CPAP or Placebo-effect?

In this issue of SLEEP, Kushida and colleagues¹ report neurocognitive performance results from a large, multi-site, double blind, controlled, randomized clinical trial on the effects of long-term continuous positive airway pressure (CPAP) therapy.^1–3 In this trial, 1,098 patients with obstructive sleep apnea were randomized to active CPAP or sham CPAP (SHAM), with follow-up neurocognitive measures at 2 and 6 months.^1–3 Their major findings were: (1) At 2 months, CPAP patients had a significantly better scores than SHAM patients on the sustained working memory test (SWMT), a measure of pre-frontal cortex executive function; being most pronounced for patients with severe OSA. There were no significant differences between groups in other types of cognitive tests. (2) There was a significant association between oxygen saturation and SWMT at 2 months. (3) There was a significant decline in SWMT from 2 to 6 months in the SHAM group but not in the CPAP group. (4) At 6 months, there were no significant differences between groups in SWMT or other measures of CF. (5) Patients on CPAP were significantly less sleepy, as measured by the Epworth Sleepiness Scale (ESS) than patients on SHAM at 2 and 6 months.¹ In addition, at baseline, the serial addition test, a simple measure of sustained working memory, was one of a few measures associated with apnea/low oxygen.^1,3 The SWMT could not be used at baseline, as it measures change over time.

Prior to this study, Beebe and Gozal⁴ had argued that sleepiness alone could not explain the cognitive impairment seen in apneic patients. They theorized a mechanistic pathway whereby apnea-related intermittent hypoxemia would lead to biochemical events that preferentially damage neurons in the pre-frontal regions of the cortex.⁴ This would affect sustained working memory and be manifested, for example, as problems mentally manipulating information, disorganization, and poor judgment, planning, and execution. The findings of Kushida et al.¹ relative to sustained working memory are consistent with this hypothesis.⁴ Relative to sham, active CPAP had a strong persistent impact on sleepiness but only minimal and transient effects on measures of most cognitive performance measures.¹ If hypoxemia rather than sleepiness were the main cause of cognitive deficits, and most of the damage from hypoxia were not reversible, then CPAP effects in a 6-month period would be minimal and perhaps transient.

Kushida et al.¹ report that only 511 of the 1,098 patients were adherent to CPAP use at 2 months, and restriction of analysis to compliant patients (i.e., ≥ 4 hours for the 2 months prior to each neurocognitive testing visit) did not lead to stronger results, with the exception that greater active CPAP adherence was associated with greater subjective alertness as measured by the ESS. On the SWMT, the difference between CPAP and SHAM in the full sample was 0.109 (P = 0.0074); in the more-compliant versus less-compliant active CPAP subsamples it was 0.088 (P = 0.0892).

While these details, particularly the change in P-value, may be partly explained by reduced samples sizes as subsets of patients are included in the data analyses, we suggest that the explanation may lie in differential placebo effects. A total of 55% of CPAP patients versus only 30% of SHAM patients believed they were on active treatment (P < 0.0001). The literature on placebos indicates that they can have very real effects on neuronal function^5,6 and that they can impact performance on cognitive tests.^7,8 Placebo effects are stronger both for patients who believe they are on active treatment and for patients who are more compliant.^9–12 Kushida and colleagues¹ adjusted for compliance, but how much of the placebo effect was removed by adjustment is unknown. Differential placebo effects could explain both the nonsignificant results among compliant patients (those compliant with SHAM may be the participants who believe they are on active treatment) and the fact that the results were short-lived, as placebo effects in both groups may be short-lived.

The limitations caused by the high IQ of the study population have been discussed by the authors and can be addressed in future studies. We note additional drawbacks worth discussing. There was little evidence of cognitive impairment at baseline.³ As noted above, cognitive deficits related to executive function might not be reversible^13–16; high-risk participants with respiratory conditions and oxygen saturation (SaO₂) below 75% for more than 10% of nights were excluded.^2,17–20 Taken together, these limitations could have attenuated treatment differences, and they have other implications.

The fact that most patients did not have cognitive impairment at baseline contradicts the concept of an “efficacy” trial. This was previously addressed by Bliwise and Greenaway²¹ in a commentary on the Apples baseline analysis, in which they suggested that eligible participants meet the definition for mild cognitive impairment (MCI). Interest seems to be focused now on how and to what extent OSA damages frontal lobe executive function,^13,22,23 so it may be more compelling to enroll subjects with deficits specifically in this area. For example, a future CPAP treatment study might randomize only those who are stable or decline on the SWMT and other measures of executive function over a 2-month period.

However, if cognitive impairment is not reversible but can be prevented, a prevention trial would be more desirable. In a prevention trial, cognitive deficits would need to be ruled out at baseline, and information on the natural history of cognitive impairment and its annual incidence would be needed for sample size considerations. Depending on the 6-month incidence of cognitive impairment, such a trial may have to enroll more than 1,100 patients, or perhaps extend for more than a year or two rather than 6 months.

Investigators need to visualize the ideal clinical trial before they can think about ways of approximating it or realize they can't do the trial for ethical reasons. One might initially accept the study exclusion of participants with lower oxygen saturation who perhaps should not be given SHAM treatment for ethical reasons, but the inability to leave these high-risk individuals untreated is the most important limitation of all. The situation is that we have a condition that can result in several debilitating outcomes, a highly effective treatment for that condition, and we want to know how well the (overall highly effective) treatment protects against one specific outcome. We require a control group of untreated patients who think they are being treated, and we want to study them until they have a good chance of developing the outcome. Knowledge may have to take a back seat to ethics on this one.

CITATION

Schwartz SW; Cimino CR; Anderson WM. CPAP or placebo-effect? SLEEP 2012;35(12):1585-1586.

DISCLOSURE STATEMENT

The authors have indicated no financial conflicts of interest.