Epidemiological analysis of host populations with widespread sub-patent infections: African trypanosomiasis

The epidemiological study of pathogens largely depends on three technologies, serology, microscopy and the polymerase chain reaction (PCR). Serological methods are unable to differentiate between current and past infections. Microscopy has historically been the mainstay of epidemiological study. In recent times the use of microscopy has been in decline, as it has been shown to have an inherent lack of sensitivity and specificity and produces many false negative results. PCR is now the method of choice for screening samples for the presence or absence of pathogens. Although PCR is widely regarded as an extremely sensitive technique, the fact that it assays a very small volume of sample is often overlooked. If the target pathogen is not present in the tiny aliquot of sample from an infected host, then a false negative results will occur. In endemic situations were the pathogen is present at low infection intensities, then the potential for false negatives results of this type is high. This intensity related false negative effect can lead to serious underestimation of diagnosed prevalence and incidence with consequent misinterpretation of the resulting data. This phenomenon has been reported in the literature for a range of pathogens and especially for epidemiological study of schistosomiasis. The extensive occurrence of false negatives during study of schistosomiasis samples was such an obstacle to epidemiological study it prompted the world health organisation to repeatedly call for quantitative methods to be employed to combat the problem. The main objectives of this thesis are to rationalise and simplify the methods of diagnosing African trypanosomes in epidemiological studies and to investigate the consequences of, and methods of dealing with infection intensity related false negative results that occur as a result of widespread sub-patent infections in the study population A new PCR assay was developed that was capable of analysing whole blood placed onto treated filter paper. The PCR assay was capable of differentiating between all the important African trypanosome species, producing a unique size of amplicon for each species of trypanosome. Initial results from repeated screening of human and cattle samples known to be parasitologically positive indicated that many false negative results occur. A more extensive analysis of thirty five bovine blood samples randomly chosen from a collection of field samples revealed that false negative results occurred regularly. The prevalence of infection after a single screening was 14.3% whereas the cumulative prevalence after over 100 repeated screenings rose to 85.7%. This showed that a severe underestimation of prevalence occurs from a single screening of the samples. In order to investigate the consequences of, and develop methods of dealing with this problem, computer based simulations were used to model the dynamics of screening samples with sub-patent infections. In order to construct the model the data obtained from repeat screening of the thirty-five bovine blood samples was fitted to a number of mathematical distributions. A negative binomial distribution best described the distribution of trypanosomes across the hosts. Exploration of the phenomenon with the resulting model showed the extensive underestimation of true prevalence that is possible. The simulations also showed that it is possible for populations with very different patterns of infection and true prevalence to all have the same diagnosed prevalence from a single screening per sample. Statistical comparison of these very different populations by diagnosed prevalence alone would conclude there was no significant difference between the populations. It was therefore concluded that the diagnosed prevalence from a single (or even multiple) screenings is an inadequate and potentially misleading measure of both infected hosts and parasite numbers. In order to deal with these problems new methods were evaluated for use in epidemiological studies. A simple method of producing quantitative measures of infection was advocated. The insensitivity of existing screening methods in detecting significant difference between populations was highlighted and a greatly improved methodology was shown. Finally, a method for inferring the true population prevalence from the data obtained from repeat screening of samples was suggested. Although some of these new methodologies have limitations, they represent a great improvement on the use of a single diagnostic test for each host. The work presented in this thesis highlights a serious potential limitation to our understanding of the epidemiology of pathogens that exist at sub-patent levels, and develops some possible methods of overcoming these limitations.