subgroup from the PROPEL trial which included participants who had received ERT for at least 2 years. Results from the PROPEL trial favoured cipaglucosidase alfa with miglustat with statistically significant differences in both 6MWD and FVC% predicted reported, mean difference: 16.8 m (95% CrI: 0.2 to 33.3) and 3.5 (95% CrI: 1.0 to 6.0) respectively. However, the reported results of this trial were judged to be at high risk of bias.

RCT extension studies

Each of the three RCTs had open-label extension studies. PROPEL was extended by 52 weeks[[26], LOTS by 26 weeks (and by 52 weeks for a subset of U.S. patients)[[27], and COMET by 48 weeks (reported in a published paper[[28]) and 96 weeks (reported in two conference abstracts[[63, 64]). Given that patients may potentially receive ERTs for many years, these extension studies are relatively short in terms of providing evidence of the long-term effects of ERTs. Results are reported in Table 14, and 15 of Additional file 2.

In PROPEL, the ERT-experienced group which continued taking cipaglucosidase alfa with miglustat had small increases in % predicted 6MWD and % predicted FVC from week 52 to week 78, followed by small declines by week 104; however, no details were reported on how missing data were handled in the analyses (11 patients discontinued treatment). The LOTS cohort also showed a small decline in 6MWD and FVC % predicted from week 78 to week 104 for the group which continued taking alglucosidase alfa (data were missing for only one patient). In the COMET extension study FVC % predicted remained relatively stable, but 6MWD and hand-held dynamometry had decreased notably by week 97. In mitigation, the authors stated that some patients missed infusions due to the COVID-19 pandemic. A further difficulty when interpreting the COMET extension results is that data were missing at week 97 for 9 (6MWD), and 8 (FVC % predicted), of the 51 patients who continued taking avalglucosidase alfa and the analyses assumed that data were missing at random. This assumption does not appear reasonable for those patients who discontinued due to adverse events.

Registry studies

Semplicini et al.[[32]followed 158 patients for a median of around five years, finding a 1.4% annual increase in % predicted 6MWD up to 2.2 years, followed by a 2.3% decline (Table 2, Additional file 3). For muscle function outcomes, the Motor Function Measurement D2 sub-score showed a progressive 1.0% decline per year, and the D3 sub-score had a slower progressive decline (0.2% per year).

Tard et al. reported on the effect of switching from alglucosidase alfa to avalglucosidase alfa in 29 patients, reporting stabilisation of 6MWD results after one year of avalglucosidase alfa (when compared to pre-switch one year data, which showed declines)[[34]. The reporting of 6MWD results data in Lefeuvre et al.’s study was somewhat unclear`, although the treated population experienced a decline in 6MWD. Martinez–Marin et al.’s Spanish registry study reported yearly 6MWD declines of between 5 and 9 m in subgroups treated for < 5 years, 5–10 years and > 10 years[[35].

All studies reported FVC % predicted, mostly as a long-term outcome. Annual declines in %FVC after up to five years of ERT ranged between 0.17% and 0.9%. Declines at time points between 5 and 13 years were similar across two studies ranging between 1.0 and 1.2%[[29, 35]. Tard et al. found no statistically significant difference in %FVC in patients who switched ERT[[34]. Three studies reported mortality, with mean or median ages at death ranging between 60 and 66 years[[29, 32, 33].

Other prospective studies

6MWD results

The reporting of the 18 studies with 6MWD results varied (Table 5, Additional file 3). Only three studies reported results as medians[[39, 48, 53]and only seven reported results as changes from baseline as absolute values;[[2, 36, 40, 47, 48, 51, 58]five studies reported results as changes in % predicted 6MWD[[41, 46, 55, 57, 58]. The remaining studies either reported results only graphically[[49], or reported baseline and end of follow up data but with the difference represented only as a p-value[[37,38,39, 52, 53].

Of the studies reporting changes from baseline, two very small studies reported improvements at 6 months of 37 m[[2]and 47 m;[[47]the result was statistically significant for the former, though level of statistical significance was not reported for the latter. One study reported a statistically significant improvement at up to one year of around 44 m, although this result was also based on a small cohort (n = 20)[[36]. For later time points, a non-statistically significant improvement of 16 m at > 3 years[[36], and a statistically significant increase of 41 m at 5 years[[48]were reported. Ravaglia et al. followed a small cohort up to 15 years; although most of the data were only reported graphically, the study’s results showed significant improvement of around 55 m at one year, a return to baseline at around three years, and continued decline up to 15 years[[50].

The studies reporting % predicted 6MWD results were generally limited by very small samples sizes or by being available only as an abstract. However, Harlaar et al., including 30 patients from the LOTS trial cohort reported initial improvements for around two years, followed by gradual decline up to 10 years[[46]. Thirteen patients had some wheelchair dependency at the end of follow-up compared to 7 patients at the start. The remaining 6MWD studies reported statistically significant improvements up to two years and of the two studies which also reported results at three years, one found a statistically significant improvement[[39]while the other did not[[52].

Other outcomes

The 16 studies which reported FVC % predicted results were broadly consistent across their results. These indicated little change after up to one or two years of ERT, thereafter followed by slow declines over up to 10 years of follow up (Table 5, Additional file 3). The most common muscle function or strength outcomes reported were the Medical Research Council (MRC) scale (8 studies), handheld dynamometry (HHD, 4 studies) and the quick motor function test (QMFT, 3 studies). The studies reporting MRC scores had heterogeneous results with two studies (both n > 50) showing small but statistically significant increases during the first 2–3 years of ERT[[36, 42]whereas other studies reporting results for up to three years did not find statistically significant improvements[[52, 53]. No statistically significant improvements were seen in any of the studies reporting MRC scores at later time points. Two quite large studies which reported HHD outcomes found statistically significant improvements after up to two years of ERT[[42, 48], with one also reporting a plateauing of effect at around three years[[48]. These two studies also reported QMFT, which did not improve significantly in either study at up to five years of follow up.

One study analysed the effect of ERT on mortality by comparing ERT patients with patients not taking ERT[[44]. It found that the use of ERT was positively and statistically significantly associated with survival. Van der Meijden et al.[[54]also compared ERT users with non-ERT users in their large survey study, finding that ERT significantly reduced the risk for wheelchair use, but not the risk of needing respiratory support. Most of the six studies reporting ERT infusion-associated reactions found the occurrence rate to be around 25% (Table 5, Additional file 3).