We conducted a retrospective cross-sectional study of consecutive, ambulatory male patients seen for screening, health care maintenance, and preventive service examinations, not specific for sleep-related symptoms, in the Executive Health Clinic at Mayo Clinic in Rochester, Minnesota, United States, from January 1, 2004 through December 31, 2010. Participant evaluations included history and physical examinations, as well as measures of height, weight, blood pressure, and fat distribution based on WHR. Regarding the latter, single measurements were made of the waist circumference (at the approximate midpoint between the lowest rib and iliac crest during end-expiration) and hip circumference (at the largest point around the buttocks) (30). An a priori WHR threshold of ≥1.0 was used to identify individuals at greatest cardiovascular risk based upon previous data (31). This study was approved by the Mayo Clinic Institutional Review Board. Given the higher prevalence of OSA in men as compared to women (32), and the potential impact of these findings on the risk of developing OSA (19, 20), we elected to exclusively study men.

Participant- or spouse-reported symptoms suggestive of sleep apnea (e.g., snoring, irregular nocturnal breathing) were recorded during each participant's medical record review as were history of diabetes, smoking, and hypertension. All participants completed an extensive health questionnaire (review of symptoms) specifically asking about fatigue, difficulties sleeping (e.g., inability to fall or stay asleep), irregular night-time breathing (e.g., gasping), and snoring along with an Epworth Sleepiness Scale (24) to identify the presence of excessive daytime sleepiness. Indications for OXI (e.g., obesity, sleepiness, snoring, fatigue) were determined based upon these physician notes and surveys.

Overnight OXI was completed in the participants' bedroom or hotel room after the participants received detailed verbal and written instructions on the oximeter (PalmSAT 2500, Nonin Medical Inc., Plymouth, MN, United States) which has excellent comparability with direct measurements from arterial blood (33). Participants maintained a sleep diary containing details on bedtime, periods of awakening, and substance use before bedtime in addition to any peculiarities about their night of sleep. Importantly, participants were not using supplemental oxygen or breathing-assistance devices during their OXI assessment.

Data were sampled every 4 s, collected at 75 Hz, stored locally, and then downloaded the following day for offline analysis (18). Variables included recording time, percentage of sleep time spent ≤ 90%SpO₂ (CT-90) (29), as well as minimum, maximum, and mean SpO₂. The oxygen desaturation index (ODI) was also calculated by indexing the number of desaturations ≥4%SpO₂ per hour of recording time. Findings were interpreted by pulmonologists with tests considered normal if mean SpO₂ ≥90% without frequent oscillatory variations. Abnormal tests indicative of OSA were identified as: hypoventilatory abnormality with desaturation, compatible with sleep-disordered breathing (oscillatory desaturation events), an oxygen desaturation index ≥15 (ODI≥15), and/or a CT-90 ≥1.0 defined as the percentage of recording time patients spent ≥90% (29).

Executive Health Program patients are, in general, medically stable (e.g., free of acute medical complications) and seen on an annual or semi-annual basis for healthcare maintenance and preventative care. The medical records of participants who underwent OXI were reviewed longitudinally to determine whether they underwent polysomnography or consultation at a sleep center. We determined whether OSA or another sleep diagnosis was made as well as if PAP or another treatment was recommended.

Data are reported as mean ± SD, odds ratio (OR) with 95%CIs, or n (%) throughout the manuscript. Univariate and multivariate logistic regression analyses were used to identify end point predictors; all data were checked for multicollinearity before proceeding with analysis. We considered the absence or presence of abnormal OXI, ODI ≥15 vs. ODI <15, and CT-90 ≥1.0 vs. CT-90 <1.0 as secondary outcome variables in our analysis. In addition, we developed then evaluated the performance of a predictive model in our database for abnormal OXI. Regression coefficients of the predictive models were transformed into scores (multiplied by 10 and rounded to the nearest integer). Percentage of participants with abnormal OXI data according to the aggregated score in the predictive model was also calculated. All statistical analyses were performed with statistical analysis software version 9.3 (SAS Institute Inc. Cary, NC, United States) with P < 0.05 considered statistically significant for all analyses.

A total of 1,085 participants underwent overnight OXI between January 1, 2004 and December 31, 2010; 1,063 consented to medical record review and 516 had concurrent cardiometabolic assessment. Some participants had overnight OXI for follow-up of known OSA (n = 50) and were excluded due to use of treatment devices during OXI. In the five participants who had duplicate OXI studies due to artifact during their first recording, only the second artifact-free tests were used. Of the 461 remaining participants, 393 were men and comprised the final study population (Figure 1). Demographical information as well as co-morbidities, prevalence of participant- and spouse-reported symptoms, indications for, and results from, OXI is shown in Table 1.

Figure 2 shows the prevalence of abnormal OXI, ODI ≥5, ODI ≥15, and CT-90 ≥1.0 categorized by the presence or absence of obesity with and without a WHR ≥1.0. Results from univariate analyses are shown in Table 2. Age, BMI, and WHR ≥1.0 were predictive of abnormal OXI as well as ODI ≥ 15 and CT-90 ≥ 1.0 (P < 0.05 for all); however, scores from the Epworth Sleepiness Scale were not (P = 0.62, 0.29, and 0.71, respectively). While diabetes mellitus was not indicative of abnormal OXI (P = 0.12), it was predictive of both ODI ≥ 15 and CT-90 ≥ 1.0 (P < 0.05 for both). Diagnosis of hypertension was predictive of all three indices of abnormal OXI (P < 0.01 for all) whereas smoking was not (P = 0.07–0.53). Surprisingly, none of the participant-reported symptoms were indicative of abnormal OXI (P = 0.06–0.85).

Findings from multivariate analyses are illustrated in Table 3; WHR ≥ 1.0 was the only measurement that predicted: abnormal OXI, ODI ≥ 15, and CT-90 ≥ 1.0 (P < 0.01 for all). Men ages 50–54 years were less likely to have abnormal OXI metrics relative to those ≥55 years (P < 0.05 for abnormal OXI and CT-90 ≥ 1.0, P = 0.08 for ODI ≥ 15). BMI followed a similar trend as age (≥55 years) was indicative of abnormal OXI and CT-90 ≥ 1.0 (P < 0.05 for both) but not ODI ≥ 15 (P = 0.21). Diagnoses of diabetes mellitus and hypertension were not indicative of abnormal overnight OXI (P = 0.72 and 0.12, respectively); although self-reported snoring was (P < 0.05). Participants were differentiated into non-obese (BMI < 29.9 kg/m², n = 198) and obese (≥30.0 kg/m², n = 195) to further assess the role of obesity in the multivariate models. From these analyses, WHR ≥1.0 was the only variable predictive of abnormal OXI in both obese and non-obese groups (P < 0.05 for both). Age ≥ 55 years was indicative of abnormal OXI in obese participants (P < 0.05) but this was not observed in their non-obese counterparts (P = 0.15). Being 50–54 years (P = 0.62 and 0.78) nor having a diagnosis of diabetes mellitus (P = 0.56 and 0.89) or hypertension (P = 0.22 and 0.42) was predictive of abnormal overnight OXI in either the non-obese or obese cohort, respectively. Similarly, self-reported snoring was not predictive of abnormal overnight OXI in either subgroup (P = 0.10 for both).

Findings from a stepwise regression analysis are summarized in Table 4. The prevalence of abnormal OXI was 47% for participants with aggregated score <4, 61% for those with the score of 4–8, 67% for those with the score of 9–14 and 89% for study participants with the aggregated score more or equal to 15. Of the variables assessed, WHR ≥1.0 had the highest score (8) followed by BMI ≥30.0 kg/m² (5) and self-reported snoring (4, P < 0.05 for all). Both age ≥55 years and hypertension received a score of 3; while the former was predictive of abnormal OXI (P < 0.01), the latter did not reach statistical significance (P = 0.10).

These findings are of immediate clinical usefulness as participant-reported symptoms, other than snoring, lack predictive utility for OSA (Table 3). Fortunately, WHR is relatively easy to calculate, interpret (e.g., 1.0 threshold), and technically simple to perform. We used a stepwise logistic regression analysis to develop a score to predict pulmonologist interpretation of OXI as normal or abnormal in male patients (Table 4). As an example, a 56-year-old man who snores and has a WHR ≥1.0 has an 89% chance for abnormal OXI. Importantly, the finding of abnormal overnight OXI in this study was reflective of OSA as confirmed in a subset of participants who underwent polysomnography; 91% received a diagnosis of OSA. Additionally, evidence suggests patients with OSA may have refractory hypertension until adequate PAP is initiated (46). To this point, patients triaged to overnight OXI via WHR measurements may avoid lengthy wait times for inpatient polysomnography. In doing so, they could initiate PAP treatment sooner and perhaps reduce the risk of future cardiovascular events. Collectively, our findings support the use of WHR measurements in frontline medical practice due to its technical simplicity, minimal cost, and capacity to predict cardiovascular risk (31), and now, abnormal nocturnal oxygen levels indicative of sleep disordered breathing. Indeed, a joint statement from the American Heart Association, American College of Cardiology, and The Obesity Society reported a limited recommendation for the use of waist circumference in a clinical setting due to a paucity of meta-data (47). The current study builds upon this void and may guide recommendations from these societies which closer align with those of the World Health Organization (30).

Our study is a retrospective cross-sectional analysis with inherent limitations. However, this study is a consecutive series of a relatively large number of participants in a general medical population being seen for screening examinations with OXI, rather than a select population from sleep clinics or patients referred for sleep-related concerns. Our study participants were generally healthy, ambulatory, and exclusively men; thus, our results may not apply to patients with complex problems nor necessarily extend to women. The overall health of participants in this study also makes the pre-test utility of our findings challenging; for instance, previous studies suggest 50% of patients with OSA are hypertensive (48) whereas only 39% of participants have hypertension in the current study. Nevertheless, the clinical significance of abnormal OXI was recently recognized as a predictor of post-operative cardiovascular events in patients undergoing non-cardiac surgical procedures (49). A second limitation is the potentially subjective interpretation of OXI data together with the clinically available information as our end point. However, the pulmonologists who evaluated all available OXI data, including ODI and CT-90, were blinded to the purpose of this manuscript (due to its retrospective nature) and synthesized this information arriving at a specific conclusion. It is the pulmonologist's interpretation of OXI that drives clinical decision making in our practice and likely in other clinical settings as well. An additional consideration is that polysomnography assessments were not rigorously used to follow up on OXI; as such, our ability to determine the rate of false positives or false negatives is attenuated. Nevertheless, in the 88 men who did complete a formal sleep study after abnormal OXI, 91% were found to have OSA.