The Relationship Between Leptin Levels and CPAP Treatment: A Cluster Analysis

Background: Leptin is an appetite-suppressing hormone, released by adipose tissues, that plays an important role in severe obstructive sleep apnea syndrome (OSAS). However, it is unclear whether leptin levels are a useful OSAS biomarker. This study assessed the effect of continuous positive airway pressure (CPAP) treatment for OSAS according to leptin levels using a cluster classication based on OSAS clinical features. Methods: Hierarchical cluster analysis was performed on 97 patients with OSAS who had been diagnosed via polysomnography. We also evaluated the adherence of CPAP data after 6 months of CPAP administration. Results: Cluster 1 (49 subjects, 50.5%) had severe OSAS, were obese, and had normal leptin levels. Cluster 2 (6 subjects, 6.2%) had the most severe OSAS, were obese, normal leptin levels, and high adiponectin levels. Cluster 3 (11 subjects, 11.3%) had the most severe OSAS, severe obesity, and the highest leptin levels. Cluster 4 (31 subjects, 32%) had the most severe OSAS, severe obesity, and high leptin levels. After CPAP treatment, EDS improved in all clusters. In Clusters 3 and 4, leptin levels were signicantly reduced after treatment. Conclusions: To establish if leptin can be a biomarker it is necessary to elucidate which further studies These results were not accompanied by BMI changes. The adiponectin levels did not change. Clusters 3 and 4 h-CRP value improved but was not signicant.

values were also calculated from PSG data. Baseline clinical features (height, body weight, Epworth Sleepiness Scale (ESS)) were assessed. Height and weight were measured at the rst medical examination, and at the same time, daytime sleepiness was evaluated using ESS in the questionnaire when the medical history was taken.

Blood tests
Blood counts and general biochemical testing data were obtained from general medical records. Serum leptin and adiponectin levels were measured in all patients. Leptin and adiponectin were measured using the Human Leptin Assay Kit (catalog number #27775; IBL, Inc., Gunma, Japan) and the Human Adiponectin ELISA Kit (catalog number #CY-8050; CircuLex, Inc., Nagano, Japan), respectively.
Spirometric predictions were obtained from the literature [18], and for the arterial blood gas analysis were taken from the radial artery in a sitting position. The criterion for obstructive ventilatory dysfunction was de ned as 70% or less per second, and the criterion for restrictive ventilatory dysfunction was 80% or less vital capacity.

ESS score
Daytime sleepiness was assessed using the ESS 19, a fully validated 8-item self-administered questionnaire. Each item is scored on a scale of 0-3, with a score of more than 10 out of a total possible score of 24 scores indicative of daytime sleepiness.

CPAP adherence
Before starting CPAP treatment, the CPAP therapist explained treatment precautions (e.g. how to put on and take off the mask, operate CPAP, and daily maintenance). Mask tting was performed to select the appropriate mask. All patients were able to use CPAP more than 70% or more than 4 hours. The residual AHI, ESS score, leptin, adiponectin, high-sensitivity C-reactive protein (h-CRP), and body mass index (BMI) were measured six months after CPAP treatment. With regard to cardiovascular risk, although long-term observation is necessary, 6 months was considered su cient for observing changes in the molecular pathology.

Statistical analysis
Ward's method was suitable for this study considering the relatively few samples were classi ed into four clusters. The cluster analysis (Ward's method) resulted in four subpopulations; the resulting data were tested for equality of variance between four clusters. For clusters with equal variance, a one-way ANOVA with Tukey's all column comparison test was used for intergroup comparisons. Otherwise, intergroup comparisons were performed using the nonparametric Games-Howell test. For F-values > 4, a p-value of < 0.05 was considered statistically signi cant. Analyses were performed using software Text Explorer module of JMP Pro 13.
The required number of subjects was determined from the independent variables while performing a multi-group comparison of each cluster, and the sample size was su cient (n > 64) even when the power was set to 80%. Normality of the variables was checked, and the results are presented as mean ± standard deviation (SD) values. We selected the four variables, AHI, BMI, leptin and adiponectin into four clusters (Table 1). AHI is considered to be the most important criterion for treatment indication [19]. High BMI is the most important factor in the severity of OSAS [20]. In obese patients, leptin and adiponectin are the most important biomarkers of lipid metabolism [21].

Patient characteristics
The cluster analysis identi ed four clusters. Compared to Clusters 1 and 2, Cluster 3 and 4 were younger, had a higher male:female ratio, a higher BMI and history of smoking. The ESS scores were elevated in Cluster 3 compared to other clusters (Table 1).    Table 3 shows the lung function test results. Although no ventilatory dysfunction was observed, forced vital capacity and forced expiratory volume were lower in Clusters 2 and 3. There were no signi cant differences regarding peripheral airway obstruction. According to the blood test results (see Table 4), the number of white blood cells and liver function parameters increased in Clusters 3 and 4. High-sensitivity test values that detect C-reactive protein (CRP) also increased, particularly in Cluster 3. these patients were obese. The mean ESS score of this cluster was 8.7 ± 4.8, which indicates nonsymptomatic excessive daytime sleepiness (EDS). The mean AHI, however, was 40 ± 15, which is below the overall average. Leptin and adiponectin levels were low compared to other clusters.
Cluster 2 (6 subjects, 6.2%) had the most severe OSAS, were obese, normal leptin levels, and high adiponectin levels. This cluster also mainly included older participants (mean age: 65 ± 13 years). Their mean BMI was 28 ± 5.8, indicating that they were mildly overweight. The mean ESS score of this cluster was 7.3 ± 3.6, which is low, indicating that this cluster had the lowest likelihood of presenting with EDS symptoms among all clusters (Table 1). However, the AHI was 60 ± 20, indicating severe apnea and hypopnea. Leptin levels were normal, and this cluster had the highest adiponectin levels of all clusters; however, the leptin/adiponectin ratio was the lowest among all clusters.
Cluster 3 (11 subjects, 11.3%) had the most severe OSAS, severe obesity, and the highest leptin levels. This cluster was centered on middle-aged patients (mean age: 52 ± 15 years). The mean BMI of patients in this cluster was 34 ± 6.7, which indicates that they had severe obesity. The mean ESS score was 9.4 ± 4.8 ( Table 1). The AHI was 72 ± 25, which is far higher than the overall mean AHI. This cluster accordingly comprised the most severe OSAS patients who had severe obesity ( Table 2). Leptin levels and the leptin/adiponectin ratio were the highest of all clusters.
Cluster 4 (31 subjects, 32%) had the most severe OSAS, severe obesity, and high leptin levels. It mainly included middle-aged patients (mean age: 51 ± 12 years). The mean BMI was 33 ± 3.5, indicating severe obesity. The mean ESS score was 8.7 ± 5.3 ( Table 1). The AHI was 84 ± 30, which is much higher than the overall AHI mean. This cluster was accordingly composed of the most severe OSAS patients, and had mild-to-moderate symptoms ( Table 2). Leptin levels were high and the leptin/adiponectin ratio was comparable to that of Cluster 3. Table 5 shows AHI, ESS, leptin, adiponectin, h-CRP, BMI before and after 6 months using CPAP. After CPAP treatment, EDS improved in all clusters. In Cluster 3, which had high leptin levels before treatment, leptin levels were signi cantly reduced after treatment. These results were not accompanied by BMI changes. The adiponectin levels did not change. Clusters 3 and 4 h-CRP value improved but was not signi cant.

Summary of results
Four variables, AHI, BMI, leptin and adiponectin were selected. They are the most important factors to OSAS and Obesity [19][20][21]. The subjects were middle-aged patients around 50 years old, and had low levels of daytime sleepiness. The cluster analysis shows that the combination of high levels of AHI, BMI, and leptin can be considered an OHS phenotype, and are considered to be the most important phenotypes requiring CPAP treatment in this study. On the other hand, there are phenotypes where leptin is not elevated even in severe OSAS and obesity, making it necessary to examine CPAP effectiveness in these individuals. For example, just as obesity and metabolic syndrome [22] depend on the presence or absence of metabolic disorders, whether apnea is directly affecting metabolic dysfunction is important in increasing cardiovascular risk. Therefore, if it can be evaluated, it will help determine the effectiveness of CPAP. However, evaluating it accurately is di cult and requires further study.

Subgroup-speci c links to adipokines
This study investigated whether adipokines, especially leptin, could be a biomarker for e cacy of CPAP.
Leptin is a hormone produced by fat cells that suppresses appetite, and may play an important role in OHS since it enhances ventilation response through the nervous system. Obesity may cause leptin resistance in the central nervous system, leading to diminished ventilatory responses [15]. Previous studies did not identify leptin as a biomarker for e cacy of CPAP as obesity could not be ruled out as a confounding factor. Therefore, we attempted to eliminate any confounding effects by separating clusters according to BMI. Consequently, Clusters 1 and 2 had normal leptin levels while Clusters 3 and 4 had high leptin levels; Clusters 3 and 4 had different leptin levels but the same degree of obesity. Adiponectin and leptin-adiponectin ratios were calculated, but the differences between the clusters were unclear.

Prevalence of OSAS in women
Leptin levels differ between men and women; however, our data did not take these differences into account, as the main purpose was to determine changes before and after CPAP use in all patients, regardless of the sex. Previous studies have reported that OSAS prevalence in women is 9% (24% for men) [23]. Although there was no signi cant difference in the number of men and women, and the number of participants was small, the proportion of women in Cluster 3 was higher than other clusters. It has been reported that OSAS frequency suddenly increases in women after menopause and that postmenopausal women's hormone changes (particularly the decrease in progesterone) suppresses respiratory stimulation [24,25]. Even then, men are more likely to suffer from OSAS, probably because of the shape of the throat and airways, respiratory stimulating hormone, and that upper body obesity is more common in men (women are more obese in the lower body) [26]. Conversely, OHS is reportedly more prevalent in women than OSAS [27]. Moreover, morbid obesity (BMI > 40) may be more common in women than men and associated with a higher OHS prevalence. Furthermore, OHS is more commonly associated with heart disease than OSAS, even at similar BMIs, and some reports have claimed that the untreated OHS mortality is 46% (over 50 months) [28]. In this study, Clusters 3 and 4 may have had more women because OHS was more common in these two clusters.

Therapeutic effects of CPAP
The therapeutic effect of CPAP on OSAS and current treatment practice are based on ndings published many years ago [29]. However, a previous randomized control trial failed to nd a therapeutic effect of CPAP on OSAS [4]. Therefore, selecting and using more cases than those selected by AHI-only criteria is necessary. Nonetheless, CPAP and non-invasive ventilation are effective treatments for OHS, the most serious form of OSAS, but blood gas analysis collection and invasive procedures are required to meet the diagnostic criteria [30]. Even if the criteria for OHS were not met, as was the case in this study, grouping by phenotype would make it possible to divide the population according to the CPAP treatment effect.
Indeed, Clusters 3 showed markedly improved leptin levels after CPAP, and Clusters 3 and 4 showed a decline in CRP although this was not signi cant. Conversely, although OSAS severity according to the AHI is high, a group with moderate obesity and low leptin levels may have low risk of abnormal lipid metabolism.
In this group, CPAP may not decrease cardiovascular disease because of normal lipid metabolism. Several previous studies have reported that CPAP improves leptin level, while others have not [5].
Assessing e cacy based on a single factor is di cult; however, it is possible to nd a population for which treatment is effective by cluster classi cation, which combines multiple factors. This study was limited by the small sample size, cluster sizes, and narrow capacity for generalization. Although the risk of cardiovascular disease was low, the patient number was small; further studies need to include a larger number of cases. In addition, we believe that the abnormal lipid metabolism in OSAS should be included as a factor in future cluster analyses. To generalize the present results, it is necessary to consider larger sample sizes, and non-hierarchical cluster classi cation should also be performed and re-evaluated.

Conclusions
To establish whether leptin could be a biomarker for CPAP treatment, it is necessary to elucidate the mechanisms of lipid metabolism, leptin, and ventilatory responses in patients with OSAS and accumulate more cases. A prospective study should examine whether the leptin level is a predictor of the CPAP effect on cardiovascular disease. Availability of data and materials; All data generated or analyzed during this study are included in this published article.