#apaperaday: Real-world and natural history data for drug evaluation in Duchenne muscular dystrophy
In today’s #apaperaday, Prof. Aartsma-Rus reads and comments on the paper titled: Real-world and natural history data for drug evaluation in Duchenne muscular dystrophy: suitability of the North Star Ambulatory Assessment for comparisons with external controls
Today’s pick us from World Muscle Society journal Neuromuscular Disorders by Muntoni et al on using real world and natural history data for drug efficacy evaluation in Duchenne patients. Collaborative effort of many clinical groups including Duchenne Center NL and cTAP. Doi 10.1016/j.nmd.2022.02.009
cTAP is the collaborative trajectory analysis project. Their goal is to provide models to predict trajectories in Duchenne in order to better design trials but also to see if natural history data can be used instead of placebo groups.
To test drug efficacy double blinded placebo controlled trials are the gold standard. However, for Duchenne, this is challenging for multiple reasons.
- Trials often take 2 or even 3 years which is a very long time to be in a placebo group.
- Sometimes placebo is not possible or rather blinded placebo is not possible e.g. due to side effect profile of the drug.
- There are many mutation specific treatments in clinical development for Duchenne. This is done in subgroups of patients, making it difficult to populate each trial with placebos.
Furthermore, even if placebo-controlled trials are done and drugs registered, beyond the trials sponsors need to collect real world longer term evidence. cTAP aims to assess whether it is possible to reduce placebo groups or even replace them with natural history data.
To do this they compare datasets from patients in a clinical trial placebo group (provided by companies) & patients in natural history studies (provided by academic centers, companies & patient organizations). Authors focused on the North Star Ambulatory Assessment scale (NSAA)
This is a scale that is used in ambulatory patients and which collects 17 items relevant for daily activity (walking, stair climbing, getting up from the floor). It is a primary endpoint in many clinical trials in ambulatory patients nowadays.
Authors collected data from 437 patients for whom 48 week data was available. 235 from placebo groups and 202 natural history. Baseline characteristics were similar for age, but the placebo group has less deflazacort use as a steroid and lower functional scores.
The placebo group was also enriched for nonsense mutations and exon 51 skippable mutations (as may patients in placebo group were involved in trials were stop codon readthrough and exon 51 skipping compounds were evaluated)
Using statistical modeling provided by James Signorovitch (second author on the paper), 28% of the variability due to prognostic factors at baseline could be explained. Additional modeling (multivariable regression) was done to adjust further. This reduced variability.
I am not a statistician but as far as I understand the modelling tries to take into account known predictive factors for disease progression (in this case NSAA trajectories) and correct for that. So the influence of known confounders (e.g. steroid use) is corrected for.
Another way to do this is propensity matching. Here the goal is to find a matched set of patients for baseline characteristics in each group and to then see whether the trajectories of these matches vary or not. Also here variability was reduced.
Using these models, authors showed that there were no differences between placebo treated patients in therapeutic intervention studies and natural history trial patients. Trajectories of the NSAA were the same – also when authors looked specifically in younger patients (5-8 years).
One could have expected the placebo patients to do better, because they knew they might get a treatment, and also the people doing the NSAA knew and might have unconsciously motivated them more. This appears not to be the case.
Clinical trials are done in very strict ways, while the natural history studies were done in multiple different centers. Still there were no geographic differences between the groups after correcting for the variables known to influence disease progression. However, the centers involved in these natural history studies are expert care sites and also have experience with conducting natural history studies. So for future natural history datasets for other centers, comparability needs to be confirmed.
Factors known to influence progression: height, weight, BMI, steroid use & functional outcomes. Including age does not provide additional information, but of course age influences many of the other factors. These factors were used to model trajectories and propensity matching.
Authors discuss that additional factors will play a role as well, e.g. steroid regimen (see also the FOR-DMD study). My own addition: genetic modifiers. Authors discuss that the modeling relies on data quality and completeness.
For now work is based on 1 year follow up. Future work will have to assess the comparability of longer term trajectories given that many placebo trials in Duchenne are now 2-3 years. Authors discuss that in the future maybe placebo groups can be replaced by natural history data.
They discuss the option of having smaller placebo groups, supplemented by data from natural history studies. However, a small placebo group would make randomization more difficult (though perhaps that can be compensated for by propensity matching?).
The last thing authors mention is how crucial collaboration is for these efforts. Without collaboration it would not be possible to make the models and to collect such a large dataset. So thanks to all the clinicians and patients involved in providing the data!
Pictures by Annemieke, used with permission.
