This prevalence suggests that the syndrome is frequently undiagnosed, and it raises questions about how the condition should be identified. The current standard of practice in the United States is to use in-laboratory overnight polysomnography [3] that monitors many variables during sleep. These variables include the electroencephalogram, submental electromyogram, and electro-oculogram (all for sleep staging); respiratory airflow; respiratory effort; oxygen saturation; electrocardiogram; snoring; body position; and tibialis electromyogram (to detect nocturnal myoclonus). The precise contribution of each datum to the final diagnosis of sleep apnea is not well established. This imprecision has led to differences in currently recommended diagnostic practices between physicians in the United States [3] and in Europe [4, 5].

The possible redundancy of some of the data, along with the high cost of the test, has led to attempts to simplify the diagnostic strategy by, for example, only using oximetry [6-11]; however, the results have been conflicting. Some studies [7-10] have shown a sensitivity as low as 50% (that is, many false-negative results) but a specificity as high as 95% (that is, few false-positive results). However, the report of Series and colleagues [11] in this issue of Annals shows the reverse: high sensitivity (98%) and low specificity (48%). Differing results are in part related to the fact that certain persons may show marked desaturations with recurrent apneic events that are easily detected using oximetry, whereas others may show little desaturation with repeated hypopneas that instead lead to arousal and sleep fragmentation. A more important reason, however, for these different results is the method of analysis of the oximetry data. Studies reporting low sensitivity [7-10] have used a precise method of quantification of overnight oxygen saturation (Sa_o2) using software developed to count the number of times per hour that brief decreases of a given magnitude (for example, a 4% decrease) in oxygen saturation occur during sleep.

In contrast, the high sensitivity that Series and colleagues [11] report is based on the overall recognition of patterns of abnormality. The authors scored records as positive for sleep apnea when they observed recurrent decreases in Sa_o2 during the nocturnal recording, without defining the magnitude of decreases in Sa_o2. Thus, the sensitivity and specificity of oximetry alone in the diagnosis of sleep apnea depends on the method of data analysis used. This provides the potential for different uses of oximetry in different clinical scenarios. At one end of the spectrum of abnormality is the patient with obvious risk factors for sleep apnea (for example, increased collar size [12]) who is observed to have snorting, gasping, and apneas during sleep and has extreme excessive daytime sleepiness. In this type of patient, the finding of more than 15 episodes of desaturation of at least 4% per hour is sufficient to confirm the diagnosis and initiate therapy [10]. At the other end of the spectrum is the patient with few, if any, symptoms but with a concerned spouse, who has observed snoring during sleep. In this type of patient, a normal overnight oximetry test based on the method of Series and colleagues [11] is sufficient to exclude the syndrome. Alternatively, in this type of patient, if the amount of time during sleep when the Sa_o2 level is less than 90% is less than 1% of the total, then the diagnosis of substantial sleep apnea is excluded [10].

If oximetry alone can be of diagnostic value, at least in certain circumstances, what additional value do the other variables measured in the polysomnogram provide? In particular, is measurement of the most time-consuming variables (such as the electroencephalogram for sleep staging) necessary? Some would say no, citing the fact that the apnea/hypopnea index (number of apneas plus hypopneas per hour of sleep) is abnormal in most patients in whom sleep apnea is diagnosed, whether the index is calculated per hour of sleep or per hour in bed [9].

Because in-home studies have the potential to reduce costs, a number of equipment manufacturers have developed various devices. These include a device that monitors heart rate variation, oximetry, body position, and snoring, primarily for screening purposes [13]; a device that monitors all respiratory variables (oximetry, airflow, effort) [14]; and devices that record all variables, including the electroencephalogram, in the home. The availability of this technology suggests the possibility of a two-stage diagnostic approach. The first stage would be an unattended study in the home, after agreement about what to monitor in this setting. The second stage would be a study in the sleep laboratory for more in-depth testing to determine the effect of such interventions as nasal, continuous positive airway pressure (CPAP), the current first-line therapy for obstructive sleep apnea. Adoption of this approach might reduce the need for in-laboratory studies, thereby reducing total cost; however, this remains to be proven. Alternatively, it might lead to more duplicate studies and to increased inappropriate use of this new technology.

At this time, patients who would most benefit from unattended in-home studies are probably those with high pretest probabilities of disease, in whom the diagnosis can be so confirmed, or patients with low pretest probabilities in whom substantial abnormality can be excluded. However, an alternative approach exists for the former group—the "split-night" study in the laboratory [15]. The first part of the night is used to establish the diagnosis, whereas the latter part of the night would be for titration of CPAP pressure. This would obviate one night in the sleep laboratory. The split-night technique appears to work in most patients [15]. It is likely, therefore, that one group of patients in whom simplified diagnostic tests might be of the most value are the same group in whom split-night studies will also be the most effective. There seems little point in doing an in-home unattended study followed by an in-laboratory study to establish a therapeutic CPAP level in patients with high pretest probabilities, if all of this could be accomplished in one night in the sleep laboratory. If, however, CPAP methods currently under development [16] to automatically adjust CPAP pressure, thereby obviating the need for in-laboratory pressure titration, become widely available, then simplified diagnostic procedures will play a much larger role.

The high prevalence of sleep-disordered respiration challenges both physicians and manufacturers to develop cost-effective methods to identify, diagnose, and treat the many affected patients. New methods to analyze oximetry data, such as that reported by Series and colleagues [11], represent one response to this challenge.