Answers are grounded in this paper and cite their sources.
BACKGROUND: Temporomandibular disorder (TMD) affects an estimated 34% of the global population, with projections suggesting further increases. A genuine but incompletely understood association with obstructive sleep apnea (OSA) has been proposed. Reported TMD-OSA co-occurrence varies from 2% to 68% across studies-a range that conflates two different measured quantities (estimands) rather than reflecting a single inconsistent phenomenon. OBJECTIVE: To determine whether apparent heterogeneity in TMD-OSA comorbidity estimates reflects true inconsistency or results from combining non-comparable measurements-studies estimating P(TMD|OSA) versus P(OSA|TMD)-alongside differences in recruitment setting, diagnostic method, severity threshold and control-group presence. METHODS: A narrative synthesis, not a systematic review, was conducted via a structured PubMed search (inception-July 2026) cross-checked against three existing systematic reviews and meta-analyses. Nine clinic-based, instrument-confirmed studies meeting predefined eligibility criteria were extracted and compared against the population-based literature underlying a 2024 meta-analysis (six studies, pooled odds ratio [OR] 2.61). Wilson 95% confidence intervals were calculated for each proportion; given the small number and clinical heterogeneity of the studies, no formal meta-analytic pooling was attempted. RESULTS: Two of the nine clinic-based studies overlapped with the existing population-based literature (one quantitatively, one qualitatively), so the two literatures are not fully independent, though most clinic-based studies lacked a control group and were excluded from pooling. Reported P(TMD|OSA) ranged from 8.6% [95% CI 4.0-17.5] to 51.7% [41.4-61.9] among studies using RDC/TMD or DC/TMD; one study using a substantially more restrictive definition of treatment-disqualifying TMJ disease reported 2.0% [0.6-7.0] and is not directly comparable to the others. P(OSA|TMD) ranged from 28% [18.0-41.6] to 68% [48.4-82.8] across three studies. The two most rigorously controlled studies did not differ significantly from each other (z = 1.14, p = 0.26), despite one reaching borderline internal significance (p < 0.05) and the other not (p = 0.83). CONCLUSION: The best-supported conclusion is that these studies measure different quantities, in different populations, using different diagnostic protocols and severity thresholds-not that they are in true biological conflict, though the relative contribution of estimand direction to the overall heterogeneity cannot be quantified from the available studies. Future research should prospectively define which estimand is being measured, apply standardised diagnostic thresholds, adjust for shared confounders (age, sex, BMI) and report control-group characteristics and overlap with existing pooled analyses in full.
No takes yet. Share an insight, caveat, or question.
John C. Comisi (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: