• E-mail
• Print
• Comment
• Font Size
• Digg
• del.icio.us
• Discuss article

Development of Composite Symptom Variables for Quantitative Analysis of Genitourinary Symptomatology in Women

Posted on: Sunday, 6 March 2005, 03:00 CST

Summary: Gonorrhoea and chlamydia infections in women are often regarded as asymptomatic. Syndromic management of sexually transmitted disease (STDs), however, is partially based on vaginal symptoms. We sought to better identify STD-associated symptoms in women by development of composite genitourinary symptom constructs. Standard symptoms were stratified, based on their descriptors (amount, frequency, severity, etc.), into pathological (likely to be STD-associated) and intermediate (unlikely to be STD-related). Simple symptoms and composite symptom constructs were significantly more common in women with STD infections (chlamydia, gonorrhoea and/ or trichomonas) than those without infection (six months later). Logistic regression confirmed the association of each pathological symptom construct individually with gonorrhoea, chlamydia and trichomonas. Composite symptom constructs improve the specificity for detecting STD infections in women.

Keywords: sexually transmitted diseases, gonorrhoea, chlamydia, genitourinary symptoms

Introduction

Urethral infection with gonorrhoea, chlamydia and trichomonas has recognized symptoms in men,1 but symptomatology in women is more controversial.2 Cervical infections with gonorrhoea and chlamydia have been regarded as asymptomatic; however, the World Health Organization (WHO) guidelines for the syndromic treatment of sexually transmitted disease (STDs) (currently utilized in developing countries) are partially based on the presence of symptoms.3 A major decision point in the flowchart for vaginal discharge, however, is the presence of STD risk factors;3 diagnosis/ treatment is thus based more on potential exposure than on self- reported symptoms.

Many factors influence women's reporting of genitourinary (GU) symptomatology, including cultural issues, disease severity and individual perception. In developing countries, full symptom disclosure may be inhibited by cultural taboos against discussing genital problems. In the USA, symptom evaluation is impacted by over- the-counter vaginal products potentially either masking or exacerbating symptoms. Even in a comfortable environment with adequate privacy and good rapport, it is difficult to assess women's symptoms in a reproducible, quantifiable manner. Analysis based on presence/absence is ineffective, since uninfected women often have GU symptoms. Symptoms are also common with non-STD infections (candidiasis and bacterial vaginosis).

In order to quantify potentially STD-related GU symptoms in women, we used common descriptors (amount, frequency, severity, etc.) of individual symptoms to develop composite symptom complexes. Descriptors for each symptom were grouped a priori, based on the theoretical likelihood of association with gonorrhoea, chlamydia and/ or trichomonas infection. Each was classified as either pathological (likely STD-associated) or intermediate (less likely STD- associated). We tested the validity of these composite measures on our well-characterized population of minority women with STDs.

Methods

Mexican-American and African-American women with an active, curable STD (gonorrhoea, chlamydia, trichomonas or syphilis) enrolled in an IRB-approved study of a behavioural-cognitive intervention to reduce STD recurrence.4 At entry, six and 12 months' follow-up, each woman was extensively questioned regarding GU symptoms and tested for gonorrhoea (GenProbe), chlamydia (GenProbe), trichomonas (culture), bacterial vaginosis (Gram stain), Candida (culture) and syphilis (Rapid Plasma Reagin). Women with gonorrhoea, chlamydia and/or trichomonas at entry were analysed for symptoms associated with active infection. Women free of infection with gonorrhoea, chlamydia, trichomonas, bacterial vaginosis and Candida at six months' follow-up were analysed for symptoms when uninfected (all had been infected initially). GU symptoms were then compared between the infected and uninfected visits. Women with non-STD infections (bacterial vaginosis and Candida) at follow-up were excluded from crude comparisons to avoid confounding effects, but were included in multivariate regression analyses. Due to potential impact on GU symptoms, pregnant women were excluded from all analyses.

Description of composite symptom variables

Nine main symptom variables were developed, including four major composite symptoms (vaginal discharge, abdominal pain, dyspareunia and urinary symptoms), two composite behaviours based on actions taken in response to symptoms, like using over-the-counter treatments (discharge action and abdominal pain action), and three simple symptoms coded as present/absent (vaginal odour, vaginal itching and fever/chills). Each of the composites is labelled as pathological, intermediate or none (participant denied the symptom). Variable construction, including specific questions used, is described in Table 1.

Analysis

Frequencies (presence/absence) of basic GU symptoms were assessed in the infected and uninfected groups. Crude rates of composite symptoms were compared between the infected and uninfected samples. Due to co-infections with multiple STD and non-STD organisms, multivariate regression analyses were performed to identify the independent association of composite symptoms with individual organisms (logistic regression for simple symptoms and multinomial regression for composites). Multinomial regression allowed the separate comparison of intermediate and pathological categories to the no symptom category, analogous to separate logistic regression analyses for intermediate and pathological levels. Regression models included gonorrhoea, chlamydia and trichomonas as well as potential confounding factors (group assignment [intervention versus control], age and non-STD infections [bacterial vaginosis and Candida]).

Symptom scores

A quantitative scoring system was developed to identify infection with gonorrhoea and/or chlamydia. Symptoms found to be associated with infection by regression analysis were incorporated into the scoring system. Receiver operating characteristic (ROC) curve analysis was performed to identify appropriate cut-offs.

Results

In all, 617 minority women with active STDs enrolled in the intervention trial. The infected cohort comprised 385 non-pregnant women with gonorrhoea, chlamydia, or trichomonas at entry (198 women excluded due to pregnancy, 31 due to having only syphilis and three due to missing pregnancy status). Among infected women, 19% had multiple STDs (n = 74) and 66% had bacterial vaginosis or candidiasis (n = 256). The uninfected group (crude comparisons) included 133 non-pregnant women free of all infections at follow- up. The uninfected group (multivariate analysis) included 188 additional women with bacterial vaginosis or Candida at follow-up (total n = 321, excluded 115 pregnant women and 72 with gonorrhoea, chlamydia or trichomonas).

Basic GU symptoms were reported more frequently by infected than uninfected women (vaginal discharge -65.7%, 27.8%; abdominal pain - 42.6%, 22.6%; dyspareunia -15.3%, 5.3%; dysuria - 20.5%, 7.5%; urinary frequency - 25.7%, 15%; vaginal itching -39.2%, 12%; vaginal odour -41.3%, 13.5%, all P < 0.01). Many uninfected women, however, reported the same symptoms. Most symptoms were reported by >10% and some by >20% of uninfected women. Although the majority of infected women reported symptoms, specificity for STDs is poor.

Based on crude rates, division of the composite symptoms into pathological and intermediate categories typically improved association with infection (Table 2). The small loss in sensitivity from subclassification was outweighed by the large increase in specificity. Specificities for pathological vaginal discharge, abdominal pain and dyspareunia exceeded 80%; however, all pathological level symptoms were present in <50% of infected women.

Based on regression analyses (Table 3), chlamydia infection was associated with vaginal discharge (both levels), abdominal pain (both levels), discharge action (both levels), pain action (both levels), urinary symptoms, itching and odour. The strongest association with chlamydia was pathological discharge (3.7-fold increase). Gonorrhoea was associated with pathological discharge, abdominal pain (both levels), intermediate dyspareunia, discharge action (both levels), pain action (both levels) and vaginal odour; urinary symptoms and itching approached significance. Trichomonas was associated with pathological discharge, an action taken in response to discharge, urinary symptoms, itching and odour. Bacterial vaginosis and Candida had no significant associations except increased odour and bacterial vaginosis (odds ratio 1.66).

Table 1 Individual symptom questions and derivation of composite symptoms (coding for each composite symptom variable in footnotes below)

Table 2 Crude rates of genitourinary composite symptoms in women infected with gonorrhoea, chlamydia or trichomonas compared with women proven to be free of any vaginal or cervical infections (uninfected)

Table 3 Association of composite symptoms with individual vaginal/ cervical infections by multinomial (3-level) or multivariate (2- level) logistic regression analysis (expressed as odds ratio with 95% confidence interval)

We developed a simple scoring method using the symptom co\mposites to predict infection with gonorrhoea and/or chlamydia. The most effective method was one point each for the presence of discharge, abdominal pain, dyspareunia, odour and itching, an additional point each for having a pathological level symptom (discharge, abdominal pain and dyspareunia), and an additional point each for an action (discharge and abdominal pain) for a maximum of 10 points. ROC curve analysis was performed to identify cutpoints for optimal sensitivity and specificity. No clear breakpoint was identified; however, cut-offs of ≥2, ≥3 or ≥4 yielded sensitivities/specificities of 74% /55%, 63%/64% and 49%/ 78%, respectively.

Discussion

We developed symptom composites to better compare GU symptoms reported by women with and without STDs. Crude symptoms were reported by many women without detectable infection. Closer examination revealed that some reported symptoms were clearly unrelated to vaginal infection ('clear' vaginal discharge). High rates of symptomatology in uninfected women have been reported in other populations (41-46% of virginal Nigerian adolescents reported vaginal discharge, but few were yellow, heavy, or malodourous).5 Mixing unrelated symptoms with potential STD symptoms makes crude measurement (presence/ absence) unreliable for identifying infection. We developed the GU symptom composites to distinguish STD symptoms from unrelated ones.

The symptom composites also allow quantification of severity of GU symptoms for comparison. Previous crude comparisons were rendered uninterpretable by high baseline symptomatology in both groups. We needed a framework for more precise comparisons. Both 'pathological level' and 'action taken' composites are measures of severity. For some symptoms, pain scores might better measure severity, but they were unavailable for this comparison.

We evaluated association of symptoms with gonorrhoea and chlamydia, classically regarded as asymptomatic STDs in women.3,5 Syndromic management is the primary means of STD detection and treatment in much of the developing world.5,6 The WHO algorithm for syndromic treatment of vaginal discharge involves treatment for gonorrhoea, chlamydia and trichomonas in women with an abnormal discharge and risk factors for STDs.3 Of all syndromic management algorithms, vaginal discharge is clearly the most problematic.5,7 The poor predictive value of this algorithm for gonorrhoea and chlamydia is well documented (positive predictive value 5-48%, sensitivity 10-89% and specificity 15-95% depending on population, practitioners and risk scores utilized),7 typically attributed to the asymptomatic nature of cervical infections.5-7 We, however, found strong associations between composite symptoms and gonorrhoea/ chlamydia (Table 3). Strong associations by logistic regression, unfortunately, do not guarantee high sensitivity and specificity in clinical settings. Clearly, women need more sensitive and specific algorithms for the diagnosis of gonorrhoea and chlamydia in resource- poor countries, where syndromic management remains the most practical (and often the only available) diagnostic method.5,6

Many factors impact GU symptoms reported by women in the presence and absence of infection, including medical, cultural, intellectual, situational and population effects. More severe disease can produce more severe symptoms, i.e. pelvic inflammatory disease versus cervical chlamydia or gonorrhoea. Cultural taboos regarding women touching, viewing or even thinking about their genitalia, much less talking to a stranger about them, can dramatically impact the reporting of symptoms. The woman's knowledge, maturity and degree of understanding of genital anatomy and function also greatly influence the degree of her sophistication in symptom reporting. Prevalence of disease in her community may also have an impact by increasing or decreasing the likelihood that she knows someone else affected similarly and has discussed her situation with them. Prior personal experience with sexually transmitted infections and other genital problems (including genital mutilation, rape or sexual abuse) may also influence perception and reporting of symptoms. In our study population, all women were infected at enrolment, so all the uninfected women (free of infection at six months' follow-up) had a history of prior infection. This would limit the generalizability of these data to women with no prior STD exposure.

Our final goal was to develop a scoring system that would predict infection with gonorrhoea and/ or chlamydia. Using a simple 0-10 point scoring system based on the composite measures, we achieved a sensitivity/specificity balance (both about 63%) at three points. This compares favourably with previously described algorithms based on either symptoms alone (sensitivity 43%, specificity 58%) or symptoms plus STD risk factors (sensitivity 69%, specificity 54%).5,6,8 As noted above, cultural and population differences and varying prevalence of disease make generalization to other populations difficult. Our scoring system, based on a brief series of simple questions, was developed and validated in our population and may require modification, for use in other settings. A simple sensitive/specific test for gonorrhoea and chlamydia could eliminate the need for syndromic management; development of such testing, however, has proven very difficult. Until a rapid, low-cost assay for gonorrhoea and chlamydia becomes available, syndromic management based on signs and symptoms of infection remains the best available diagnostic method for limited-resource health-care settings.

Acknowledgements: This study was supported by grants No. SG 2P01 A122380-403 and No. 5U01 A140029-04 from the National Institute of Allergy and Infectious Diseases.

References

1 Dallabetta G, Behets F, Lule G, et al. Specificity of dysuria and discharge complaints and presence of urethritis in male patients attending an STD clinic in Malawi. Sex Transm Infect 1998;74:S34-7

2 Van Dam CJ, Beeker KM, Ndowa F, Islam MQ. Syndromic approach to STD management: where do we go from here? Sex Transm Infect 1998;74:S175-8

3 World Health Organization. Guidelines for the Management of Sexually Transmitted Infections. Chapter 2, Treatment of STI- associated Syndromes. Geneva: WHO, 2000

4 Shain RN, Piper JM, Newton ER, et al. Behavioral intervention to prevent sexually transmitted disease among minority women: results of a controlled randomized trial. N Engl J Med 1999;340:93- 100

5 Brabin L. Clinical management and prevention of sexually transmitted diseases: a review focusing on women. Acta Tropica 2000;75:53-70

6 Bosu WK. Syndromic management of sexually transmitted diseases: is it rational or scientific? Trop Med Int Health 1999;4:114-19

7 Pettifor A, Walsh J, Wilkins V, Raghunathan P. How effective is Syndromic management of STDs? Sex Transm Dis 2000;27:371-85

8 Mayaud P, Grosskurth H, Changalucha J, et al. Risk assessment and other screening options for gonorrhea and chlamydial infections in women attending rural Tanzanian antenatal clinics. Bull World Health Org 1995;73:621-30

(Accepted 9 October 2003)

Jeanna M Piper MD1, Jeffrey E Korte PhD1, Alan E C Holden MA1, Rochelle N Shain PhD1, Sondra Perdue DrPH2, Jane D Champion PhD3 and Edward R Newton MD4

Departments of 1 Obstetrics and Gynecoiogy, 2 Microbiology, 3 Family Nursing, University of Texas Health Science Center at San Antonio, 7703 Floyd Curl Drive, San Antonio, TX 78229; 4 Department of Obstetrics and Gynecoiogy, East Carolina University, Greenville, NC, USA

Correspondence to: Dr Jeanna M Piper

Email: piperj@uthscsa.edu

Copyright Royal Society of Medicine Press Ltd. Feb 2005

Source: International Journal of STD & AIDS

More News in this Category