Flavor Disorder Is Associated with Obstructive Sleep Apnea: A Nationwide Population-Based Cohort Study

Background: Flavor sensation was the joint perception of smell and taste sensation was associated with the pleasure of life. Obstructive sleep apnea (OSA) had been demonstrated the presence of upper airway remodeling which altered sensory and motor function due to hypoxia or snore vibration. This study aims to investigate whether OSA is associated with the risk of avor disorder (FD). Methods: We conducted a nationwide cohort study using the Taiwan National Health Insurance Research Database, one million subjects were sampled with data collected from 1999 to 2013 and 9,191 identied diagnosed OSA patients were included. Each patient was matched with non-OSA controls from the general population by propensity score matching (1:1) based on age, gender, hypertension, hyperlipidemia, ankylosing spondylitis, and Charlson comorbidity index, 8,037 OSA and an equal number of non-OSA subjects were used to compared in this study. The incidence of FD was assessed at the end of 2013 and cumulative incidences, hazard ratios (HRs) and 95% condence intervals (CIs) were calculated. Results: The adjusted HR (aHR)of FD for the OSA group was 2.082-fold (95% CI = 1.149–3.773, p=0.0155) higher than non-OSA group. The stratied analyses revealed aHR of FD for the hyperlipidemia group was 2.264-fold (95% CI = 1.061–4.832, p=0.035). Subgroup analysis showed the subjected of female OSA had more risk to develop FD (aHR:2.345, 95% CI = 1.026–5.357, p=0.043). Conclusion: A higher risk of developing FD was found among the newly diagnosed OSA cohort during the longer than 10-year follow-up period greater risk of subsequent development of FD among OSA patients than the general population, and 2.264-fold greater riskto develop FD amount group with hyperlipidemia .Furthermore, the stratied analyses revealed signicant effects of female OSA patient with exposure 2.345-fold greater risk than OSA male,OSA patients without hypertension were developing 2.440-fold risk to develop FD than non-OSA patients without hypertension. Instead, there was no signicant difference in FD between the sub group of hyperlipidemias with OSA and non-OSA, it is speculated that the side effects of anti-hyperlipidemia agent are greater than the impact of OSA on avor sensation.


Introduction
Compared to visual, auditory, painful sensation or motor defects, loss of taste, somatosensory, and olfactory damage are generally regarded less to cause immediate attention. However, taste and olfactory disorder diminish the pleasures of life. Flavor sensation (FD) is de ned as a unitary perception derived from multiple sensory afferents, including the joint perception of taste, smell and somatosensory [1].
Taste sensation provides important sensory cues for regulating complex autonomic nerve activities. Such as sour taste has been shown to be associated with parotid salivary ow for initiating digestion [2], Salty taste stimulates the pancreatic and hepatic branches of vagus nerve [3], Sweet taste increases gastric acid secretion by exciting the vagus nerve [4], umami sensation elicited by L-glutamate(glu) stimulates palatability and induces pleasure in gustation [5]. In terms of nutrient homeostasis and modulation of digestion, taste sensation plays multiple important roles.

Page 3/13
The general role of smell sensory is guiding our attention to avoiding hazards (e.g., smoke and microbial threats) and mainly in food intake, the discrepancies between expectation and perceived avor can lead to appetite regulation. [6,7]. Smell sensory also be involved in social communication, via the odors, the mate selection and reproductive behavior can be proceeded. [8] Smell disorder directly reduce food enjoyment, affected in social security and professions depending on olfactory ability (e.g., medical staff, sommelier, cooks, perfumers) may lead to disturbance in important areas and make persons more prone to depression [9]. However, unawareness of olfactory loss was common in general population and prevalence of smell and taste impairment were seem be estimated lower than true prevalence rates [10,11].
Recent study had shown that patients with OSA were under the risk of hyposmia even anosmia due to hypoxia and lower nasal ow [12]. Some report had discussed the association about neuropathies on olfactory, cold sensory over the upper airway [13][14].As a chronic disease syndrome, OSA had been characterized by recurrent nasal, oropharyngeal or hypopharyngeal obstruction and vigorous snoring [15][16], which provokes local in ammation as leukocyte accumulated in oral mucosa or peripheral nerve injury due to snore related vibration [17][18], cause hypoxia, hypoventilation and hypoxemia in system [19], impaired vascular endothelial function and peripheral perfusion [20],affects integrity of vascular endothelium and coagulation thus linked to autoimmune disease [21][22][23] , in addition ,emotion, judgment, and cognition in mental were involvement in human [24][25][26].
In 1993, the prevalence of OSA were reported about 2~4% in North America by the de nition of apnea-hypopnea index, AHI ≥5 [27]. As the time going, even with the more stringent de nition of AHI ≥15 events per hour, the estimated prevalence is increased to around 15 % in males and 5 % in females with the same Wisconsin sleep cohort in 2013 [28]. A recent population-based study demonstrated a need to revise the de nition of this disease in Switzerland, and proposed a high prevalence of moderate to severe OSA (AHI≥15) (23.4% for women, 49.7% for men) [ 29]. In addition to increasing numbers of consultations, sleep-related surgery was become popular and all above factors may affect the sense of taste, smell and somatosensory [1,30]. Flavor is a perception of anatomically separated functionally united, a complete sense of avor is based on a complete taste and smell, even somatosensory and memory. Overall, taste and oral somatosensory cues combine centrally with retro nasal olfactory to produce the composite experience of avor sensation. Therefore, this longitudinal nationwide cohort study was conducted to explore the risk of FD between OSA and general people, whether patients with newly diagnosed of OSA was prone to develop (FD) than general population subsequently, which was not been discussed in a large-scale population-base cohort of single race.

Data source
The data source for this cohort study was reimbursement claims data from Taiwan's National Health Insurance (NHI) program. The reliability of the NHIRD diagnostic codes has been supported by a recent validation study [31].
Mandatory insurance policies mean that more than 99% of the 23 million people of Taiwan participate in the universal health insurance program [32]. One million subjects were sampled from the population and collected the data from 1999 to 2013. The incidence of FD was assessed at the end of 2012.

Patient and Public involvement
From 2000 to 2012, we identi ed 9,191 newly diagnosed OSA, excluded 1,149 subjects with diagnosis of disturbances of sensation of smell and taste or cancer before index date and 5 patients death before censor date, nally match with 8,037 non-OSA subjects included in this study after propensity score matching (1:1) based on age, gender, hypertension, hyperlipidemia, ankylosing spondylitis and Charlson comorbidity index. There was long follow-up time with high similarity between both groups after propensity matching. The study was approved

Propensity Scores
Generalized propensity scores based on the probability of being in both groups were used to decreased the potential confounders, rstly, a 1:4 age and gender matching was used to give the index date corresponding to the non-OSA cohort. The propensity score matching (1:1) was used to establish non-OSA cohort based on age, gender and preexisting comorbidities including hypertension, hyperlipidemia, ankylosing spondylitis, and Charlson comorbidity index.

Identi ed patients of avor disorder (FD)
The date of rst diagnosis of avor disorder (FD) (ICD-9CM codes 781.1) during the follow-up period was de ned as the primary endpoint. ICD-9 CM codes used in this study included those for avor disturbance including disturbances in sensation of smell or taste. All patients were tracked from the index date to the date of diagnosis of avor disorder, withdraws from the Mandatory insurance or death before censor date during the period of follow up have been eliminated before the screening for inclusion. The matched samples are also eligible for those who did not withdraw or death within the tracking period.

Statistical analysis
Demographic data, between quali ed OSA and non-OSA groups were analyzed with Chi-square test or independent T test. Kaplan-Meier method was used to describe the cumulative incidences of avor disorder among the two groups with differences between the groups evaluated by log rank test. Cox's proportion hazard model was conducted to measure the effects of OSA on the risk of avor disorder.

Results
After applying the inclusion and exclusion criteria and carrying out age-sex matching, 8,037 OSA and equal number of non-OSA subjects were enrolled (Figure 1), table 1 indicates that most of the patients included in the study were age 40-65(OSA group ,n=4044,50.3%),44.4% were female, 24.1% were with hypertension, 11.7% were with hyperlipidemia and 0.3% were with ankylosing spondylitis, among whom 35.7% of the patient' Charlson comorbidity index(CCI) ≧1.Totally ,14.5% (1162/8037) OSA patient accept surgery, including tonsillectomy, adenoid tonsillectomy and UPPP (surgical code :71006,71008,66025). Table 1 compared the demographic characteristics of OSA cohort and non-OSA cohort after propensity score matched.   person-years) and in 40-65 age group (compared to <20 age, 0.69 vs 0.43). Among the preexisting comorbidity subgroups, OSA and hyperlipidemia showed signi cantly higher aHR. Table 3 shows the results of subgroup analysis. The female OSA patient exposure increased risk to develop FD (aHR was 2.345, 95% C.I.=1.026-5.357, p=0.043) for female of non-OSA, on strati ed analysis of non-hypertension, there were 2.440-fold of aHR (95% C.I.=1.210-4.919, p=0.013) to develop FD on OSA group with non-hypertension than non-OSA group with non-hypertension. On hyperlipidemia strati ed analysis (p for interaction was 0.083), OSA with hyperlipidemia were with 3.952 aHR of develop FD (95% C.I.=0.838-18.648) than non-OSA with hyperlipidemia, but no statistically signi cant (p=0.083). Among the preexisting OSA subgroups, female and non-hypertension showed signi cantly higher aHR to develop FD.

Discussion
This is the rst study to determine that patients with OSA exhibited a 2.082-fold greater risk of subsequently developing avor disorder than did the general population after propensity score matching by using a nationwide population-based cohort study. Furthermore, the effects of OSA were found to be signi cant in subgroups of nonhypertension patients.
This population-based follow-up study possesses several unique characteristics. First, our OSA subjects were newly diagnosed by otolaryngologist or pulmonologist and t the criteria of ≧3 outpatient visits (including the index visit) or at least 1 inpatient admission. This is a more rigorous method of sample collection than following examination codes of polysomnography (PSG, examination codes: 17008A and 17008B in HNIRD) as OSA patients accept therapy from otolaryngologist or pulmonologist, while many non-OSA patients accept PSG also. Second, the selection of our control group was based on propensity score matching. Therefore, comorbidities were controlled. Third, 99% of the population in Taiwan is insured and almost all patients are comprehensively followed up with medical and even surgical services.
There are several possible reasons for the increase in risk of FD in OSA. First, the common theory is that inherent vibration injury to soft tissue over oropharynx leads to peripheral neuropathy or hypoxia both [17][18][19]. Second, surgery induces avor disorder [33]. Due to high availability and low cost, one population-based study using NHIRD from 2002-2010 showed that 17.9% (922/5139) of patients accept surgery for OSA including oropharyngeal surgery and nasal surgery [32]. Our data showed 14.5% OSA patient accept surgery over oropharynx, including tonsillectomy, adenoid tonsillectomy and UPPP (surgical code :71006,71008,66025) which may impact to gustatory. Subjected with persistent taste disorders and smell disorders after surgery would be recorded in the diagnostic code. Previous research demonstrated that the percentage of temporal subjective gustatory changed after tonsillectomy is nearly 60/188 (32%) at 2 weeks and 15/181 (8%) at more than 6 months [34]. There may have some effect to our result. Therefore, we roughly estimate that will increase the 4.64~1.16% (14.5% times 32% or 14.5% times 8%) risk of FD in OSA group.
There is some evidence to suggest that there are no fundamental taste differences associated with hypertension and patients with hypertension have higher recognition threshold for salt [35]. An animal study has shown an association between hypertension and salt preference when Ang-II type II angiotension receptors in rat brain are stimulated [36][37]. Moreover, some studies have illustrated that hypertensive humans have a preference for salty as salt-sensitive rats [38][39]. Although some species of antihypertensive drugs may alter taste (e.g., ACEI inhibitors, some calciumchannel blockers), some agents can produce a prolonged bitter taste (e.g., β-adrenergic receptor blockers, rilmenidine and antiarrhythmic agents) [40], diuretics and α/β adrenergic blockers were less impact to altered taste or induce ageusia [41]. Therefore, from the characteristics of hypertension itself and the enhancement or reduction of antihypertensive agents to chemoreceptor, hypertension is not a risk factor for avor disorder. Our data showed that aHR (0.871) is lower than in control groups in hypertension, indeed is lower, but not signi cant(p=0.708).
Many studies have demonstrated associations for hyperlipidemia and antihyperlipidemic drug use with development of FD. One large cross-sectional study showed that avor dysfunction is related to higher concentration of LDL-C (Ptrend =0.07), the development of dyslipidemia may be a harbinger of future diabetes and may in uence chemoreceptors [42]. A limited number of studies have illustrated that 70% of patients have altered taste after taking hydroxymethylglutaryl-coenzyme A(HMG-CoA) reductase inhibitors(e.g. atorvastatin calcium, uvastatin sodium, simvastatin, lovastatin and pravastatin sodium) [41].
The group of hyperlipidemias had taken HMG-CoA commonly in Taiwan, table 2 showed 2.264-fold of FD for risk of hyperlipidemias to non-hyperlipidemia(p=0.035), which were contained the patient with hyperlipidemia who had taken or had not taken HMG-CoA. However, the subgroup analysis of table 3 showed there were no difference for FD development between hyperlipidemia with OSA and hyperlipemia without OSA (p for interaction was 0.083). If there persist a hypothesis that the side effect of anti-hyperlipidemia agents were more stronger than OSA to develop FD? It's may due to the interference of drug side effects, high availability and low cost of medical environment. This result can be left to subsequent data con rm.
Gender difference in taste sensitivity has been reported, with females being more sensitive to avor change. Many researchers have demonstrated that women have greater taste perception than men based on different taste scores [43][44][45][46]. Table 2 showed no risk of gender difference in general condition (aHR:1.448, p=0.195) but subgroup analysis illustrated woman with OSA had 2.345-fold aHR (p<0.05), whether women's avor sensation was more sensitive to external factors such as hypoxia of OSA, vibration trauma of oropharyngeal mucosa from snoring may need further research to con rm.
Several limitations of this study should be considered when interpreting the results. First, although we obtained a large population with OSA, surgery and medical-seeking behavior were confounding factors due to accessible and convenient medical consumption. For example; 1162 (14.5%) OSA patient accept surgery , estimated literature on the incidence of FD after surgery, that will increase the 4.64~1.16% risk of FD in OSA group, of the same reason to the group of hyperlipidemias had taken HMG-CoA commonly, Therefore, there may be selection bias and this study cohort may not be entirely representative of the population in other country. Second, the severity of OSA is unavailable in the NHIRD, including treatment statuses and improvements in avor sensation. Third, our study utilized a nationwide administrative claims database and the diagnoses of medical comorbidities were solely based on ICD-9-CM codes from the NHIRD. The validity of diagnosis needs to be checked where misclassi cation might exist.
Although there are no uni ed criteria for the diagnosis of OSA, most sleep centers in Taiwan strictly follow the guidelines of AASM (American Academy of Sleep Medicine) for the diagnosis of OSA, meaning the combination of clinical symptoms and PSG test.
However, arranging polysomnography exam were not expensive or even too popular under Taiwan's health insurance system. People of non-OSA (AHI<5) people may be mis-arranged to OSA group. Due to the result about polysomnography were unavailable in the NHIRD database, it may not be the most appropriate way to use polysomnography as a diagnosis of OAS. There for, we introduce a more rigorous method(≧3 outpatient visits or at least 1 inpatient admission) to increase robustness of diagnostic accuracy of OSA cohort. The same methodology was used in some population base research as below for more rigorous diagnosis as ankylosing spondylitis [47].

Conclusion
This 12-13-year population-base cohort study demonstrated a higher risk of developing FD in patient with OSA and women with OSA may be more sensitive. Clinicians need to provide appropriate monitoring for the high-risk patients.

Declarations
Con ict of interest: The authors declare that there is no con ict of interests regarding the publication of this paper.