When Is It Effective to Offer Self-Sampling to Non-Attendees—Response

We appreciate Dr. Castle's concern that a loss in CIN2⁺ sensitivity, when using self-sampling instead of office-based sampling, may still be an issue (1), even though data of a recently published meta-analysis reported otherwise (2).

Data of the Dutch PROHTECT study strongly suggested that the CIN2⁺ sensitivity of HPV self-sampling was non-inferior to that of HPV office-based sampling (3). As self-sampling was offered to non-attendees at home, it is unlikely that the sensitivity profile in this study was biased because of “in office procedures,” although the PROHTECT study estimates may have been biased due to higher risk profiles in the non-attendees. Therefore, studies are needed to validate self-sampling test characteristics in regular responders with lower risk profiles in case they switch from office-based sampling to self-sampling.

In our article, we showed that even if the loss in CIN2⁺ sensitivity is 10%, and 20% to 30% of the regular attendees switch to self-sampling, it is still both effective and cost-effective to offer self-sampling to non-attendees as long as unscreened women attend, and the extra attendance rate is at least 6 percentage points (4). Results of the Dutch PROHTECT studies, where self-sampling was offered via an opt-out procedure (i.e., a self-sampling kit was sent to all non-attendees except when they opted-out via a letter), showed that both assumptions were realistic (3, 5). However, as opt-in procedures (i.e., involving a request for a self-sampler) may reduce response rates (6), the chosen strategy could be crucial in whether or not offering self-sampling is (cost-)effective. Therefore, we fully concur with Dr. Arbyn and Dr. Castle that the introduction of self-sampling strategies should be carefully prepared and evaluated in pilot studies integrated in well-organized settings before general rollout (6).

See the original Letter to the Editor, p. 1295

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.