Study Design
This was a cross-sectional, descriptive study involving a one-time, Web-based discrete choice experiment (DCE) survey completed by injection-naïve T2DM patients.
DCE, a type of conjoint analysis, is a robust methodology to elicit consumer preferences and is increasingly being used to examine patient preferences for healthcare interventions [20, 21]. DCE assumes that any product or service can be described by its characteristics (or attributes) and that an individual values the product or service on the basis of the levels of these attributes. In a DCE, individuals are presented with hypothetical descriptions of products and are asked to choose their preferred product on the basis of the levels of attributes presented. Response data are modelled within a benefit (or satisfaction) function, which provides information on whether or not the given attributes are important, the relative importance of attributes, the rate at which individuals are willing to trade between attributes, and overall benefit scores for alternative products [20].
In this study, the DCE survey was completed by adult T2DM patients in the UK, Germany, China, Brazil, and Japan. Eligibility was determined by the following criteria: patient-reported diagnosis of T2DM; age 18 years or older; adequate written and oral fluency in the target language; willing and able to complete a 30-min online survey to answer questions about treatment experiences and preferences; have access to the Internet; willing to provide informed consent; currently on an OAD; and no experience with injection treatment for T2DM (i.e., injection-naïve). Eligible participants were identified by a market research recruitment agency using multiple methods including searching databases of patients who had agreed to be contacted for participation in such research studies, referrals from clinicians and patient associations, and advertising in targeted publications. This study aimed to recruit a diverse sample with respect to age, gender, and education status.
The survey was programmed and hosted by Global Perspectives (Berkshire, UK) on a secure server. Interested participants were provided with an Internet link to the online Web-based DCE survey. All participants were screened online to ensure that they met the specific eligibility criteria. All eligible patients provided online consent through the Web link prior to completing the survey. Those who completed the survey received compensation for their participation.
The study protocol was approved by the Salus International Review Board (Austin, TX, USA). All procedures followed were in accordance with the ethical standards of the responsible committee on human experimentation (institutional and national) and with the Helsinki Declaration of 1964, as revised in 2013.
Survey Development
A literature review and qualitative research with clinicians and T2DM patients were conducted to inform attribute selection and survey development. Open-ended, one-on-one interviews were conducted with 50 T2DM patients (from the five target countries) with experience taking GLP-1RAs or who were injection-naïve to identify and examine treatment attributes important to T2DM patients [22].
On the basis of the findings from the literature and qualitative research, eight attributes related to treatment efficacy, safety, tolerability, and convenience were included in the DCE survey: efficacy [improvement in glycated hemoglobin (HbA1c)], common GLP-1RA side effects (nausea, diarrhea, vomiting, and injection site nodules), injection device size, needle size, need for titration, required injection preparation (associated with vial/syringe, single-use pen, multi-use pen, or auto-injector), evidence of long-term efficacy/safety, and dosing frequency (daily or weekly). The US Food and Drug Administration-approved labels of two common GLP-1 RAs [23, 24] were reviewed to provide profiles reflecting two hypothetical treatments (Fig. 1). One treatment was a profile of a daily injectable, approximating liraglutide QD, the most commonly prescribed daily GLP-1RA treatment. The other treatment was an injectable approximating exenatide QW, the first QW GLP-1RA treatment. Exenatide QW is delivered once weekly, does not require titration, and uses a thicker, longer needle. Delivery of exenatide was specified as either using a single-use pen device or via an auto-injector. The single-use pen device contains exenatide microspheres and lipid-based diluent in separate chambers, requiring careful preparation of the medication, while the auto-injector has the exenatide microspheres and lipid-based diluent prefilled in one chamber. Exenatide QW can also be delivered in a single-dose tray (vial and syringe). Liraglutide QD is delivered once daily via a multiple-use pen, requires careful preparation of the medication, requires titration of the medication, and uses a thinner, shorter needle.
The attributes and levels were combined into choice sets using a published orthogonal array [25]. An orthogonal fractional factorial design was used to identify the minimum specification of the DCE experimental design in order to fairly represent combinations of the attributes and levels. These combinations were paired using a fold-over design. Each choice question presented two hypothetical treatments (Treatment A or B) and participants were asked to indicate their preferred choice. On the basis of the orthogonal fractional factorial design, a total of 32 pairs of choice sets were generated, with each respondent being presented with a set of 16 unique paired choice sets [20]. An example of a choice set included in the survey is provided in Fig. 1.
The first section of the survey consisted of questions on past and current medication use, adherence, clinical history, and demographic information. The adherence questions asked about frequency of missing, skipping, or being late in taking their oral medications and what aspect of their current medications made it most difficult or inconvenient to take as prescribed. In addition, participants were asked hypothetical questions about the potential impact of dosing frequency on their adherence.
The second section of the survey included the DCE questions. The DCE began by presenting participants with descriptions of the attributes and the levels included in the survey, followed by the 16 paired comparison questions. The English-language survey was pilot-tested with five T2DM patients in the UK, and revisions were made to improve clarity and assess cognitive and overall burden. The survey was then translated into Portuguese (Brazil), Chinese (China), German (Germany), and Japanese (Japan). Two native speakers in each country reviewed the translations for accuracy and clarity.
Data Analysis
Descriptive statistics using SAS (v.4.3) were performed to calculate means, standard deviations, ranges, frequencies, and percentages as appropriate, for the demographic and clinical characteristics data, as well as self-reported data on adherence to medication. SAS (v.4.3) was also used to analyze the DCE data using a conditional or mixed logit model, with each attribute and level included as a separate variable in the model. The choice of treatment was the dependent variable and was analyzed conditionally for each choice set (i.e., each treatment choice the respondent had to make). The conditional logistic model provided regression coefficients (i.e., utility estimates), which were then presented as odds ratios (ORs) and their 95% confidence intervals (CI) in order to indicate the importance of a given attribute level versus a reference attribute level, as well as the likelihood of a respondent choosing a treatment with a given level of an attribute rather than a treatment with the reference level of the attribute. All analyses were performed for the total sample and separately for each country.
Sensitivity analyses were conducted to examine how much a product’s overall preference was improved or worsened by changing certain attributes, while holding all other attributes at constant base case levels. For example, the exenatide QW profile was compared to the liraglutide QD profile assuming better efficacy for liraglutide QD versus exenatide QW. The exenatide QW profile assuming an auto-injector device size and auto-injector preparation was also compared to the liraglutide QD profile.