The Demise of Portable Monitoring to Diagnose OSA? Not So Fast!

Healthcare costs money—a lot of money. In 2008, $7538 per capita or 16% of the gross domestic product was spent on healthcare in the United States.¹ Not surprisingly, interventions are being increasingly scrutinized to assess whether they provide good value for money (i.e., are cost-effective).

Cost-effectiveness is assessed by the incremental cost-effectiveness ratio (ICER), which is the ratio of the incremental cost and change in quality adjusted life years (QALY) that follows from the adoption of a treatment compared to no treatment.² The QALY takes into account both added years of life by therapy and improvements in quality of life. An ICER of less than $50,000 per QALY is oft quoted as the threshold for cost-effectiveness, although many believe that this value should be greater.³

In this issue of SLEEP, Pietzsch and colleagues should be congratulated for doing a comprehensive Markov model to assess the cost-effectiveness ratio of different diagnostic and therapeutic strategies in the management of patients with suspected obstructive sleep apnea (OSA).⁴ The authors used a 10-year and a lifetime time horizon, and included various outcomes, including motor vehicle crashes (MVC), strokes, and heart attacks in their models. The base case was a 50-year-old male with a 50% pre-test probability of OSA.

Continuous positive airway pressure (CPAP) treatment, compared to no treatment, had an ICER of $15,915/QALY; this ICER is far less than the threshold commonly used to indicate whether a therapy is cost-effective. These findings are consistent with other investigators that have also demonstrated that CPAP therapy for OSA is a very cost-efficient use of resources.^5,6

Of more interest is the analysis of various strategies for the initial diagnosis of suspected OSA. The authors compared three different strategies (full-night polysomnography [PSG], split-night PSG, and unattended portable home monitoring). In the base case analysis, the most cost-effective strategy for diagnosing patients with suspected OSA was, surprisingly, full-night PSG.

These findings are counterintuitive. Home studies are much less costly than PSG and have very good sensitivity and specificity.⁴ What then are the explanations for the study's findings? The crux of the argument is that PSG—though a more expensive test up front—ends up costing the healthcare system less over time by minimizing the number of patients with false negative tests (i.e., patients with OSA who were incorrectly identified as not having OSA and were thus not provided treatment) as well as minimizing patients with false positive tests (i.e., patients without OSA who were incorrectly identified as having OSA and thus provided with unnecessary treatment).

However, as in any modeling exercise, the underlying assumptions of the model used by Pietzsch and colleagues⁴ need to be examined, as these may have inflated the impact of false negative and false positive test results. First, the authors assumed that OSA treatment would dramatically reduce the risk of cardiovascular events. For example, the lifetime rate of myocardial infarction was reduced from 56% to 39% with CPAP. The question is whether this is a reasonable assumption. The benefits of CPAP were extrapolated from observational studies and short-term impact on hypertension rather than large randomized controlled trials of robust clinical endpoints. One could argue that it is controversial whether treatment of patients with OSA have this magnitude of effect.⁷ Including these unproven benefits of CPAP in the models increases the economic impact of false negative patients (patients with OSA not provided therapy due to a negative home study).

Second, the authors assumed that patients incorrectly diagnosed with OSA by a home study would use CPAP long term, with a compliance (and therefore CPAP costs) similar to that of patients with OSA. This seems unlikely, as patients without OSA would not improve symptomatically with CPAP, causing one to readdress the original diagnosis (and obtain full PSG or discontinue CPAP) rather than continuing therapy. This assumption magnifies the impact of false positive tests.

Third, it was assumed that 22% of patients with a technical failure or a negative ambulatory study would not return for a follow-up PSG. This is the source of patients with false negative tests in the model. Even if the value of 22% were correct (though it seems high), patients who do not return are likely not be representative of the group at large. Specifically, patients who are not motivated to have a follow-up PSG would likely be less compliant with therapy, less symptomatic, and more likely not to have substantial OSA. Assuming that this group is representative makes the PSG arm appear better, as it inflates the rates of false negative tests. Indeed, as pointed out by the authors, when it is assumed that all negative home studies have full PSG and patients with false positive home studies receive no benefit from CPAP, ambulatory studies were cost-effective compared to PSG.

To the authors' credit, they did extensive sensitivity analyses to determine whether these and other factors affected the ICER (Table 3 in the publication). For instance, even when costs and benefits from MVC and CV disease were excluded, PSG still appeared to be the most cost-effective. However, it is unclear whether changing all the factors noted above might change the ICER in favor of ambulatory studies. In other words, they did conduct extensive univariate analysis but did not simultaneously examine joint uncertainty by sampling from probability distributions using Monte Carlo methods. This incomplete uncertainty analysis creates only point estimates for the ICERs without any confidence limits.

What should we take from the study by Pietzsch et al.?⁴ As with all economic modeling studies, the end results are highly dependent on the underlying assumptions and the data used to populate the model. Given the multiple assumptions, this study should not be considered definitive evidence that home studies should not be used. However, it should give us pause to think about whether home studies should automatically be considered the most cost-effective test just because the upfront costs are lower than PSG. At the very least, we need to be careful about advocating for the broad use of home studies, especially under conditions where the frequency of false negative and false positive tests may be increased. This could include patients with moderate (as opposed to high) pre-test disease probability or patients with underlying cardiopulmonary disease.

In the end, the fundamental issue is not whether ambulatory studies are better or worse than PSG, but rather under what conditions use of home studies are appropriate. Large randomized trials of different diagnostic strategies in different groups of patients with collection of concomitant economic information are required to address this important issue. It is unlikely that one diagnostic strategy will be superior in all the different clinical scenarios that are encountered.

DISCLOSURE STATEMENT

Dr. Ayas has received research support from Phillips/Respironics Inc. Dr. Pack is the John L. Miclot Professor of Medicine. Funds for this endowment have been provided by Phillips/Respironics Foundation. Dr. Marra has indicated no financial conflicts of interest.