Introduction

Cervical cancer is the fourth most common cancer among women in the world and the second most common in low income countries (LICs) [1]. In 2012, it was estimated that 580,000 new cases were diagnosed globally, of which 85% were from LICs [1,2]. With an annual incidence rate of 34.8 per 100.000 women and a mortality rate of 22 per 100.000 women, sub-Saharan Africa is the region with the highest burden of cervical cancer [3]. The corresponding figures for North America are 7.5 per 100,000 being diagnosed and 2.3 per 100.000 dying from the disease [4].

Cervical high-grade lesions and cervical cancer are caused by persistent infection with human papillomavirus (HPV). It has been shown that immunocompromised individuals who are positive to human immunodeficiency virus (HIV), have an increased risk of HPV infection [5]. Furthermore, a higher HPV prevalence has been documented in women with decreasing CD4 count [6]. Finally, studies have shown that HIV positive women are at increased risk of cervical cancer. Cervical cancer is preventable, and curable if detected at an early stages. In high income countries (HICs), screening against cervical cancer using cervical cytology has reduced the incidence of cervical cancer with more than 50% during the past 40 years [7]. In contrast, there has been observed no decline or only little decline in LICs [8]. Moreover in sub-Saharan Africa, it appears as the incidence has increased during the past years [9,10]. The high incidence of cervical cancer in sub Saharan Africa reflects a low coverage of cervical cancer screening as well as high prevalence of HPV and HIV [11].

Visual inspection with acetic acid (VIA) is used as a primary screening test in many LICs since it is cheap and easily accessible. One of the drawbacks of VIA is the subjectivity of the diagnosis. This is reflected in, a meta-analysis that reported a sensitivity range between 41% and 92% for the detection of high-grade cervical lesions [12]. In an increasing number of HICs, HPV DNA testing is used as the primary screening method followed by e.g. cytology as a triage test. HPV DNA testing has good sensitivity but the specificity is not as optimal due to the occurrence and detection of transient infections with no concomitant cervical lesions. However, most HPV DNA tests such as Hybrid Capture 2 (HC2) are expensive and require sophisticated infrastructure and qualified laboratory staff, which makes them unsuitable for most LICs. As a response, efforts have been made to manufacture less expensive and simple HPV DNA tests such as the careHPV test, which can be used at low costs, with minimal expertise and at the point of care [13]. The performance of careHPV has been evaluated in a recent meta-analysis and the test was found to have good sensitivity for the detection of CIN2+ and CIN3+ [14]. Based on these encouraging findings, leading health agencies, including the WHO, are increasingly recommending that simple and cheap HPV testing is used as primary screening modality in LICs followed by VIA [15,16]. While the performance of careHPV has been evaluated in specific focused studies in a few selected countries, it is not yet established how the test perform in a routine screening setting in a Tanzanian context.

The aim of this study was to examine the performance of careHPV, HC2, and VIA, the standard screening method in Tanzania, for detection of cytologically diagnosed cervical high grade lesions or cancer, overall and according to age and HIV status. Furthermore, we examined the value of VIA as a triage strategy following careHPV testing.

Materials and method

Results

In all, 4080 women were enrolled in the CONCEPT project. A total of 37 women were excluded from the analysis due to incomplete questionnaire or registration forms (n = 14), lack of cervical specimens (n = 17) and invalid study number (n = 6). We also excluded 26 women where the cytology result was lacking, and 377 with no matched careHPV and HC2 results, leaving 3640 women for analysis (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Schematic flow chart of the study population.

https://doi.org/10.1371/journal.pone.0218559.g001

Socio-demographic characteristics of the study population are displayed in Table 1. The mean age of the study women was 40 years (±4sd). Most women were married or cohabiting (71.1%), and the majority (64.7%) had attended primary school. Of the 3640 women included in this paper, 3209 (88.1%) had normal cytology, 299 (8.2%) had ASCUS/LSIL (of these 66% were ASCUS), and 135 (3.7%) had a cytological diagnosis of HSIL+.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Socio-demographic and cytology characteristics of the study participants (N = 3640).

https://doi.org/10.1371/journal.pone.0218559.t001

Among all women, 23.6% (859/3640) tested positive for careHPV, 19.1% (696/3640) tested positive for HC2 and 6.3% (231/3640) tested positive for VIA. For all three methods, the prevalence of test positive women decreased with increasing age. The proportion of women positive to respectively careHPV, HC2 or VIA was higher among HIV positive women, with positivity rates of 36.4%, 33.8% and 10.6, respectively. The corresponding figures among HIV negative women were 20.0%, 15.9% and 5.4% respectively. Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of careHPV, Hybrid Capture 2 (HC2), and visual inspection with aceto acid (VIA) in detection of cervical high-grade squamous intraepithelial neoplasia or cancer (HSIL+) among all women, and stratified by age, and HIV status.

https://doi.org/10.1371/journal.pone.0218559.t002

The test performance of careHPV, HC2 and VIA in detecting HSIL+ lesions are shown in Table 2. Women in the age group 25–29 showed slightly higher sensitivity of both for CareHPV and HC2compared to age group 30–34, but a lower sensitivity of VIA. However, there was no significant differences in sensitivity in the overall analysis between the two groups 25–64 years and 30–64 years old women. Among all women, careHPV had a sensitivity of 88.9% (95% CI: 82.4–93.9) and a specificity of 78.9% (95% CI: 77.5–80.3). The HC2 test had a slightly higher sensitivity of 91.1% (95% CI: 85.0–95.3) and a slightly higher specificity of 83.7% (95% CI: 82.4–84.9). In contrast to these findings, VIA was found to have a low sensitivity of 31.1% (95% CI: 23.4–39.6) but a comparatively higher specificity of 94.6% (95% CI: 93.8–95.3). The sensitivity of both careHPV, HC2, and VIA tended to decrease with increasing age. For careHPV, and HC2, the specificity increased with older age, whereas for VIA the specificity was constantly high. Among HIV positive women, the sensitivity of careHPV (92.4%), and HC2 (95.5%) was higher than among HIV negative women (85.5%), and (87.0%), whereas the sensitivity did not vary according to HIV status for VIA. In contrast, specificity was lower among HIV positive women than among HIV negative women for all three tests.

We also examined the performance of VIA as a triage test among the 859 careHPV positive women (Fig 2). In all, 98 women were both careHPV positive and VIA positive, among these 37 women (38.9%) had HSIL+, 38 (38.8%) had normal cytology, 23 (23.4%) had an ASCUS/LSIL diagnosis. Altogether 761 women were careHPV positive but VIA negative, and among these, 523 women (68.7%) had normal cytology and 155 (20.4%) had ASCUS/LSIL, whereas 83 women (10.9%) had HSIL+.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Overview of women’s cytology results distribution according to VIA results among careHPV positive.

https://doi.org/10.1371/journal.pone.0218559.g002

Discussion

In more than 3600 women from a Sub-Saharan country (Tanzania), we evaluated test performance of careHPV, HC2 and VIA to detect cervical high-grade lesions with expert reviewed liquid-based cytology as the gold standard. Compared to HC2, careHPV had an only slightly lower sensitivity. In contrast, the test performance of VIA was poor. When using VIA as a triage test among careHPV positive women, the sensitivity decreased even further, whereas the specificity increased substantially.

Screening is an essential component in the prevention of cervical cancer, however, in LICs conventional screening approaches are hampered by complex infrastructure and lack of expert staff. The World Health Organization therefore advocates a policy of 'screen and treat' for cervical screening in LICs. This involves VIA followed by cervical cryotherapy or LEEP if indicated. The performance of VIA as a primary screening tool for the detection of cervical pre-cancer and cancer has, however, shown rather inconsistent results. It is therefore increasingly being questioned whether VIA can stand alone as a scaled up screening procedure due to its high inter-operator variability and low sensitivity [22], and it has been suggested that HPV testing can be provided at point of care in LICs.

In the present study, the sensitivity of careHPV (88.9%) was slightly lower than that of HC2 (91%) in detecting HSIL+, and a similar picture was found for the specificity (80% for careHPV and 84% for HC2). This level of accuracy is similar to that from previous studies in Africa and other low resource settings [14,23,24] and is only slightly lower than in a recent study of more than 7,500 women from rural China [25]. The finding of the present study is promising, especially when keeping in mind that careHPV analyses were done at the point of care in a LIC by newly trained and inexperienced laboratory technicians, while HC2 analyses were performed in a sophisticated environment with experienced laboratory staff in a HIC.

In contrast, we found a poor sensitivity for VIA (~31%) to detect high-grade cervical precancer lesions; VIA failed to detect 97 out of 135 HSIL+ cases, and among women with a positive VIA test, only 18.2% had HSIL+ on cytology. These VIA results are similar to a recent study that we performed among more than 3000 Tanzanian women (sensitivity: 28%) [14] and in some other studies [24,26] but the sensitivity is lower than that from a recent meta-analysis reporting a pooled sensitivity estimate of 82% for detection of CIN2+ [27]. The low VIA sensitivity in our study may reflect the nature of the test itself, which is subjective and related to the competence of health provider. The health providers involved in this study were experienced and had refresher training just before implementing the study, however, the VIA test was still found to have a low sensitivity in our hands. The findings from the present study show that careHPV testing is efficient as a primary screening method for cervical cancer in Tanzania. However, the test cannot stand alone and HPV positive women should be offered triage. This information is important for Tanzanian policymakers and will help them in their efforts in better controlling cervical cancer.

The sensitivity varied with age for all three tests, and was lower in the oldest age group (55–64 years) (~82% for careHPV and HC2; 0% for VIA) than in the youngest (25–34 years) (~95% for careHPV and HC2; 38% for VIA). This is in line with some studies [14,28] but not all [29]. For VIA this could be due to the fact that as women age there is a tendency of the transformation zone to move inside the cervical channel, making it more difficult to visualize which may imply a decrease in sensitivity of the test. This underlines that one of the limitations of VIA is the low sensitivity, especially for older women [22,30]. The currently available results regarding the test performance of careHPV in relation to age are equivocal as some find an association with age and cut off value [31], whereas others do not [24], so more studies are needed to solve this. The sensitivity of both tests was higher among HIV positive women compared to HIV negative. Higher HPV positivity is noted among HIV positive women probably due to immune suppression that causes HPV to infect and grow easily. These findings are in line with other studies from LICs [14,32].

Although in our study careHPV was a little less sensitive than HC2, it performed substantially better than VIA, and would have treated the majority of HSIL+ cases. However, at the same time a large proportion of careHPV positive women had normal cytology, or ASCUS/LSIL lesions that in the majority of cases will regress. Consequently, there is a need for a triage test to minimize referral and overtreatment. When we applied VIA as a triage strategy, we substantially increased the positive predictive value and specificity, but at the cost of sensitivity, that dropped to an unacceptably low level (27%). Our results are in agreement with a recent, large study in a low-resource setting in China [24]. As performance of VIA is so variable in different settings, one solution does not seem to fit all, and in areas with poor VIA performance, alternative triage strategies such as different cut off value on the HPV test or use of other biomarkers should be considered. Finally, in contrast to VIA, careHPV testing allows for homebased testing which may help overcome many of the personal and system-level screening barriers, which are especially pronounced in low income countries.

This study has several strengths. First, it has a large sample size making it possible to conduct stratified analyses according to age and HIV status. In addition, the study was performed in both an urban and a rural area, which enhances the representativity of the study population. We used presence of HSIL+ as threshold, since lower grade lesions are common abnormal cytological findings which usually regress spontaneously. Hence if women with lower grade lesions are considered as test positive it would lead to a high number of false positive findings and risk of overtreatment and treatment complications. In low-income countries, such an approach will have great negative impact on the scarce resources available. We did not have access to histological diagnosis and had to rely on a cytological diagnosis of HSIL+ as gold standard. This could be considered a limitation, as it may be argued that our findings are not fully comparable with studies where the diagnosis has been histologically verified. Consequently, our results should be interpreted with caution.

VIA is an inexpensive and simple test, but it is subjective and highly dependent of the skills and experience of the user, and therefore the accuracy is not optimal and inter-observer variability is high. Still, the implementation of VIA in many LICs may have paved the way for a screening infrastructure. However, the poor performance of the VIA test in some settings may indicate a need for rethinking its use in the future. As VIA mostly has a good specificity, it has been suggested that screen positive women subsequently could be triaged by VIA. However, if such an approach is to be successful, it requires that the health staff undergo thorough and continued training in how to interpret VIA results. With the increasing use of the Internet and smart-phones also in low-resource areas, a future solution for improved cervical cancer screening could be to consider computer assisted visual evaluation such as digital cervicography where pictures of the VIA test are shared with different experts for 2^nd opinion to optimize quality. This program has been implemented elsewhere and has shown promising results for rapid scale up of cervical cancer screening [33,34]. Another option is the use of telepathology where cytologies/biopsies are obtained among HPV positive women, processed into slides and stained. The pictures are subsequently shared via the Internet [35].

In conclusion, our results show that the careHPV and HC2 tests had high sensitivity whereas VIA had much lower sensitivity in detecting cytologically diagnosed high-grade lesions or cancer (HSIL+). Based on our findings, implementation of an HPV-based test such as careHPV as a primary screening test should be considered in Tanzania. An additional advantage of introducing careHPV testing is the possibility of obtaining self-collected specimens, which makes it possible to reach more women with the screening program and focus resources only on those who test HPV positive. However, to improve the specificity and positive predictive value of HPV-based screening tests, it is necessary to have a secondary test in screen positive women. Moreover, it is important that the triage test does not significantly lower the sensitivity of the primary test. Finally, in areas with high prevalence of HIV positive women, it is important that the performance of the tests is as good in HIV positive women as in HIV negative women.

Supporting information

Acknowledgments

We thank the nursing staff at ORCI, KCMC and Mawenzi hospital cervical cancer screening clinics for their assistance and devotion in performing the interviews and obtaining cervical samples from the women. We also thank all PhD students in the CONCEPT project for the coordination and smooth running of the project.