A survival analysis of trypanosomosis diagnostic-test performance under natural infection challenge

Little is known about the time-to-first detection and the time difference (TD) between first parasitological and first serological diagnosis of Trypanosoma spp. infections under natural infection challenge in cattle. The objective of our study was to estimate these measures of "longitudinal aspects" of diagnostic performance and to investigate potential biological factors. Emphasis was on diagnosis at the genus level (Trypanosoma spp.). Twelve N'Dama, 12 Gobra zebu and 12N'DamaxGobra (F1) crossbred cattle (all animals non-infected at the start of the experiment, six male and six female animals in each cohort) were exposed to natural high tsetse challenge in the Niamina East area in The Gambia Acta Trop. 71 (1998) 57]. The animals were investigated parasitologically (detection of trypanosomes by buffy-coat technique), serologically (detection of T. brucei, T. congolense and T. vivax antigen by enzyme-linked immunosorbent assay (ELISA)) and clinically (packed-cell volume, PCV) over a period of 180 days. The time-to-first detection of trypanosomes, trypanosomal antigen (cut-off as suggested by test supplier) and drop in PCV (subject-based cut-off values) were recorded as outcomes of interest. Thus, incidence was both parasitologically (I(p)), serologically (I(s)) and clinically (I(c)). Recurrent events were not considered. The TD between first parasitological and first serological detection was established as I(s) time minus I(p) time. The effect of breed and sex on the time-to-first detection and on TD was investigated using Cox (proportional hazard) regression and ANOVA, respectively.We found that time-to-first parasitological detection of trypanosomosis in N'Dama animals was significantly longer than in the two other breeds (Cox regression, P=0.002). A similar but less-strong (P=0.063) effect of breed on time-to-first detection of trypanosomal antigen was found, whereas no breed effect was observed for clinical detection (P=0.432). Sex had no effect in all detection systems. The TD varied between -56 and 115 (mean 28). Marked differences among breeds and between sexes were not observed (ANOVA, P=0.8). We suggest that incidence studies are more suitable for detecting risk factors for animal trypanosomosis than prevalence-based (cross-sectional) studies because the latter often result in misinterpretation of factors that increase the survival time with infection as risk factors.