Applying causal models to explore the mechanism of action of simvastatin in progressive multiple sclerosis

Arman Eshaghi, Rogier Kievit, Ferrán Prados, Carole H. Sudre, Jennifer Nicholas, M. Jorge Cardoso, Dennis Chan, Richard Nicholas, Sébastien Ourselin, John Greenwood, Alan J. Thompson, Daniel C. Alexander, Frederik Barkhof, Jeremy Chataway, Olga CiccarelliProceedings of the National Academy of Sciences 2019 · DOI

TL;DR

This study uses advanced causal modeling to investigate how simvastatin benefits patients with secondary progressive multiple sclerosis (SPMS). The findings reveal that simvastatin's positive effects on brain atrophy and disability are largely independent of its cholesterol-lowering action, suggesting a more complex mechanism of action involving other pathways. This work highlights the power of computational models in dissecting drug mechanisms in clinical trials.

What Problem Does This Paper Address?

Understanding how a drug works is crucial for optimizing its use and developing new therapies, but this is often a significant challenge in clinical trials, especially for complex neurodegenerative diseases like multiple sclerosis (MS). Conventional trial analyses typically focus on whether a drug has an effect, but not necessarily the precise biological pathways through which that effect is achieved. In the case of simvastatin, a drug known to lower cholesterol, it was observed to have beneficial effects on brain atrophy and disability in patients with SPMS. However, the key question remained: were these benefits primarily due to the reduction in cholesterol, or did simvastatin act through other mechanisms? This paper aims to disentangle these potential causal pathways.

Methods

The researchers employed structural equation modeling (SEM), a powerful statistical technique that allows for the simultaneous analysis of multiple variables and the testing of hypothesized causal relationships. They applied this to data from the MS-STAT trial, which involved patients with SPMS randomized to receive either simvastatin or a placebo.

The core of their approach involved:

  • Hypothesizing two competing models:
    • Cholesterol-mediated model: Simvastatin's benefits are primarily driven by its ability to lower serum cholesterol, which then impacts brain atrophy and disability.
  • Cholesterol-independent (pleiotropic) model: Simvastatin has direct beneficial effects on brain atrophy and disability, independent of its impact on serum cholesterol.
  • Model Selection: They used various goodness-of-fit measures (e.g., CFI, SRMR, RMSEA) and information criteria (AIC, BIC, Akaike/Schwarz weights) to determine which of the two hypothesized models best explained the observed data.
  • Bayesian Mediation Analysis: To quantify the extent to which simvastatin's effects were mediated by intermediate variables (like brain atrophy) or acted directly, they used Bayesian methods. This approach provides credible intervals (CIs) for parameter estimates, allowing for a more nuanced interpretation of both significant and non-significant findings.
  • Regional Brain Atrophy Analysis: They also investigated which specific brain regions were most affected by simvastatin's action, using mixed-effects models to compare regional atrophy rates between treatment groups.

Datasets

The study utilized data from the MS-STAT trial, a phase 2, double-blind, randomized controlled trial involving 140 patients with secondary progressive multiple sclerosis (SPMS). The data collected included:

  • Brain magnetic resonance imaging (MRI) at baseline, 1 year, and 2 years.
  • Cognitive assessments (Block Design test, PASAT, FAB).
  • Physical disability assessments (Expanded Disability Status Scale - EDSS, MSIS-29v2 physical subscore).
  • Serum total cholesterol levels.

The researchers also made their computer codes and simulated data publicly available on GitHub.

Key Insights and Ideas

  • Simvastatin's benefits are largely independent of cholesterol reduction: This is the central finding, challenging the intuitive assumption that simvastatin's primary mechanism in MS is through lowering cholesterol.
  • Direct effects on brain atrophy and disability: Simvastatin appears to directly slow down brain atrophy and reduce the progression of disability (measured by EDSS), bypassing the need for cholesterol reduction.
  • Brain atrophy as a mediator: While direct effects are prominent, brain atrophy does play a significant role in mediating the treatment's effect on disability, accounting for about 31% of the total treatment effect on EDSS.
  • Pleiotropic effects are key: The findings strongly suggest that simvastatin's beneficial actions in SPMS are due to its "pleiotropic" effects – actions beyond cholesterol lowering, likely involving other downstream metabolites of the cholesterol synthesis pathway.
  • Causal modeling is essential for understanding drug mechanisms: The study demonstrates that advanced statistical techniques like SEM can move beyond correlational analyses to provide causal insights into treatment effects, which is invaluable for complex diseases.

Key Results

  • The cholesterol-independent model was significantly more likely to explain the data than the cholesterol-mediated model, with Akaike weights suggesting it was over 42 times more probable.
  • Simvastatin had a direct effect on delaying EDSS worsening, explaining 69% of the overall treatment effect.
  • Simvastatin had a direct effect on reducing brain atrophy, which in turn mediated 31% of the treatment effect on EDSS.
  • The direct effect of simvastatin on slowing brain atrophy was independent of cholesterol, with non-overlapping credible intervals for the direct and cholesterol-mediated pathways.
  • The transverse temporal gyrus was identified as a region showing a significantly faster rate of atrophy in the placebo group compared to the simvastatin group, suggesting a specific regional impact.

Limitations and Future Work

  • Post hoc analysis: The study acknowledges that its analysis is post hoc, meaning it was conducted after the original trial data was collected and analyzed. While valuable for generating hypotheses, pre-planned analyses are considered the gold standard for definitive conclusions.
  • Focus on specific outcomes: The study focused on specific clinical and imaging outcomes that had shown significant or marginally significant changes in previous analyses. Other potential outcomes might have been influenced differently.
  • Medulla oblongata as a proxy: The study used medulla oblongata volume as a proxy for spinal cord atrophy due to the absence of direct spinal cord imaging. This is an indirect measure.

The authors suggest that future research should focus on investigating the role of upstream intermediate metabolites of the cholesterol synthesis pathway (beyond cholesterol itself) as potential mediators of simvastatin's beneficial effects. They also propose that mechanistic approaches like SEM can be applied to other trials of neurodegenerative disorders to gain deeper insights into treatment mechanisms.

Summary updated

Abstract

Understanding the mode of action of drugs is a challenge with conventional methods in clinical trials. Here, we aimed to explore whether simvastatin effects on brain atrophy and disability in secondary progressive multiple sclerosis (SPMS) are mediated by reducing cholesterol or are independent of cholesterol. We applied structural equation models to the MS-STAT trial in which 140 patients with SPMS were randomized to receive placebo or simvastatin. At baseline, after 1 and 2 years, patients underwent brain magnetic resonance imaging; their cognitive and physical disability were assessed on the block design test and Expanded Disability Status Scale (EDSS), and serum total cholesterol levels were measured. We calculated the percentage brain volume change (brain atrophy). We compared two models to select the most likely one: a cholesterol-dependent model with a cholesterol-independent model. The cholesterol-independent model was the most likely option. When we deconstructed the total treatment effect into indirect effects, which were mediated by brain atrophy, and direct effects, simvastatin had a direct effect (independent of serum cholesterol) on both the EDSS, which explained 69% of the overall treatment effect on EDSS, and brain atrophy, which, in turn, was responsible for 31% of the total treatment effect on EDSS [β = -0.037; 95% credible interval (CI) = -0.075, -0.010]. This suggests that simvastatin's beneficial effects in MS are independent of its effect on lowering peripheral cholesterol levels, implicating a role for upstream intermediate metabolites of the cholesterol synthesis pathway. Importantly, it demonstrates that computational models can elucidate the causal architecture underlying treatment effects in clinical trials of progressive MS.
BibTeX
@inproceedings{eshaghi2019applying,
  author = {Arman Eshaghi and Rogier Kievit and Ferrán Prados and Carole H. Sudre and Jennifer Nicholas and M. Jorge Cardoso and Dennis Chan and Richard Nicholas and Sébastien Ourselin and John Greenwood and Alan J. Thompson and Daniel C. Alexander and Frederik Barkhof and Jeremy Chataway and Olga Ciccarelli},
  title = {{Applying causal models to explore the mechanism of action of simvastatin in progressive multiple sclerosis}},
  booktitle = {Proceedings of the National Academy of Sciences},
  year = {2019},
  doi = {10.1073/pnas.1818978116},
}
Figure