To the Editors: Lennon et al present a meta-analysis of school or community-based studies involving treatment of streptococcal sore throat (some including treatment of throat carriage as well) and claim that the summary estimate of efficacy (59%) is likely to represent a “true treatment effect.”1 We question this result and indeed the validity of the meta-analytic methodology used. Because of the paucity of randomized controlled trials (RCT), the inclusion criteria were broadened to include before/after and observational studies. As a result, only one RCT was included—the authors' own study (the NZ randomized controlled trial (ref. 17 in the Lennon et al paper) is listed as published in the Pediatric Infectious Diseases Journal, but does not appear to have been published at the time of submission of this letter), a trial of very high quality involving almost 87,000 person-years of observation, which found no significant reduction in acute rheumatic fever (ARF) incidence in the treatment group. That trial, the scope of which is unlikely to be repeated, surely qualifies as the most important ARF research in at least the past 2 decades. However, in this meta-analysis, the results of this RCT are pooled with 2 observational studies (comparing ARF incidence in schools with interventions to schools without the intervention) and 3 before/after studies (comparing ARF incidence in a region before the intervention and after the intervention), none of which involved any blinding or randomization. Interestingly, in the Forest plots of the study results, the study demonstrating the largest treatment effect, and with the greatest impact on the pooled efficacy result, was the least able of all the studies to determine the specific effectiveness of a sore throat treatment intervention.2 This report documents changes in the incidence of ARF in a region of Cuba over 10 years, during which a comprehensive ARF/RHD public health intervention was instituted, which included not only primary prophylaxis but also “secondary prevention of ARF/RHD, training of health personnel, healthcare education via dissemination of information, community involvement, and epidemiological surveillance.” Comprehensive programs such as this have been associated with ARF incidence reductions over a long period previously in the Caribbean,3 which may speak to their overall effectiveness, but does not permit one to estimate the effectiveness of any single component of the intervention, particularly in the absence of any attempt to control for multiple other confounding factors that may be present in a single region over a 10-year period. Lennon et al also do not address the issues of practicality of such school or community based programs, particularly in resource-poor countries, where the burden of ARF/RHD is the greatest. One of the studies included in the analysis pointed out that it cost US $12 per child enrolled in school per year (in 1970s currency),4 and we would be interested to know the cost of the New Zealand RCT, which we suspect to have been expensive. In low and middle income countries, there are additional barriers including access to microbiological diagnosis, training of primary care staff, transport to and from remote areas, availability of antibiotics, and ability to follow-up patients that would require considerable additional resources to overcome.5 As a result, such programs would be even less affordable in these settings. Meta-analysis can be a powerful tool, but only if appropriate studies are selected for inclusion in the pooled analysis. Although the 5 observational or before/after studies included in the analysis provide some encouraging nonrandomized data in favor of community-based primary prophylaxis, they are not of sufficient quality to include in a meta-analysis. For an intervention where the only true efficacy study does not demonstrate a significant treatment effect, and which has significant cost implications for countries with the highest disease burden, we believe it is unreasonable to conclude, as Lennon et al do, that “in communities with high rates of RF (we suggest greater than 50 per 100,000 children per year), the school- and/or community-based programs be actively considered.” Jonathan Carapetis, MB BS, PhD Menzies School of Health Research and Charles Darwin University Darwin, Australia Andrew Steer, MB BS, FRACP Department of Pediatrics University of British Columbia Vancouver, Canada
Read full abstract