N-terminal pro-brain natriuretic peptide: a potential follow-up biomarker of mandibular advancement device efficacy on cardiac function in obstructive sleep apnea

Interrelationships between obstructive sleep apnea (OSA) and cardiovascular diseases are now recognized, but some underlying pathophysiological mechanisms remain controversial. Circulating cardiac biomarkers are diagnostic tools that can help understand them, in particular the N-terminal pro-brain natriuretic peptide (NT-proBNP), a marker of myocardial stretch, and a potential indicator of subclinical cardiac stress in OSA. Continuous positive airway pressure (CPAP), the first-line treatment of moderate to severe OSA, may be considered as uncomfortable, resulting in poor adherence, and reduced effectiveness. In this case, mandibular advancement devices (MAD) are an effective alternative therapy, more comfortable, and generally well accepted, with higher compliance. To date, few studies have compared the cardiovascular effects of CPAP and MAD. From recent literature reviews, it emerges that both therapies are effective in blood pressure reduction. However, the effects of MAD on other cardiovascular outcomes are conflicting, in particular as regards to its impact on circulating cardiac biomarkers. In a recent ancillary study from a randomized controlled trial, Recoquillon et al concluded that two months of MAD treatment had no effect on NT-proBNP plasma levels in patients with severe OSA. The present discussion analyses this result from a biological, statistical, and analytical standpoint, in light of results from other studies evaluating natriuretic peptides in MAD-treated OSA, with the aim to support further longitudinal studies designed with a high methodological quality.

In a recent randomized controlled trial, Recoquillon et al. evaluated the effect of two months of MAD treatment on N-Terminal Pro-Brain Natriuretic Peptide (NT-proBNP) plasma levels in patients with severe OSA 4 . Compared to a sham device, the high treatment adherence (6.6 hours/night) significantly reduced the mean apnea-hypopnea index (AHI), and the oxygen desaturation index. Nevertheless, according to their model, the authors stated that MADs had no effect on NT-proBNP levels, nor on other inflammatory and metabolic biomarkers. To our knowledge, to date only two studies have investigated the natriuretic peptides in such contexts 5,6 . Given their scarcity, any type of study looking at relevant cardiac biomarkers of MAD efficacy must be encouraged, and designed in as detailed and robust a manner as possible. In this way, some issues have to be discussed regarding the evaluation of NT-proBNP from Recoquillon et al.

1) Biological standpoint.
After two months of MAD use, the NT-proBNP plasma concentrations decreased from 296.8 to 252.5 pg/mL (-14.9%) in treated patients, whereas they decreased from 189.8 to 184.3 pg/mL (-2.9%) in patients with the sham device, resulting in a mean adjusted intergroup difference of 12.0 pg/mL (-40.9 to 64.9, 95%CI; P =0.65). The question arises as to whether a NT-proBNP decrease of about 15% after treatment is biologically significant. Indeed, according to the specifications of the desirable biological variation database 7 , this decrease should be considered as significant according to the within-subject biological variation (CVi 10%), but not significant according to the betweensubject biological variation (CVg 16%). In any case, this decrease should be considered as analytically significant since it exceeds the allowable limit of total error, which combines the analytical imprecision and the inter-method inaccuracy, fixed at 13% for NT-proBNP. Moreover, one could argue that a longer treatment period, even one extra month, could be sufficient to significantly lower its circulating level. In support of this assumption, Hoekema et al. showed a significant decrease in NT-proBNP (-58%, P =0.035) in ten patients with moderate to severe OSA treated by MAD (adherence 6.8 hours/night, 6.9 nights/week) after a period of 69 to 82 days 5 . For these ten patients, baseline (52 pg/mL, interquartile range (IQR): 13-105), and follow-up NT-proBNP values (22 pg/mL, IQR: 15-33) were within or close to the reference intervals established according to the method 8 , and were thus in accordance with exclusion criteria discarding patients with a history of cardiovascular disease (CVD). Unlike Hoekema et al., NT-proBNP values from Recoquillon et al. reached 500 to 700 pg/mL, i.e. much higher than the normal values announced by the manufacturer (median 47 pg/mL, min-max: 3.9-155 pg/mL 9 ). This is somewhat in contradiction with the exclusion criteria supposed to discard patients with a history of CVD, including heart failure 4 . Another study showed a significant decrease of plasma BNP levels (-24%) after 6 months of MAD therapy in patients with stable, mild to moderate congestive heart failure (CHF), and OSA 6 . Although less stable in plasma than NT-proBNP 10 , BNP, the other biomarker of CHF, is still widely and routinely assayed on analyzers in hospital laboratories, and thus remains of potential interest for the follow-up of cardiac function under MAD treatment.

2) Statistical standpoint.
In the supplemental statistical section, Recoquillon et al. mentioned that variables with non-continuous distributions are described as median (IQR) 4 . However, NT-proBNP results were expressed as mean (standard deviation (SD)), and reached 296.8 (401.6) pg/mL. Such a wide SD suggests a strong skewness of distribution, with possibly the presence of extreme outliers at baseline and/or at follow-up, in the low values and/or in the high values, which could ultimately false the observed difference in concentrations. The impact of outliers is well-known in research, be they of clinical or laboratory origin 11,12 ; they must be detected through appropriate methods, possibly with adjustment for skewness 13 , and their presence requires investigation, especially for small groups. In the study from Recoquillon et al., a median (IQR) expression would therefore have been more appropriate than mean (SD), and a graph detailing the scatter dot plots for both groups, with connecting lines before and after treatment, would have been required. Moreover, a linear regression analysis was used for the adjustment of baseline values and potential covariates: age, gender, body mass index, and baseline AHI. Nevertheless, these covariates were used for the adjustment of all biomarkers, but no statistical proof was provided as regards to their degree of correlation with NT-proBNP specifically. Furthermore, given the limited number of patients (n ±55), if NT-proBNP results were not normally distributed (as seems to be the case), nonparametric ANCOVA or robust regression methods would probably have been more appropriate 14-17 .

3) Analytical standpoint.
Recoquillon et al. assayed plasma NT-proBNP using a multiplex electrochemiluminescent immunoassay on a MESO QuickPlex® SQ120 analyzer (MSD, Rockville, USA). Using this technology for assaying this cardiac biomarker is somewhat unusual. Indeed, as reminded by the manufacturer, this method is for research use only, but not for use in diagnostic or therapeutic procedures. It involves three incubation steps, interspersed with three wash sequences, requiring at least five hours of preparation for one 96-well plate. To our knowledge, no studies based on

Amendments from Version 1
As suggested by the referee 2, this new version reminds the lack of consensus regarding the interest of natriuretic peptides in the evaluation of cardiac function in OSA, or regarding the impact of CPAP in this context. From a statistical standpoint, the potential effect of outliers on final results is also reminded, especially when groups include few subjects and when biomarkers concentrations display too wide standard variations.

REVISED
this assay have been published up to now, not even the eight references cited in the MSD technical sheet of the human NT-proBNP assay kit 9 . Given the long and tedious assay protocol impractical in hospital laboratory routine, and given the absence of hindsight about its analytical performance, the authors should rather have used a most widespread, reliable, and rapid automated method, like the electrochemiluminescent immunoassay method on Roche analyzers (Roche Diagnostics, Mannheim, Germany) 18 , which was, moreover, available in the laboratory from one of the five hospital centers participating to the trial (Poitiers Hospital Center 19 , NCT01426607 20 ).
The strong association between the severity of OSA and a higher prevalence and incidence of cardiovascular events has been evidenced for a long time 21 . However, where cardiac biomarkers are useful for the diagnosis of many cardiovascular diseases, the interest of natriuretic peptides in the evaluation of OSAassociated cardiac dysfunction and CPAP effects remain controversial, especially because of uncontrolled co-morbidities, as summarized by Maeder et al. 22,23 . To date, no study has specifically evaluated whether NT-proBNP reflects cardiac dysfunction in OSA or whether NT-proBNP is under the influence of obesity in this context. An interesting perspective is the ongoing MOSAIC study, whose main objective is to assess the impact of three months of MAD therapy on AHI in Asian patients with heart failure and OSA 24 . Indeed, one of the planned secondary objectives is the evaluation of cardiac remodeling and of cardiac biomarkers, including NT-proBNP. Meanwhile, the present standpoints remind and emphasize the need for close collaborations between sleep specialists and laboratory practitioners to strengthen the methodological quality and robustness of studies involving biomarkers.

Data availability
All data underlying the results are available as part of the article and no additional source data are required.

Grant information
The author(s) declared that no grants were involved in supporting this work. With regards to the statistical standpoint I fully agree that BNP and NT-proBNP always display a skewed distribution, and that this should have been taken into account. When using non-parametric tests, the result is even more unlikely to be statistically significant.
With regards to the analytical standpoint I also agree: the use of an established easy to use NT-proBNP assay would have made it easier to check the plausibility of the data and to compare it with the existing literature.
In general, the comments were written well, as the cited paper is a randomised control trial design and the author made a critical appraisal. Since many readers might not have time to review the original paper, it might be helpful to write a summary before the comments. Otherwise no further comments were made; and the report is suitable to be accepted in current version.
Is the rationale for commenting on the previous publication clearly described? Yes

Are any opinions stated well-argued, clear and cogent? Yes
Are arguments sufficiently supported by evidence from the published literature or by new data and results? Yes

Is the conclusion balanced and justified on the basis of the presented arguments? Yes
No competing interests were disclosed.

Competing Interests:
Reviewer Expertise: Sleep medicine and cardiovascular disease I have read this submission. I believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
The benefits of publishing with F1000Research: Your article is published within days, with no editorial bias You can publish traditional articles, null/negative results, case reports, data notes and more The peer review process is transparent and collaborative Your article is indexed in PubMed after passing peer review Dedicated customer support at every stage For pre-submission enquiries, contact research@f1000.com