Evaluating diagnostic indicators of urogenital Schistosoma haematobium infection in young women: A cross sectional study in rural South Africa

Abstract

Background: Urine microscopy is the standard diagnostic method for urogenital S. haematobium infection. However, this may lead to under-diagnosis of urogenital schistosomiasis, as the disease may present itself with genital symptoms in the absence of ova in the urine. Currently there is no single reliable and affordable diagnostic method to diagnose the full spectrum of urogenital S. haematobium infection. In this study we explore the classic indicators in the diagnosis of urogenital S. haematobium infection, with focus on young women.

Methods: In a cross-sectional study of 1237 sexually active young women in rural South Africa, we assessed four diagnostic indicators of urogenital S. haematobium infection: microscopy of urine, polymerase chain reaction (PCR) of cervicovaginal lavage (CVL), urogenital symptoms, and sandy patches detected clinically in combination with computerised image analysis of photocolposcopic images. We estimated the accuracy of these diagnostic indicators through the following analyses: 1) cross tabulation (assumed empirical gold standard) of the tests against the combined findings of sandy patches and/or computerized image analysis and 2) a latent class model of the four indicators without assuming any gold standard.

Results: The empirical approach showed that urine microscopy had a sensitivity of 34.7% and specificity of 75.2% while the latent class analysis approach (LCA) suggested a sensitivity of 81.0% and specificity of 85.6%. The empirical approach and LCA showed that Schistosoma PCR in CVL had low sensitivity (14.1% and 52.4%, respectively) and high specificity (93.0% and 98.0, respectively). Using LCA, the presence of sandy patches showed a sensitivity of 81.6 and specificity of 42.4%. The empirical approach and LCA showed that urogenital symptoms had a high sensitivity (89.4% and 100.0%, respectively), whereas specificity was low (10.6% and 12.3%, respectively).

Conclusion: All the diagnostic indicators used in the study had limited accuracy. Using urine microscopy or Schistosoma PCR in CVL would only confirm a fraction of the sandy patches found by colposcopic examination.

Fig 1. Flow chart showing the inclusion of participants.

a. We were only able to calculate the school prevalence after all the pupils from a given school had been examined. Therefore we excluded a lot of the young women who visited the clinic. b. All were invited to provide samples. Only a selection of samples (34.6%) was sent for PCR analyses.

Fig 2. Venn diagram showing the overlap between positive findings in the four diagnostic indicators.

The four diagnostic indicators were: (1) Urine microscopy, (2) Schistosoma PCR in cervico-vaginal lavagae, (3) sandy patches identified using clinical photocolposcopic examination or by computerised colourimetric image analysis and (4) self-reported urogenital symptoms: abnormal discharge colour, abnormal discharge smell, burning sensation in the genitals, bloody discharge, genital ulcer, red urine, pain on urination, stress incontinence and urge incontinence).