Skip to content

Abstract Search

LATEBREAKER

COVID-19 Pandemic

Flatland: 2-dimensional fixed effect models in meta-analyses of non-pharmaceutical interventions need to be upgraded to a dynamic, non-linear, high-dimensional world Brian Gurbaxani* Brian Gurbaxani

  • Forest plots for the meta-analysis of interventions are flat, 2-dimensional, and tend to use linear fixed and random effects models as modes of analysis. By contrast, the effects of non-pharmaceutical interventions (NPI) on an epidemic live in a high-dimensional, non-linear world.  These facts have been brought into dramatic relief by the recent COVID-19 pandemic, and in particular by the Cochrane Review’s recent meta-analysis of NPI such as facemasks, handwashing, and social distancing for respiratory infections1.  The Cochrane Review’s conclusions for facemasks and COVID-19 in particular were inconclusive, and could not provide evidence for their efficacy, whereas a plethora of other studies have shown efficacy.  A more nuanced review of Cochrane’s results in the light of non-linear, dynamic modeling2 and a lively exchange on these issues3,4 has recently occurred in the pages of the American Journal of Public Health and the Journal of the American Medical Association5, an important result of which is that linear fixed and random effects models are not the best way to approach the epidemiological data on NPI, as can be seen in even very basic dynamic models.  In this talk we expand on this theme that dynamic models are needed to evaluate some interventions like facemasks and other NPI, because the interventions do not have a fixed effect over the variables of time, percentage of use, and attack rate of the pathogen.  Importantly, the results show that if the design, duration, and statistical power of a randomized controlled trial of an NPI are chosen poorly, the study will likely show no effect, even when the NPI are performing exactly as predicted.

 

  1. Jefferson T, Dooley L, Ferroni E, et al. Physical interventions to interrupt or reduce the spread of respiratory viruses. Cochrane Database Syst Rev. 2023;1(1):CD006207
  2. Gurbaxani BM, Hill AN, Patel P. Unpacking Cochrane’s update on masks and COVID-19. Am J Public Health. 2023;113(10): 1074–1078
  3. Kivelä, JM. Unpacked Cochrane review on masks – a further look. Am J Public Health.  2024:114(2): 251-252
  4. Gurbaxani BM, Hill AN, Patel P. Gurbaxani Et Al. Respond. Am J Public Health. 2024;114(2): 252–253
  5. Cash-Goldwasser S, Reingold AL, Luby SP, et al. Masks During Pandemics Caused by Respiratory Pathogens—Evidence and Implications for Action. JAMA Network Open. 2023;6(10): e2339443