Page 156 - Strategies for non-invasive managementof high-grade cervical intraepithelial neoplasia - prognostic biomarkers and immunotherapy Margot Maria Koeneman

Chapter 8
Methods
Participants
The study was performed in five Dutch hospitals. The target population consisted of women diagnosed with premalignant abnormal cervical cytology (PAP 2/ASCUS to PAP 3b/HSIL), for whom colposcopy was planned and who would thus potentially need treatment for high-grade CIN.
Discrete choice experiment
Patient preferences were investigated with a discrete choice experiment (DCE).[11] In a DCE, two or more interventions are compared based on the attributes of these specific interventions. Attributes represent the characteristics of the intervention, such as treatment efficacy or side effects. The preference for either intervention is then based on the levels of the different attributes: the rate of successful treatment or the rate of side effects. Respondents choose between different hypothetical scenarios described in terms of their attributes and associated levels. Several choice sets are presented that consist of two or more scenarios in which the levels of the attributes differ. The assumption is that a respondent will choose the scenario that provides the highest utility or benefit.
Attributes and levels
Selection of the most relevant attributes and their levels was based on literature research followed by a systematic discussion with members of the research team, consisting of a gynecologist (AK), a resident-gynecologist (MK), and two DCE experts (CD and BE). Three reviews and three clinical trials reporting on outcome measures of LLETZ treatment and vaginal imiquimod treatment were identified.[3-5, 8, 9, 12] Subsequently, six interviews were conducted with women who had undergone colposcopy (with or without diagnostic biopsies or LLETZ treatment) to verify the relevance and relative importance of the selected attributes. The interviews consisted of qualitative, open-ended questions and discrete interview questions. Ages differed from 30 to 55, and three women desired a future pregnancy. The patient interviews did not lead to the identification of additional attributes. The selected attributes and their levels are illustrated in Table 1. Because the evidence on the attributes and levels of imiquimod treatment is limited, we designed a DCE with two hypothetical imiquimod options per choice set, in which levels were varied.
Study procedures
An efficient main-effect design was created, using Ngene software, to select a subset of all possible choice sets.[13] An efficient design refers to the precision with which the different model parameters can be estimated.[14] Prior parameter values were based on the importance ranking that was provided in patient interviews. Based on the S-estimate (sample size) of the efficient design, 50 patients were sufficient to estimate the significant parameters for a main-effect model. We aimed for the inclusion of 100 patients in order to be able to explore potential interaction effects.
154