DISCUSSION

Section:

ChooseTop of pageAbstractMETHODSRESULTSDISCUSSION <<ReferencesCITING ARTICLES

This large follow-up study with a high rate of ascertainment shows that continuing CPAP usage is related to the severity of subjective sleepiness and to the apnea plus hypopnea frequency. Continuing CPAP usage at 3 yr occurred in 86% of patients with an AHI ⩾ 30 and an Epworth sleepiness rating of > 10. Our previous studies have shown that patients with mild SAHS and AHI between 5 and 15/h benefit from treatment (22), but this study showed that many such patients abandon CPAP, especially if they are not “subjectively” sleepy. Further, long-term use can be predicted most reliably by the average nightly use of the CPAP machine during the first 3 mo. The study also showed that patients using CPAP for less than 2 h per night at 3 mo are unlikely to continue with long-term treatment. Overall CPAP use in those taking CPAP home was 68% at 5 yr. Patients continued to abandon CPAP during the first 4 to 5 yr of treatment, in contrast to our clinical impression that those who continued using CPAP after the first few months tended to use it long-term. Although patients are less likely to abandon CPAP the longer they have been using it, many (n = 63; 29%) stop the treatment after 1 yr. Similarly, Krieger and colleagues (17) found that 20% of those who stop treatment do so after 1 yr.

In the analysis of continued use the outcome of interest is deliberate stopping of CPAP treatment and includes stopping because the patient abandoned treatment (15.5%) or because the machine was reclaimed for reasons of poor compliance or serious side effects. CPAP machines were reclaimed on 50 occasions (4.5%). Patients were warned about their poor use with the threat of having their machine reclaimed if use did not improve. This threat led to improved use in some patients, whereas others remained unwilling to increase use. The choice of 2 h as a cutoff is a somewhat arbitrary value as there is no known lower limit of average nightly use below which CPAP is ineffective (17, 22, 23). Kribbs and colleagues (16) used a criterion of minimal acceptable duration of use as at least 4 h use per night on ⩾ 70% of nights monitored. This equates to an average nightly use figure of 2.8 h. Krieger and colleagues (17) used a 1-h cutoff (after attempts to improve compliance) for withdrawal of CPAP treatment.

A number of smaller studies, some using only subjective reports of CPAP use (24, 25), have attempted to establish the important determinants of long-term CPAP usage with variable results (14, 17, 24, 25). Our results confirm that the severity of breathing abnormality during sleep (14, 17) and the degree of subjective sleepiness (24, 25) are predictors of continued CPAP use and shows that age, arousals, BMI, coexistence of COPD, and CPAP pressure are also relevant factors. As a lack of snoring history was rarely found (2%) among our patients this is not a very helpful discriminator in routine clinical practice. The study also shows for the first time that AHI and Epworth sleepiness score are the independent predictors of long-term CPAP use. These predictors are relatively weak, as can be seen from their hazard ratios (Table 3), but they can be combined to give a better prediction (Figure 2).

We found that continued CPAP use was greater in patients with higher CPAP pressures. One might expect increased pressure would increase mouth leak and thus nasal symptoms (26), thus worsening compliance. Because CPAP pressure was not an independent predictor of continuing use, presumably the relationship between pressure and use resulted from covariance of CPAP pressure with AHI, which was an independent predictor of use. The 95th centile of CPAP pressure was 14.2 cm. We are unable to comment with confidence on the effects of pressure on compliance at high pressures (higher than 14 cm). Further, there was no significant correlation between the CPAP pressure used and mean nightly CPAP use (p = 0.09).

Lack of benefit and side effects from CPAP were cited by patients as the main reasons for stopping treatment. Although side effects are reported equally in compliant and noncompliant groups (25, 27), tolerance of these side effects may differ. Evidence for this view was found by Pepin and colleagues (27) as CPAP was considered to be noisier by patients with mild SAHS. It is probable that the balance of benefits to side effects from CPAP is important in determining continued CPAP use.

In contrast to baseline information, data on rates of CPAP use obtained within 3 mo of starting CPAP is strongly predictive of long-term use. Our findings extend those of Kribbs and colleagues (16) who found that the pattern of CPAP use during the first 3 mo is established in the first month for most patients. Although early rate of use data are predictive as a continuous variable, an arbitrary cutoff of 2 h per night will separate most long-term CPAP users from those who have stopped (Figure 3). The likelihood of stopping is much higher for those who at 3 mo use CPAP for less than 2 h per night than for those who use it for 2 h or more, with a hazard ratio of 13. Long-term use of CPAP occurs in only 22% of patients with early time clock readings less than 2 h compared with 78% of the remaining patients. The importance of early use data in predicting continuing CPAP use may mean that there is a window of opportunity, shortly after starting CPAP, to influence future compliance. Increased patient education and support may be one way in which this could be done as suggested by a number of studies showing improved compliance with more intensive educational strategies (28, 29). There may, however, be a number of patients who are unable to tolerate CPAP treatment and who will not continue using CPAP long-term. Early identification of these patients would be useful so that resources could be deployed elsewhere and alternative treatments tried instead.

Only 4.5% of patients refused CPAP treatment, less than the 15 to 50% reported by other investigators (17, 25, 30). In this study we used comprehensive pre-CPAP education. As the acceptance rate was better than that reported by others (17, 25), where there was unknown or limited pretreatment education, it is postulated that the education provided by our center may be a reason for the high acceptance rate found. Free CPAP provision could also be a factor in initial acceptance, but Rauscher and colleagues (30) offered CPAP free to unselected patients with AHI > 15 with an acceptance rate of only 50%. Thus, patient selection may be another important factor as all our patients had to be symptomatic to be offered CPAP. In this study and in another (25), some patients may have been advised to begin CPAP treatment in the clinic but refused, without therefore being booked for CPAP titration. The exact number of these patients remains unknown in this study; however, either N.J.D. or T.M. directly made the decision to offer CPAP in all patients, and this number was very small (estimate < 10 patients).

A higher number of female patients refusing CPAP treatment has not been reported by others (17, 25, 30), and the reason for this is unknown. Witnessed apneas may provoke a degree of anxiety in the patient about their disease that results in them persevering at least beyond a CPAP titration night. The finding that CPAP refusers were more often specialist- referred probably results from 38% of such patients being referred from ENT specialists, and many of these presented with snoring as the main complaint. Presumably, many such patients will be less concerned about their own symptoms, perhaps decreasing motivation for therapy (17, 24). CPAP refusers also drank less alcohol and were more often current smokers, but the reasons for these relationships are unclear.

The median average nightly use of CPAP determined from clocks built into the machine was 5.7 h in those who continue with CPAP therapy. No lower threshold for effectiveness of CPAP is known; however, Engleman and colleagues (31) were able to show improvements in objective daytime sleepiness, symptom scores, mood, and cognitive function in a group of consecutive patients with SAHS with an average CPAP use of 3.7 h per night. Seventy-six percent of our patients used their CPAP machine for an average of 3.7 h or more per night. The rates of use quoted above refer to time the machine is running, not time at the preset therapeutic pressure. These data are not known for most of our patients; however, a number of studies have shown that a correction factor of 0.9 will give a reasonable estimation of time at therapeutic mask pressure (13, 15, 16). As demonstrated by Kribbs and colleagues (16), average nightly CPAP use figures may result from a wide range of patterns of actual nightly use and do not indicate whether use has been regular or irregular.

Automatic CPAP titration using the Autoset was performed on 106 of the patients studied. Autoset CPAP titration produces an accurate assessment of the optimal CPAP pressure (32). All of the remaining patients (n = 1,105) had manual CPAP titrations. These two groups of patients were both predominantly male and similar in terms of age (median = 50 yr, Autoset versus 50 yr; manual), BMI (median = 30 versus 30.5), AHI (median = 32 versus 31), and Epworth score (median = 13 versus 12). Two of the patients receiving Autoset CPAP titration refused to take CPAP home. Kaplan-Meier survival plots of continued use were similar (log-rank test, p = 0.11) and the median CPAP run time at the last visit for those continuing CPAP was similar (5.0 versus 5.7 h). Thus, the patients titrated using Autoset were similar to those titrated manually and had similar compliance and are therefore included in the analysis.

One of the potential criticisms of this study is that it involved 46 comparisons. Therefore, by chance alone, two to three relationships would be expected to be significant at the 0.05 level, whereas 23 significant relationships were found. There is no universally accepted statistical solution to this problem, and most statisticians feel that the Bonferroni correction for multiple comparisons is an overly conservative approach for such correlational data. However, even if this were used, there are still significant univariate relations between continued CPAP use and age, BMI, Epworth score, AHI, symptoms of SAHS, and CPAP pressure, with the independent determinants being unchanged except that snoring history is no longer significant.

This study was conducted in a health care system where there was no treatment cost for the patient. Although cost is a potential determinant of compliance in systems where treatment is not free the available evidence suggests that it may not significantly affect continued use. Studies in cost-free systems of subjective continued use (12) and objective short-term use (13) had similar levels of use to those in the United States, arguing against a significant effect of cost. Furthermore, Rolfe and colleagues (24) assessed reasons for discontinuing CPAP treatment in Australia where patients were asked to buy their CPAP machine and cost was not commonly given as a reason for discontinuing CPAP treatment (six of 61 patients). It is therefore likely that most of the findings of this study are also applicable to health systems that involve costs to the patient for CPAP treatment.

The effects of a strongly supportive system on long-term use are uncertain. Waldhorn and colleagues (25) reported their findings of subjective compliance in a system with limited follow-up support (“A follow-up home visit was provided by a respiratory care company on initiation of home therapy”). They found 76% of patients starting CPAP were still using it at 14 mo, comparable to the 82% in our center at 14 mo. Hence, our findings on long-term use of CPAP may apply to centers with a less supportive system than our own.

Other limitations of this study include the failure to contact 4.6% of patients and uncertainty about the number of patients who declined a CPAP trial before it was ever booked. However, we believe that selection bias has been minimized by the availability of data from nearly all patients in the study group. Furthermore, the remaining patients do not appear to be different from the study group in important baseline determinants of CPAP use. All of the objective data used in the study were accrued prospectively, thus avoiding potential bias from retrospective evaluation of data.

Other limitations include the facts that the Epworth sleepiness scale has only been in use since 1991 (33), and objective data from in-built clocks were not available in the early years of the study (pre-1990), although patients were all subsequently changed over to these CPAP machines. There are only five patients who started CPAP before 1990 and then stopped before being changed to machines with built-in time clocks. We therefore have objective data on compliance on the vast majority of patients starting CPAP before 1990. In addition, as we successfully followed up only 281 patients at 3 yr and 61 patients beyond 5 yr, longer-term studies will be required in the future.

In conclusion, this study found that in a cost-free system, with comprehensive pre-CPAP education very few symptomatic patients refused CPAP treatment, and they were more often women, were more often specialist-referred, and less often had a history of witnessed apneas. Just over two thirds of patients who took CPAP home continued with CPAP at 5 yr, with a median nightly use of 5.7 h. Although AHI, subjective daytime sleepiness, and snoring history have an independent role in predicting long-term CPAP use, they are weak predictors and should not be used in isolation to determine who merits a trial of CPAP treatment. Early CPAP use data are strongly predictive of continuing use and may help decisions about perseverance with CPAP treatment and allow early identification of patients who might benefit from more intensive education and support. It also suggests (28) that efforts to improve CPAP compliance should be targeted to the initial 3 mo.