Table 5

Research gaps for self-sampling for human papilloma virus (HPV) infection

Clinical, programmatic, values-based or human rights question		Current WHO recommendation	Strength of recommendation	Source or quality of evidence	Key GER considerations	Ideal study design(s)	Feasibility and practical constraints	Alternative study design	Main methodological issues arising in the alternative study design
Benefits versus harms	PICO: should HPV self-sampling be made available to adult women as an additional approach to clinician-based sampling and cervical services?	HPV self-sampling should be made available to adult women as an additional approach to clinician-based sampling and cervical services.	Strong.	Moderate.	No data from LMIC. Availability and accessibility.	RCT conducted in LMIC with inclusion of outcomes to measure availability and accessibility.	As a strong WHO recommendation exists, it would be unethical to repeat an RCT of the effectiveness of the intervention.	Prospective non-controlled cohort study to be conducted in several LMIC where the intervention is or will be implemented. Analysis to combine quantitative survey supplemented with focused in-depth qualitative interviews with women to explore GER outcomes of availability and accessibility. Cost analyses can be nested in the cohort study.	Confounding: participants in cohort self-select to self-sample and may have other similar characteristics that may influence their ability to access HPV self-sampling. Measurement bias: self-reporting bias an issue for qualitative component. This will require well trained researchers and validated tools. The triangulation of the qualitative data with the quantitative survey results provides an element of additional validation of outcomes.
Values and preferences	What is the optimal way(s) to engage women to self-sample, for example, via text or via community-based means?	Not applicable.		GDG.	Accessibility.	Comparative effectiveness research to compare different types of methods to engage women.	A controlled trial that may or may not be randomised is feasible. Identification or development of methods of engagement during a formative stage will be required prior to testing in a controlled manner. Current methods employed in HIV adherence programmes may be informative.		Detection bias: the outcome will be number of women engaged for each method. Ascertainment of engagement will likely be via self-report and as women will be aware of what method they received, there is a risk of detection bias. This can be reduced by ensuring that outcome assessors are masked to the allocated group, but a risk of detection bias remains.
Human rights and equity	What are the optimal methods to encourage women living in humanitarian settings to self-sample?				Availability, accessibility, acceptability, non-discrimination, privacy and confidentiality	Qualitative key informant study of users in humanitarian settings regarding their experiences of healthcare and desires for self-sampling.	This is a feasible study.		Selection bias: as with all qualitative studies, participants are purposively selected. However, this is an advantage in this study design as users and potential users living in humanitarian settings will have rich experiences and observations to share.
	How can linkage to care be ensured following HPV self-sampling?			GDG	Accessibility, acceptability and non-discrimination.	Comparative effectiveness research of a package of care that includes self-sampling combined with different strategies such as training of clinic staff to encourage women and facilitate linkage to care compared with self-sampling only.	A controlled trial that may or may not be randomised is a feasible design. Linkage to care needs to include links to care for women self-sampling and sensitisation to HPV vaccination and potential messaging to partners.		Social desirability bias: women will be aware of the different strategies they have received and may feel compelled to report positive experiences from staff, rather than discrimination. To address this, outcome assessment should be done by those blinded to the group and who are not part, or perceived to be part, of the clinic staff.
	What are the optimal methods to access homeless women?				Accessibility, acceptability and non-discrimination.	This is an under-researched area. Initial research should include identification of current health-seeking practices of homeless women via qualitative interviews with those working with the homeless and the homeless to find out how they would like to be reached and to develop strategies prior to further evaluation.	This is a feasible design.		Lack of generalisability: the results will be context specific and may not be generalisable to other regions; this is especially true of access to services. However, the desire for approaches to outreach may be generalisable to other settings.
Resource use	What is the cost-effectiveness of self-sampling when linkage to care is included as an outcome in the analysis?	Not applicable.	Not applicable.	GDG.	Availability and accessibility.	Modelling studies of cost- effectiveness across broader range of settings, taking both a healthcare provider and societal perspective.	This is a feasible study, but its quality and usefulness depend on the reliability of the parameters for unit costs and effectiveness, which would be drawn from costing studies and the RCTs effectiveness data. It will also require more long-term modelling to capture downstream effects on linkage to care.		Measurement bias: the results are dependent on selection of variables included in the model, assumptions regarding discount rates, cost-effectiveness thresholds and accurate estimates of current and projected costs. For example, vulnerable populations with limited access to care are likely to generate different results based on the probability that they would have limited or no access to care without self-sampling.
Resource use	What are the differences in out-of-pocket expenditures for self-sampling between high-income and low-income regions?	Not applicable.	Not applicable.	GDG.	Accessibility.	Empirical costing studies to estimate user and provider costs, and financing sources for each (users pay, or health system pays)	This is a feasible study.		Measurement bias: the results are dependent on user recall and self-reported resource use, opportunity costs, as well as on assumptions regarding shared costs in health facilities.

GDG, Guideline Development Group; GER, gender, equity and human rights; PICO, Population, Interventions, Comparison and Outcomes; RCT, randomised controlled trial.