For qualitative analyses, patients with a confirmed diagnosis of CAD or cold agglutinin syndrome (previously known as secondary CAD) and their caregivers in the US were identified for inclusion in a survey panel. Quantitative analyses were based on data from two Phase 3 clinical trials evaluating sutimlimab as a treatment for CAD. Sutimlimab (ENJAYMO^TM) is the only approved treatment for patients with CAD; it was first approved in 2022 in the USA (9). In the CARDINAL trial (NCT03347396), 24 adult patients with CAD with a recent history of blood transfusion and Hb ≤ 10.0 g/dL received sutimlimab for 6 months during Part A of the trial (3). In the CADENZA trial (NCT03347422), adult patients with CAD without a recent history of blood transfusion and Hb ≤ 10.0 g/dL were randomized 1:1 to receive sutimlimab (22 patients) or placebo (20 patients) for 6 months during Part A of the trial (10). The primary efficacy endpoint in both trials was a composite that required a Hb increase compared with baseline, no blood transfusions received, and no forbidden medications received. Fatigue was assessed as a secondary efficacy endpoint in both these trials using the FACIT-Fatigue questionnaire. In addition, other patient-reported outcome (PRO) tools, such as the 12-Item Short-Form Health Survey - Version 2 (SF-12v2®), the Patient Global Impression of [fatigue] Severity (PGI-S), and the Patient Global Impression of Change (PGI-C), provided exploratory efficacy endpoints.

Data from CARDINAL and CADENZA were pooled (N = 66) for a better CAD population representativity given the small sample size of both trials and the different inclusion criteria in terms of recent history of blood transfusions.

The SF-12v2 is a multipurpose, short-form health survey with 12 questions categorized into eight domains (subscales) of functioning and well-being, including two psychometrically based Physical Component Summary (PCS) and Mental Component Summary (MCS) measures. Each domain is scored on a T-score metric, which has a mean of 50 and a standard deviation of 10 (15). Higher scores (and positive change scores) on SF-12v2, PCS, and MCS indicate better daily functioning of patients. General health item (GH01) is the general health domain including only one question: In general, would you say your health is: Excellent, Very Good, Good, Fair, Poor?

The PGI-S is a self-reported, one-item questionnaire designed to assess participant’s impression of fatigue severity over the past 7 days using a five-point scale ranging from 1 to 5 (where 1 = none; 2 = mild; 3 = moderate; 4 = severe; and 5 = very severe). Higher scores indicate more severe fatigue.

The PGI-C measures self-reported overall health change from baseline in a seven-point, single-item scale ranging from “very much worse” to “very much improved.”

The internal consistency reliability of FACIT-Fatigue scores was evaluated using Cronbach’s alpha coefficient. Cronbach’s alpha was computed at baseline and Week 26 using listwise deletion of a single item at a time. The test–retest reliability of the FACIT-Fatigue was not investigated since neither CARDINAL nor CADENZA was designed to evaluate the stability of scores over time.

Convergent validity was examined by evaluating the direction and magnitude of correlations between FACIT-Fatigue scores and scores from instruments measuring conceptually related PROs or clinical measures. Several such instruments were used to evaluate the convergent validity of FACIT-Fatigue scores, including SF-12v2, PCS and MCS, PGI-S, and Hb level. Convergent validity between FACIT-Fatigue scores and each of the instrument measures was calculated at baseline and Week 26 using Spearman rank-order correlation coefficient, where both FACIT-Fatigue and the respective instruments were concurrently assessed. Correlations of ≥0.40 were interpreted as strong evidence for convergent validity. Correlations of 0.20–0.39 were considered to provide more limited support for convergent validity.

Known-group validity tests whether a PRO scale can discriminate between groups of patients who differ on a conceptually related criterion measure in the direction and magnitude that is consistent with a prespecified hypothesis (16). Hb and PGI-S were used to assess known-group validity of the FACIT-Fatigue scale in CAD.

The evidence to support the ability of FACIT-Fatigue to detect change was produced in two ways. The responsiveness of FACIT-Fatigue was investigated using Spearman rank-order correlations between changes in FACIT-Fatigue scores and changes in SF-12v2 PCS and MCS scores, changes in PGI-S scores, and changes in Hb levels from baseline to Week 26 (or to turn around time for Hb). FACIT-Fatigue’s ability to detect change was also evaluated by comparing mean changes in scores between groups of patients with different Hb levels and between groups of patients with different PGI-C responses.

The Meaningful Within Patient Change (MWPC) reflects a score change in a PRO instrument that indicates a meaningful treatment benefit for individual patients. Anchor-based and distribution-based methods were used to estimate the MWPC of the FACIT-Fatigue in patients with CAD. Anchors used in this analysis were: the GH01 of SF-12v2, PGI-S, PGI-C, and Hb level (g/dL). Three different anchor-based analyses were applied: mean score change, receiver operating characteristic (ROC) curve (model-based adjusted), and logistic regression (model-based). A detailed description is available in the Supplementary Material .

For the qualitative study, 22 patients with CAD expressed interest in participating in the concept elicitation interviews. Out of the 22 patients, 18 were contacted and screened, and 16 patients were found eligible. The 16 patients were white, and their mean age was 67.8 years. Most of these patients were female (62.5%) with three-fourths having a college or professional degree. The patient population was geographically diverse, with half of the patients residing in the South or Midwest of the USA. The average time since CAD diagnosis was approximately 6 years; 68.8% of these patients had been diagnosed with CAD within the last 1.5 years. Fatigue was the most common and bothersome symptom associated with CAD in the concept elicitation interviews and had a significant impact on a patient’s ability to perform routine daily activities. All patients, except one (15/16; 93.8%), reported feeling fatigued due to CAD. Among these, 12 patients (80%) experienced fatigue daily, while three patients (20%) reported feeling fatigued few days per week. Overall, 10 (66.7%) of these 15 patients described their fatigue/tiredness/lack of energy as physical compared with two patients (13.3%) who described it as mental and three patients (20%) who described it as both physical and mental.

Following the concept elicitation portion of the interview, participants reviewed and provided feedback on the FACIT-Fatigue scale, which was sent to each participant before their scheduled interview. The impacts identified by patients in the concept elicitation interviews mapped well with the item content measuring fatigue impact in the FACIT-Fatigue questionnaire. Overall, patients were able to understand and interpret the FACIT-Fatigue scale instructions, items, response options, and recall period without any problem. Nearly all patients (n = 15; 93.8%) reported that it was easy for them to recall the past 7 days. Each item of the FACIT-Fatigue questionnaire was reported to be clear and easy to understand by most of the patients (n = 15; 93.8%). All patients could answer all 13 items; however, one patient reported that she was unsure what “listless” or “washed out” meant.

All patients (n = 16; 100%) reported that the FACIT-Fatigue questionnaire could capture their experiences with CAD-related fatigue. All patients (n = 16; 100%) reported that FACIT-Fatigue measured a concept important to them and was not missing any important elements. However, none of the items were identified as most important by the majority of patients ( Supplementary Table 1 ). Two items “I feel fatigued” and “I have trouble starting things because I’m tired” were the most frequently selected items and considered most important to patients. In contrast, three items were not identified by any of the patients as being important to them: “I am too tired to eat,” “I need help doing my usual activities,” and “I need to sleep during the day” ( Supplementary Table 1 ).

The cognitive debriefing interviews that were conducted with an additional set of 21 patients with CAD confirmed that the items of the FACIT-Fatigue questionnaire were easy to understand and relevant to the experience of patients with CAD.

The internal consistency reliability of the FACIT-Fatigue scale far exceeded the minimum threshold for acceptable reliability. Cronbach’s alpha coefficient was 0.94 at baseline and 0.96 at Week 26. Deletion of one item at a time did not change the magnitude of Cronbach’s alpha coefficient.

Data from the pooled analytic sample reported a correlation of 0.24 between FACIT-Fatigue scores and Hb levels at baseline and that of 0.39 at Week 26 ( Table 2 ). The lower correlation at baseline reflected the truncated distribution of Hb levels due to the trial inclusion criteria (Hb ≤ 10.0 g/dL in both CARDINAL and CADENZA studies).

FACIT-Fatigue scores were highly correlated with the PCS and MCS measures of SF-12v2 and PGI-S at baseline and Week 26. All correlations exceeded the recommended (absolute value) level of 0.40 for “strong” convergent validity ( Table 2 ).

Data from the pooled analytic sample were utilized for known-group validity tests of the FACIT-Fatigue scale in patients with CAD at baseline and Week 26 based on Hb levels and PGI-S outcomes. The mean FACIT-Fatigue score was significantly higher (indicating less fatigue) among patients in the higher Hb level group (≥12 g/dL) than the lower Hb level group (<12 g/dL) at Week 26, while it was numerically higher in patients in the ≥9 g/dL Hb group than the ≤9 g/dL Hb group at baseline ( Table 3 ).

For PGI-S at baseline, patients reporting “severe” fatigue had the lowest FACIT-Fatigue score, mean (SD): 16.8 (3.4), followed by patients reporting “moderate” fatigue severity: 30.6 (10.3). Patients reporting “none/mild” fatigue severity had the highest FACIT-Fatigue score – none: 47.7 (3.6); mild: 37.5 (5.7). Similarly, mean FACIT-Fatigue scores differed significantly (P <0.001) across patients who differed in the patient global assessment of fatigue severity at Week 26. Patients reporting “severe” fatigue had the lowest FACIT-Fatigue score, mean (SD): 19.2 (8.2), followed by patients reporting “moderate” fatigue severity: 35.7 (6.2). Patients reporting “none/mild” fatigue severity had the highest FACIT-Fatigue score – none: 47.3 (6.4); mild: 45.3 (3.2) ( Table 3 ).

Changes in FACIT-Fatigue scores were moderately correlated with changes in Hb levels (r = 0.52, P <0.001), SF-12v2 PCS (r = 0.45, P <0.001), and MCS (r = 0.59, P <0.001) measures, and PGI-S (r = −0.76, P <0.001). The differences in mean changes in FACIT-Fatigue scores between Hb responders and non-responders were large and statistically significant (11.3 vs. 1.5, P = 0.0045) ( Table 4 ). Mean FACIT-Fatigue score changes were also statistically significantly different across categories of the PGI-C (P = 0.0027) ( Table 4 ). The mean improvement in FACIT-Fatigue scores was largest among patient groups reporting the largest improvement in PGI-C, while patients who reported that their PGI-C was the same as or worse than baseline showed little change or worsening in FACIT-Fatigue scores.

Anchor-based estimates: Polychoric correlations between FACIT-Fatigue scores and each of the four anchors (GH01 of SF-12v2, PGI-S, PGI-C, and Hb level) were computed at baseline, at Week 26, and for the change in score (from baseline to Week 26). These correlations mostly exceeded 0.40, indicating moderate-to-strong associations and supporting the use of these anchors to estimate the MWPC for FACIT-Fatigue in CAD ( Supplementary Table 3 ).

MWPC thresholds for FACIT-Fatigue in CAD, as estimated from the anchor-based methods, are presented in Table 5 . In the mean change approach, the subset of patients ranged between 10 and 30 across the four anchors. The lowest mean (SD) change estimate in FACIT-Fatigue score was 9.2 (16.0) for PGI-C, and the highest was 15.7 (10.6) for PGI-S. The analyses of mean change produced the highest MWPC estimates of any of the analyses. In the ROC curve approach, the MWPC estimates ranged from 2.0 for Hb level to 8.0 for GH01. In the logistic regression approach, the MWPC estimates were highest for GH01 (8.9) and lowest for PGI-C (4.2).

Distribution-based estimates: Using effect size, the MWPC based on one-half of SD of FACIT-Fatigue at baseline was 5.6. The MWPC based on the standard error of measurement, which incorporates the SD and the internal consistency reliability of the FACIT-Fatigue at baseline, was estimated to be 2.8.

Retained MWPC: FACIT-Fatigue MWPC estimates produced from the anchor-based methods ranged from 2.0 to 15.7 (interquartile range [IQR]: 4.8–9.7) with a median of 7.5. The mean score change analyses were highly sensitive to distributional characteristics of data (outliers) given that only a subset of the patient data was used (the most significant limitation of the mean change method). When considering the ROC and regression-based analyses only, the IQR of MWPC estimates was 4.1–7.3, and individual responder definitions were in the range of 5–8 points, where “5” is the lowest recommended MWPC threshold for FACIT-Fatigue in CAD, and “8” is the most stringent threshold.

The CARDINAL and CADENZA trials had small patient numbers. To mitigate the impact of this, and the single-arm design of the CARDINAL trial, data were pooled from both trials. Nevertheless, the pooled sample size remained small with regard to the reported analyses, in particular the fit statistics. Although three of the four pre-specified fit statistics suggested uni-dimensionality, one (RMSEA) did not meet the prespecified criterion. Traditional cutoffs and standards for CFA fit statistics are not universally recommended when evaluating dimensionality. Cook et al. concluded that an investigative approach such as this is favored over reliance on published criteria, and that bifactor analysis is appealing when one wishes to account for evaluation of the relative impact of secondary dimensions such as experience and impact. These results would therefore benefit from replication in a larger sample (23).

The post hoc nature of these analyses remains a limitation.