Contributions to the literature
* Given stark global inequities in cervical cancer, there is a critical need to identify effective implementation strategies to ensure timely access to evidence-based care in low- and middle-income countries.
* This study seeks to address this critical need and advance global implementation science by testing the effectiveness of adaptive strategies on timely treatment adoption using a SMART design and evaluating contextual mechanisms contributing to the success or failure of each strategy.
* In addition to testing novel strategies, this study seeks to advance capacity building and infrastructure related to cancer care and implementation science in Botswana through strong and sustained partnerships.
Background
Delays and missed opportunities for timely treatment contribute significantly to inequities in cervical cancer mortality in low- and middle-income countries (LMICs) compared to high-income countries (HICs) [1, 2]. Cervical cancer is the fourth leading cause of cancer mortality in females globally, with an estimated 604,000 new cases and 340,000 deaths in 2020 alone [1]. The vast majority (approximately 90%) of new cases and deaths occur in low- and middle-income countries (LMICs), particularly those with high rates of HIV as cervical cancer is an AIDS-defining malignancy [3]. This global inequity is partly driven by successful efforts in HICs to increase implementation and adoption of evidence-based cancer care [2, 4, 5].
In Botswana, a LMIC with a particularly high prevalence of HIV (18.5%), and cervical cancer incidence (34.4 per 100,000) and mortality (20.1 per 100,000), we identified substantial delays in cervical cancer care from diagnosis to treatment in a cohort of nearly 1,000 patients, driven by myriad individual- and system-level barriers [1, 4, 6, 7]. Despite a robust HIV care infrastructure and a national program in cervical cancer screening, the majority of patients with cervical cancer in Botswana present with locally advanced disease, driven partly by suboptimal implementation and delays in evidence-based care across the cancer control continuum [6, 8]. This is particularly concerning given that cervical cancer is considered to be nearly completely preventable and often curable if treated early [8, 9].
Despite this great need, implementation strategies to improve timely cancer care are limited in LMICs. Due to a number of barriers, improving evidence-based cancer care in LMICs can be challenging [10]. In Botswana and other LMICs, a substantial amount of important work has focused on improving cancer prevention and screening [7, 11,12,13]. Other studies have shown high touch strategies such as patient navigators and community health workers can be feasibly implemented [7, 14], yet there are concerns regarding their sustainability even in higher-resource settings [14,15,16].In other clinical contexts including HIV, there is growing evidence of the positive impact of patient- and system-level implementation strategies to increase access and adherence to recommended treatment in low resource populations and settings [17,18,19]. However, no published implementation studies to date have specifically targeted treatment adoption following diagnosis of HIV-associated malignancies in LMICs, leaving a critical gap in how best to improve and sustain cancer control across the continuum for countries with a high burden of both cervical cancer and HIV.
Study objectives
To help fill this critical gap, we are testing the effectiveness of adaptive strategies on cervical cancer treatment initiation (adoption) using a hybrid (type III) Sequential Multiple Assignment Randomized Trial (SMART) design (Aim 1), complemented by mixed-methods evaluation via patient and clinician interviews (Aim 2) [10, 20]. By randomizing patients at two stages, this design allows for the assessment of multiple strategies in the same trial and the identification of the minimum level of support needed to help patients who may face greater barriers to care [21]. The adaptive strategies are guided by key principles in behavioral economics (e.g., reduction of system-level friction) and are designed to target contextual determinants, including delayed communication of diagnostic results, individual and structural barriers to accessing treatment, and suboptimal care coordination between referring and cancer treatment clinics identified in our prior work [4]. The study is guided by an integrated theoretical model, drawing from the Capability, Opportunity, Motivation, and Behavior (COM-B) model and the refined Consolidated Framework for Implementation Research (CFIR) for use in LMICs [17, 18].
Methods
Theoretical causal pathway model
To guide our study, using the approach proposed by Lewis and colleagues [19], we developed an integrated causal pathway model by which we hypothesize the proposed adaptive strategies will improve timely treatment adoption (Fig. 1). The rationale is that timely treatment adoption will increase by reducing system-level friction and providing patient-level support to enhance capability and motivation. The first phase strategies (clinic and patient outreach) focus on decreasing friction that results in poor communication of positive results and delays in treatment scheduling. Second stage strategies (framed messages and navigation) are designed to increase opportunity for patients who face additional barriers to care by targeting capability (e.g., individual knowledge, structural barriers to care) and motivation (e.g., cancer and treatment beliefs). We hypothesize that while the strategies will be effective overall, there may be individual and contextual determinants that impact reach and effectiveness for different patients.
In this study, we will empirically test this model by comparing the effectiveness of four adaptive telehealth strategies on timely treatment adoption in a hybrid (type III) trial using a SMART design and evaluating multilevel determinants contributing to the effectiveness of these strategies. In Aim 1, we will conduct a 36-month pragmatic and hybrid (type III) trial to determine the effectiveness of first stage strategies (clinic outreach with or without patient outreach) in combination with second stage strategies (low touch messaging or high touch navigation) to increase treatment adoption within 90 days. The primary implementation outcome is adoption, defined as the initiation of treatment within 90 days from randomization. Secondary implementation and clinical outcomes include fidelity, reach, acceptability, implementation costs, and cancer- and HIV-related clinical outcomes. In Aim 2, we will conduct surveys and semi-structured interviews with an embedded cohort of patients and clinicians to evaluate multilevel mechanisms contributing to the effectiveness of each strategy using qualitative comparative analysis [22].
Study site & enrollment of participants
Female citizens of Botswana older than age 18 with a pathology-confirmed cervical cancer diagnosis are eligible to participate in the study. The study will primarily take place in two interconnected settings in Botswana: the Multidisciplinary Team (MDT) clinic at the public tertiary care center Princess Marina Hospital (PMH) where the MDT and oncology clinics are located and the National Health Laboratory (NHL) [23]. The majority of patients treated for cervical cancer in Botswana are seen at the MDT clinic, which is managed by a multidisciplinary oncology and nursing team, who assist with scheduling, treatment, and staging logistics for all patients. Patients with confirmed cancer diagnosis are staged at MDT and then referred for chemoradiotherapy or surgery as needed. After completion of treatment, patients are referred back to the MDT clinic for follow-up and surveillance care. The NHL is a government referral laboratory offering services to all citizens through the Ministry of Health and Wellness. Nearly 75% of all cervical cancer specimens are evaluated at NHL and as such it is a promising hub to deliver Stage I interventions. In Botswana, all aspects of cervical cancer care are fully funded by the government for all citizens, and therefore, not provided as part of this study.
Pathology team members embedded in the NHL (where referral sites send pathology specimens as part of routine practice) will monitor diagnostic results weekly in the integrated patient management system (IPMS), an existing national data system that tracks, monitors, and reports laboratory results, to identify eligible patients. Once a patient is identified as eligible (including clinic referral), necessary patient contact and clinical information will be extracted from the IPMS and entered in the secure study database (REDCap) for randomization, monitoring, and analysis by trained research staff. Randomization will be generated using a centralized, computer-generated allocation sequence and administered using secure platform. Study investigators not directly involved in delivering the strategies and outcome assessors will be blinded, unless there is a serious adverse event that requires unblinding to be assessed.
Study implementation strategies
A Sequential Multiple Assignment Randomized Trial (SMART) design will be used to assign individual participants to intervention groups (Fig. 2) [24]. In the first phase of randomization, eligible patients will be randomized to either the clinic or enhanced outreach arm (1:1 allocation using permuted blocks of 4 or 8). For patients in both arms, the pathology team will contact the referring clinic where the patient had the diagnostic procedure to communicate positive results and provide a scheduled appointment at the MDT treatment clinic. In the enhanced outreach arm, the patient will also be notified directly by the pathology team; the pathology team will not disclose the specific results to patients but notify them that their diagnostic results are complete and encourage them to contact the referring clinic to discuss immediately. At 30 days following outreach, the study team will assess response to the initial strategy, defined as a completed initial visit at MDT within 30 days from outreach. If a patient has not completed an initial visit (non-responders), they will be randomized again (1:1 using permuted blocks of 4 or 8) to receive asynchronous nudges using framed messages alone (low touch strategy) or in combination with synchronous patient navigation (high touch strategy). Patients who have not yet received results from their referring clinic before 90 days will not receive the second stage strategy.
In the second phase of randomization, non-responders will be randomized to either Low Touch Strategy (Framed Text Messaging) or High Touch Strategy (Framed Text Messaging + Patient Navigation). All participants will be sent gain-framed messages highlighting the importance of timely treatment via text message within 1 week of second stage randomization. Text messages (SMS) will be sent in Setswana using an existing text-based platform (OpCare) [25]. In the High Touch Strategy group, participants will also be contacted directly via telephone by an embedded nurse navigator at the MDT Clinic. They will be called within 1 week of second stage randomization to talk through any questions they have about their diagnosis and to help mitigate any barriers to attending their scheduled visit including providing transportation or other support for patients who may need it. The study will not restrict clinical care or concomitant non-study interventions received by participants regardless of study arm.
Sample size calculation
To calculate power for this study, we identified eligible patients in our longitudinal cohort and national estimates of cervical cancer and calculated the proportion who initiate treatment within 90 days (55%). Using a conservative estimate, we anticipate that at least 426 patients will be eligible and randomized during the 36-month trial period (Fig. 2). Using these estimates and assuming 65% initiate treatment by day 90 in the clinic outreach arm, our sample will provide over 85% power (two-sided alpha = 0.05) to detect a minimum difference in probability of treatment initiation of 13% between the first-stage strategies averaging over the second stage. For comparing strategies for non-responders, we assume an overall response rate at the first stage of 60% and 5% loss to follow-up, leaving approximately 148 patients randomized to the low and high touch strategies in stage two of the SMART. With this number of patients, our sample will provide over 80% power (two-sided alpha = 0.05) to detect a difference in probability of treatment adoption in non-responders of 20% between the second stage strategies (assuming 65% of non-responders start treatment by day 90 in the low touch arm). Recruitment for this study started on September 18, 2023, and at time of protocol submission, 157 participants had been enrolled, with recruitment ongoing.
Primary and secondary outcomes
The primary outcome is adoption, defined as the initiation of treatment within 90 days from randomization (first stage outreach). Secondary outcomes include fidelity defined as completion of evidence-based cancer treatment according to international guidelines; reach defined as the proportion of patient engagement at each SMART stage; clinical outcomes related to cancer and HIV care; implementation costs of each strategy; and acceptability, appropriateness, and feasibility of each strategy. Outcomes will be assessed and managed via a secure platform (REDCap).
Data analyses
All primary analyses will use an intent-to-treat (ITT) approach, including all randomized participants. Analysts will be blinded to study arm assignment and data quality will be assessed prior to analysis. We will fit logistic regression models with treatment adoption as the outcome. To compare the four adaptive strategies, we will use the standard weighted and replicated approach. Specifically, we will fit a generalized estimating equation (GEE) with a logistic link function to a data set where responders are replicated (i.e., once for each of the two embedded regimes) and non-responders are included once since they will only be consistent with one of the embedded regimes. The GEE will then be fit using design weights that account for over- and under-representation of outcomes from responders and non-responders, respectively, due to the restricted randomization of the SMART. The primary implementation outcome (treatment adoption) will be coded as a 1 for all patients who start treatment by 90 days and 0 otherwise. Given the centralization of treatment in the setting, we do not anticipate missing data but if treatment data are missing, we will classify them as 0 for primary analysis. The GEE model will include indicators for the first and second stage strategies and a term for their interaction. We will test the appropriate linear contrasts of these terms to compare the four adaptive strategies. We will also investigate potential heterogeneity of effects using individual- and system-level characteristics (e.g., HIV-status) to help inform future studies in addition to scalability and sustainability in LMICs with high HIV rates.
Mixed method evaluation
To help understand contextual factors shaping implementation strategies and outcomes, we will conduct mixed methods interviews with patients and clinicians, including nurses, physicians, and community health workers, involved in the care of patients with cervical cancer in Botswana. We will purposively stratify interview sampling by implementation outcome and study arm to understand factors contributing to both success and failure of the four adaptive strategies. Upon completion of the full trial, clinicians will be purposively sampled based on clinical role and clinic type and invited to be interviewed. The structured interview guide will be designed to assess across the six domains of CFIR, including: system characteristics, outer setting, inner setting, characteristics of individuals involved, intervention characteristics, implementation process. Interviews will be conducted by a research team member in Botswana and will be supervised by Dr. Rendle (MPI), who has extensive experience in qualitative research [26]. Interviews will be conducted in the preferred language of the participant and take place in a private setting at PMH aligned with clinical visits to reduce burden.
The constant comparative method, guided by modified grounded theory, will be used to iteratively identify a priori domains of interest (guided by our conceptual model and hypotheses) and to inductively explore emergent themes [27, 28]. Two trained coders will first independently read through each transcript to identify themes within each domain. We will then use this list to develop a coding dictionary and apply it to subset of the data. We will measure inter-rater reliability to document and improve coding consistency. Once high reliability is achieved (kappa > 0.7), we will apply the full coding dictionary to the interview data using NVivo and produce thematic reports summarizing our findings. We will then use qualitative data to expand upon and triangulate quantitative patterns identified in trial and surveys. We will use convergent mixed methods analysis to code contextual conditions (inner setting, outer setting, and individual characteristics) and implementation conditions (characteristics of specific strategy and process) [28]. These data will be used to conduct qualitative comparative analysis (QCA) to identify how contextual factors shaped effectiveness of each adaptive strategy [22, 29]. Results from both the mixed methods analysis and trial outcomes will be disseminated broadly through existing partnerships with the Ministry of Health, clinical institutions, and partners in Botswana.
Discussion
Global inequities in cervical cancer are stark but in places like Botswana, evidence-based practices are available but often not received. This highlights the critical need and opportunity to apply implementation science to ensure all patients receive timely treatment and cervical cancer outcomes are improved. In this study, we are working closely with partners across Botswana to test integrative and pragmatic strategies designed to target key determinants of delays specific to this context. Beyond local impact, the use of a SMART design is innovative and will help to advance the field of implementation science locally and globally and have implications for other clinical contexts if successful.
This study also seeks to help advance global implementation science and partnerships. At its core, the project is driven by priority areas identified by the people and leaders of Botswana and leverages existing strengths and partners within the country in its design and execution. The project also builds upon long-standing partnerships between clinicians, researchers, and leaders in Botswana and the United States, in which implementation science expertise serves to complement the depth and range of contextual and clinical knowledge in Botswana. This collaboration is central to the work and will help to ensure sustainment long after the project is complete.
Data availability
Not applicable.
Sung H, Ferlay J, Siegel RL, et al. Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin. 2021;71(3):209–49. https://doi.org/10.3322/caac.21660.
Bhatia RK, Rayne S, Rate W, et al. Patient factors associated with delays in obtaining cancer care in Botswana. J Glob Oncol. 2018;4:1–13. https://doi.org/10.1200/JGO.18.00088.
Stelzle D, Tanaka LF, Lee KK, et al. Estimates of the global burden of cervical cancer associated with HIV. Lancet Glob Health. 2021;9(2):e161–9. https://doi.org/10.1016/S2214-109X(20)30459-9.
Ogden SN, Glanz K, Schapira MM, Bocage C, Eriksen W, Rendle KA. Patient perspectives on primary HPV testing for routine cervical cancer screening. J Gen Intern Med. 2020;35(3):956–8. https://doi.org/10.1007/s11606-019-05175-4.
Ferreira da Silva I, Ferreira da Silva I, Koifman RJ. Cervical cancer treatment delays and associated factors in a cohort of women from a developing country. J Glob Oncol. 2019;5:1–11. https://doi.org/10.1200/JGO.18.00199.
Friebel-Klingner TM, Luckett R, Bazzett-Matabele L, et al. Clinical and sociodemographic factors associated with late stage cervical cancer diagnosis in Botswana. BMC Womens Health. 2021;21(1):267. https://doi.org/10.1186/s12905-021-01402-5.
Johnson LG, Armstrong A, Joyce CM, Teitelman AM, Buttenheim AM. Implementation strategies to improve cervical cancer prevention in sub-Saharan Africa: a systematic review. Implement Sci. 2018;13(1):28. https://doi.org/10.1186/s13012-018-0718-9.
MacDuffie E, Bvochora-Nsingo M, Chiyapo S, et al. Five-year overall survival following chemoradiation therapy for locally advanced cervical carcinoma in women living with and without HIV infection in Botswana. Infect Agent Cancer. 2021;16(1):55. https://doi.org/10.1186/s13027-021-00389-w.
Griesel M, Seraphin TP, Mezger NCS, et al. Cervical cancer in Sub-Saharan Africa: a multinational population-based cohort study of care and guideline adherence. Oncologist. 2021;26(5):e807–16. https://doi.org/10.1002/onco.13718.
Collins LM, Murphy SA, Strecher V. The Multiphase Optimization Strategy (MOST) and the Sequential Multiple Assignment Randomized Trial (SMART): New Methods for More Potent eHealth Interventions. Am J Prev Med. 2007;32(5 Suppl):S112–8. https://doi.org/10.1016/j.amepre.2007.01.022.
Means AR, Kemp CG, Gwayi-Chore MC, et al. Evaluating and optimizing the consolidated framework for implementation research (CFIR) for use in low- and middle-income countries: a systematic review. Implement Sci. 2020;15(1):17. https://doi.org/10.1186/s13012-020-0977-0.
Johnson LG, Ramogola-Masire D, Teitelman AM, Jemmott JB, Buttenheim AM. Assessing nurses’ adherence to the see-and-treat guidelines of botswana’s national cervical cancer prevention programme. Cancer Prev Res Phila Pa. 2020;13(3):329–36. https://doi.org/10.1158/1940-6207.CAPR-19-0348.
Bouvard V, Wentzensen N, Mackie A, et al. The IARC perspective on cervical cancer screening. N Engl J Med. 2021;385(20):1908–18. https://doi.org/10.1056/NEJMsr2030640.
Dalton M, Holzman E, Erwin E, et al. Patient navigation services for cancer care in low-and middle-income countries: a scoping review. PLoS ONE. 2019;14(10): e0223537. https://doi.org/10.1371/journal.pone.0223537.
Nelson HD, Cantor A, Wagner J, et al. Effectiveness of patient navigation to increase cancer screening in populations adversely affected by health disparities: a meta-analysis. J Gen Intern Med. 2020;35(10):3026–35. https://doi.org/10.1007/s11606-020-06020-9.
O’Donovan J, Newcomb A, MacRae MC, Vieira D, Onyilofor C, Ginsburg O. Community health workers and early detection of breast cancer in low-income and middle-income countries: a systematic scoping review of the literature. BMJ Glob Health. 2020;5(5): e002466. https://doi.org/10.1136/bmjgh-2020-002466.
Michie S, van Stralen MM, West R. The behaviour change wheel: a new method for characterising and designing behaviour change interventions. Implement Sci. 2011;6(1): 42. https://doi.org/10.1186/1748-5908-6-42.
Means AR, Kemp CG, Gwayi-Chore MC, et al. Evaluating and optimizing the consolidated framework for implementation research (CFIR) for use in low- and middle-income countries: a systematic review. Implement Sci. 2020;15(1):1–19. https://doi.org/10.1186/s13012-020-0977-0.
Lewis CC, Klasnja P, Powell BJ, et al. From classification to causality: advancing understanding of mechanisms of change in implementation science. Front Public Health. 2018;6: 136. https://doi.org/10.3389/fpubh.2018.00136.
Kahwati L, Jacobs S, Kane H, Lewis M, Viswanathan M, Golin CE. Using qualitative comparative analysis in a systematic review of a complex intervention. Syst Rev. 2016;5(1):82. https://doi.org/10.1186/s13643-016-0256-y.
Glasgow RE, Fisher L, Strycker LA, et al. Minimal intervention needed for change: definition, use, and value for improving health and health research. Transl Behav Med. 2014;4(1):26–33. https://doi.org/10.1007/s13142-013-0232-1.
Thiem A. Using qualitative comparative analysis for identifying causal chains in configurational data: a methodological commentary on baumgartner and epple (2014). Sociol Methods Res. 2015;44(4):723–36. https://doi.org/10.1177/0049124115589032.
Grover S, Chiyapo SP, Puri P, et al. Multidisciplinary gynecologic oncology clinic in botswana: a model for multidisciplinary oncology care in low- and middle-income settings. J Glob Oncol. 2017;3(5):666–70. https://doi.org/10.1200/JGO.2016.006353.
Nahum-Shani I, Qian M, Almirall D, et al. Experimental design and primary data analysis methods for comparing adaptive interventions. Psychol Methods. 2012;17(4):457–77. https://doi.org/10.1037/a0029372.
Grover S, Shah S, Bhatia R, et al. Development and usability of a smartphone application for tracking oncology patients in gaborone. Botswana Methods Inf Med. 2020;59(1):31–40. https://doi.org/10.1055/s-0040-1713129.
Rendle KA, Abramson CM, Garrett SB, Halley MC, Dohan D. Beyond exploratory: a tailored framework for designing and assessing qualitative health research. BMJ Open. 2019;9(8): e030123. https://doi.org/10.1136/bmjopen-2019-030123.
Glaser BG, Strauss AL. The discovery of grounded theory: strategies for qualitative research. Chicago: Aldine Publications; 1967.
Creswell JW, Clark VLP. Designing and conducting mixed methods research. Thousand Oaks: SAGE Publications; 2011.
Kane H, Lewis MA, Williams PA, Kahwati LC. Using qualitative comparative analysis to understand and quantify translation and implementation. Transl Behav Med. 2014;4(2):201–8. https://doi.org/10.1007/s13142-014-0251-6.
© 2024. This work is licensed under http://creativecommons.org/licenses/by-nc-nd/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.