ROC analysis determines optimal cutoff and doubtable interval of the Bath ankylosing spondylitis metrology index in a multicenter case control study

Abstract

In clinical practice, the Bath Ankylosing Spondylitis Metrology Index (BASMI) score of healthy individuals is more than zero. However, the lower limit of the BASMI score range for patients with ankylosing spondylitis (AS) is 0, which is inconsistent with that observed in actual clinical settings. Based on findings from a multicenter case-control study, the receiver operating characteristic (ROC) curve was used to determine the optimal cutoff value and doubtable value interval of BASMI, thereby providing a reference to facilitate the evaluation of patients with AS.Patients with AS and healthy controls were recruited, and their demographic characteristics were examined. Spinal mobility was assessed by our standardized and trained measurement team. We used the SPSS26.0 statistical software to categorize the pertinent BASMI scores of the patients and healthy controls by age and sex. We subsequently prepared the BASMI ROC curve, calculated the Youden index, and determined the optimal cutoff value and doubtable value interval. In the present study, 246 patients with AS (patient group) and 246 healthy controls (control group) were recruited. Only three (0.9%) and four (1.18%) individuals in the control and patient groups had a BASMI score of 0, respectively. The optimal cutoff value of BASMI in the general population was 1.5 points, the doubtable value interval was 0.7-2.1 points, and the area under the ROC curve was 0.867 (95% confidence interval, 0.835-0.899; P < 0.001). The optimal cutoff value of BASMI was determined as part of a preliminary analysis.A BASMI score > 1.5 may signal spinal mobility limitations; however, its clinical relevance requires validation against functional and structural progression endpoints.Furthermore, when the BASMI score of an individual range from 0.7 to 2.1, it is crucial to determine whether the individual has AS.

Keywords: Ankylosing spondylitis; BASMI; Doubtable value interval; Optimal cutoff value; ROC curve.

Fig. 1

Histograms of BASMI scores distribution in the patient and healthy control groups.The histograms illustrate the BASMI score distributions of 246 ankylosing spondylitis (AS) patients and 246 healthy controls. The median BASMI score was 2.0 (IQR: 1.6–3.4) in the patient group and 1.2 (IQR: 0.8–1.4) in the healthy group. Only 4 patients (1.6%) and 3 controls (1.2%) had a BASMI score of 0. The x-axis represents BASMI scores (range: 0–10), and the y-axis shows frequency. The right-skewed distribution in the patient group indicates greater spinal mobility limitations in AS patients.

Fig. 2

Receiver operating characteristic (ROC) curves of BASMI stratified by sex and age.ROC curves demonstrate the diagnostic performance of BASMI across subgroups (stratified by sex and age). Subgroups are differentiated by color and line style. The area under the curve (AUC) is labeled in the legend (e.g., total population: AUC = 0.867,95% CI 0.835–0.899). The x-axis represents 1-specificity (false positive rate), and the y-axis represents sensitivity (true positive rate).

Fig. 3

Analysis of the BASMI doubtable value interval in the general population.The doubtable value interval (0.7–2.1) was determined by analyzing the first derivative changes in sensitivity and specificity curves. Key inflection points are labeled: a (lower limit: 0.7, corresponding to the first significant decline in sensitivity), b (optimal cutoff value: 1.5, maximizing the Youden index), and c (upper limit: 2.1, where the specificity curve slope markedly decreases). The shaded area represents the doubtable interval, encompassing potential misdiagnosis (false positives in healthy individuals) and missed diagnosis (false negatives in AS patients).The interval (0.7–2.1) is exploratory and should not replace comprehensive diagnosis.The optimal cutoff (b) requires validation against disability/progression outcomes.

Fig. 4

BASMI doubtable value intervals stratified by sex and age.Subplots display doubtable intervals for subgroups. Each subplot specifies subgroup sample size(e.g., Males aged 25–34 years, N = 41),age, sex, and interval bounds. Dashed lines indicate optimal cutoff values, and shaded areas denote doubtable intervals. For example, females aged 25–34 years have an interval of 0.9–1.7,optimal cutoff value: 1.5.