RESULTS

Section:

ChooseTop of pageAbstractMETHODSRESULTS <<DISCUSSIONReferencesCITING ARTICLES

Of the 296 participants enrolled (95% males), the home testing group (n = 148) included 44.2% African Americans and the in-laboratory testing group (n = 148) had 34.5% African Americans. Table 1 compares baseline characteristics between the two groups of participants in the efficacy evaluable cohort. Baseline characteristics comparing treatment groups in the intent-to-treat sample are provided in Table E1 in the online supplement. The mean MAP index at baseline, a measure of OSA likelihood based on symptom experience, body mass index (BMI), age, and sex, was 0.78 ± 0.13 in the in-laboratory group and 0.76 ± 0.14 in the home group (P = 0.55). There were no differences between the two groups in FOSQ total and component scores at baseline (P values ≥ 0.22). The AHI was 47.3 ± 29.4 events/hour on the diagnostic PSG (full-night or split) in the in-laboratory group; in participants in the home testing group who were diagnosed with OSA on the home unattended sleep study, the AHI was 42.9 ± 23.2 events/hour. The home group was slightly older (55.1 ± 10.3 yr) than the in-laboratory group (51.8 ± 10.1 yr) (P = 0.02). The SF-12 mental health component score at baseline was higher for the home group (44.4 ± 10.8) compared with the in-laboratory group (41.1 ± 10.7) (P = 0.02). Current and previous cigarette smoking was more prevalent in the home group (P values < 0.01). In participants who initiated treatment with CPAP, the time span from signing participant consent to CPAP set-up was 96.2 ± 64.0 days for the in-laboratory group and 103.1 ± 81.1 days for the home group (P = 0.48).

Of the 148 participants randomized to home testing, 113 (76.4%) were initiated on CPAP treatment (Figure 1). When we compared baseline characteristics of the subjects who initiated CPAP treatment with those of the 35 who did not initiate CPAP, we found a significant difference in MAP index (0.78 ± 0.13 vs. 0.65 ± 0.23, respectively; P < 0.001). Of the 143 participants who performed a home unattended sleep study, 3 were diagnosed with central sleep apnea and were referred to the sleep center for appropriate clinical management. A total of 35 participants were scheduled for an in-laboratory PSG: 27 participants (18.9%) with an AHI less than 15 events/hour on the home recording and 8 (5.6%) with technically unsuccessful home recordings on two attempts. Only 8 of the 35 subjects who had confirmatory in-laboratory PSG were found not to have sleep apnea. Home unattended autoCPAP titration studies were performed in 119 participants. The autoCPAP titration study was unsuccessful in only 18 participants (15.1%): 12 (10.1%) due to insufficient use of the device on two attempts and 6 (5.0%) due to an AHI of 15 events/hour or more at the 90th percentile pressure. During the in-laboratory titration PSG, 4 of the 18 participants required bilevel positive airway pressure and/or supplemental oxygen and were excluded from the study. Five participants who initiated CPAP withdrew before the 1-month follow-up. Of the 133 participants diagnosed with OSA, 113 (85.0%) were initiated on CPAP, 103 (77.4%) completed the 1-month follow-up, and 96 (72.2%) completed the 3-month follow-up. With application of last observation carried forward, sample size at the 3-month time point increased to 105 (78.9%). A total of 19 participants (12.8%) randomized to the home testing group withdrew from participation for non–protocol-related reasons before initiation of CPAP treatment.

Figure 1. Cohort diagram of participants randomized to the home versus in-laboratory management pathways. In the intent-to-treat analysis of participants initiated on continuous positive airway pressure (CPAP) treatment, results from participants assigned to the home testing group who required in-laboratory testing were included in the home group. BIPAP = bilevel positive airway pressure; Dx = diagnosis; FU = follow-up; OSA = obstructive sleep apnea; PSG = polysomnography.

Download Figure | Download Powerpoint

[More] [Minimize]

Of the 148 participants randomized to in-laboratory testing, 110 (74.3%) initiated CPAP treatment (Figure 1). There was no significant difference in MAP index or other baseline characteristics between the subjects initiated and not initiated on CPAP treatment. Of the 141 participants who underwent PSG recordings, a split-night study was performed in 42 participants (29.8%) and a full-night diagnostic PSG in 99 participants (70.2%). Fourteen (9.9%) of the 141 participants were excluded from participation because of an AHI less than 5 events/hour (n = 8) or a diagnosis of central sleep apnea (n = 6). Twelve participants required supplemental oxygen or bilevel positive airway pressure and were withdrawn from the study. After initiation of CPAP, 5 participants withdrew before the 1 month of follow-up. Of the 127 participants diagnosed with OSA, 110 (86.6%) were initiated on CPAP, 92 (72.4%) completed the 1-month follow-up, and 86 (67.7%) completed the 3-month follow-up. With application of last observation carried forward, sample size at the 3-month time point increased to 96 (75.6%). A total of 22 participants (14.9%) randomized to in-laboratory testing withdrew from participation for non–protocol-related reasons before initiation of CPAP treatment.

The mean FOSQ total score and all of the FOSQ component scores improved significantly within both groups after 3 months of CPAP treatment (P values < 0.0001; Table 2). The improvement in mean FOSQ total score was 1.74 ± 2.81 in the home group and 1.85 ± 2.46 in the in-laboratory group with effect sizes of 0.62 and 0.77, respectively. The improvement in both groups was similar (P = 0.77). The adjusted group difference (home minus in-laboratory) in mean change in FOSQ total score from baseline to Month 3, controlling for baseline FOSQ and investigative site, was −0.004 ± 0.33 (SEM) (P = 0.90; Table 3). The lower bound of the 95% noninferiority confidence interval was −0.54. Because −0.54 is greater than −1.0, the null hypothesis that home testing is clinically inferior to in-laboratory testing was rejected (P < 0.05). Supporting the robustness of the findings, similar results were obtained when an intent-to-treat analysis was performed using all participants diagnosed with OSA. In addition, nearly identical results were obtained in sensitivity analyses using observed cases (home testing, n = 96; in-laboratory testing, n = 86). The scores on the ESS, CES-D, mental health component of the SF-12, and FOSQ component scores improved significantly after 3 months of CPAP treatment within both groups (Table 4), and the changes were similar between groups (Table 5).

The mean CPAP setting was 11.1 ± 3.2 cm H₂O in the home tested group and 9.4 ± 2.9 cm H₂O in the in-laboratory tested group (P < 0.001). Over the 3-month treatment period, mean CPAP use was 3.49 ± 2.45 hours/day in the home tested group and 2.92 ± 2.32 hours/day in the in-laboratory tested group (P = 0.08). Controlling for investigative site, the adjusted group difference (home minus in-laboratory) in mean CPAP daily use was 0.55 ± 0.32 (SEM) hr (P = 0.085). The lower bound of the 95% noninferiority confidence interval was +0.03. Because 0.03 is greater than −0.75, the null hypothesis that home testing is clinically inferior to in-laboratory testing was rejected (P < 0.05). Figure 2 shows the distribution of mean daily CPAP use over the 3-month treatment period in the two groups. The mean percentage of days that CPAP was used for at least 1 hour/day was 82 ± 29% in the home tested group and 78 ± 33% in the in-laboratory tested group (P = 0.28), and the median percentage of days with at least 1 hour of use was 95 and 94%, respectively (Wilcoxon rank sum P = 0.70). The mean percentage of days that CPAP was used for at least 4 hours/day was 52 ± 34% in the home tested group and 49 ± 35% in the in-laboratory tested group (P = 0.42), and the median percentage of days with at least 4 hours of use was 55 and 52%, respectively (Wilcoxon rank sum P = 0.17). The mean (SD) AHI on the CPAP downloads over the 3-month intervention was 4.3 ± 3.4 events/hour in the home group and 4.7 ± 3.4 events/hour in the in-laboratory group (P = 0.51). The latter results indicate that the methods for determining the CPAP settings in both the home and in-laboratory groups achieved excellent treatment efficacy.

Figure 2. Distribution of mean hours of continuous positive airway pressure (CPAP) use per day during the first 3 months of CPAP treatment in participants randomized to in-laboratory (top) and home (bottom) testing. Over the treatment period, mean CPAP use was 3.49 ± 2.45 hours/day in the home group and 2.92 ± 2.32 hours/day in the in-laboratory group (P = 0.08).

Download Figure | Download Powerpoint

[More] [Minimize]

DISCUSSION

Section:

ChooseTop of pageAbstractMETHODSRESULTSDISCUSSION <<ReferencesCITING ARTICLES

The results indicate that improvement in daytime function after 3 months of CPAP treatment and adherence to CPAP during that period in patients evaluated by a management pathway using home testing with portable monitors are not clinically inferior to the results in patients receiving standard in-laboratory PSG testing. The within-group improvements in FOSQ total score, the primary functional outcome measure, were highly statistically significant (both P values less than 0.0001) and likely of clinical significance given the respective effect sizes of 0.62 and 0.77 for the home and in-laboratory tested groups. Significant improvements within groups, which were similar between groups, were observed for the ESS, CES-D, mental health component of the SF-12, and FOSQ component scores.

We selected a −1 point change in FOSQ score as our noninferiority delta based on the work of Weaver and colleagues (9) and were reassured that similar conclusions would have resulted if an even more conservative threshold had been selected. The lower bound of our 95% confidence interval for change in FOSQ score between the home and in-laboratory groups was −0.54, which is considerably greater than the a priori threshold. Similarly, the a priori noninferiority delta for CPAP adherence of −0.75 hour/day was selected with the assumption that the mean daily CPAP adherence over the 3-month intervention period would be at least 4 hours/day. Although −0.75 hour/day is relatively large considering the actual group means of 3.5 and 2.9 hours/day, the observed lower bound of the 95% confidence interval was +0.03 hour/day, which is considerably larger than −0.75 hour/day. These results indicate that even if a more conservative noninferiority delta had been chosen for CPAP adherence, our interpretation of the results would not have changed.

The study has a number of significant strengths. Previous studies comparing the effectiveness of ambulatory versus in-laboratory testing on functional outcomes in patients referred to a sleep center for evaluation of OSA used a screening procedure to preselect those patients with a high pretest likelihood of OSA (6–8). The current study evaluated a larger cohort of subjects than the previous studies and took a less restrictive approach to enrollment by recruiting from all new patients referred to the sleep center who were thought to need testing for OSA. Even using this approach, however, 88% of individuals enrolled in the study were diagnosed with OSA on home and in-laboratory testing. This high percentage is likely due to the 95% male cohort and the primary care providers recognizing and referring patients at high risk for sleep apnea, who were then evaluated by a sleep specialist before enrollment. Different results may be obtained if the home testing algorithm employed in this study is applied to patients at lower risk for sleep apnea. The home diagnostic study records fewer signals than the in-laboratory PSG and therefore carries a greater possibility of producing false negative results. In the current study, 25% of the participants performing home sleep testing needed in-laboratory testing either due to an AHI less than 15 events/hour on the home study or technical failure, the latter occurring in only 5.6% of the participants. Regardless of the AHI threshold selected for the diagnosis of OSA on the home sleep study, application of the home testing algorithm to a broader range of patients, for example, women and men in a primary care setting, would likely result in an even greater percentage of negative studies requiring in-laboratory testing or repeat home testing. These considerations underscore the importance of still having access to in-laboratory testing when conducting home testing with portable monitors.

An important strength of the home testing algorithm used in this study was the safeguards used to ensure that participants received appropriate clinical care. Parameters for technically acceptable studies were established for both the home unattended sleep study and home unattended autoCPAP titration study. The type 3 portable monitor used for the home unattended sleep study was able to distinguish obstructive from central sleep apnea/Cheyne-Stokes respiration, ensuring that patients with the latter diagnosis were identified and scheduled for in-laboratory PSG rather than a home auto-CPAP titration study to determine appropriate treatment. The autoCPAP device used in the study estimated the AHI at individual pressure settings that assisted in selecting an efficacious fixed pressure setting for CPAP treatment. In addition, some patients experienced persistent or increased frequency of apneas while wearing autoCPAP; this finding triggered an evaluation for possible complex sleep apnea and the need for an alternative form of positive airway pressure treatment via an in-laboratory PSG.

In contrast to previous comparative effectiveness studies that found no difference in fixed pressure setting for CPAP treatment based on autoCPAP versus in-laboratory CPAP titration (6–8, 20), the setting selected for the fixed pressure for CPAP treatment in the current study was significantly higher in the home versus in-laboratory groups. Although the reason for the difference in CPAP settings between groups is unclear, CPAP adherence was similar in both groups and the mean AHI less than 5 events/hour on the CPAP downloads in both groups indicates excellent treatment efficacy.

We decided to perform testing on a fixed- rather than a real-time schedule to enhance the applicability of our results to other sleep centers in the Veterans Health Administration and private sector. Wait times vary widely across sleep centers, especially in the private sector. Use of a fixed interval minimizes bias by removing this confounding variable. Finding similar CPAP adherence and functional outcomes between home and in-laboratory testing pathways when similar patient wait times are employed across the two groups makes it likely that even more beneficial effects of home testing, in terms of cost-effectiveness, will be present if patients are forced to wait longer times for in-laboratory testing, that is, go undiagnosed and therefore untreated for more than half a year before starting treatment.

Our study has a number of limitations that may limit the generalizability of the findings. Instead of using a recognition strategy, such as the MAP index, to preselect patients with a high pretest likelihood of OSA, we wanted to evaluate the home testing pathway under “real life” conditions in which any patient referred to the sleep centers who was being scheduled for sleep testing to rule out OSA was asked to participate. Although consecutive patients referred to the sleep centers for evaluation of OSA were recruited, the subjects were almost exclusively middle-aged obese men, and most of the veterans enrolled at both sites turned out to have OSA. Although our results are directly applicable to the more than 3 million veterans enrolled nationwide in the Veterans Health Administration, our results may differ when applied to patient populations with a wider range of pretest probability. In such populations, one might consider incorporating some type of recognition strategy to identify those individuals who are at greater risk of the diagnosis.

A second limitation of this study was the relatively low adherence to CPAP compared with that reported in other trials (6–8, 21, 22). A number of measures were included in our protocol to promote the participants' adherence to CPAP. Before randomization, all participants were evaluated by a sleep specialist and they were shown a video and provided brochures on OSA and CPAP treatment. Additional education and support were provided by the home health care company therapist who delivered the CPAP apparatus to the participant's home. Last, participants were scheduled for a 1-month follow-up visit with a health care provider in the sleep center.

We believe that two factors may explain the low CPAP adherence in our participants compared with other studies. In our calculation of mean daily hours of CPAP use over the 3-month intervention, missing values were counted as no use of the treatment. Few if any previous studies comment on how missing adherence data were handled. The low CPAP adherence of our participants may also be explained by the lower socioeconomic status of veterans receiving care through the Veterans Health Administration. A retrospective study at one of the sites participating in the current study found that CPAP adherence in the first week of treatment, a predictor of long-term CPAP adherence, was directly related to the patients' neighborhood socioeconomic status (23). It is important to note that these patients were not charged for their clinical care, including the cost of the CPAP apparatus and supplies, removing those potential financial barriers to treatment. The results of that previous study indicate that a patient's decision concerning whether to adhere to CPAP is influenced by factors other than the education and support provided by their health care providers. Other factors contributing to lower adherence may include sleep patterns (including presence of insomnia or night shift work), living conditions (such as availability of electricity, threats to personal safety), and competing health concerns that are perceived as more urgent. Despite the relatively low adherence of our cohort, this randomized controlled trial shows that, compared with standard in-laboratory testing, comparable CPAP adherence can be achieved using alternative methods of testing.

The primary analysis used to test our noninferiority hypotheses included just those participants diagnosed with OSA and initiated on CPAP. The type 3 portable monitor used in this study did not possess sleep-staging capabilities and the AHI on the home unattended sleep study was based on recording time rather than total sleep time. The difference in AHI calculation on the home versus in-laboratory diagnostic tests might have resulted in different percentages of enrolled participants being diagnosed with OSA in the two groups, potentially biasing the results. Two approaches were taken to address this concern. First, the study was designed to minimize false negative results on the home diagnostic test by performing an in-laboratory PSG in participants with an AHI less than 15 events/hour on the home study. Using this strategy, a similar percentage of randomized subjects were initiated on CPAP in the home and in-laboratory groups (76.4 and 74.3%, respectively). Second, an intent-to-treat analysis of the entire cohort revealed similar results to the modified intent-to-treat analysis that included participants diagnosed with OSA and initiated on CPAP.

It is possible that the results were influenced by the criteria used to score abnormal respiratory events on the in-laboratory and home sleep studies. For both types of recordings, nasal pressure, rather than thermometric flow, was used to identify apneas. The study was initiated before publication of the latest American Academy of Sleep Medicine standards for scoring apneas based the absence of airflow from an oronasal temperature sensor. In addition, different criteria were used to score hypopneas on the in-laboratory PSG versus home portable monitor recording (see the online supplement). The criteria for scoring hypopneas on PSG were a 30% reduction from baseline in a respiratory signal for at least 10 seconds associated with a 4% or greater oxygen desaturation and/or an arousal. On the home sleep studies, hypopneas were defined as either a 30% reduction in a respiratory signal for at least 10 seconds associated with a 4% or greater oxygen desaturation or a greater than 50% reduction in airflow for at least 10 seconds. The latter criterion was added to offset the inability to detect arousals on the portable monitor study given the absence of sleep-staging signals. It is possible that adding this additional criterion may have overestimated the numerator in the AHI calculation for portable monitor studies, which may have compensated for using study time rather than the unobtainable sleep time in the denominator.

The difference in AHI calculation on the home versus in-laboratory diagnostic tests prevented a comparison of OSA severity between the two groups at baseline. The original analysis plan specified that the MAP index would be included in the a priori specified model to account for between-patient differences in apnea severity. However, six evaluable subjects in each group were missing baseline MAP values. In order not to decrease the power of the study, the baseline MAP index was not included in the model. However, in those participants with available data, the mean MAP index was similar between home (n = 107) and in-laboratory (n = 104) groups: 0.78 ± 0.13 and 0.76 ± 0.14, respectively (Wilcoxon rank sum P = 0.14; standardized effect size = 0.16). These results suggest that the two groups had equal OSA severity at randomization and support our decision not to include MAP index as a covariate.

In summary, the results of this two-site, open-label, randomized, parallel groups, noninferiority study indicate that veterans with OSA randomized to a home unattended testing algorithm have functional improvement and adherence to CPAP over a 3-month treatment period that is not clinically inferior to that of individuals receiving standard in-laboratory PSG testing. The ability to use portable monitor testing to manage patients with OSA should improve their access to care. The findings extend results reported by previous studies by evaluating a larger cohort of patients and enrolling individuals referred to the two sleep centers without the use of additional prescreening procedures to increase the pretest likelihood for the diagnosis of OSA. Nevertheless, a high proportion of the participants had OSA, and different results might be obtained when such an ambulatory management pathway is applied to patients recruited directly from a primary care setting. Prospective studies are needed to evaluate the cost-effectiveness of home versus in-laboratory management pathways in this and other patient populations.