CORONADO, CALIF. –
“Currently, the gold standard for diagnosis is a polysomnography (PSG), which is basically a sleep study test,” one of the study authors,, said in an interview in advance of the Triological Society’s Combined Sections Meeting. “While it is very thorough, it is expensive, intrusive, and gives data on only 1 night of sleep and cannot give information on a patient’s sleep pattern over several days. Using wrist-worn devices such as Fitbit products or smartphone applications such as Sleep Cycle would be much cheaper alternatives with little distraction that can provide continuous data over several days if accurate.”
According to Dr. Chung, an otolaryngology head and neck surgery resident in the department of surgery at the George Washington University, Washington, previous studies about this topic are inconclusive.
“Most have been studied only in the pediatric population, and in general, show that Fitbit overestimates PSG measurements,” he said. “Also, prior studies have looked into 20-60 patients, depending on the study.”
In an effort to address these limitations, he and his colleagues prospectively evaluated 180 adult patients who were already scheduled to undergo a PSG with or without CPAP testing from the sleep lab. The overnight test was performed with a Fitbit Alta HR on their wrist of choice and a smartphone at their bedside. Each smartphone was randomly assigned and had a popular sleep application installed, with Sleep as Android on the Android phone and Sleep Cycle on the Apple iPhone. For the main outcomes of interest, the researchers collected the total sleep time, sleep efficiency, and apnea-hypopnea index (AHI) from the PSG and compared them with their equivalents from Fitbit and smartphone applications. For statistical analysis, they performedplots, paired t-tests, and regression lines to assess R2, slope, and Y-intercept.
Dr. Chung and his colleagues found that both Fitbit Alta HR (P = .0014) and smartphone applications (P less than .0001) significantly overestimated the total sleep time, compared with PSG, while moderate correlation for sleep efficiency was observed between PSG and Fitbit (r = 0.38). Other findings of note were that the number of times awake recorded by Fitbit significantly underestimated the PSG AHI (P less than .0001), the snoring noise measurements from Sleep as Android had strong associations with PSG AHI (R2 = 0.43), and those from Sleep Cycle had modest associations with PSG AHI (R2 = 0.12). However, Bland-Altman plots showed wide limits of agreement for total sleep time and sleep efficiency, though as sleep efficiency approached 100%, the discrepancy between the PSG and Fitbit decreased.
“While there are various tools out right now that claim to measure sleep in a more comforting setting, we cannot recommend any product to act as a screening device or replacement for the PSG at this time,” Dr. Chung said.
He acknowledged certain limitations of the study, including the fact that Fitbit is unable to fully record sleep data if the wearer is asleep for fewer than 3 hours, which can happen in patients with severe OSA. In addition, the smartphone applications require the user to begin and conclude the recording. “This was performed only by the sleep technicians and not the patients, and can be a significant source of error, particularly with the Sleep as Android application, as it had a less intuitive interface,” he noted. “This lead to multiple unusable data points. There were also incidences where the Fitbit Alta HR failed to record sleep data even though it was fully charged and correctly placed, which led to only a smaller number of patients having both measurements at the same time.”
Dr. Chung reported having no financial disclosures. The meeting was jointly sponsored by the Triological Society and the American College of Surgeons.
SOURCE: Chung D et al. Triological CSM, .