Quantitative Testing of the Myelofibrosis Symptom Assessment Form Version 4.0, a Harmonized Patient-Reported Outcome Measure for Collecting Key Secondary Endpoint Data in Myelofibrosis Clinical Trials

Background: The original 20-item Myelofibrosis Symptom Assessment Form (MFSAF) was developed to assess patient-reported symptom improvement in MF treatment trials (Mesa et al, 2009). The MFSAF was expanded to 27 items for use in myeloproliferative neoplasms (MPNs) including essential thrombocythemia (ET) and polycythemia vera (PV; MPN-SAF, Scherber et al, 2011), and later reduced to a 10-item version called the MPN-10 (Emanuel et al, 2012; scale 0-100). At the same time, investigators developed a modified version of the MFSAF called the MFSAF v2.0 diary (6 items, scale 0-60) for use in an MF treatment trial (Falcone and Levy, 2012). As part of next generation JAK1/2 inhibitor trials, various pharma companies collaborated with the US FDA to develop other versions (e.g., MFSAF-revised; 7 items, scale 0-70). Variability across versions (e.g., item wording, recall periods, response scales) increases uncertainty when selecting a measure for use in trials and when comparing findings across trials. To mitigate this uncertainty, the 7-item MFSAF v4.0 (scale 0-70) was recently developed through a harmonization process including stakeholders and the developers of the legacy measures (Gwaltney et al, 2017).

Methods: To compare the MFSAF v4.0 to previous versions, an online REDCap survey was developed to assess symptoms (four symptom measures using 24-hour recall), physical functioning (PROMIS® Physical Function-Short Form 10a [PROMIS-PF]), and preference-based health status (EQ-5D-5L Index). Subjects were randomized to complete the (1) MFSAF v4.0 twice (for test-retest reliability) and MFSAF v2.0; or the (2) MFSAF v4.0, MFSAF v2.0, and MPN-10+MFSAF-revised. The three measures in each arm were administered in random order. PROMIS-PF and EQ-5D-5L were used as distractors between the 1st/2nd and 2nd/3rd symptom measures, respectively. The survey was promoted via multiple MPN-related websites and their affiliated Facebook pages June-July 2017. Differences between groups were assessed by ANOVA F-tests and within subjects by paired t-tests. Pearson correlations were computed for the entire sample and then within MF patients only. Cronbach's alphas assessed internal consistency and intraclass correlation coefficients assessed test-retest reliability.

Results: 636 subjects accessed the survey of which 513 (81%) were eligible, consented, and attempted at least one of the symptom measures. Respondents with ET (37%), PV (41%), or MF (22%) were 81% female, 95% white, 68% residing in US, and 14% high school education or less with median age 59 (range 27-90). Among 511 with spleen information, 10/511 (2%) reported a prior splenectomy and 175/511 (34%) reported an enlarged spleen (91 [18%] unsure). Similar to other versions, MFSAF v4.0 Total Symptom Score (TSS; overall mean 20.4, SD 14.5, range 0-69) did not significantly differ across disease groups (ET mean 21.1, SD 15.4; PV mean 20.6, SD 14.4; MF mean 18.8, SD 12.9; p=.43) but did significantly differ between subjects with (mean 23.9, SD 14.8) and without enlarged spleen (mean 15.7, SD 12.1; p<.001). MFSAF v4.0 TSS was significantly correlated with the MFSAF v2.0 TSS (overall r=.96, MF r=.97), MPN-10 TSS (overall r=.92, MF r=.85), MFSAF-revised TSS (overall r=.97, MF r=.96), EQ-5D-5L Index (overall r=-.63, MF r=-.61), and PROMIS-PF (overall r=-.60, MF r=-.65; all p<.001). For individual symptoms, correlation among symptoms within the MFSAF v4.0 ranged from .40-.54 (MF .36-.63). Correlations between corresponding items on the different versions were similarly high (.84-.97). Three items of the MPN-10 and 1 item of the MFSAF v2.0 have higher mean scores, and 1 item of each of the MFSAF v2.0 and MFSAF-revised have a lower mean score relative to the MFSAF v4.0 (all p<.05). Internal consistency (overall .86, MF .84) and test-retest reliability (overall .98, 95% CI .97-.98; MF .98, .97-.99) of the MFSAF v4.0 TSS were high.

Conclusions: The content validity of the MFSAF v4.0 was established through the harmonization process. Results of this study demonstrate favorable reliability and cross-sectional validity of the MFSAF v4.0 and a high level of comparability to the previous versions, both overall and for individual symptoms. The uniform use of this harmonized version in clinical trials may reduce the uncertainty of assessing symptom improvement and enable comparison across trials. The MFSAF v4.0 is now publicly available for use through the Critical Path Institute.