Introduction

Obstructive sleep apnea (OSA) is a sleep-related breathing disorder characterized by recurrent partial or complete collapse of the upper airway during sleep, resulting in chronic intermittent hypoxia, respiratory cessation, recurrent micro-arousal, and daytime sleepiness. Untreated OSA patients are at increased risk for cardiovascular disease, stroke, atherosclerosis, hypertension, and venous thromboembolism [1,2,3,4]. Although the pathophysiological mechanisms behind the association are still not thoroughly clarified, several studies have proposed that OSA predisposes to a multifactorial prothrombotic state, in which endothelial dysfunction, exaggerated platelet activity, impaired fibrinolysis, as well as chronic, low-grade systematic inflammation are known to play a critical role[5,6,7,8].

P-selectin, as a member of the selectin family of adhesion molecules, is localized in the membranes of alpha-granules of platelets and the Weibel–Palade bodies of endothelial cells [9]. It is expressed upon stimulation by thrombin, hypoxia, histamine, or cytokines by both platelets and endothelial cells [10,11,12]. It is a critical mediator of platelet–leukocyte interaction and plays an important part in leukocyte recruitment, attachment, rolling, and diapedesis into the vessel wall in early stages of atherosclerosis [13,14,15]. It is also a key molecule in hemostasis and thrombosis mediating platelet rolling, generating procoagulant microparticles, enhancing fibrin deposition, and initiating the extrinsic pathway of the coagulation cascade indirectly [9, 16,17,18]. Taken together, P-selectin should no longer be regarded only as a simple marker of platelet or endothelial activation, but also as a direct inducer of procoagulant activity associated with vascular and thrombotic diseases [9].

Soluble P-selectin (sP-selectin) is the soluble form of P-selectin, which is a circulating protein derived from the membrane form of P-selectin on both platelet and endothelium cell by proteolytic cleavage [9, 19]. It is elevated in pathologic processes involving activated platelets and endothelial cells. Elevated levels of sP-selectin have been demonstrated in a variety of cardiovascular diseases, including coronary artery disease, myocardial infarction, stroke, hypertension, atrial fibrillation, and venous thromboembolism, with some relationship to prognosis [14, 19]. In addition, sP-selectin levels have been found to be positively correlated with apnea hypopnea index (AHI) and respiratory arousal index, inversely related to minimum oxygen saturation during sleep in some studies [20]. Based on the above findings, a growing number of studies have tried to assess sP-selectin levels in patients with OSA. Unfortunately, whether sP-selectin levels were elevated in OSA patients remain controversial, since previous studies yielded inconsistent results. And most of these studies had relatively small sample sizes. Hence, we conducted a meta-analysis of published studies to determine the levels of sP-selectin in OSA, aiming to further the understanding of the potential role of sP-selectin in patients with OSA.

Methods

We carefully followed the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) in this paper [21].

Data sources and search strategy

The authoritative databases including PubMed, Embase, Web of Science, and Cochrane Library were searched systematically by us to find out eligible studies. Each database was searched from inception through October, 2020. The search terms included: (“OSA” or “Obstructive Sleep Apnea” or “OSAS” or “Obstructive sleep apnea syndrome” or “sleep apnea”) and (“P selectin” or “CD62P”). Apart from the above databases, additional studies were also identified from the references cited in relevant articles.

Selection criteria

The inclusion criteria: (1) All participants were adults (age > 18 years). (2) All participants were examined by polysomnography (PSG). (3) The diagnosis of OSA was based on AHI ≥ 5 events/h. (4) The studies must have included at least two separate groups. Those with AHI ≥ 5 were allocated into OSA group and those with AHI < 5 were allocated into the control group. (5) All OSA patients were newly diagnosed and untreated. (6) Studies reported sP-selectin levels in both OSA group and control group quantitatively. (7) The number of participants for all groups must have been reported. (8) sP-selectin levels were expressed as mean ± standard deviation, mean ± standard error, median and range, or median and interquartile range. (9) Only English language studies were included.

The exclusion criteria: (1) conference abstracts, editorials, reviews, and case reports; (2) studies that did not provide available data for calculating effect estimate; (3) studies were not conducted in humans.

Quality assessment

The Newcastle–Ottawa quality assessment scale (NOS) was adopted to evaluate the quality of each study according to the following aspects: selection of participants, comparability of groups and exposure assessment. The quality score ranges from 0 to 9. Two authors (D.Z and Z.B.X) participated in the process of quality assessment independently. Any dispute was discussed with a third author (T.T.L) and resolved by consensus.

Data extraction

Two authors (D.Z and Z.B.X) identified articles that may be qualified by conducting a screening of titles and abstracts independently. After that, a further screening based on full-text review was undertaken. Only studies satisfied the above selection criteria were included. Data extraction was also carried out by above two authors independently. Then, another author (T.T.L) checked the extracted data for completeness and accuracy. Any discrepancy during data extraction between two independent authors was resolved by consensus with a third author (T.T.L) if necessary. The following data were extracted from the articles: name of first author, publication year, region, sample size, age, gender, body mass index (BMI), AHI of OSA group and control group, type of blood sample, time point of blood drawing, sP-selectin levels in OSA group and control group, and assay methods for sP-selectin levels. If the included studies provided data of median and range or median and interquartile range, the data were transformed to mean and standard deviation using an online computing tool (http://www.math.hkbu.edu.hk/~tongt/papers/median2mean.html) [22, 23].

Statistical analysis

Since the differences in assay methods or units, standardized mean difference (SMD) was chosen to calculate the effect size and results were presented as SMD and 95% confidence intervals (CI) for sP-selectin levels. We chose a random-effects model rather than a fixed-effects model because of this takes into account heterogeneity among multi-studies. We also performed subgroup analysis in this article. Heterogeneity across the studies was assessed using the I2 statistic, with a value of 25–49% representing low heterogeneity, 50–75% moderate heterogeneity, and > 75% high heterogeneity. Sensitivity analysis was conducted to assess the stability of pooled results. All analyses were performed with Review Manager (Version 5.2, The Cochrane Collaboration) and Stata (Version SE12.0, Stata Corporation, USA). A P value of less than 0.05 was judged as statistically significant.

Results

Identification of relevant studies

The detailed steps of the study selection process are shown in Fig. 1. A total of 513 records were initially obtained from electronic databases by the mentioned web-search strategies. After checked for duplicates, 167 articles remained. Among them, 143 articles were discarded on the basis of titles and abstracts, leaving 24 articles for further assessment in full texts. After reading full text thoroughly, 15 articles were excluded due to the following reasons: lack of basic information, OSA was not defined as AHI ≥ 5, controls were not non-OSA, without measuring soluble P-selectin level, and study population was children. Eventually, a total of nine eligible studies were included for analysis [15, 24,25,26,27,28,29,30,31].

Fig. 1
figure 1

Flow diagram of literature search and study selection

Full size image

Characteristics of the studies

Totally, nine studies involving 494 OSA patients and 446 controls were enrolled. The information such as the first author’s name, publication year, region, OSA severity, type of blood sample, time point of blood drawing, assay methods for sP-selectin levels, and NOS Score are summarized in Table 1. And the information of sample size, AHI, age, gender, BMI, and sP-selectin levels are presented in Table 2.

Table 1 The characteristics of included studies
Full size table
Table 2 Main data of included studies
Full size table

Meta-analysis

By combining all included studies, the pooled results exhibited that sP-selectin levels in OSA patients were significantly higher than that in controls (SMD = 0.54, 95%CI 0.29–0.78, I2 = 66%, p < 0.0001). Due to the evidence of moderate heterogeneity (I2 = 66%), a random-effects model was adopted (Fig. 2).

Fig. 2
figure 2

Forest plot for sP-selectin levels between OSA patients and non-OSA controls based on controls with BMI matched or without BMI matched

Full size image

Subgroup analysis

As is known to all, BMI is a main index to measure the degree of obesity. Several scholars have proposed that elevated sP-selectin levels observed in OSA group in their studies may be due to comorbid obesity rather than representing a direct effect of OSA itself [32,33,34]. Hence, it seems to be reasonable to select the studies which contain BMI matched OSA groups and control groups as a new subgroup for analysis. It showed that in this new subgroup, sP-selectin levels were still significantly higher in OSA patients than that in controls (SMD = 0.52, 95% CI: 0.27–0.76, I2 = 23%, p < 0.0001). Meanwhile, the heterogeneity (I2) decreased from 66 to 23% when we conducted subgroup analysis based on control groups with BMI matched, which indicated that BMI might be a potential source of heterogeneity. Besides, the remaining studies which contain BMI unmatched OSA groups and control groups were also combined into another subgroup. And in this subgroup, the average BMI of OSA group in each study was all higher than control group. The pooled result of this subgroup revealed that sP-selectin levels were higher in OSA patients than that in non-OSA controls with lower BMI (SMD = 0.55, 95% CI 0.15–0.95, I2 = 77%, p = 0.007) (Fig. 2).

Considering the impact of blood sample type on measuring sP-selectin levels, subgroup analyses based on different types of blood sample (serum or plasma) were performed. Four studies used serum as blood sample for measurement [24, 26, 28, 31]. The results demonstrated that OSA patients had significantly higher serum sP-selectin levels compared with their controls (SMD = 0.74, 95% CI: 0.28–1.19, I2 = 75%, p = 0.002). And another four studies used plasma as blood sample for measurement [15, 25, 27, 30]. Similarly, the pooled results of them indicated that plasma sP-selectin levels in OSA patients were also significantly higher than that in controls (SMD = 0.51, 95% CI 0.20–0.82, I2 = 56%, p = 0.001) (Fig. 3).

Fig. 3
figure 3

Forest plot for sP-selectin levels between OSA patients and non-OSA controls based on types of blood sample (serum or plasma)

Full size image

Subgroup analyses stratified by different OSA severity were also performed. It showed that compared with control group, the moderate-to-severe OSA (AHI ≥ 15 events/h) group had elevated sP-selectin levels (SMD = 0.80, 95% CI 0.45–1.15, I2 = 67%, p < 0.00001). However, the sP-selectin levels in mild OSA (5 ≤ AHI < 15 events/h) group were similar as that in control group (SMD = 0.14, 95% CI −0.29 to 0.57, I2 = 54%, p = 0.51) (Fig. 4).

Fig. 4
figure 4

Forest plot for sP-selectin levels between OSA patients and non-OSA controls based on different OSA severity (5 ≤ AHI < 15 or AHI ≥ 15)

Full size image

Sensitivity analysis

The removal of each study did not alter the pooled results significantly, indicating that our results were reliable and stable (Fig. 5).

Fig. 5
figure 5

Sensitivity analysis of studies on sP-selectin levels for OSA patients versus controls

Full size image

Publication bias

Owing to the limited number (below 10) of studies included in our meta-analysis, publication bias was not evaluated.

Discussion

The previous systematic review and meta-analysis performed by Nadeem et al. first reviewed and evaluated the selectins levels in OSA [35]. And their results were based on the pooled data of the whole selectins family including P-selectin, E-selectin, and L-selectin. Although significant elevation of the whole selectins levels was detected in OSA, the conclusions on P-selectin were still unknown, because Nadeem’s paper failed to analyze P-selectin independently. In addition, the number of selected studies at that time was too small to conduct subgroup analysis. So far, our paper is the first systematic review and meta-analysis to assess the association between sP-selectin levels and OSA. It provides convincing evidence that patients with OSA have significantly elevated levels of sP-selectin compared with controls. And after subgroup analysis, this conclusion remains unaltered in both BMI matched population and BMI unmatched population.

Most studies have concentrated primarily on obese subjects, because OSA is more common in obese/overweight populations. However, they ignored the fact that the subjects in control groups had normal weight or lower degree of obesity [15, 28, 30, 31]. While previous studies demonstrated that overweight and obese healthy subjects had higher plasma sP-selectin concentrations compared to normal-weight population [36, 37]. Ziccardi et al. also confirmed that weight reduction could result in a significant decrease of sP-selectin levels in obese subjects [37]. Robinson et al. detected significantly higher P-selectin levels in patients with OSA than their unmatched controls and found only BMI was correlated with levels of sP-selectin after multiple linear regression [32]. Even some scholars proposed that possibly obesity was a more potent factor maintaining high P-selectin concentrations than OSA. Therefore, the results of our included studies were probably confounded by the factor of obesity. To control the confounded factor, we performed a subgroup analysis based on BMI matched groups, which became a major merit of this meta-analysis. It showed that sP-selectin levels were still significantly higher in OSA patients than that in controls (SMD = 0.52, 95%CI 0.27–0.76, I2 = 23%, p < 0.0001). And the decrease of heterogeneity (I2) after subgroup analysis also indicated that BMI might be a potential source of heterogeneity.

And according to another subgroup analysis, moderate-to-severe OSA patients with AHI ≥ 15 events/h had significant higher sP-selectin levels, while mild OSA patients with 5 ≤ AHI < 15 events/h showed no significant difference with controls. In view of these results, we inferred a possible explanation that mild OSA patients probably suffered from relative lower degree of hypoxia and the mild injury caused by intermittent hypoxia also could be timely compensated or repaired by human body itself. It might not be sufficient to form an obvious prothrombotic state in vivo. As a result, it was no surprise that sP-selectin levels were similar in mild OSA group and non-OSA control group. Nevertheless, this was just a speculation. Reaching a convincing conclusion was difficult because of too small number of studies in the subgroup analysis. More studies are needed regarding the population of mild OSA.

Clinical implication

As is known to all, PSG is a gold standard modality for diagnosis and screening of OSA. The parameters of PSG (such as AHI, minimum oxygen saturation, mean oxygen saturation, and oxygen desaturation index) were also widely used in estimating curative effect of OSA patients before and after receiving treatments. However, the time-consuming process and expensive cost of PSG greatly restricted its wide application. And PSG required subjects to sleep in hospital wearing devices at least one night, which may affect sleep quality and comfort level resulting in failure of monitoring. Besides, limited by the space and the devices, PSG failed to be implemented in most primary hospitals, especially in developing countries [38]. Therefore, it is urgent to find readily measurable, comparatively cheap biomarkers to be helpful for screening OSA and evaluating the curative effect.

According to our findings in this meta-analysis, sP-selectin levels were significantly higher in OSA patients than non-OSA controls on the whole. We infer that if we try to put sP-selectin measurement before PSG, it may help clinician to identify people who are more likely to suffer from OSA. Then, the clinician can apply PSG more targetedly and avoid unnecessary PSG. However, this is only a preliminary hypothesis. Whether sP-selectin level can be served as a biomarker for screening or therapeutic monitoring of OSA is unknown, which needs further study. However, it was still a potential possible biomarker, because some single-armed studies with small sample size have found that OSA patients showed a fall in sP-selectin level after treatment of continuous positive airway pressure (CPAP), although the evidence grade was not strong [26, 39]. Besides, compared with measurement of P-selectin expression on membrane surface using flow cytometry, measurement of sP-selectin in blood using ELISA is easier, cheaper, and more reliable [40]. Therefore, from the medical economic point of view, measuring sP-selectin before PSG may be benefit.

Possible mechanisms

Although the results of this meta-analysis have important clinical significance, the potential mechanism needs to be further explored. It is well known that P-selectin does indeed reflect some aspect of platelet function or activity. And P-selectin was increasingly used as an important clinical marker of platelet activation [9]. The important pathophysiological consequences of OSA were intermittent hypoxemia and sleep fragmentation, which could result in an increased sympathetic activity shown by increased epinephrine levels in OSA patients [41]. And even when OSA patients were awake, they still exhibited high levels of sympathetic nerve activity [42]. Some scholars have found evidence that increased levels of P-selectin expression were observed in the situation of sympathetic activation [43, 44]. Although this sympathetic activation in their studies was induced by hypoglycemia, depression, or anxiety, we inferred that the increased sympathetic activity caused by OSA also might promote an increase in platelet activation and P-selectin level. In addition, animal experiments showed that a marked overexpression of P-selectin was observed in the larynx tissue and soft palate tissue in sleep apnea rat model [45]. Another study also found the expression of P-selectin on the endothelium in colonic venules was up-regulated only in the rat model of apnea group [46]. A vitro experiment conducted on human umbilical vein endothelial cells showed that P-selectin expression could be enhanced in both hypoxic condition and the stage of hypoxia reoxygenation [11]. Besides, hypoxia led to the inflammation and oxidative stress in endothelial cells, which triggered the release of P-selectin [47, 48]. In summary, elevation of P-selectin in OSA patients may be due to multiple mechanisms.

Limitation

Several limitations of our meta-analysis should be acknowledged. First, the number of included studies and their sample sizes were relatively small, which may lower the statistical significance. More studies with larger sample size would be needed. Second, although all the blood samples from subjects were detected timely, it could not be guaranteed that the processes were conducted with identical method among different studies. The difference of detection methods and the diversity of detection kits might influence the accuracy of sP-selectin level. Third, the uses of common drugs such as statin, aspirin, and clopidogrel have been found to reduce the levels of sP-selectin, which may interfere with our results [49,50,51]. We could not guarantee that all participants were free from the therapy of these drugs. Because, the concrete medication histories of participants were not provided in most of the included studies. Fourth, we could not read and understand the studies published in non-English language since the language barrier. Therefore, only studies published in English language were searched and included in our meta-analysis. Now, English has gone global. In non-English native speaking countries, those studies with positive results were relatively easier to be published in English language; while those studies with negative results were more likely to be published in their native language journals instead of English journals, which might introduce a language bias. Therefore, we should acknowledge that this bias was unavoidable in this meta-analysis.

Suggestions for further research

Since all included studies were preliminary studies, the consequence of this paper was regarded as a proposal. We make four suggestions for further research in this field.

First, oxygen desaturation index (ODI), as another measure of OSA severity, has been adopted to classify patients by some scholars in their studies [48, 52]. Jurado-Gamez et al. found an increase in sP-selectin in severe “desaturators” (ODI > 30%) when compared to mild-to-moderate “desaturators” (ODI 5–30%) [48]. Compared to AHI, ODI is more selective of hypoxemic events in OSA patients. And ODI provides a direct measure of intermittent hypoxemia. It has been proved that intermittent hypoxemia causes increased platelet activation. And a correlation between increased platelet activation and the severity of intermittent hypoxemia was also observed [52]. Hence, ODI-based grouping is worth exploring in future studies.

Second, not all the studies in this field used the unified cut-off values of AHI (5 events/h) as criterion to distinguish OSA from non-OSA. For example, Lo´pez-Cano et al. took 10 events/h of AHI as cut-off value [53]. O’Brien et al. took 1 event/h of AHI as cut-off value in the population of children [20]. And Harsch et al. even selected the OSA patients according to unknown criteria without giving a concrete cut-off value of AHI [54]. The above inconsistent criteria limited the persuasiveness of their studies, which became a reason for us to exclude them from our meta-analysis. Besides, the division criteria for OSA severity were also different among various studies. Consequently, the unified criteria of diagnosis and severity classification should be established in future studies.

Third, some of included studies surveyed the serum level of sP-selectin and others surveyed the plasma level of sP-selectin. Therefore, we conducted a subgroup analysis stratified by blood source (serum or plasma). Although the same conclusion was drawn both in serum subgroup and plasma subgroup, we still suggested that blood sample type should be unified in future studies to avoid confounding factor.

Fourth, whether sP-selectin could be served as a biomarker to evaluate the curative effect of CPAP in OSA patients was still controversial, since some studies show a fall in sP-selectin after CPAP treatment and some not [26, 32, 39]. However, the designs of most studies were single-armed. The absence of either a placebo-controlled arm or a matched control significantly weakened the credibility of their findings. In addition, the treatment duration of CPAP varied from studies to studies. For example, the CPAP treatment duration in Robinson’s study was only 1 month which might be too short to produce a detectable curative effect of CPAP [32]. Therefore, the randomized-controlled trials with longer treatment period, larger sample size are needed to confirm this issue in the future.

Conclusion

On the whole, the pooled results reveal that OSA patients have higher sP-selectin levels than non-OSA controls. And after subgroup analysis, this conclusion still remains unaltered in all subgroups other than the subgroup of mild OSA patients. Only sP-selectin levels in mild OSA subgroup do not exhibit a statistically significant increase when compared to controls. Additional studies are warranted to better identify the role of sP-selectin as a potential biomarker in OSA patients.