Literature reviews  >  Articles for review > Hoyo et al. Improving the accuracy of syndromic... 

Unweighted algorithms can be used to improve the accuracy of syndromic diagnosis of genital ulcer disease.

Improving the accuracy of syndromic diagnosis of genital ulcer disease in Malawi.
Hoyo C, Hoffman I, Moser BK, Hobbs MM, Kazembe P, Krysiak RG, Cohen MS.
Sexually Transmitted Diseases 2005;32:231-37. 
 
Summary:

Question
Should syndromic algorithms for the management of genital ulcer disease change to accommodate changes in the prevalence of pathogens responsible for this syndrome?  Do unweighted algorithms perform as well as weighted algorithms?

Design

This article describes a cross-sectional study conducted at an STD clinic in Malawi to monitor the prevalence of HSV, H. ducreyi, and T. pallidum infection so as to reevaluate the syndromic guidelines.  Based on the current prevalence and strength of associations between PCR-based diagnoses and factors independently predictive of HSV and H. ducreyi, new algorithms were developed to guide clinicians in the management of ulcers without dependence on laboratory testing. 

Participants

One hundred thirty-seven patients with genital ulcers from among 599 consecutive STD patients who were treated syndromically at the Lilongwe Central Hospital STD Clinic in Malawi were included in this study. 

Description of Tests and Diagnostic Standard

Demographic and clinical data were collected during a clinical examination. Genital ulcers were cleaned and material was taken from the base of the ulcer using a swab, which was placed in transport medium (Amplicor, Roche), frozen, and shipped to the testing laboratory.  DNA was extracted from each swab specimen and tested by multiplex PCR for detection of T. pallidum, HSV, and H. ducreyi.  The digoxigenin labeled amplicons were captured and detected in three wells of an enzyme-linked immunosorbent assay plate using pathogen specific oligonucleotide probes labeled with biotin (PCR DIG ELISA detection kit, Roche Diagnostic Systems).  A specimen was considered positive for a specific pathogen if duplicate wells yielded an absorbance at 450 nm >0.25.  Discordant duplicates were reamplified and reanalyzed until the replicates were concordant. 

Blood was obtained for syphilis and HSV serology.  Serum was tested using the MacroVue RPR card test (Becton Dickinson, Cockeysville, MD); positive results were confirmed by Serodia-TPPA (Fujirebio, Tokyo, Japan).  Serum was tested for HSV type 2 IgG antibodies using Focus Technologies HSV-2 EIA; results were confirmed using the HSV immunoblot kit to identify antibodies to HSV-1, HSV-2, or both. 

Odds ratios for the association between demographic and clinical characteristics of participants with PCR–based diagnoses of HSV and H. ducreyi were calculated.  Individual risk scores were calculated by assigning to the independent clinical and demographic predictors of HSV or H. ducreyi a weighted score equivalent to 10 times the absolute value of the regression coefficient (natural log of odds ratios).    Unweighted scores were estimated by summing up all independent predictors.  When a factor was present, a value of 1 was assigned.  The sensitivity and specificity of diagnosis were calculated for a variety of possible risk score cutoff values and a receiver operating characteristic (ROC) curve was plotted for the weighted and unweighted scores.

Main Outcome Measures

The performance of risk scores in predicting the etiologic diagnoses of HSV and H. ducreyi was determined by comparing the areas under the ROC curves for the weighted and unweighted scores. 

Main Results

Thirty-three percent of patients reported a history of genital ulcers.  T. pallidum, HSV, and H. ducreyi were detected by PCR in 5 (3.7%), 47 (34.3%), and 41 (29.9%) of 137 patients, respectively.  Twelve (9%) patients were positive by RPR and TPPA; 107 (78.1%) and 128 (93.4%) had antibodies to HSV-2 and HSV-1, respectively.  Seventy-two percent of the 47 patients positive for HSV by PCR also had HSV-2 antibodies detected.   

Seven risk factors independently predicted HSV infection.  Patients with HSV were more likely to be married, be older than 25 years, be female, consult within 7 days of symptoms, have sexual contact in the 7 days before consultation, have prior genital ulcer disease, have superficial ulcers, and lack inguinal lymph node enlargement.  Two factors, age less than 25 years and deep ulcers, were independently predictive of H. ducreyi infection.  The 7 predictors of HSV infection and the 2 of H. ducreyi infection were used to derive weighted and unweighted risk scores, for which the sensitivity and specificity, compared to a diagnosis by PCR detection, were determined at several cutoff values.  There was no significant difference in the performance of the weighted and unweighted risk scores.  An unweighted algorithm with 2 demographic and/or clinical features was 70% sensitive and 55% specific for the diagnosis of HSV.  The new algorithm derived from this study is shown in the figure.  Empiric treatment of syphilis is reduced to reflect the current infrequency of this infection (3.7%).

Authors' Conclusions

The sizable temporal shift in the prevalence of GUD-related pathogens required a change in syndromic management.  The criteria used to establish the new algorithm did not need to be weighted because the difference in accuracy between weighted and unweighted algorithms was small.  Unweighted algorithms are easier to implement because they use data available in clinic notes, require little data manipulation, and only periodic laboratory support to estimate prevalence of etiologic pathogens. 

Source of funding:  NIDDK and the UNC STD Cooperative Research Center, and the HPTNC

about SDI | newsletters | grants | publications | literature reviews

WHO Home - WHO Search - TDR Home - SDI Home - SDI Contact us
(c) WHO/OMS 2001