Presentation (CTI)

OBJECTIVES: FDA guidance discourages proxy reporting in favor of observer reporting. We evaluated the feasibility of using the Patient-Reported Outcomes Measurement Information System (PROMIS) item banks to build custom short-forms (SF) that could qualify as observer reported outcomes (ObsROs) per FDA Patient-Focused Drug Development (PFDD) fit-for-purpose guidance.
METHODS: PROMIS Parent Proxy and Early Childhood Parent Report item banks for anxiety and physical activity were evaluated with 24 and 10 unique items, respectively. Two independent raters coded items as “observable” (assessing observable signs/indicators) or “proxy” (assessing internal states). Codebook revision added “ambiguous” to accommodate items with multiple interpretations. Items were also coded as subject to third-party information and as asking for attribution of indicators to internal states. Reconciliation resulted in creation of custom SFs from “observable” items, which were evaluated against PFDD validity evidence guidance.
RESULTS: For anxiety, the ObsRO SF retained 3 items; none were likely subject to third-party information; 2 of these asked for attribution of indicators to internal states. Of excluded items, 2 were coded “ambiguous”, and 19 were coded as “proxy”. For physical activity, the ObsRO SF retained 7 items; all 7 were likely subject to third-party information, and 2 items asked for attribution of indicators to internal states. Of excluded items, 1 was coded “ambiguous”, and 2 were coded as “proxy”. Evaluation of the SFs against PFDD guidance suggested risk of insufficient conceptual coverage and error variance related to increased focus on strictly measuring observable indicators.
CONCLUSIONS: Dichotomization (Observed versus Proxy reporting) may not be a practical theoretical approach for building SFs. Furthermore, proxy- and observer reporting-type items contribute different potential sources of error variance (e.g., attribution of indicators versus restricted conceptual coverage. Balancing the sources of measurement error for each study and its population is required to build fit-for-purpose instrumentation that captures meaningful information reliably.