WHAT IS ALREADY KNOWN ON THIS TOPICWHAT THIS STUDY ADDS

As a triage method, visual inspection with acetic acid (VIA) with iodine was better at diagnosing cervical intraepithelial neoplasia (CIN2+), but not significantly better than VIA alone. Being HIV positive was a superior predictor of CIN2+ than VIA with or without iodine.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICYIntroduction

When introducing human papilloma virus (HPV)-based screening for cervical cancer, the management of those who are HPV positive arises as a new obstacle for the research community.1 The service providers working in high HPV prevalence areas are faced with many testing positive during primary screening. It is known most will heal from their HPV infection, and the high sensitivity of HPV DNA tests comes at the expense of a somewhat lower specificity.1 Thus, an ideal triage test would identify women who are HPV positive and have a pre-cancerous lesion. It would also help to stratify who is in need of follow-up and/or treatment, keeping specificity high, without hampering sensitivity of the primary screening test to a significant extent.

Cytology (including dual stain) as a triage test has had success primarily in high-income countries, but is resource heavy and subjective.1 Visual inspection with acetic acid (VIA) or with Lugol’s iodine requires a pelvic exam and a trained professional and has been implemented in many low-middle income countries as a primary screening method.2 Pre-cancerous cells have high protein content, which acetic acid stains white, and low glycogen content, which iodine stains yellow. The use of VIA as a triage method was demonstrated by Bigoni et al 3 with a sensitivity of 25% compared with cytology (sensitivity 90%). Specificity remained high for both methods, 74.2% and 85.2%, respectively. For co-testing (VIA with iodine), one study found a very high sensitivity of 80.8%, but a low specificity of 31.2%.4

When selecting a triage strategy, it is important to consider logistic and cost factors, especially in a low-resource setting.5 VIA, because of the above, is one of the recommended triage tests by the World Health Organization (WHO). However, most studies on its effectiveness were based on it being a primary screen-and-treat test.6 The accuracy of VIA as a triage test has therefore been questioned.3 7 One limitation of previous studies is that pathology results were not available for all women going through triage, hence inflating sensitivity. Moreover, the use of combining VIA with Lugol’s iodine has not been compared with only VIA in a triage setting.

The aim of this study was to examine the sensitivity and specificity of VIA with and without Lugol’s iodine as a triage test for women who tested HPV positive in a self-sample. We hypothesized that the diagnostic accuracy of VIA would improve with the addition of Lugol’s iodine.

MethodsStudy design

This two-armed randomized controlled trial (RCT) was based on a prospective cohort established in Adama, Ethiopia between 2015 and 2018.8 Women were recruited from local antenatal clinics and followed by Adama Lund Research Station (ALURS), a joint venture between Lund University and Armauer Hansen Research Institute (AHRI) in Addis Abeba, Ethiopia. The study was initiated in December 2021 and the last follow-up was done in October 2023. Study data were collected and managed using REDCap electronic data capture tools, hosted by Lund University servers.9 10 Clinical trials ID: NCT05125380.

Ethical considerations

The study was approved by the institutional review board (IRB) of AHRI (ref number 005437/0014) and the National Research Ethical Review Committee (NRERC) in Ethiopia (14.2/8328/21).

In line with the Declaration of Helsinki, all women received oral, visual and written information about the study before accepting to be part of the trial through a written consent form.

Study participants

Women from the ALURS cohort8 were recruited to this trial at two health centers, Geda and Adama. Inclusion criteria were women older than 18 years, including those who were pregnant. Exclusion criteria were previous hysterectomy or ongoing treatment for dysplasia of the cervix.

Data collection

The women first watched a short animation that was created by the research team together with a local production company (shared on request). The animation narrated in either Amharic or Afaan Oromo language how HPV is related to cervical cancer, the importance of preventative screening and how to obtain a self-sample from the vagina. The research nurses then answered any remaining questions the women might have and asked questions from a short questionnaire in RedCap (supplement). Finally, they brought up consent to participate in the trial. Women who gave written informed consent each received two eNAT swabs (COPAN, USA) and then went to the bathroom for self-sampling from the vagina.

The self-samples received barcodes before being sent to the HPV lab at Tikur Anbessa Teaching Hospital (TASH), Addis Abeba. The self-samples first underwent DNA extraction using the Nimbus system, and then through a quantitative polymerase chain reaction (qPCR) technique, HPV was detected. Each sample was analyzed for 14 high-risk genotypes using Anyplex™ II HPV HR Detection kits (Seegene, South Korea).

Women who tested positive for any of the high-risk HPV types (16, 18, 31, 33 35, 39, 45, 51, 52, 56, 58, 59, 66, 68) were called and booked in for follow-up at the local public hospital (Adama Medical College Hospital, AMCH). The same women were stratified according to age and pregnancy status, and then randomized 1: 1 to VIA with or without iodine using the RedCap tool for randomization. The women were blinded to the allocation test, but not the midwife or the gynecologist. Because of a technical error with the RedCap application, three women were randomized manually using a random integer generator online (https://www.random.org/integers/). Four blocks were created in order to stratify according to age and pregnancy status.

One experienced midwife with a certificate of WHO training for trainers for cervical cancer screening, who had facilitated more than 40 rounds of basic training and more than five training of trainers, saw all the women who tested HPV positive after self-sampling. The women were examined in the lithotomy position with a metal speculum. A cervical HPV test was taken before acetic acid was sprayed on the cervix. One minute was timed before making an assessment. If randomized to VIA with iodine, iodine was applied after the acetic acid. The midwife judged the triage test as being positive, negative, inconclusive, or with suspicion of cancer according to WHO training.11 A positive test for visual inspection with Lugol’s iodine meant any well defined yellow area in the transformation zone, whereas staining brown was considered a negative test.12 A cervical punch biopsy was taken either from the positive lesion or a random biopsy from the 12 and 6 o’clock position by a gynecologist. All women who were positive at triage were treated either with cryotherapy or thermal ablation. The cervical specimen was transported in 5% formalin solution to the ONCO Pathology Clinic, PLC in Addis Abeba. Assessment was done by a pathologist with experience in gynecological pathology. If the biopsy revealed cervical intraepithelial neoplasia (CIN) 2 or more, but had not been diagnosed at triage, the women were re-appointed for either loop electrosurgical excision procedure (LEEP), or referred to a gynecological oncology clinic for treatment.

In accordance with the journal’s guidelines, we will provide our data for independent analysis by a team selected by the Editorial Team for the purposes of additional data analysis or for the reproducibility of this study in other centers if such is requested.

Statistical methods

Descriptive statistics were used for all the participants’ characteristics, but also to compare the women who were HPV positive in each randomized arm. χ2 testing was used to compare the two arms. Using Fleming’s one stage procedure (type I error, α=0.05 and power=0.8) to achieve in a randomized trial, a sensitivity of >80% of VIA with iodine that would be significantly superior to 60% for VIA, 82 patients would have to be recruited in each arm, a total of 164 patients. The sensitivity and specificity for each triage method was calculated using cross-tabulation. For each HPV positive study participant, the outcome of the triage method and the biopsy result was added. CIN2, 3, carcinoma and endocervical glandular atypia were grouped and considered a positive biopsy for the calculations of sensitivity and specificity. Calculations were done according to per protocol and intention to treat.

If the women were pregnant or had no biopsy taken, they were excluded from the diagnostic accuracy calculation. Histopathology was gold standard . For hypothesis testing, confidence intervals were calculated. Area under the curve (AUC) and receiver operating curve (ROC) were calculated for each arm and compared. The diagnostic accuracy was calculated according to per protocol and intention to treat. Logistics regression was used to calculate OR for proportions. Statistics were done using STATA statistical software.13

Results

A total of 878 women participated in the study (Figure 1), and a majority 64.1% (n=648) had previously heard about cervical cancer screening. From those women, 17.5% had previously been screened. Baseline patient characteristics are presented in Table 1. The proportion of women who were HPV positive was 22.4% (197/878). The median age was 32 years (range 22–47). Of these women, 146 came for follow-up (74.1%). The median time to follow-up was 5 months (range 1–22). Most women had cleared their HPV infection at follow-up, 53.5% (n=38) in the VIA arm and 53.6% (n=37) in the VIA with iodine arm. The two groups did not have any major differences in participant characteristics (Table 2). Cross-tabulation for VIA and VIA with iodine are presented separately in Table 3 with the total number of women who came for follow-up and had cervical biopsies taken according to per protocol.

Figure 1

Consort 2010 flow diagram.

Table 1

(A) Participants’ demographics. (B) Characteristics of women who are HPV DNA positive by randomization arm

Table 3

Cross-tabulation for VIA with and without iodine and their respective triage results (according to per protocol)

Sensitivity and specificity done per protocol for VIA was 25.0% (1/4) (95% CI 0.6 to 80.6) and 82.7% (43/52) (95% CI 69.7 to 91.8), respectively. Whereas for VIA with iodine the sensitivity was 50.0% (2/4) (95% CI 0.7 to 93.2) and specificity 84.6% (44/52) (95% CI 71.9 to 93.1). The prevalence of disease was 7.1% in each arm. Sensitivity and specificity results from the per protocol and intention to treat calculations are presented in Online supplemental tables 4 and 5. The AUC for VIA was 0.52 (95% CI 0.41 to 0.62), and for VIA with iodine 0.57 (95% CI 0.44 to 0.71), p=0.5 Figure 2. The null hypothesis was not rejected

Figure 2

Area under the curve for visual inspection with acetic acid (VIA) with and without iodine (VILI). ROC, receiver operating curve.

The odds of detecting CIN2+ were 5.5 times higher (95% CI 0.67 to 44.9, p=0.1) if the woman tested positive for VIA with iodine (2/8 vs 2/44), whereas the odds were 1.6 higher (95% CI 0.15 to 17.1, p=0.7) if testing positive for VIA without iodine (1/9 vs 3/43). Adjusting for HIV status in both arms, the odds of detecting CIN2+ were 4.0 times higher (95% CI 0.4 to 36.0, p=0.2) in the VIA with iodine arm, and 1.4 higher (95% CI 0.1 to 17.2, p=0.8) in the VIA alone arm.

The odds of detecting CIN2+ in women who were HIV positive were 5.1 times higher (95% CI 1.2 to 22.5) than in those who were HIV negative (p=0.03).

DiscussionSummary of main results

This is, to the best of our knowledge, the first randomized trial comparing VIA to VIA with iodine (co-testing) as a triage method for women who are high-rsik HPV DNA positive in a self-sample. VIA with iodine was more sensitive than VIA alone, but not significantly better than VIA without iodine. The OR shows that the probability of detecting disease (CIN2+) using only acetic acid is low. In fact, being HIV positive was more predictive of having disease (CIN2+). Moreover, part of the association seen between the triage test and biopsy result is related to HIV status, as can be seen in the adjusted OR. HIV status had more effect on outcome in the VIA only arm. On the other hand, in the VIA+iodine arm, the odds of having dysplasia when positive in the triage test was still four times more likely, even when adjusting for the patient’s HIV status.

Results in the context of published literature

VIA as a triage test has been evaluated in Cameroon, where the sensitivity for CIN2+ was 34.4% and specificity 87.4%.14 In Tanzania, VIA as a triage test reached a sensitivity of 65.6% and a specificity of 75.5%.15 For co-testing, one study found a very high sensitivity of 80.8%, but a low specificity of 31.2%.4 However, the co-testing study used a lower threshold than original protocols for judging an exam as positive, by describing an ABCD approach.4 The ABCD approach, where A stands for acetowhite, B for bleeding on touch, C for coloring with Lugol’s iodine, D for diameter, seems to give an improved sensitivity but with more women being overtreated.16 Similarly, the use of point-of-care digital imaging of the cervix together with co-testing (VIA with iodine) for triage of women testing HPV positive achieved a high sensitivity of 79.5% but lower specificity of 43.4%.17

The characteristics of participants coming for triage affects the prevalence of the disease. Being multipara increases the risk of cervical cancer,18 and HIV positivity increased the odds of having CIN2+. Even though sensitivity and specificity are treated as being independent from prevalence, unlike positive predictive value and negative predictive value. A large meta-analysis demonstrated a positive correlation between sensitivity and prevalence,19 and negative correlation between specificity and prevalence. Therefore, in our results, the low prevalence of disease may be one reason for the calculated low sensitivity.

Strengths and weaknesses

In our study one examiner performed all triage examinations, minimizing the inter-observer variability but reducing the reproducibility. Another strength was to test HPV in the cervix at follow-up. The results show that about half of those positive in a self-test have cleared the infection or may never have had it in the cervix. Previous knowledge confirms that HPV more commonly clears than persists.20 Nevertheless, one WHO screening strategy is that all women who are HPV positive receive treatment.2 This HPV test and treat strategy may however pose a major issue for follow-up after treatment, associated costs for treatment, and reproductive consequences since many low-income and middle-income countries are dealing with a high HPV prevalence. In our study, HIV status was more predictive of disease than both triage tests. Genotyping high-risk HPV for women who are HIV positive and treating them if positive for genotypes 16, 18, 31, 33, 45, 52 or 58 gave in one study a sensitivity and specificity for CIN2+ of 94.7% and 77.3%, respectively.21

The low prevalence of disease and difficulties reaching women for follow-up were reasons for not reaching statistical power. We could not increase statistical power by including more women in the study because the ethical approval was only sought for women from the ALURS cohort.8 Another major limitation was the time from primary screening to follow-up, which ranged from 1 to 22 months. When women were called back, many phones did not work because of either network issues or they had lost their phones. During the long time to follow-up many healed from their HPV infection and some dysplasias may have regressed as well.22 It is known that a majority of CIN2 lesions resolve spontaneously23 and this may be the cause for the low prevalence of disease.

Certain problems with VIA and co-testing are essentially the same as with colposcopy. They are all subjective methods, and not possible to evaluate with a transformation zone type 3 (not visible transformation zone),24 which occurred in 4% of patients. This was a limitation in relation to accurate histopathological diagnosis, as we did not perform cervical curettage nor conization for them. If still testing HPV positive at follow-up, these women were scheduled to be re-screened after 1 year.

Implications for practice and future research

Prevalence of cervical cancer as well as risk factors such as HIV need to be taken into account when implementing a triage strategy in different geographical areas. A larger sample size for future studies would be needed to clearly know the added value of iodine to VIA. However, since most HPV infections resolve spontaneously, our recommendation is that despite its low sensitivity, VIA with or without iodine is used as triage for the general population until future research that focuses on triage strategies such as DNA methylation is cost-effective. Methylation triage would take into consideration the potential of progression, not simply the presence of dysplasia. This would decrease the need for treatment, lessen the individual burden for women traveling far for follow-up, as well as its associated psychosocial effects. It would also be efficient for women with a transformation zone type 3.25

HPV genotyping and treatment of women who are HIV positive can be implemented and studies focusing on scaling up screening and treatment through HIV clinics are necessary. The HPV genotypes to include in triage should be tailored to each region. This will be possible if each country comits to implementing databases for the screening system.

Conclusion

VIA with iodine improved detection of CIN2+ in our sample of women who tested positive for HPV DNA through self-sampling but was not significantly better than VIA without iodine.

Data availability statement

Data are available upon reasonable request.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

This study involves human participants and was approved by the institutional review board (IRB) of Armauer Hansen Research Institute, ref number 005437/0014, National research ethical review committee (NRERC) in Ethiopia, 14.2/8328/21. Participants gave informed consent to participate in the study before taking part.

Acknowledgments

Takele Deressa Negera, midwife and trainer of trainers instructor at Adama Medical College Hospital: for his efforts in improving women’s health in Ethiopia. Dr Mesfin Assefa at ONCO Pathology diagnostic center in Addis Aebeba, Ethiopia: for gracious work with pathology specimens. Pär-Ola Bendahl PhD at Lund University, Sweden: for help with statistical queries.