Symptom questionnaire shows strong agreement with autonomic neuropathy tests in Zanzibar diabetes clinic

A brief questionnaire asking people with type 2 diabetes about dizziness and faintness when they stand up has performed surprisingly well as a screening check for cardiac autonomic neuropathy in a Tanzanian hospital, according to a cross-sectional diagnostic accuracy study published in Cardiovascular Diabetology – Endocrinology Reports. The symptom domain flagged 260 participants, or 71.4%, as abnormal, while standardized cardiovascular autonomic reflex tests classified 291 participants, or 79.9%, as having neuropathy based on a composite score of at least 0.5.

The study, conducted from December 2023 to March 2024, addressed a practical gap in diabetes care. Cardiac autonomic neuropathy is a frequent but under-recognized complication of type 2 diabetes that is strongly associated with arrhythmia, silent myocardial ischemia, and cardiovascular mortality. The objective reference tests, which measure heart rate variability during deep breathing, heart rate responses to standing, blood pressure changes during the Valsalva maneuver, postural hypotension, and blood pressure responses during sustained handgrip, require dedicated equipment, standardized procedures, and trained personnel. In many low-resource settings these cardiovascular autonomic reflex tests are rarely incorporated into routine care, which leaves the condition largely undiagnosed despite its clinical importance.

Participants were recruited using systematic random sampling at the main tertiary referral hospital in Zanzibar. All were adults aged at least 18 years, permanent residents of Zanzibar, with confirmed type 2 diabetes receiving at least one antidiabetic medication. People with conditions likely to affect autonomic function independently of diabetes, and those with serious illnesses of the autonomic nervous system, were excluded. Before assessment, participants were instructed to avoid smoking, caffeine, intense exercise, and drugs that would impair cardiac autonomic function for 24 hours. Objective autonomic function was measured with an automated CAN analyser, the CAN 504 from Kody Medical Electronics, which computes autonomic indices from ECG-derived R–R intervals and blood pressure readings according to standard Ewing definitions, such as the ratio of maximum to minimum heart rate during deep breathing and the ratio of longest to shortest R–R intervals during the Valsalva maneuver.

The subjective side of the comparison came from the orthostatic intolerance domain of COMPASS-31, in which participants self-reported the frequency and severity of symptoms such as dizziness, lightheadedness, and near fainting upon standing. The domain consists of four items. The first acts as a screening question worth one point if answered yes and zero if no; a no response ends the domain with a score of zero. A yes response opens three further items covering symptom frequency from rarely to almost always, severity from mild to severe, and progression over the past year, with higher scores assigned to worsening symptoms. The raw domain score ranges from 0 to 10 and is multiplied by a weighting factor of 4.0, giving a maximum weighted score of 40. For the accuracy analysis, any weighted score above zero counted as positive.

Against the reflex-test reference standard, the questionnaire domain achieved a sensitivity of 85.9% and a specificity of 86.3%. The positive predictive value reached 96.2%, meaning that a positive symptom screen was very likely to coincide with objective neuropathy in this high-prevalence clinic population, while the negative predictive value was 60.6%, so a negative screen left a considerable chance of underlying disease. Positive and negative likelihood ratios were 6.27 and 0.16 respectively. Internal consistency of the orthostatic intolerance domain was good, with a Cronbach’s alpha of 0.82. Among the 260 participants reporting symptoms on standing, 46.2% reported them rarely, 34.2% occasionally, 17.3% frequently, and 2.3% almost always; most described the symptoms as mild, 78.5%, while 19.2% reported moderate and 2.3% severe symptoms.

Beneath the strong headline numbers, however, the item-level picture was far less impressive. Spearman’s rank correlation showed weak associations between individual questionnaire items and the analyser findings, and only the symptom progression item reached statistical significance, with a weak positive correlation of r = 0.113 and p = 0.040. Receiver operating characteristic analyses of individual components produced areas under the curve of only 0.51 to 0.56, close to the 0.5 that would indicate discrimination no better than chance. The authors attribute this apparent contradiction to the difference between agreement and discrimination: Cohen’s kappa evaluates concordance in binary classification, whereas ROC analysis assesses the ability to separate individuals across thresholds. The very high prevalence of neuropathy in the clinic, 79.9%, likely inflated agreement while leaving discrimination poor.

The study population itself helps explain the high prevalence and the performance profile. The mean age of participants was 56.0 ± 10.9 years, most were aged 40 to 60 years, and 66.8% were female. The mean body mass index was 28.3 ± 6.4 kg/m², with 68.1% of participants overweight or obese, and mean systolic and diastolic blood pressures were 151.1 ± 26.0 mmHg and 87.8 ± 15.3 mmHg, with 84.1% classified as hypertensive. Median sedentary time was 45.0 minutes per day. Reported prevalence estimates for cardiac autonomic neuropathy vary widely depending on diagnostic criteria and assessment methods, ranging from 22% to 62% in the literature, and a tertiary referral clinic would be expected to sit toward the higher end of that range.

The authors also emphasize the conceptual gap between what the two measurements capture. COMPASS-31 records the subjective experience of autonomic symptoms, whereas the reflex tests detect physiological responses reflecting cardiovascular autonomic dysfunction; the constructs are related but not identical. Some individuals may have physiological autonomic impairment without prominent symptoms, while others may report dizziness or light-headedness for reasons unrelated to autonomic dysfunction. Autonomic dysfunction in diabetes may initially affect vagal fibers before producing overt hemodynamic manifestations, which may partly explain why physiological abnormalities can precede clinically recognizable symptoms. Symptom recognition and reporting can also be shaped by language, health literacy, sociocultural interpretation of bodily symptoms, and the normalization of chronic dizziness or fatigue, factors the authors flag as especially relevant in low-resource settings, although formal evaluation of cultural or linguistic influences on responses was not performed.

Placed against earlier validation work, the findings are broadly consistent with studies in other regions in which COMPASS-31 showed fair correlations with objective autonomic measures and low diagnostic accuracy, although the agreement observed in Zanzibar appears slightly higher than in some previous validation studies, which the authors suggest may reflect differences in study populations, disease duration, and autonomic testing protocols. The authors note that symptom-based questionnaires cannot replace objective autonomic testing, and that cardiovascular autonomic reflex tests remain the reference standard for diagnosing the condition. They point to proposed stepwise strategies in which an abnormal symptom screen prompts further autonomic testing, while concordant normal results make neuropathy unlikely, and note that people who score poorly on the questionnaire must still be assessed with the full reflex battery to distinguish early disease from no disease.

The study has limitations that temper the conclusions. The single tertiary referral setting may limit generalizability and may overestimate disease burden through referral bias. The analysis was restricted to the orthostatic intolerance domain without assessing the total COMPASS-31 score, and the domain captures only one dimension of autonomic dysfunction. The high prevalence of hypertension and overweight or obesity in the population may have influenced autonomic measures, and stratified accuracy across these subgroups was not performed. Self-reported symptoms may have introduced reporting bias and misclassification, reflected in the observed false-positive and false-negative results, and the cross-sectional design precludes evaluation of the temporal relationship between symptom development and physiological dysfunction. Even so, the authors argue that a low-cost, symptom-based first-line screen could help identify people who may benefit from further cardiovascular autonomic evaluation, and they call for future studies of the full instrument, locally relevant cutoffs, and cultural or linguistic adaptation in sub-Saharan African settings.

Subject of Research: Diagnostic accuracy of the COMPASS-31 orthostatic intolerance symptom questionnaire for screening cardiac autonomic neuropathy in adults with type 2 diabetes in Zanzibar

Article Title: Diagnostic accuracy of the COMPASS-31 orthostatic intolerance domain for detecting cardiovascular autonomic neuropathy in resource-limited settings

Article References: Haji, H. T., Ali, R. M., Khamis, A. G., Njelekela, M., Ramaiya, K., & Mashili, F. L. (2026). Diagnostic accuracy of the COMPASS-31 orthostatic intolerance domain for detecting cardiovascular autonomic neuropathy in resource-limited settings. Cardiovascular Diabetology – Endocrinology Reports, 12(1), Article 46. https://doi.org/10.1186/s40842-026-00313-4

Image Credits: AI Generated

DOI: 10.1186/s40842-026-00313-4

Keywords: COMPASS-31, cardiac autonomic neuropathy, type 2 diabetes, orthostatic intolerance, diagnostic accuracy, cardiovascular autonomic reflex tests, Zanzibar, screening, Cohen's kappa, resource-limited settings, autonomic dysfunction, questionnaire