A Limited Time to Act: Cervical Cancer Stage Progression Prior to Treatment

Background: Cervical cancer is the leading cause of cancer death for women in sub-Saharan Africa where access to timely diagnosis and treatment remains a major challenge. Several investigators have documented the adverse impact of delays during curative chemoradiation on survival: 2%-5% decreased survival per week delay. However, the clinical impact of delays in initiating treatment has not previously been estimated and is key to direct allocation of resources and patient care. Aim: To estimate the duration treatment initiation delay leading to a one-stage progression in cervical cancer. Methods: Radiation is available in Botswana from a single linear accelerator and services are provided free of cost to Botswana citizens. Two prolonged service interruptions (late 2014 due to mechanical failure/upgrades, and late 2015 for machine replacement) contributed to substantial additional delays in accessing treatment of several months after service resumed. Using these natural experiments and the prospectively enrolling Thabatse Cancer Cohort (included cases enrolled from September 2014 to January 2018), we used instrumental variable methods to estimate the mean number of weeks to progress one stage (e.g., IIIA to IVA). Our instrument was a binary variable indicating treatment initiation occurred during service interruption. We used inverse probability weighting to adjust for possible bias in exposure to the instrument by HIV status, age, and income. Cancer stage was considered ordinal and we used Poisson regression to calculate weeks for stage progression. We calculated confidence limits using bootstrapping from 500 samples. Analyses were performed in SAS 9.4. Results: A total of 301 cervical cancer cases were included with 138 (46%) exposed to service disruption and 163 (54%) not exposed. The majority, 217 (72%), were HIV-infected with 86% on ART prior to cancer diagnosis. At time of treatment, 27 (9%) stage I, 108 (36%) stage II, 113 (38%) stage III, and 53 (18%) stage IV. Median time from diagnosis to initiation of treatment of all patients was 9.7 weeks (IQR 6.3-13.7), and was longer in those exposed to service disruption (11.0 weeks [IQR 7.7-15.1]) than those not exposed (8.0 weeks [IQR 5.3-11.7]. Time for progression of one cancer stage was 8.9 weeks (95% CI 4.6-15.8). Among HIV-infected women, estimated time for one-stage progression was 6.5 weeks (95% CI 2.3-12.3), significantly faster than HIV-uninfected women ( P < 0.001). The small number of HIV-uninfected women prevented separate estimation. Conclusion: Stage progression is rapid in locally advanced cervical cancer with typical treatment delays of two months associated with a full stage increase. These estimates may not generalize to contexts with lower HIV prevalence; however add urgency to effort to improve efficient access to treatment and avoidance of waiting lists.