Reader Comments

Post a new comment on this article

Statistical Analysis and Diagnostic Tests Used

Posted by torger on 07 Jan 2013 at 11:30 GMT

A primary hypothesis of this paper is with regard to tapeworm carriers having a geographical association with pigs with cysticercosis cannot be supported by the data. Although of course an association of such pigs with human carriers, it could be argued, is inevitable as humans are the only definitive host of T. solium and pigs are nearly always the intermediate host.
The authors find 6 coproantigen positive individuals in a sample of 402. However coproantigen tests rarely have a specificity of 100%. The authors cite Allan et al (Allan JC, Velasquez-Tohom M, Torres-Alvarez R, Yurrita P, Garcia-Noval J (1996) Field trial of the coproantigen-based diagnosis of Taenia solium taeniasis by enzyme-linked immunosorbent assay. Am J Trop Med Hyg 54: 352–6.) as their methodology for coproantigen diagnosis. From reading Allan et al, this test has a specificity of 99.2%. Therefore one would expect approximately 3 false positives in a population of 402 individuals, none of whom had tapeworm infections. This leaves only 3 individuals which one could do statistics with! Furthermore there would be an 11% probability of getting 5 false positives from 401 non infected individuals given a specificity of 99.2%: the only individual “proved” infected was the one with Taenia eggs in the faeces. And you cannot come to any conclusions with one positive (or even 3!).
The only way to solve this problem would be to undertake a much larger study such that there would be a sufficient number of true positives to be able to undertake appropriate statistical analysis.

No competing interests declared.

RE: Statistical Analysis and Diagnostic Tests Used

oneals replied to torger on 11 Jan 2013 at 02:44 GMT

We thank the previous poster for their careful reading of our manuscript and for their thoughtful comments. The main concern regards the imperfect specificity of the coproantigen ELISA which introduces the possibility of false positive taeniasis results into the study. However, we do not use the standard cutoff which results in 99.2% specificity as reported by Allen et. al. Instead, we use a more conservative cutoff based on percent positivity (PP greater than or equal to 14%; PP is defined as the OD of the sample relative to the OD of the positive control) which is approximately twice the standard cutoff. This approach greatly decreases the overall likelihood of false-positive results for Taenia sp. and also decreases the likelihood that T. saginata infection will produce positive results. Nevertheless, even assuming 100% specificity we concede that this is a small study with a limited number of taeniasis cases to work with. As we state in the discussion, our results and conclusions should be validated in a larger study. Ultimately, however, the question we are interested in is whether targeted interventions around heavily-infected pigs can reduce transmission of T. solium in endemic communities. Rather than dedicating resources to conduct a larger descriptive study to validate these findings, we are moving forward with a pilot intervention trial to test the effectiveness of targeted screening in reducing transmission. We hope to be in a position to share our results later this year.

No competing interests declared.