Obesity modulates the association between sleep apnea treatment and CHI3L1 levels but not CHIT1 activity in moderate to severe OSA: an observational study

A subset of 97 participants were selected from a larger 284 patients study [24]. All the participants had been diagnosed with OSA in Iceland and referred for PAP treatment to the Landspitali—The National University Hospital in Iceland from February 2010 to December 2013. The initial diagnosis of OSA was defined by AHI ≥ 15 events/h and oxygen desaturation index (ODI) ≥ 10 events/h. When sleep studies were rescored, there were however some subjects (n = 9) who had AHI between 10 and 15 events/h but they were not excluded from the study. More than 90% of approached subjects agreed to participate. Selection criteria for the subset used in this study were as follows: (1) participation in a sleep study before starting PAP treatment with a recorded apnea–hypopnea index (AHI) and oxygen desaturation index (ODI). All participants had moderate to severe OSA (defined as AHI ≥ 15) as that was a requirement for PAP treatment; (2) having blood samples taken while untreated and at follow-up; (3) used PAP treatment for ≥ 20 days and ≥ 4 h per day on average for the previous 4 weeks; and (5) completion of a 4-month follow-up as of May 13, 2014. This study was approved by the National Bioethics Committee and the Data Protection Authority of Iceland (10-048). Written consent was obtained from all subjects.

A core questionnaire on sleep, health, and medication use was answered by subjects at baseline and at a 4-month follow-up (e.g., if they had been diagnosed with hypertension, diabetes or cardiovascular disease, or other diseases by a physician). Height and weight were measured for all participants with the same instruments (a ruler and a scale), after having removed shoes as well as objects from clothes. Blood was drawn in the morning from the antecubital vein after overnight fasting, both at baseline from untreated participants and at follow-up. Prior to referral for PAP treatment, all subjects had a type 3 sleep study with a T3 device (Nox Medical, Reykjavik, Iceland), an Embletta type 3 portable monitor, or an Embla 12-channel system (Natus Medical Inc., San Carlos, CA, USA). Type 3 portable monitors were used for home sleep apnea testing (HSAT) as this is the clinical practice in Iceland as it is in many other countries, e.g., in Europe [25], and accepted by the AASM in the USA [26]. The study therefore lacks electroencephalographic recording and assessment of arousals. However, the validity of the Nox T3 portable monitoring system for assessing sleep-disordered breathing has been shown when compared with polysomnography with similar AHI and ODI levels [27, 28]. Also, the respiratory scoring rules used (requiring 4% oxygen desaturation for hypopnea events) do not require arousals to assess hypopneas. The same signals were recorded on all studies. Nasal airflow was recorded through a cannula. Chest and abdominal movements were measured using respiratory inductance plethysmography belts. Pulse and oxygen desaturation were measured by a finger probe oximeter based on a four-beat exponential average (Nonin Medical Inc., Plymouth, MN, USA). Body position and activity were measured using sensors situated on the chest. The sleep studies were scored by trained sleep technologists. Studies had to have ≥ 4 h of scorable oxygen saturation and more than two out of three respiratory traces: cannula flow, thorax and respiratory inductive plethysmography belts. Total analysis time was assessed based on questionnaires and the sleep technologist’s review of the study. Sleep studies were scored in accordance with the American Academy of Sleep Medicine (AASM) 2007 manual [29], using the recommended hypopnea classification requiring a ≥ 30% drop in respiratory flow for ≥ 10 s with ≥ 4% oxygen desaturation.

PAP adherence at follow-up was measured objectively by downloading the mask-on time stored by the PAP unit in the previous 4 weeks (S8 machines, ResMed, San Diego, CA, USA). Participants who used PAP for ≥ 20 days and ≥ 4 h per day on average for the previous 4 weeks were considered full users.

The CHI3L1 levels in human plasma samples were measured using a commercially available Sandwich Elisa Duoset from R&D systems (cat. DY2599, Minneapolis, MN). The assays were run in 96-well plates (Nunc 442404, MaxiSorp) using 100 μl of plasma (1:800 dilution) as per manufacturer’s instructions. For standards, human recombinant CHI3L1 protein was used at concentrations ranging from 31.25 to 2000 pg/ml. All samples were measured in duplicates.

Chitotriosidase enzyme assay was based on the method described by Hollak et al. [30] with minor modifications. Briefly, chitotriosidase activity was determined by incubating 5 μl of plasma with 100 μl of 22 mol/l fluorogenic substrate 4-methylumbelliferyl β-d-N,N′,N′′-triacetylchitotrioside (Sigma M5639) in 0.1 M/0.2 M citrate–phosphate buffer (pH 5.2) for 15 min at 37 °C. The reaction was stopped with 200 μl of 1 M glycine–NaOH buffer (pH 10.6) by mixing at room temperature. The substrate hydrolysis by chitotriosidase produces the fluorescent molecule 4-methylumbelliferone, which was quantified with a fluorometer (Spectramax M3 instrument), excitation at 360 nm and emission at 450 nm, and compared with a standard 4-methylumbelliferone (Sigma M1508) calibration curve. Plasma chitotriosidase activity was measured in triplicates. Inter-assay coefficient of variation (CV) for both CHI3L1 and CHIT1 was within 20% and intra-assay CV was within 10%.

Descriptive group comparisons were performed using one-way analysis of variance and chi-square tests for continuous and categorical variables, respectively. CHI3L1 and CHIT1 were natural log-transformed for all analyses (based on assessment of residual error distribution, thereby permitting parametric analysis). The strengths of linear associations among biomarker values (CHI3L1 levels, CHIT1 activity), BMI, and OSA severity were assessed using Pearson’s correlation and multiple linear regression. Paired t test was used to compare the means of biomarker values before and after PAP treatment. Independent t test was used to compare the means of biomarker values between men and women. A two-way repeated measures ANOVA with interaction effects was performed to evaluate interaction effects between PAP treatment and BMI groups on biomarker values. All statistical analyses were performed using SPSS v.23.0 (IBM Corp., Armonk, NY, USA).

To avoid extrapolation beyond the scope of the data, sensitivity analyses were performed excluding those participants who had OSA severity not found in the other BMI groups [3]. All of the analyses led to the same conclusions as the original ones, and therefore all participants were used for the final analyses of the study.

Clinical and biochemical characteristics of OSA patients at baseline, both overall and within different BMI categories, are presented in Table 1. Severely obese subjects (BMI ≥ 35 kg/m²) had more severe OSA (higher AHI and ODI) as well as higher levels of CHI3L1, compared to patients with lower BMI (< 30 kg/m² and 30–35 kg/m²). Furthermore, OSA patients within the highest BMI group also had a higher prevalence of type 2 diabetes compared to the other groups.

The range in values for AHI was between 15 and 116 events/h; for ODI, between 4 and 113 event/h; and for BMI, between 24.8 and 53.3 kg/m².

The correlation between chitinases and OSA severity markers (AHI, ODI) as well as obesity levels (BMI) was assessed. BMI was positively correlated with CHI3L1 levels (r = 0.21, p = 0.04) but not with CHIT1 activity (r = 0.06, p = 0.53). For OSA severity markers, a tendency for a weak positive correlation was observed with CHI3L1 levels, although not significant, and no relationship was found with CHIT1 activity (Fig. 1).

The relationships between the OSA severity markers and the chitinases were also assessed within the three different BMI categories. An indication of a specific pattern appeared for CHI3L1 levels. Although all insignificant (p > 0.05), both OSA severity markers correlated negatively with CHI3L1 levels within the lower BMI groups (< 30 kg/m² and 30–35 kg/m²) but positively within the highest BMI group (BMI ≥ 35 kg/m², Fig. S1a and Fig. S2a in supplement). No effect between OSA severity and CHIT1 activity was found when assessed within the three the same BMI categories (Fig. S1b and Fig. S2b in supplement).

To further evaluate how the level of BMI affects the relationship between OSA severity and chitinases, multiple linear regression analysis with interaction was performed (Table 2). The analysis demonstrated independent association of BMI and age with CHI3L1 levels. Significance (p < 0.05) was reached, independent of AHI (model 1) or ODI (model 2) being included in the analysis. Neither AHI nor ODI associated independently with CHI3L1 levels and the interaction between the markers and BMI was also non significant (p > 0.05). No independent association was found with CHIT1 activity.

Overall, no significant differences in CHI3L1 levels (3.87 ± 0.05 vs. 3.79 ± 0.05, p = 0.08) and CHIT1 activity (3.17 ± 0.09 vs. 3.20 ± 0.09, p = 0.08) were found among OSA patients before and after PAP treatment. There was a trend towards an interaction between treatment and BMI, although the interaction effect did not reach significance (p = 0.08, Fig. S3 in supplement). For an increased power, we also ran the analysis by combining the two lower BMI groups. Figure 2 presents differences in CHI3L1 levels and CHIT1 activity before and after PAP treatment when stratified by two BMI groups (< 35 kg/m² and ≥ 35 kg/m²). A two-way repeated measures ANOVA revealed a significant interaction (p = 0.028) between the two factors for CHI3L1 levels showing a significant decrease in CHI3L1 levels for OSA patients with PAP treatment with BMI ≥ 35 only, not those with BMI < 35 (Fig. 2a). No significant interaction between PAP treatment and BMI groups (p = 0.976) was found for CHIT1 activity (Fig. 2b).

Four additional repeated measures ANOVA analyses were also performed, with each analysis excluding a different group of patients from the study, to assess the potential confounding effects of other diseases. Each analysis excluded patients with hypertension, cardiovascular disease, stroke, or type 2 diabetes. Interaction effects between PAP treatment and the two BMI groups reached significance for CHI3L1 levels (p < 0.05) but not for CHIT1 activity, independent of which patient group was excluded. This is in accordance with the primary analysis. No significant differences were found between men and women in CHI3L1 levels at baseline (3.86 ± 0.06 vs. 3.87 ± 0.10, p = 0.99) or after treatment (3.78 ± 0.06 vs. 3.82 ± 0.10, p = 0.72). Despite this lack of difference, sensitivity analysis was performed including men only (n = 72). The interaction effect for CHI3L1 levels showed a similar trend as for the main analysis but did not reach significance (p = 0.20, Fig. S4 in supplement).