Screening for Sleep Apnea-Hypopnea

Cost-effectiveness Analysis of Nocturnal Oximetry as a Method of Screening for Sleep Apnea-Hypopnea Syndrome* Lawrence J. Epstein, MD, FCCP;f and Gina R. Dorlac, MDt Study objective: Determine the utility of nocturnal oximetry as a screening tool for sleep apnea-hypopnea syndrome (SAHS) compared with polysomnography (PSG). Design: Cost-effectiveness analysis based on retrospective review of overnight sleep studies. Setting: United States Air Force tertiary teaching hospital. Patients: One hundred consecutive patients evaluated for SAHS by overnight sleep study. Intervention: Participants underwent PSG and oximetry on the same night. Patients with obstructive sleep apnea had a continuous positive airway pressure trial. Measurements: Oximetry was abnormal when ^10 events per hour occurred. Two criteria were evaluated. A "deep" pattern of >4% change in oxyhemoglobin saturation to <90%, and a "fluctuating" pattern of repetitive short-duration fluctuations in saturation. The diagnostic accuracy of both methods was compared with PSG. Cost-effectiveness of screening oximetry was compared with PSG alone and use of split-night studies. Results: The fluctuating pattern had a greater sensitivity and negative predictive value, while the deep pattern had a greater specificity and positive predictive value. Oximetry screening using the fluctuating pattern was not as sensitive as PSG for detecting patients with mild disease; 17 of 28 patients (61%) with normal oximetry results had treatable conditions detected by PSG. Cost analysis showed that screening oximetry would save $4,290/100 patients but with considerable loss of diagnostic accuracy. Conclusion: Screening oximetry is not cost-effective because of poor diagnostic accuracy despite increased sensitivity using the fluctuating pattern. Greater savings, without loss of diagnostic accuracy, may be achieved through increased utilization of split-night PSGs. (CHEST 1998; 113:97-103) Key words: cost-effectiveness; obstructive sleep apnea; polysomnography; pulse oximetry Abbreviations: AHI=apnea-hypopnea index; BMI^body mass index; CPAP=continuous positive airway pressure; OSA=obstructive sleep apnea; PSG=polysomnogram; SAHS sleep apnea-hypopnea syndrome; Sa02=oxyhemoglobin = saturation; SDB sleep disordered breathing; UARS = = upper airway resistance syndrome {\ bstructive sleep apnea (OSA) is a common ^-^ disorder with significant cardiovascular morbid ity and effects on daytime performance. Recent studies have estimated the prevalence of the sleep apnea-hypopnea syndrome (SAHS), OSA plus clini- *From the Department of Pulmonary and Critical Care Medi cine, Division of Medicine, Wilford Hall Medical Center, Lackland Air Force Base, San Antonio, Tex. fcurrently at the Pulmonary and Critical Care Medicine Divi sion, Brockton/West Roxbury VA Medical Center and Harvard Medical School, Cambridge, Mass. +Currently at the Department of Pulmonary and Critical Care Medicine, David Grant Medical Center, Travis Air Force Base, Calif. The views expressed in this article are those of the authors and do not reflect official policy of the Department of Defense or other Departments ofthe US Government. Manuscript received January 7, 1997; revision accepted July 1. cal symptoms, at 4% of men and 2% of women in a middle-aged population,1 with higher rates in elderly populations.2 Overnight polysomnography (PSG) is the standard for diagnosis of OSA, but it is expensive and resource intensive. For this reason, investigators are searching for alternative means for diagnosing OSA. Pulse oximetry has been proposed as a useful diagnostic and screening tool. The lack of airflow during apneic periods can lead to recurrent episodes of hypoxemia that can be detected on oximetry as fluctuations in oxyhemoglobin saturation (Sa02). Williams et al3 described the use of nocturnal pulse oximetry to screen patients for OSA. Using a de crease in oxygen saturation >4% and to a value <90% as their criteria for significant desaturation, CHEST / 113 / 1 / JANUARY, 1998 97

they found that pulse oximetry was specific but not sensitive for detecting OSA. The addition of a clinical score improved sensitivity but did not eliminate false-negative results. Series et al4 reported improve ment in detection by using different criteria for positive screening oximetry: the presence of repeti tive, short-duration fluctuations in Sa02 without any absolute value decrease in saturation. They con cluded that SAHS could be ruled out by normal and that pulse oximetry results using their criteria, patients with abnormal oximetry screens should go on to PSG for definitive diagnosis. The improvement in detection led to an elevation in the false-positive rate, which resulted in a large number of additional PSGs. To be an effective screening tool for SAHS, pulse oximetry must be able to screen out patients without disease and detect patients with all levels of disease severity in a manner that is less expensive than current diagnostic procedures. We investigated the utility of pulse oximetry as a screening tool for SAHS by comparing both diagnostic criteria directly with PSG and evaluating the cost-effectiveness of a nostic diag algorithm employing screening oximetry. Patients Materials and Methods We retrospectively reviewed the sleep studies of 100 consec utive patients referred for evaluation for possible sleep-disor dered breathing (SDB) who underwent PSG in our sleep center. We studied 93 men and seven women, ages 19 to 72 years (mean, 39.3 years), with a mean body mass index (BMI) of 28.2±4.4 kg/m2 (mean±sd). Our referral base consists primarily of active duty military personnel, but also includes family members and retired military members. Patients were referred to the sleep center by their primary care physicians. After examination by sleep medicine physician, a a sleep study was ordered because of suspicion of the presence of SAHS. Protocol Sleep studies were performed in a monitored facility. All measurements were recorded on a 16-channel polygraph (Grass Instruments; Quincy, Mass) running at 10 mm/s and included the determination of sleep stages (with EEG, electro-oculogram, submental electromyogram), nasal/oral airflow with thermistor (Grass Instruments), ECG, thoracoabdominal movements by inductive plethysmography (Ambulatory Monitoring Inc; Ardsley, NY), limb electromyogram, and Sa02 with pulse oximeter (Ohmeda; Madison, Wis). The oximeter signal was also sent to a strip chart recorder that provided a separate real-time tracing of Sa02. Sleep stages were defined in 30-s epochs according to standard criteria.5 An apneic event was defined as a cessation of nasal/oral airflow for at least 10 s, while a hypopnea was defined as a >50% decrement in nasal/oral airflow for at least 10 s associated with either an arousal or oxyhemoglobin desaturation >4%. An arousal was defined by an EEG frequency shift to alpha range for at least 3 s.6 An apnea-hypopnea index (AHI) was calculated using the following formula: number of apneas+number of hypopneas/total sleep time (hours). Split-night PSGs were performed on patients with significant SDB. After the first episode of rapid eye movement sleep or 3 h in bed, whichever came first, the sleep technicians evaluated the records for OSA. Patients with an AHI >20 were started on a regimen of continuous positive airway pressure (CPAP) that was titrated to a pressure level adequate to eliminate OSA and snoring. Data Analysis PSG recordings were scored manually by a single registered PSG technician. The diagnosis of SAHS was made when the AHI on PSG was >10/h. The oximetry strip chart tracings were examined by single physician interpreter blinded a to the results of the PSG. Oximetry was scored by two methods: the standard "deep" pattern of >4% desaturation to below 90%3 (Fig 1), and the "fluctuating" pattern of either large desaturations or fluctuations low-amplitude periodic using neither a minimum decrease in Sa02 levels nor a threshold minimum saturation level4 (Fig 2). Oximetry results were classified as abnormal by the presence of repetitive desaturations (>10/h) followed by a rapid return to baseline Sa02 levels. The interpretations were verified by a second scoring physician, blinded to the PSG results and the results of the first scorer, on 10 randomly selected tracings, with a 100% concurrence rate. Scoring of both oximetry and PSG was performed only on die pre-cpap diagnostic portions ofthe split-night studies. Other types of SDB diagnosed by PSG included mild OSA (5^AHI<10), and upper airway resistance syndrome (UARS), defined by the presence of a discrepancy between the arousal index and the AHI with evidence of a respira tory etiology for the arousals, such as crescendo snoring or paradox ical abdominal/thoracic movements. Cost Analysis Two diagnostic algorithms were compared using the data from the initial PSGs and oximetry (Fig 3). In the first algorithm, described by Series et al,4 oximetry was used as the initial screening test, and all patients with abnormal oximetry results went on to PSG for definitive diagnosis. For the second algorithm, all patients had an initial PSG. In both algorithms, all patients diagnosed as having SAHS underwent a CPAP titration trial for treatment. The CPAP titration trial is equiv alent to PSG in performance and cost. Assumptions made for the cost analysis wrere as follows: the results of the most sensitive oximetry criteria (deep vs fluctuating) would be used in the cost analysis, and CPAP titration would require only single additional night. The a cost analysis was performed using price estimates obtained by telephone survey of local sleep centers. The mean charge reported for oximetry wras $294 (range, $125 to 464); the mean charge for interpreted PSG and CPAP titration trial was $1,123, (range, $1,000 to 1,217). The costs were calculated at both the mean and the lowest prices. Previous studies have shown that certain patients can undergo both diagnosis and CPAP titration in a single night of study ("split-night" study).7-8 In a split-night study, CPAP titration is initiated if a predetermined OSA severity threshold is crossed during the initial portion of the study. Sanders et al9 have shown a high correlation between indexes of OSA obtained in the first 2 h of a study compared with those calculated over the entire night.9 We performed an additional analysis to evaluate the effect on cost of utilizing split-night studies. The cost ofthe initial PSG algorithm was recalculated based on the number of patients with an AHI >20 who had split-night studies. The cost of a split-night study is the same as a diagnostic PSG or a CPAP titration. For each split-night study, only one PSG charge was added to the total instead of two. 98 Clinical Investigations

Figure 1. Deep pattern. Example of oximetry tracing scored as abnormal by deep pattern; repetitive desaturations >4% to below 90%. Figure 2. Fluctuating pattern. Example of oximetry tracing scored as abnormal by the fluctuating pattern; low-amplitude periodic fluctuations without threshold levels for change in Sa02 levels or minimum saturation level. In particular, note variations on the right side of the tracing. Results A diagnosis of SAHS was made by PSG in 53 ofthe 100 patients studied, with a mean AHI of 32.4±22.1. The ability of oximetry to detect SAHS varied ac cording to the diagnostic criteria used, and is shown in Table 1. A comparison ofthe diagnostic accuracy of the two oximetry criteria for detecting an AHI CHEST / 113 / 1 / JANUARY, 1998 99

A. Initial screening oximetry abnormal normal?psg<abnormal-? CPAP titration normal-?no further evaluation -*? no further evaluation B. Initial PSG abnormal normal CPAP titration no further evaluation Figure 3. Diagnostic decision trees. A: algorithm utilizing oximetry as initial screening Algorithm using PSG as initial test. test. B: >10/h, and the effect of weight on diagnostic accu racy, is summarized in Table 2. Use ofthe fluctuating pattern for oximetry interpretation resulted in a higher sensitivity and negative predictive value, but decreased specificity and positive predictive value. Roth methods were more sensitive in patients with an increased RMI. The diagnostic algorithms applied using the fluctuating pattern because of were the greater sensitivity of this method. The resulting values were utilized in the cost analysis. The results of the cost analysis are summarized in to PSG Table 3. Use of screening oximetry prior would have saved $4,290 per 100 patients evaluated compared to initial PSG testing if all patients with SAHS had required a second night study for CPAP titration. The cost savings calculated with the lowest prices increased to $17,500 per 100 patients. How ever, despite the cost savings, 17 of 28 patients with normal oximetry results (61%) had a treatable disor der missed by screening oximetry alone. Two pa tients had an AHI >10 but normal oximetry results (false negatives) and 15 others had an elevated arousal index identified on PSG, indicating a disor der of sleep fragmentation (mean arousal index of 18.5±5.8/h). Diagnoses in this group included eight patients with mild OSA (5/h>AHI<10/h) and addi tional respiratory related arousals due to increased upper airway resistance, four with UARS, one pa tient with periodic limb movements of sleep, and two with periodic limb movements of sleep plus mild OSA. Table 1.Contingency Tables Comparing Oximetry and PSG Using Two Oximetry Diagnostic Criteria PSG Results Oximetry Abnormal 51 Normal 2 Fluctuating Pattern Deep Pattern AHI >10 AHI <10 AHI >10 AHI <10 21 26 39 14 5 42 Split-night studies were successfully performed in 21 of 33 patients with an AHI>20. The decrease in the number of PSGs performed due to the use of split-night studies resulted in the greatest cost sav ings (Table 3), with a reduction of $19,293/100 patients compared to the initial oximetry algorithm. Discussion This study compared two different diagnostic cri teria for interpreting nocturnal pulse oximetry with standard PSG evaluation for the detection of SAHS. In addition, we evaluated the cost-effectiveness of pulse oximetry as a screening tool for the detection of SAHS. We found that the use of screening pulse oximetry resulted in only a small cost savings with a significant reduction in diagnostic accuracy. Our study confirms the findings of Series et al4 that using a less rigid criterion for interpreting oximetry improved its ability to detect SAHS, in creasing both the sensitivity and negative predictive value by almost 20% to >90% each. However, this improvement was achieved with a significant reduc tion in specificity, with 29% of all abnormal oximetry test results being false-positives. Despite the im provement in SAHS detection, 61% of patients with normal results of oximetry, 17% of all the patients, had a treatable sleep disorder that would have remained undetected. Applying a diagnostic algorithm utilizing oximetry as a screening tool to guide decisions on which patients should undergo PSG did result in a small cost savings compared to using PSG alone. However, the consequence of using screening oximetry was that a significant number of patients with sleep disorders that cause excessive sleepiness would re main undiagnosed and untreated. The high cost of sleepiness-related motor-vehicle, work-related, and home-based accidents10 could negate these small cost savings. 100 Clinical Investigations

Sensitivity, % Specificity, % Positive PV, % Negative PV, % Table 2.Diagnostic Accuracy of Oximetry for Detecting an AHI >10* BMI, kg/m2 All Patients <30 Deep Fluctuating Deep Fluctuating Deep 73.6 89.4 88.6 75.0 96.2 55.3 70.8 92.9 *Deep=>4% desaturation to <90%; fluctuating=repetitive fluctuations in saturation; PV=predictive value. 65.6 84.6 77.8 75.0 93.8 56.4 63.8 91.7 85.7 87.5 94.7 70 >30 Fluctuating 100 37.5 80.8 100 Screening oximetry was most successful in detect ing SAHS in patients with a high likelihood of having OSA (those with an elevated RMI) or those with more severe disease (frequent desaturations >4%). However, this is the subset of SAHS patients likely most to undergo successful split-night studies, which we found to give the greatest cost savings. In patients with milder disease, normal results of nocturnal oximetry did not mean the absence of SDR. As a result, additional testing would be necessary, negat ing the usefulness of the screening tool. There are several reasons for the discrepancy between our results and those of Series et al.4 One is difference in methods. The Series et al study did not compare oximetry and PSG on the same night. There could have been discrepancies between the oximetry and the PSG due to first-night in laboratory effects or positional effects11 if studied on different nights. All of our studies were performed in the sleep laboratory, while the Series et al study utilized home oximetry. This should have improved the quality, and thus accuracy, of the oximetry tracings in our study, since technicians could correct problems giving poor signals. Our definition of hypopnea was slightly different in that a reduction in airflow was scored as a hypopnea if it was accompanied by a desaturation or an arousal, in contrast to only being scored if accompanied by a desaturation in the Series et al study. This may have led to a higher number of abnormal studies by PSG in our study. The number of studies affected by this difference is likely small; however, this highlights the problem of detecting patients with mild disease using oximetry. Guilleminault et al12 have shown that clinically significant sleep fragmentation, amenable to treatment, can Table 3.Cost Analysis Algorithm Initial oximetry Initial PSG Initial PSG/split-night Total Cost $167,529 $171,819 $148,236 occur in the absence of desaturation. Another possi ble source of bias was the use of split-night PSGs that reduced the amount of time available to detect desaturations by oximetry to <3 h. However, this did not influence the outcome of the study since oxim etry results were abnormal in all of the patients who had split-night studies. A major difference between the two studies is the populations investigated. Although both popula tions were referred for suspected SDR, the charac teristics of the two groups in were different in one major factor, body weight. The RMI for the Series et al4 group was 31.7±0.8 vs 28.2±0.4 kg/m2 (mean±se) in ours. The population we studied, drawn from military personnel and their families, tended to be thinner and, as we have shown previ ously, have milder disease.13 As a result, the average AHI for the patients with SAHS was greater in the Series et al study, 38.1±2.5, than in ours, 32.4±2.2 (mean±se). The effect of body mass on the accu racy of nocturnal oximetry can be seen in Table 2; the sensitivity and negative predictive values are less in the RMI <30 group compared to the >30 group. In both studies, the false-negative oximetry occurred tests in patients with lower RMIs, 23.3±1.0 in ours and 29.8±2.7 kg/m2 in the Series et al study. It is important to distinguish between the use of overnight oximetry as a diagnostic vs a screening tool. This study shows that the fluctuating for scoring oximetry criterion is accurate for detecting an AHI >10, as shown by the high sensitivity and positive predictive value. However, its usefulness as a diagnostic tool is limited by the high rate of false-positives, which requires that all patients with abnormal oximetry results undergo PSG anyway to confirm the diagnosis. Svanburg et al14 similarly found that OSA could be diagnosed by oximetry with a high sensitivity but a low specificity, requiring that patients with abnormal results of studies needed further evaluation with PSG. Douglas et al15 found to two thirds of that oximetry could diagnose up patients suspected of having SAHS, but could not detect disease in the rest, mostly patients with mild CHEST/113/1 /JANUARY, 1998 101

and moderate severity SAHS. The specificity of oximetry is improved using the deep criteria of Williams et al3 but at the cost of significant reduction in ability to detect disease. Using oximetry as the initial diagnostic test might significantly reduce cost if oximetiy could be used as the only diagnostic tool and the patients could go directly to CPAP on the basis of oximetry alone. However, the reduced spec ificity means some patients without OSA would be started on a regimen of CPAP. Detection by oxime try is most accurate in patients with severe disease, the same group manageable by split-night studies. Use of split-night studies in a significant portion of these patients allows diagnosis and treatment with one test and reduces the need for preliminary night oximetry. These findings over raise doubts about the utility of oximetry as a diagnostic tool. Its appropriate role may be when PSG is not readily available and alternative methods are needed for diagnosis. The value of oximetry as a screening tool is also questionable. A useful screening test should have several features; it should detect almost all patients with disease (high sensitivity), a normal test result should eliminate the possibility of disease (high specificity), screening errors requiring additional costly or invasive testing should be minimized (low false-positive and false-negative rates), and it should be widely applicable, reducing cost and resource utilization compared with other tests.1618 Screening tests should err on the side of higher sensitivity, detecting all disease, with the cost being a higher false-positive rate requiring further testing. This will reduce cost if conducting a large number of screen ing tests saves performing a significant number of the more expensive tests. Screening pulse oximetry does not meet these standards. Nocturnal oximetry scored by the fluctuating pattern has the highest sensitivity for detecting SAHS and potentially could be a screening device. The sensitivity of nocturnal oximetry is high if the only criterion for SDR is an AHI >10/h. All of the patients studied in this evaluation were referred with complaints of daytime sleepiness or disturbed sleep, giving a high pretest likelihood of a sleep disorder. Yet, 17 of 28 patients with treatable sleep disorders had normal oximetry results. Using a screening tool that does not detect mild forms ofthe disease, such as oximetry, will lead to a large number of patients with undetected and untreated disease. To detect these disorders, patients with normal oximetry results would still need to have a PSG, thereby eliminating the cost savings of using a screening test. This study found that more money can be saved, with greater diagnostic accuracy, by increased use of split-night studies, eliminating the need for screening oximetry by combining diagnosis and treatment into one night. Multiple studies have compared PSG and oxime try in an attempt to find a screening test for SAHS. Several studies have been able to obtain high sensi tivity for detecting OSA by manipulating the AHI level detected and criteria for an abnormal test result.1922 However, they all reported low specificity values, did not take into account milder forms of disease, and concluded that all abnormal oximetry results required PSG for confirmation. Three other studies evaluated oximetiy vs PSG with respect to severity of disease and found that oximetry suitable was a screen for patients with moderate and se vere OSA but was inadequate for patients with milder cases.15'2324 Only one study looked specifi cally at patients with milder forms of disease, includ ing UARS. Yamashiro and Kryger25 studied patients with all the varieties of SDR and compared results from same night oximetry and PSG. Oximetry de tected all patients with moderate and severe OSA. However, 30% of the patients who were normal diagnosed by oximetry were found to have UARS on PSG. Similar to us, they concluded that oximetry was a poor screening tool since normal results did not rule out disease and all abnormal results required further PSG evaluation. To our knowledge, our is the study only one to apply a cost-effectiveness analysis to the issue of screening for SAHS with oximetry. We conclude that the sensitivity of oximetry screening as a test for SAHS can be improved with the use of less rigid criteria for interpretation of an abnormal test result. However, the use of nocturnal oximetry as a screening tool does not appear to be justified on the basis of our cost-effectiveness anal ysis. Only minor cost savings are achieved by ing patients with oximetiy before PSG with screen a icant loss of signif diagnostic accuracy. When compared with oximetiy7 screening, greater savings, without loss of diagnostic accuracy, may be achieved through increased utilization of split-night studies. Methods of screening for OSA other than oximetiy, such as multiple parameter monitors or portable PSG, should be evaluated for cost-effectiveness. ACKNOWLEDGMENTS: The authors thank William Beninati, MD, and Isabel Diaz for assistance in data collection, and James H. Henderson, MD, and Dorothy Cunningham, MD, for their editorial assistance. References 1 Young T, Palta M, Dempsey J, et al. The occurrence of sleep disordered breathing among middle-aged adults. N Engl J Med 1993; 328:1230-35 2 Ancoli-Israel S, Kripke DF, Mason W, et al. Sleep apnea and periodic movements in an aging sample. J Gerontol 1985; 40:419-25 3 Williams AJ, Yu G, Santiago S, et al. Screening for sleep apnea using pulse oximetry and a clinical score. Chest 1991; 100:631-35 102 Clinical Investigations

4 Series F, Marc I, Cormier Y, et al. Utility of nocturnal home oximetry for case finding in patients with suspected sleep apnea hypopnea syndrome. Ann Intern Med 1993; 119: 449-53 5 Rechtschaffen A, Kales A. A manual of standardized termi nology, techniques, and scoring system for sleep stages of human subjects. Los Angeles: Brain Information Service, Brain Information Institute, University of California, 1968 6 American Sleep Disorders Association. EEG arousals: scoring rules and examples. Sleep 1992; 15:174-84 7 Sanders MH, Kern NB, Costantino JP, et al. Adequacy of prescribing positive airway pressure therapy by mask for sleep apnea on the basis of a partial-night trial. Am Rev Respir 1993; 147:1169-74 Dis 8 Iber C, O'Brien C, Schluter J, et al. Single night studies in obstructive sleep apnea. Sleep 1991; 14:383-85 9 Sanders MH, Black J, Costantino JP, et al. Diagnosis of sleep-disordered breathing by half-night polysomnography. Am Rev Respir Dis 1991; 144:1256-61 10 Leger D. The cost of sleep-related accidents: a report for the national commission on sleep disorders research. Sleep 1994; 17:84-93 11 Swieca J, Westbrook PR. Relationship between body position dependence of apnea and hypopnea and overall severity of sleep-disordered breathing [abstract]. Am J Respir Crit Care Med 1994; 49:A54 12 Guilleminault C, Stoohs R, Clerk A, et al. A cause of excessive daytime sleepiness: the upper airway resistance syndrome. Chest 1993; 104:781-87 13 Atwood C, Strollo P, Epstein L, et al. Sleep disordered breathing in the young adult: analysis of an active duty US military cohort. Chest 1994; 106:149S 14 Svanburg E, Larsson H, Carlsson-Nardlander B, et al. A limited diagnostic investigation for obstructive sleep apnea: oximetry and static charge sensitive bed. Chest 1990; 98: 1341-45 15 Douglas NJ, Thomas S, Jan MA. Clinical value of polysom nography. Lancet 1992; 339:347-50 16 United States Department of Health and Human Services. Screening in health fairs, a critical review of benefits, risks, and costs. JAMA 1984; 254:1492-98 17 Morrison AS. Screening in chronic disease: monographs in New York: Oxford University epidemiology and biostatistics. Press, 1985; 139-48 18 Macfarlane A, Sefi S, Cordeiro M. Child health, the screening tests. Oxford: Oxford University Press, 1989:18 19 Farney RJ, Walker LE, Jensen RL, et al. Ear oximetry detect to apnea and differentiate rapid eye movement (REM) and non-rem (NREM) sleep: screening for the sleep apnea syndrome. Chest 1986; 89:533-39 20 Rauscher H, Popp W, Zwick H. Model for investigating snorers with suspected sleep apnoea. Thorax 1993; 48:275-79 21 Gyulay S, Olson LG, Hensley MJ, et al. A comparison of clinical assessment and home oximetry in the diagnosis of obstructive sleep apnea. Am Rev Respir Dis 1993; 147:50-53 22 Levy P, Pepin JL, Deschaux-Blanc C, et al. Accuracy of oximetry for detection of respiratory disturbances in sleep apnea syndrome. Chest 1996; 109:395-99 23 Cooper BG, Veale D, Griffiths CJ, et al. Value of nocturnal oxygen saturation as a screening test for sleep apnoea. Thorax 1991; 46:586-88 24 Rodriguez JM, de Lucas P, Izquierdo JL, et al. Nocturnal oximetry in the screening of obstructive sleep apnea syn drome. Chest 1994; 106:147S 25 Yamashiro Y, Kryger MH. Nocturnal oximetry7: ing tool for sleep disorders? Sleep 1995; 18:167-71 is it a screen CHEST / 113 / 1 / JANUARY, 1998 103