Human immunodeficiency virus (HIV)-infected women in India and other developing country settings are living longer on antiretroviral therapy, yet their risk for human papillomavirus (HPV)-induced cervical cancer remains unabated because of lack of cost-effective and accurate secondary prevention methods. Visual inspection after application of dilute acetic acid on the cervix (VIA) has not been adequately studied against the current standard: conventional cervical cytology (Pap smears) among HIV-infected women. We evaluated 303 nonpregnant HIV-infected women in Pune, India, by simultaneous and independent screening with VIA and cervical cytology with disease ascertainment by colposcopy and histopathology. At the cervical intraepithelial neoplasia (CIN2+) disease threshold, the sensitivity, specificity and positive and negative predictive value estimates of VIA were 80, 82.6, 47.6 and 95.4% respectively, compared to 60.5, 59.6, 22.4 and 88.7% for the atypical squamous cells of undetermined significance or severe (ASCUS+) cutoff on cytology, 60.5, 64.6, 24.8 and 89.4% for the low-grade squamous intraepithelial cells or severe (LSIL+) cutoff on cytology and 20.9, 96.0, 50.0 and 86.3% for high-grade squamous intraepithelial lesion or severe (HSIL+) cutoff on cytology. A similar pattern of results was found for women with the presence of carcinogenic HPV-positive CIN2+ disease, as well as for women with CD4+ cell counts <200 and <350 μL−1. Overall, VIA performed better than cytology in this study with biologically rigorous endpoints and without verification bias, suggesting that VIA is a practical and useful alternative or adjunctive screening test for HIV-infected women. Implementing VIA-based screening within HIV/acquired immunodeficiency syndrome care programs may provide an easy and practical means of complementing the highly anticipated low-cost HPV-based rapid screening tests in the near future, thereby contributing to improve program effectiveness of screening.
Recent research has suggested that human papillomavirus (HPV) testing is a superior strategy than the most widely used method, viz., cervical cytology (Pap smears) for age appropriate screening for cervical cancer.1, 2 However, HPV testing is still very costly in most developing countries, which bear more than 80% of the global burden of HPV-induced invasive cervical cancer (ICC).3, 4 Public health and medical professionals worldwide earnestly await the introduction and wide availability of low-cost, rapid and easily implementable HPV testing methods.5, 6 In the interim, however, it is imperative to evaluate and consider implementing cheaper and more practical modalities than cytology for cervical cancer screening, especially for women at high risk of cervical disease.
Human immunodeficiency virus (HIV)-infected women constitute one of the highest risk population subgroups for increased incidence and rapid progression of HPV-induced cervical intraepithelial neoplasia (CIN) and ICC.7 India faces large burdens of HIV/acquired immunodeficiency syndrome (AIDS; third highest number of HIV-infected persons in any single nation in the world, with more than a million HIV-infected women) and ICC (almost 30% of the global burden with more than 130,000 new cases and 75,000 deaths annually).4, 8 Although HIV-infected women in India and elsewhere continue to live longer due to improved access to antiretroviral therapy, most women lack access to adequate cervical cancer prevention services.9
Visual inspection after application of dilute (3–5%) acetic acid on the cervix (VIA) has been widely researched as an alternative to cytology for use in resource-limited settings.10, 11 However, there are very few studies reporting test accuracy of VIA, without verification bias or with biologically salient diagnostic endpoints, among HIV-infected women.12–14 Given this research gap and the need for providing better models for preventive clinical care for HIV-infected women in resource-limited settings, we conducted a cross-sectional study to rigorously evaluate the clinical accuracy of VIA and cytology among HIV-infected women in Pune, India.
Material and Methods
The study protocol, approved by the institutional research ethics committees of the collaborating institutions, has been previously described.15 Briefly, consenting nonpregnant and previously unscreened HIV-infected women were enrolled in this study. Two nurses, trained in a 2-week didactic and hands-on training course on cervical cancer prevention, performed clinical examination. This included collection of samples for cervical cytology and HPV testing and independent conduct and recording of VIA screening test. A positive VIA result implied definite acetowhite lesions visualized as touching or abutting the squamocolumnar junction or if the entire cervix or a growth on the cervix turned acetowhite, as per International Agency for Research on Cancer (IARC)/World Health Organization (WHO) guidelines.16 Two trained gynecologists, blinded to the VIA status, performed standardized diagnostic colposcopy examination. Samples for histopathology were obtained by colposcopically guided cervical punch biopsies, endocervical curettage (ECC) and loop electrosurgical excision procedures (LEEP), as clinically indicated. Histopathology samples were analyzed by two independent pathologists, and consensus diagnoses were reported, as previously described.15 Cervical cytology slides were stained conventionally, interpreted and reported as per revised (2001) Bethesda guidelines by an independent pathologist in a routine clinical laboratory setting. Cytology was considered to be positive at three cutoff levels: presence of atypical squamous cells of undetermined significance or severe (ASC-US+); low-grade squamous intraepithelial cells or severe (LSIL+) and high-grade squamous intraepithelial lesion or severe (HSIL+). HPV DNA testing in cervical cytobrush samples was performed by Digene Hybrid Capture-2™ (HC2) assay (Qiagen, Gaithersburg, MD) as per manufacturer's guidelines for aggregate presence of carcinogenic HPV types.
We compared the accuracy measures of VIA and cytology at the CIN2 or worse (CIN2+) cervical disease state threshold on colposcopy and histopathology (as CIN2 is the most commonly used threshold for diagnostic referral). Sensitivity, specificity, positive predictive values (PPV) and negative predictive values (NPV) and their exact 95% confidence intervals were calculated using standard formulae.17 We also analyzed test accuracy for women having CIN2+ lesions with concurrent presence of carcinogenic HPV (as carcinogenic HPV-negative CIN2+ has very low potential for progression). Secondary analyses were conducted with exclusion of women with missing histopathology results (if either no histopathology was done or histopathology results were unavailable). Similar analyses were repeated at the CIN3+ diagnostic threshold. Thus, our disease evaluation included colposcopic, molecular (carcinogenic HPV testing) and histopathologic endpoints. The role of combination testing of screening tests was evaluated by comparing test performance measures using two scenarios of (i) both VIA and cytology positive and (ii) either VIA or cytology positive (at all three cytology positivity cutoffs) at the CIN2+ threshold for all participants. Data were analyzed using IBM SPSS Statistics v18.0 (SPSS/IBM, Chicago, IL), R statistical software (Version 2.9.1; http://www.r-project.org) and Stata IC 10.1 (Stata Corp LP, College Station, TX).
Between September 2006 and February 2007, a total of 303 eligible, consenting HIV-infected women were enrolled. Their sociodemographic information is summarized in Table 1. VIA results were recorded for all 303 women, 84 (27.7%) were VIA positive and 219 (72.3%) were VIA negative. Cytology results revealed 150 (49.5%) women with normal cytology (including 19 with inflammatory smears), 11 (3.6%) with ASC-US, 87 (28.7%) with LSIL and 18 (5.9%) with HSIL. A total of 37/303 (12.2%) slides were reported as inadequate for evaluation. There were no statistically significant differences between demographic and clinical characteristics of women screening positive by VIA and cytology (all p > 0.05; Table 1).
Table 1. Demographic, sexual/reproductive and clinical characteristics and their distribution for the entire study population and women testing positive by VIA and cytology (ASC-US+, LSIL+ and HSIL+ cutoffs)
Colposcopic–histopathologic diagnoses have been reported previously.15 Briefly, colposcopic impressions served as a final diagnosis in 246/303 (81.2%), whereas histopathology results were available for the rest (57/303, 18.8%). Histopathology was based on LEEP results in 21/57, punch biopsy results in 34/57 and ECC in 2/57 results. The composite colposcopic–histopathologic diagnosis revealed that ICC was present in 1 (0.3%), CIN3 in 18 (5.9%), CIN2 in 31 (10.2%) and CIN1 in 33 (10.9%), whereas no CIN abnormalities were present in 220 (72.6%) of the 303 women. Sixteen women had CIN2+ colposcopic impression but were missing histopathology results (this included two women with inconclusive histopathology results and 14 women for whom no histopathology was obtained) and were excluded from the secondary analysis. Thus, of the 287 women included in the secondary analysis, ICC was present in 1 (0.3%), CIN3 in 11 (3.8%), CIN2 in 22 (7.2%), CIN1 in 33 (10.9%) and no CIN abnormalities were present in 220 (76.7%) women.
Carcinogenic HPV by HC2 assay was present in 124/297 (41.75%) women (HPV test results were unavailable in six women). A total of 19 histopathologically CIN2+ women were carcinogenic HPV negative; thus the prevalence of carcinogenic HPV-positive CIN2+ lesions was 10.23% (31/303). Additionally, carcinogenic HPV was present in 33.3% of women with normal cytology results, as well as 35.8% women with VIA negative results.
The proportion of positive and negative screening test results classified as “true” and “false” against the various disease state thresholds (reference standards) are shown in Table 2. VIA had higher sensitivity than cytology at all cytology positive cutoffs (80 vs. 60.5%, 60.5 and 20.9%, at the cytology ASCUS+, LSIL+ and HSIL+ cutoffs, respectively), although the differences reached statistical significance only between VIA and HSIL+ cytology. On the other hand, the specificity of VIA (82.6%) was higher than cytology at ASC-US+ (59.6%) and LSIL+ (64.6%) cutoffs but lower than cytology at the HSIL+ (96%) cutoff (all p < 0.01; Table 2). The PPV for HSIL+ cutoff cytology was also higher than that of VIA and the two other cutoffs of cytology. On the other hand, the NPV of VIA was greater than cytology at all three cytology positive cutoffs. Similar trends in accuracy estimates were observed in disease states defined by women with CIN2+ with concurrent presence of carcinogenic HPV, as well as at both these thresholds after excluding women without histopathology results (Table 2).
Table 2. Test positive results and measures of screening test performance (individual and in combination) at various disease state (diagnostic) thresholds among HIV-infected women in India
Cytology results in n = 37 of 303 women were “inadequate for evaluation,” hence excluded from this analysis.
The simulated combinations of VIA and cytology (all three cutoffs) with an “either test positive” definition had higher sensitivity, whereas the simulated combination with a “both tests positive” definition had higher specificity among corresponding combination and individual test performance measures. The NPV for the individual tests and both their simulated combinations were uniformly higher than 85% (with many over 95%), whereas their PPV were under 54% (with most under 40%).
Among women with CD4+ cell counts <200 and <350 μL−1, VIA had higher sensitivity than cytology (all three cutoffs), whereas the specificity of HSIL+ cutoff cytology was higher than VIA or positivity on cytology at the ASC-US+ and LSIL+ cutoffs (Fig. 1). When compared to their counterparts with higher CD4+ cell counts, women with lower CD4+ cell counts (CD4 <200 μL−1vs. CD4 ≥200 μL−1, and CD4 <350 μL−1vs. CD4+ ≥350 μL−1) had higher sensitivity but lower specificity for VIA and HSIL+ cutoff-cytology, whereas lower sensitivity and specificity at the ASC-US+ and LSIL+ cutoffs on cytology (Fig. 1).
Test accuracy estimates using CIN3+ and carcinogenic HPV-positive CIN3+ were limited by few analyzable endpoints (i.e., only n = 19 of 303 women had CIN3+, and only n = 15 women had carcinogenic HPV-positive CIN3+). Yet, a similar pattern of VIA having a higher sensitivity and lower specificity in comparison to HSIL+ cutoff on cytology and both higher sensitivity and higher specificity than ASC-US+ and LSIL+ cutoffs on cytology was noted at both disease positivity thresholds, before and after excluding women without histopathology results (Supporting Information).
The independent and simultaneous comparison of VIA and cytology against biologically salient ascertainment of true disease (colposcopically and histopathologically confirmed CIN2+ outcomes, further refined by adding results of carcinogenic HPV testing) has allowed us to compute and directly compare test accuracy while limiting verification bias in this cross-sectional study among HIV-infected women in India.
VIA had higher sensitivity than cytology, although statistical significance was reached only for cytology screen positivity at the HSIL+ cutoff. Cytology had significantly higher specificity than VIA at the HSIL+ cutoff, although VIA had significantly higher specificity compared to cytology at the ASCUS+ and LSIL+ cutoffs. Although multiple studies in resource-limited settings have evaluated and compared VIA and cytology in the general population, most have either not tested or not reported HIV serostatus of the screened population.10 To date, there are only a few studies directly comparing the performance of cytology against alternatives such as VIA among HIV-infected women.12–14 Our study adds to this literature with additional strengths of limiting verification bias and refinements of disease endpoints. Indeed, the higher sensitivity of VIA among HIV-infected women as reported in these studies and in our study may possibly reflect the more pronounced, larger and well-demarcated (and hence easier to identify) acetowhite lesions secondary to local and systemic immune suppression, further supported by our finding of higher sensitivity of VIA in HIV-infected women with lower CD4+ cell counts compared to those with higher CD4+ cell counts.
As a point-of-care clinical test, VIA is easy and inexpensive to perform, can be taught to nonphysician health workers and can link screening and diagnostic or treatment interventions in the same clinic visit. On the other hand, there are significant real-world challenges in implementing cytology-based screening in resource-limited settings. Adequacy of samples is hampered by challenges in sample collection and slide preparation. Even in the hands of highly qualified pathologists, cytological interpretation is precariously modest in its detection rate and sensitivity.18, 19 Besides, the costs and inconvenience of multiple clinic visits and the resulting poor compliance in cytology-based screening programs are significant barriers to successful cervical cancer control for most resource-limited settings.10
Although all women underwent simultaneous and independent diagnostic assessment by colposcopy and when necessary by histopathology,15 there remains a possibility of residual misclassification bias due to potential overcalling in colposcopic impressions and histopathology reporting and thereby misclassification of low grade (CIN1) results as high grade (CIN2 or above; especially as 19 of 50 CIN2+ lesions and 4 of 19 CIN3+ lesions were carcinogenic HPV negative). Yet, this is indicative of an ongoing challenge in diagnostic research on cervical cancer screening tests where balancing the necessity for full disease ascertainment has to be balanced against the ethical necessity of avoiding unnecessary biopsy procedures, especially in the context of immunocompromised HIV-infected women. Our additional analyses to quantify the extent of this bias by exclusion of women without confirmatory histopathology (either not done or had inadequate results) demonstrates that this did not significantly alter the trends in the comparison of test accuracy measures.
While acknowledging the obvious benefits of VIA, it is equally important to highlight its limitations, as it is a rater-dependent subjective test and has challenges with its reproducibility. Although a range of training materials is now available for conducting visual screening competently, the inherent nature of the in vivo testing presents logistical challenges for standardizing quality assurance procedures for visual screening to ensure consistent high performance under field conditions.20 Furthermore, it requires constant monitoring and frequent retraining of test providers.21 Not surprisingly, the sensitivity of VIA has ranged widely between 41 and 79% in multiple studies worldwide, whereas specificity has varied widely between 49 and 86% (both ranges overlap with the 95% confidence intervals of the corresponding estimates of VIA in our study).10, 22
Our results need replication in other settings and if proven generalizable will have important programmatic and policy implications. Both international (UNAIDS/WHO) and Indian national guidelines for HIV/AIDS care and treatment continue to recommended screening for HIV-infected women with cytology. Yet, as shown in our study, one-time screening with cytology is limited in its sensitivity and detection rates of high-grade cervical neoplasia. Additionally, the fraction of women who receive “inadequate for evaluation” results is nontrivial (12% in our study). This necessitates a repeat cytology and another return visit, which is often a drain on resources and time for the patient and the healthcare system. On the other hand, in addition to being optimally sensitive, VIA can also link screening results to future diagnostic or treatment approaches during the same patient visit, thereby providing better clinical and programmatic outcomes.
Currently, cytology continues to be used in many centers in India as the only method of screening. In settings where cytology is available, modest investments through training nurses/clinicians in conducting VIA as an adjunct method and restricting cytology screening to VIA-negative women (like the “either test positive” simulation results in our study; Table 2) may substantially improve the programmatic sensitivity with cost savings. In other severely resource-constrained settings with limited or no availability of cytology or histopathology, a VIA-only screening approach will still provide optimal outcomes, where positive VIA can be followed by same-visit cryotherapy for treatment of eligible precancerous lesions, without diagnostic confirmation by histopathology.11, 23, 24
We believe that it is important to take advantage of the unique attributes of the VIA screening approach that can provide a suitable clinical care platform for implementing the eventual roll-out of highly anticipated low-cost, rapid and highly sensitive HPV screening assays.5 Future algorithms that use VIA to triage HPV-positive women for either referral for physician evaluation (extensive lesions) or undergoing same-visit cryotherapy (for eligible lesions) could maximize accuracy of screening and effectiveness of treatment intervention.25 In the interim, however, national programs should consider incorporating VIA-based screening as an independent (where no other screening tests exist) or adjunctive (where cytology is available) approach, depending on setting-specific availability and requirements. The importance of operational and implementation research to develop and implement such guidelines for integrating cost-effective cervical cancer screening methods in clinical care for HIV-infected women cannot be overstated.
The authors thank Dr. Groesbeck Parham, Dr. Robert Bollinger and Dr. Eric Chamot for advice and guidance for protocol development; Dr. Philip Castle for comments on a previous version of the manuscript; Dr. Santosh Jaybhaye and Dr. Nayana Dhage for clinical care of participants and Mr. Avinash Gaikwad for data management support. They also thank all HIV-infected women who participated in this study.