Study flow
Following FDA, ISPOR, and EMA best-practice guidelines on PRO development and measurement [6,7,8,9,10], we used an iterative approach to conduct patient interviews and develop and evaluate the PRO measure (Fig. 1). Data collection and analysis were also done following FDA, ISPOR, and EMA best-practice guidelines. The various steps in the development process are as follows:
Step 1
We conducted a targeted review of literature articles published between January 2006 and March 2016 using the following search string: (CKD OR “Chronic Kidney Disease”) AND (symptoms OR anemia OR fatigue OR energy OR cognition OR memory) AND questionnaire, to identify concepts to explore during subsequent qualitative patient interviews, and to identify existing CKD and anemia-specific PRO measures. Our targeted literature review identified 5 existing PRO measures (The Kidney Disease Quality of Life Instrument [KDOQOL], the Functional Assessment of Cancer Therapy – Anemia [FACT-an], The Fatigue Assessment Scale [FAS], Patient-Reported Outcome Measurement Information System Fatigue [PROMIS Fatigue], and Dialysis Symptom Inventory [DSI]) with potential relevance in the anemia of CKD population (Item S3 and Table S2) [12,13,14,15,16]; however, none of these specifically examined anemia in the context of CKD, nor were they developed with input from patients with anemia of CKD. Consequently, we decided to conduct qualitative interviews and to potentially develop a new measure specific for this patient population.
Step 2
We developed a semi-structured interview guide, which included open-ended questions designed to facilitate discussion, with additional probing to further explore concepts as needed. We made minor revisions to the interview guide after the first few interviews. Qualitative concept elicitation (CE) interviews (N = 14) were conducted in accordance with good research practices [6] by telephone (Wave 1) to identify relevant symptoms and impacts as reported by patients with anemia of CKD. All interviews were conducted in English by interviewers experienced in conducting qualitative interviews in a manner that encouraged participant engagement and open communication. Interviewer characteristics and sample interview questions are summarized in items S1 and S2 and Table S1.
Step 3
An initial item generation process resulted in the development of 2 PRO measures, a 23-item daily symptom diary, and an 18-item weekly impact questionnaire, based on a thematic analysis of the interview transcripts to identify the most relevant symptoms and impacts and the frequency of symptoms reported by study participants during the CE interviews. To the extent possible, questions were constructed using language expressed by participants.
Step 4a
A second wave of interviews (N = 14) was conducted in a separate set of participants to further explore concepts of interest in patients with anemia of CKD and to assess the content, clarity, and relevance of the 2 draft PRO measures developed in Step 3. These in-person interviews combined CE and cognitive debriefing (CD) components [10] and were conducted one-on-one at either the dialysis center or nephrology clinic. Participants were first asked an abbreviated set of the CE questions from the initial interviews and then asked to provide feedback on the draft measures. The draft PRO measures were updated as interviews were conducted. All revisions to the measures and the rationale for revising were documented using an item-tracking matrix. Because interview time was limited (~ 60 min), not all participants were asked all interview questions.
Step 4b
In parallel to Wave 2 interviews (Step 4a), a translatability assessment was conducted by experienced translators in Hindi, Russian, and Spanish to assess the feasibility of translating the draft items into other languages for use in global studies [17]. The translatability assessment was completed to determine where difficulties would be encountered in subsequent translation efforts for the new PRO measures. The languages were selected because they represent diverse language families spoken by large populations.
Step 5
Based on results from the CD interviews (Wave 2) and input from the research team, it was determined that a 7-day recall period would increase the practicality for using the new PRO measures in clinical trials. As such, the content of the 2 draft PRO measures was combined into a single questionnaire, with a 7-day recall period for all symptoms (except for bruising, which uses a 1-month recall period).
Step 6
An additional wave of CD interviews (N = 8) (Wave 3) [10] was conducted via telephone to obtain feedback on the content, relevance, and clarity of the newly combined PRO measure and to confirm the revisions in Step 5, including the relevance of the 7-day recall period for the items previously included in the daily diary.
Step 7
Following completion of Wave 3 interviews (Step 6), the content of the PRO measure was finalized.
Step 8
The measure was translated into additional languages. All translations underwent either full linguistic validation, including dual forward translations and dual backward translations or linguistics review. All translations were subsequently reviewed by a clinician and underwent cognitive debriefing with 5 patients, as well as proofreading and quality control (QC) steps throughout. The translation process included full linguistic validation interviews.
Study population
Human subjects research approval for this project was provided by an independent scientific review committee (The Copernicus Group, Cary, NC). All participants provided informed consent before enrolling in the study.
Study participants were recruited from Fresenius Medical Care North America (Research by Design site in Evergreen, IL) and DaVita Clinical Research network practices in the United States via telephone and in person. Recruitment was designed to select a diverse range of patients with anemia of CKD rather than replicating real-world demographics. Therefore, an effort was made to diversify recruitment, including a heterogeneous sample across stages of CKD and type of dialysis, sex, age, educational levels, and race [6]. Although sampling relied on a convenience sample, a recruitment target was used whereby ≥3 patients were sought in each subgroup of interest, including patients treated with erythropoietin-stimulating agents (ESAs), those receiving intravenous iron, treatment naïve patients, dialysis-dependent patients, and those not currently receiving dialysis.
To be included in the study, participants were required to be US residents, aged ≥18 years, and have a confirmed diagnosis of CKD. All study participants were required to have hemoglobin (Hb) levels ≥8.0 g/dL and < 12.0 g/dL. However, an effort was made to include patients with both low and high Hb levels, with low Hb defined as 8.0 to 9.9 g/dL in dialysis-dependent patients and 8.0 to 8.9 g/dL in non-dialysis patients and high Hb defined as 10.0 to 11.0 g/dL in dialysis-dependent patients and ≥9.0 g/dL in non-dialysis patients. Participants also needed to speak and read English fluently, provide consent to participate, and be willing to participate in a single audiotaped interview (in person or by telephone) of approximately 60 min.
Patients not on dialysis were eligible regardless of whether they were being treated with an ESA. However, any change in ESA use (initiation or discontinuation) must not have occurred within the past 12 weeks. Patients receiving dialysis were eligible if they were currently receiving an ESA for ≥12 weeks and were on dialysis for ≥12 weeks. Patients who had initiated dialysis within the past 4 weeks were also eligible if they were not currently receiving an ESA. Patients undergoing hemodialysis were required to receive dialysis ≥2 times weekly. Patients undergoing peritoneal dialysis had to be on daily dialysis to be included in the final round of CD interviews (Wave 3).
Patients with medical or psychiatric conditions or those being treated for a condition that resulted in a cognitive or other (e.g., visual, hearing) impairment that would interfere with study participation (based on the investigator’s opinion) were excluded.
Analysis
All interviews were audio recorded and transcribed. All data from the interviews was then coded using MAXQDA (version 11.1.2). A code book was developed iteratively to categorize concepts of interest from the interviews and included descriptions and examples for each code to ensure consistency across coders. Each transcript was coded by 1 coder, then reviewed, summarized, and analyzed by a second coder for accuracy. A saturation table was developed to document table to document the point at which no new concepts were mentioned by subsequent participants for each symptom mentioned during Wave 1 and the CE portion of Wave 2 of the study. Analyses for subgroups were also conducted of interest including dialysis vs non-dialysis patients, hemodialysis vs peritoneal dialysis patients, and patients with Hb level <10.0 g/dL vs those with >10.0 g/dL. Subgroup analyses were descriptive only, and no formal statistical testing was done.