Are arms-based methods required to test the stable unit treatment value assumption (SUTVA) among Cochrane reviews of contagious infection prevention interventions?

Herd (spillover) effects of infection prevention interventions are of increasing interest and are usually presumed beneficial. Their presence invalidates the Stable Unit Treatment Value Assumption (SUTVA) but testing for them is not simple. Moreover, in deriving causal inference from randomized concurrent controlled trials (RCCT) SUTVA, whilst difficult to verify, must still be considered. The effects of various interventions for preventing intensive care unit (ICU) acquired infections in high quality RCCTs were summarized in thirteen Cochrane reviews. Topical antibiotic prophylaxis or Selective digestive decontamination (TAP/SDD) appears the most effective intervention towards preventing pneumonia and mortality but what of SUTVA? Originally, TAP/SDD was presumed to mediate both direct and spillover (herd) effects with both presumed beneficial but four studies to estimate the TAP/SDD spillover effects were underpowered. TAP/SDD is mediated through decolonization which, unlike vaccination interventions, rebounds on TAP/SDD cessation. Are TAP/SDD spillover effects harmful? A trial to demonstrate that TAP/SDD herd effects are safe, being logistically complex and ethically fraught, would be infeasible. This commentary explores whether the 13 Cochrane reviews might present a natural experiment with which to evaluate the spillover (herd) effects of TAP/SDD to establish its population safety. Failure to consider SUTVA within Cochrane and similar systematic reviews of prevention interventions for contagious infections represents a major threat as any apparent summary effect size might have conflated spillover and direct effects. Summary effects derived from RCCTs with harmful spillover effects would cause profoundly misleading inferences.