MINIMAL CLINICALLY IMPORTANT DIFFERENCE AND CLINICAL INTERPRETABILITY OF THE CERVICOGENIC HEADACHE SEVERITY QUESTIONNAIRE (CEH-SEQ)

Abstract

Background: The Cervicogenic Headache Severity Questionnaire (CeH-SeQ) is a novel patient-reported outcome measure (PROM) developed to screen for and grade the severity of cervicogenic headache (CeH). Its content validity and test–retest reliability have previously been established, but the minimal clinically important difference (MCID) and the clinical interpretability of its severity bands had not yet been empirically defined.

Objective: To derive an anchor-based and distribution-based MCID for the CeH-SeQ Part 2 severity score, and to evaluate the clinical interpretability of its predefined severity bands, using data from a preliminary prospective cohort.

Methods: Thirty adults with clinically confirmed CeH (CeH-SeQ Part 1 score ≥ 8) completed the CeH-SeQ, Patient Global Impression of Change (PGIC), Numerical Pain Rating Scale (NPRS), Neck Disability Index (NDI), and Headache Impact Test-6 (HIT-6) at baseline and at 4-week follow-up. Anchor-based MCID was derived via receiver operating characteristic (ROC)/Youden Index analysis using PGIC-defined ‘minimally improved’ status as the external anchor. Distribution-based estimates (half standard deviation [SD], standard error of measurement [SEM], smallest detectable change [SDC]) were calculated from a co-located 72-hour test–retest reliability sub-study (n = 30, ICC(2,1) model). Agreement between anchor- and CeH-SeQ-defined improvement was quantified with Cohen’s kappa, and concurrent responsiveness was assessed with Spearman correlations against NPRS, NDI, and HIT-6 change scores.

Results: Mean baseline CeH-SeQ severity was 76.48% ± 13.96% (Very Severe band in 20/30 participants), improving to 74.33% ± 14.11% at 4 weeks (mean change 2.14 ± 1.21 percentage points). The anchor-based MCID was 2.0 percentage points (sensitivity 0.75, specificity 0.50, Youden J = 0.25, AUC = 0.61). Distribution-based estimates were higher than the anchor-based value: half-SD = 6.98%, SEM = 0.88%, SDC = 2.44%. Test–retest reliability was excellent (ICC(2,1) = 0.996). Agreement between PGIC- and CeH-SeQ-defined improvement was fair (κ = 0.25). Correlations between CeH-SeQ change and NPRS (ρ = 0.11, p = 0.577), NDI (ρ = 0.33, p = 0.071), and HIT-6 change (ρ = 0.32, p = 0.081) were weak to moderate and did not reach statistical significance.

Conclusion: In this preliminary cohort, the empirically derived anchor-based MCID for the CeH-SeQ (≈ 2 percentage points) was substantially smaller than the a priori projected range of 10–15%, and concurrent responsiveness against established pain and disability measures was weaker than hypothesised, despite excellent test–retest reliability. These findings support the CeH-SeQ’s measurement precision but indicate that its MCID and severity-band thresholds require confirmation in the full-scale, adequately powered validation cohort before adoption as a definitive clinical or trial-endpoint threshold.