Publications

Cervical cancer remains the leading cause of cancer death among women in sub-Saharan Africa and is more severe in high HIV-burdened countries due to persistent high-risk human papillomavirus (hrHPV). In 2021, the World Health Organization recommended primary hrHPV testing for cervical cancer screening; however, optimal triage strategies following positive hrHPV tests remain unclear. We conducted a prospective cost analysis of triage methods for positive hrHPV results among women living with and without HIV in Gaborone, Botswana. We used a micro-costing approach from the perspective of the healthcare provider. The main outcomes were the implementation costs associated with three triage strategies following hrHPV testing: 8-type HPV genotype restriction, visual inspection with acetic acid (VIA), and colposcopy. We also compared the strategies by measuring the change in costs divided by the change in number of true cases of cervical intraepithelial neoplasia (CIN) 2 or worse (CIN2+) identified, based on the results of a prospective cohort study. Results indicated that the 8-type HPV genotype restriction strategy was the most cost-efficient, requiring no additional costs beyond hrHPV testing and identifying the highest number of true CIN2 + cases. VIA and colposcopy triage identified fewer true cases of CIN2+ and incurred additional costs, with colposcopy being the most expensive. Results were consistent in women with and without HIV. Sensitivity analysis highlighted personnel and hrHPV test kit cartridge costs as significant drivers of overall costs. Post-hoc analysis incorporating average treatment costs for precancer demonstrated that genotyping remained dominant at lower treatment costs but became less favorable as treatment costs increased. We found that 8-type genotype restriction was optimal compared to hrHPV screening combined with VIA or colposcopy. Cost estimates can inform future studies that examine the long-term costs and health outcomes of HPV-based two-stage screening algorithms.

OBJECTIVE: Guidelines for effective triage following positive primary high-risk human papillomavirus (HPV) screening in low- and middle-income countries with high human immunodeficiency virus (HIV)-prevalence have not previously been established. In the present study, we evaluated the performance of three triage methods for positive HPV results in women living with HIV (WLHIV) and without HIV in Botswana.

METHODS: We conducted baseline enrollment of a prospective cohort study from February 2021 to August 2022 in South-East District, Botswana. Non-pregnant women aged 25 or older with an intact cervix and no prior diagnosis of cervical cancer were systematically consented for enrollment, with enrichment of the cohort for WLHIV. Those who consented completed a questionnaire and then collected vaginal self-samples for HPV testing. Primary HPV testing for 15 individual genotypes was conducted using Atila AmpFire® HPV assay. Those with positive HPV results returned for a triage visit where all underwent visual inspection with acetic acid (VIA), colposcopy, and biopsy. Triage strategies with VIA, colposcopy and 8-type HPV genotype restriction (16/18/31/33/35/45/52/58), separately and in combination, were compared using histopathology as the gold standard in diagnosing cervical intraepithelial neoplasia (CIN) 2 or worse (CIN2+).

RESULTS: Among 2969 women enrolled, 1480 (50%) tested HPV positive. The cohort included 1478 (50%) WLHIV; 99% were virologically suppressed after a mean of 8 years on antiretroviral therapy. In total, 1269 (86%) women had histopathology data for analysis. Among WLHIV who tested positive for HPV, 131 (19%) of 688 had CIN2+ compared with 71 (12%) of 581 in women without HIV. Screening by 8-type HPV genotype restriction was more sensitive as triage to detect CIN2+ in WLHIV 87.79% (95% CI: 80.92-92.85) and women without HIV 85.92% (95% CI: 75.62-93.03) when compared with VIA (WLHIV 62.31% [95% CI: 53.39-70.65], women without HIV 44.29% [95% CI: 32.41-56.66]) and colposcopy (WLHIV 70.77% [95% CI: 62.15-78.41], women without HIV 45.71% [95% CI: 33.74-58.06]). However, 8-type HPV genotype restriction had low specificity in WLHIV of 30.88% (95% CI: 27.06-34.90) and women without HIV 37.06% (95% CI: 32.85-41.41). These results were similar when CIN3+ was used as the outcome. When combining 8-type HPV genotype restriction with VIA as the triage strategy, there was improved specificity to detect CIN2+ in WLHIV of 81.65% (95% CI: 78.18-84.79) but dramatically reduced sensitivity of 56.15% (95% CI: 47.18-64.84).

CONCLUSIONS: Eight-type HPV genotype restriction is a promising component of effective triage for HPV positivity. However, novel triage strategies in LMICs with high HIV prevalence may be needed to avoid the trade-off between sensitivity and specificity with currently available options.

CLINICAL TRIALS REGISTRATION: This study is registered on Clinicaltrials.gov no. NCT04242823, https://clinicaltrials.gov/ct2/show/NCT04242823.

OBJECTIVE: The World Health Organization (WHO) recommended addition of local anesthetic to reduce the intense pain of intramuscular injection of 50% Magnesium Sulphate (MgSO4) salt solution has been found to be ineffective. We tested whether giving the local anesthetic 5 min before the MgSO4 injection would reduce pain.

METHODS: We conducted a prospective cross-over trial where each participant with pre-eclampsia or eclampsia received sequential and mixed injection methods in random sequence during sequential MgSO4 administrations. Pain and preference were assessed using descriptive words, a numeric pain scale and direct comparison between the two injection methods. Differences were measured using the Wilcoxon signed rank test, risk ratios with 95% confidence intervals and the Chi squared or Fisher's test. The administration techniques were refined based on an initial pilot of 8 participants.

RESULTS: We enrolled 49 consented participants and analysed data from 41 post-pilot participants The sequential injection method had a non-significantly lower mean pain score than the mixed injection method (3.1 vs. 3.3, p = 0.44). Severe pain was reported for 3/41 vs. 9/41, p = 0.12. The sequential injection method was perceived to be more painful by 13 (37%) vs. 22 (63%) participants (p = 0.03). The sequential injection was preferred by 21(60%) vs. 14 participants (40%) (p = 0.1).

CONCLUSION: Our results consistently favoured the novel sequential injection method. The lack of statistical significance for most results is not surprising given the small sample size. Given the potential for clinically important benefits to women, a larger study to confirm these results is justified.

CLINICAL TRIAL REGISTRATION: https://pactr.samrc.ac.za/, Identifier (PACTR202201521544765).

BACKGROUND: Cervical cancer associated with human papillomavirus has the highest cancer incidence and mortality for women in Botswana because of a high HIV prevalence and limited screening. This study investigates the significance of HIV on the overall survival (OS) of patients with locally advanced cervical cancer by various treatment categories (curative chemoradiation, definitive radiation [RT] alone, or palliative RT alone).

METHODS: This study included patients diagnosed with cervical cancer between 2013 and 2020, prospectively enrolled in the Botswana Prospective Cancer Cohort. OS based on HIV status and completion of planned treatment regimen was estimated by the Kaplan-Meier method. Comparisons of 2-year OS by HIV status was performed by the log-rank test, univariate and multivariable Cox analyses adjusting for cancer stage, RT dose, number of chemotherapy cycles, and baseline hemoglobin levels.

RESULTS: Of 1131 patients diagnosed with stage IB-IVB cervical cancer, 69.8% were women living with HIV (n = 789). For patients receiving curative chemoradiation, HIV status was not significantly associated with OS in unadjusted (p = .987) and adjusted (p = .578) analyses. For RT only treatment and definitive (high-dose) RT alone, HIV status was significantly associated with OS in unadjusted analysis (HR = 1.77, p = .002; HR = 1.95, p = .014), but not in adjusted analysis (p = .227, p = .73). For patients receiving palliative (low-dose) RT, HIV status was not associated with OS in unadjusted (p = .835) or adjusted analysis (p = .359).

CONCLUSIONS: In Botswana, a resource-limited setting, HIV status had no significant effect on 2-year OS in patients with cervical cancer with well-managed HIV receiving chemoradiation, RT alone, or palliative RT. This demonstrates that patients living with HIV receiving antiretroviral treatment can receive clinically appropriate treatment with no evidence that HIV may lead to poorer outcomes.

SARS-CoV-2 infection during pregnancy was associated with maternal mortality and adverse birth outcomes in the pre-Omicron era, including a stillbirth rate of 5.6% in Botswana. We re-evaluated these outcomes in the Tsepamo Study during the Omicron era. We assessed maternal mortality and adverse birth outcomes for all singleton pregnancies from mid-November 2021 (the start of the Omicron era) to mid-August 2022 at nine Tsepamo sites, among individuals with documented SARS-CoV-2 screening PCR or antigen tests and known HIV status. Of 9,705 women routinely screened for SARS-CoV-2 infection at delivery (64% of deliveries at these sites), 373 (3.8%) tested positive. Women with HIV were as likely to test positive for SARS-CoV-2 (77/1833, 4.2%) as women without HIV (293/6981, 4.2%) (p = 1.0). There were 5 recorded maternal deaths (0.03%), one occurring in a woman with a positive SARS-CoV-2 test result. In contrast, maternal mortality was 3.7% and 0.1% in those with and without SARS-CoV-2, respectively, during the pre-Omicron era. In the Omicron era, there were no differences among infants exposed or unexposed to SARS-CoV-2 in overall adverse birth outcomes (28.1% vs 29.6%; aRR 1.0, 95%CI 0.8-1.1), severe adverse birth outcomes (11.9 vs 10.6%; aRR 1.1, 95%CI 0.8-1.5), preterm delivery (15.1% vs 14.9%; aRR 1.0, 95%CI 0.8-1.3), or stillbirth (1.9% vs 2.3%; aRR 0.8, 95%CI 0.4-1.7). Adverse outcomes among those exposed to both HIV and SARS-CoV-2 were similar to those exposed to HIV alone (31.2% vs. 33.1%; aRR 0.9, 95%CI 0.6-1.3; p = 0.5). Maternal mortality was far lower in Botswana during the Omicron era than in the pre-Omicron era, and adverse birth outcomes were no longer significantly impacted by exposure to SARS-CoV-2 either overall or with HIV co-exposure. Increased population immunity to SARS-CoV-2, less stress on the hospital systems in the Omicron era, and possible differences in viral pathogenicity may combine to explain these findings.

A number of challenges hinder artificial intelligence (AI) models from effective clinical translation. Foremost among these challenges is the lack of generalizability, which is defined as the ability of a model to perform well on datasets that have different characteristics from the training data. We recently investigated the development of an AI pipeline on digital images of the cervix, utilizing a multi-heterogeneous dataset of 9,462 women (17,013 images) and a multi-stage model selection and optimization approach, to generate a diagnostic classifier able to classify images of the cervix into "normal", "indeterminate" and "precancer/cancer" (denoted as "precancer+") categories. In this work, we investigate the performance of this multiclass classifier on external data not utilized in training and internal validation, to assess the generalizability of the classifier when moving to new settings. We assessed both the classification performance and repeatability of our classifier model across the two axes of heterogeneity present in our dataset: image capture device and geography, utilizing both out-of-the-box inference and retraining with external data. Our results demonstrate that device-level heterogeneity affects our model performance more than geography-level heterogeneity. Classification performance of our model is strong on images from a new geography without retraining, while incremental retraining with inclusion of images from a new device progressively improves classification performance on that device up to a point of saturation. Repeatability of our model is relatively unaffected by data heterogeneity and remains strong throughout. Our work supports the need for optimized retraining approaches that address data heterogeneity (e.g., when moving to a new device) to facilitate effective use of AI models in new settings.

OBJECTIVES: International guidelines recommend cervical screening cessation at age 50 following two consecutive negative screens. However, many women aged 50 and older in low-income and middle-income countries (LMICs) have not had prior opportunity to screen. We examine the prevalence of cervical dysplasia and cervical cancer stage in Botswana women aged 50+ compared with 30-49, stratified by HIV status.

DESIGN: Secondary analysis of data from two prospective cohort studies.

SETTING: The screening cohort was recruited at health facilities in South East District. The cancer cohort was recruited from the primary public tertiary referral hospital and a private hospital in Gaborone, Botswana.

PARTICIPANTS: The screening cohort included 2570 women aged 30 and older recruited from February 2021 to August 2022. Screening eligibility included anyone with a cervix and without a prior history of cervical cancer. The cancer cohort included 1520 patients diagnosed with cervical cancer who sought care at the facilities where recruitment took place from January 2015 to December 2022.

PRIMARY AND SECONDARY OUTCOME MEASURES: The prevalence of cervical intraepithelial neoplasia (CIN)2+ and cancer stage at diagnosis was compared across age groups, stratified by HIV status. Prevalence ratios were calculated for the association between age and CIN2+/CIN3+via log-binomial regression.

RESULTS: The prevalence of CIN2+ was similar between 30-49 years old and 50+, both among women with HIV (WWH, 15.9% and 19.3%, respectively) and without HIV (13.3% and 10.4%, respectively). Similar findings were found when CIN3+ was used as the outcome. There were no statistically significant differences in prevalence ratios (PRs) across age groups for CIN2+ (adjusted PR (aPR) WWH 1.1 (95% CI 0.80 to 1.6); aPR HIV- 0.78 (95% CI 0.45 to 1.4) nor CIN3+ (aPR WWH 1.1 (95% CI 0.70 to 1.6); aPR HIV- 0.81 (95% CI 0.40 to 1.7)). Nearly half of cervical cancer diagnoses were made in women 50+; three-quarters of cases in women without HIV were diagnosed at 50+ years.

CONCLUSIONS: Our findings demonstrate the prevalence of high-grade cervical dysplasia and cervical cancer remains high beyond age 50 in both women with and without HIV in an LMIC context with high HIV prevalence. Screening women 50+ will allow treatment for cervical dysplasia and may provide early diagnosis of curable cervical cancer. These findings support the rapid introduction of high-performance cervical screening to increase access for women 50+.

TRIAL REGISTRATION NUMBER: NCT04242823.

Women living with HIV (WLWH) are at an increased risk of developing HPV-related high grade cervical dysplasia and cervical cancer. Prior World Health Organization (WHO) screening guidelines recommended starting screening at age 30. We assessed characteristics of women diagnosed with cervical cancer to further inform and refine screening guidelines. We prospectively enrolled women diagnosed with cervical cancer from January 2015 to March 2020 at two tertiary hospitals in Gaborone, Botswana. We performed chi-square and ANOVA analyses to evaluate the association between age upon diagnosis and HIV status, CD4 count, viral load, and other sociodemographic and clinical factors. Data were available for 1130 women who were diagnosed with cervical cancer and 69.3% were WLWH. The median age overall was 47.9 (IQR 41.2-59.1), 44.6 IQR: 39.8 - 50.9) among WLWH, and 61.2 (IQR 48.6-69.3) among women living without HIV. There were 1.3% of women aged <30 years old, 19.1% were 30-39 and 37.2% were 40-49. Overall, 20.4% (n = 231) of cancers were in women <40 years. Age of cervical cancer diagnosis is younger in countries with higher HIV prevalence, like Botswana. Approximately 20% of the patients presented with cancer at <40 years of age and would have likely benefited from screening 10 years prior to cancer diagnosis to provide an opportunity for detection and treatment of pre-invasive disease.

OBJECTIVE: To evaluate the combined association of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and human immunodeficiency virus (HIV) infection on adverse birth outcomes in an HIV-endemic region.

METHODS: The Tsepamo Study abstracts data from antenatal and obstetric records in government maternity wards across Botswana. We assessed maternal mortality and adverse birth outcomes for all singleton pregnancies from September 2020 to mid-November 2021 at 13 Tsepamo sites among individuals with documented SARS-CoV-2 screening tests and known HIV status.

RESULTS: Of 20,410 individuals who gave birth, 11,483 (56.3%) were screened for SARS-CoV-2 infection; 4.7% tested positive. People living with HIV were more likely to test positive (144/2,421, 5.9%) than those without HIV (392/9,030, 4.3%) (P=.001). Maternal deaths occurred in 3.7% of those who had a positive SARS-CoV-2 test result compared with 0.1% of those who tested negative (adjusted relative risk [aRR] 31.6, 95% CI 15.4-64.7). Maternal mortality did not differ by HIV status. The offspring of individuals with SARS-CoV-2 infection experienced more overall adverse birth outcomes (34.5% vs 26.6%; aRR 1.2, 95% CI 1.1-1.4), severe adverse birth outcomes (13.6% vs 9.8%; aRR 1.2, 95% CI 1.0-1.5), preterm delivery (21.4% vs 13.4%; aRR 1.4, 95% CI 1.2-1.7), and stillbirth (5.6% vs 2.7%; aRR 1.7 95% CI 1.2-2.5). Neonates exposed to SARS-CoV-2 and HIV infection had the highest prevalence of adverse birth outcomes (43.1% vs 22.6%; aRR 1.7, 95% CI 1.4-2.0).

CONCLUSION: Infection with SARS-CoV-2 at the time of delivery was associated with 3.7% maternal mortality and 5.6% stillbirth in Botswana. Most adverse birth outcomes were worse among neonates exposed to both SARS-CoV-2 and HIV infection.