The prediction of obstructive sleep apnea severity based on anthropometric and Mallampati indices
Babak Amra1, Mohsen Pirpiran2, Forogh Soltaninejad1, Thomas Penzel3, Ingo Fietze4, Christoph Schoebel5
1 Bamdad Respiratory and Sleep Research Center, Isfahan University of Medical Sciences, Isfahan, Iran
2 Department of Internal Medicine, Isfahan University of Medical Sciences, Isfahan, Iran
3 Center of Sleep Medicine, Charité – Berlin University of Medicine, Berlin, Germany
4 Department of Cardiology and Pulmonology, Center of Sleep Medicine, Charité – Berlin University of Medicine, Berlin, Germany
5 Department of Cardiology and Angiology, Center of Sleep Medicine, Charité – Berlin University of Medicine, Berlin, Germany
|Date of Submission||30-Aug-2018|
|Date of Decision||26-Oct-2018|
|Date of Acceptance||02-May-2019|
|Date of Web Publication||24-Jul-2019|
Asst. Prof. Forogh Soltaninejad
Bamdad Respiratory and Sleep Research Center, Isfahan University of Medical Sciences, Khorshid Hospital, Ostandari Street, Isfahan
Source of Support: None, Conflict of Interest: None
Background: Obstructive sleep apnea (OSA) is a common health issue with serious complications. Regarding the high cost of the polysomnography (PSG), sensitive and inexpensive screening tools are necessary. The objective of this study was to evaluate the predictive value of anthropometric and Mallampati indices for OSA severity in both genders. Materials and Methods: In a cross-sectional study, we evaluated anthropometric data and the Mallampati classification for the patients (n = 205) with age >18 and confirmed OSA in PSG (Apnea–Hypopnea Index [AHI] >5). For predicting the severity of OSA, we applied a decision tree (C5.0) algorithm, with input and target variables considering two models (Model 1: AHI ≥15 with Mallampati >2, age >51 years, and neck circumference [NC] >36 cm and Model 2: AHI ≥30 with condition: gender = female, body mass index (BMI) >35.8, and age >44 years or gender = male, Mallampati ≥2, and abdominal circumference (AC) >112 then AHI ≥30). Results: About 54.1% of the patients were male. Mallampati, age, and NCs are important factors in predicting moderate OSA. The likelihood of moderate OSA severity based on Model 1 was 94.16%. In severe OSA, Mallampati, BMI, age, AC, and gender are more predictive. In Model 2, gender had a significant role. The likelihood of severe OSA based on Model 2 in female patients was 89.98% and in male patients was 90.32%. Comparison of the sensitivity and specificity of both models showed a higher sensitivity of Model 1 (93.5%) and a higher specificity of Model 2 (89.66%). Conclusion: For the prediction of moderate and severe OSA, anthropometric and Mallampati indices are important factors.
Keywords: Anthropometry, body mass index, decision trees, gender, obstructive sleep apnea, polysomnography
|How to cite this article:|
Amra B, Pirpiran M, Soltaninejad F, Penzel T, Fietze I, Schoebel C. The prediction of obstructive sleep apnea severity based on anthropometric and Mallampati indices. J Res Med Sci 2019;24:66
|How to cite this URL:|
Amra B, Pirpiran M, Soltaninejad F, Penzel T, Fietze I, Schoebel C. The prediction of obstructive sleep apnea severity based on anthropometric and Mallampati indices. J Res Med Sci [serial online] 2019 [cited 2019 Aug 17];24:66. Available from: http://www.jmsjournal.net/text.asp?2019/24/1/66/263364
| Introduction|| |
Obstructive sleep apnea (OSA) is a common sleep disorder, and the prevalence of OSA among adult men and women is 14% and 5%, respectively. The prevalence of this disorder has been reported to be 4.98% in Iranian population whereas a higher prevalence of OSA in the USA, Pakistan, and Jordan. The daytime sleepiness, impaired daytime performance, lack of concentration, neurocognitive impairment and hemodynamic, and cardiovascular instability are some consequences of OSA.,
Early diagnosis as well as proper measures are essential to prevent serious complications. Although more attention has been paid to OSA in recent years, this disorder could be remained unrecognized and untreated in more than 80% of cases., Polysomnography (PSG) is the most definitive approach to detect OSA, but it requires expert technical service and is considered as a costly procedure and not readily available. Therefore, a simpler approach with lower cost is required.
To screen this disorder, various instruments have been designed such as Berlin, STOP, and STOP-Bang questionnaires. Although these questionnaires were useful, the results for various races and populations are different.,
On the other hand, numerous studies have reported significant effects of anthropometric indices and obesity on OSA severity.,, The findings suggest that body mass index (BMI), hip circumference (HC), abdominal circumference (AC), neck circumference (NC), and modified Mallampati index as well as waist-to-hip ratio, waist-to-height ratio, and neck-to-waist ratio are associated with OSA severity.,,,,,,,,, However, no prediction analysis has performed with respect to these variables. Therefore, we evaluated the predictive value of anthropometric and Mallampati indices for OSA severity to establish an accurate predictive model.
| Materials and Methods|| |
Study design and participants
The population of this cross-sectional study included patients with confirmed OSA in PSG from March 2016 to September 2017. PSG was performed in Bamdad Respiratory and Sleep Research Center in Isfahan. Cochran's sample size formula was used to determine the optimal number of sampling at 95% confidence interval, a test power of 80%, OSA prevalence of 0.49, and the Type I error rate of 0.1. Finally, a sample size of 205 was selected. The inclusion criteria were age >18 years with confirmed OSA and being consent to participate in the study. The patients with underlying disease (cardiovascular, neurologic, and severe psychological diseases) were excluded.
The ethical code (IR.MUI.REC.1396.3.493) was obtained from the Ethics Committee of Isfahan University of Medical Sciences, and written informed consent was collected from all patients. Overnight, PSG had been performed in all participants. Participants were connected to a set of electrodes including electroencephalography, electrooculography, electrocardiography, chin electromyogram (EMG), leg EMG, respiratory flow sensor, respiratory effort sensor, and SpO2 sensor, and after the onset of sleep, a set of information was recorded and analyzed by computer. Apnea (respiratory airflow collapse) and hypopnea (respiratory airflow reduction) during sleep were recognized and measured as well. Scoring was done according to the American Association of Sleep Medicine guideline 2015. Patients were classified based on the Apnea–Hypopnea Index (AHI) values without OSA (AHI < 5), mild OSA (5 ≤ AHI <15), moderate OSA (15 ≤ AHI < 30), and severe OSA (AHI ≥30). Patients with AHI >5 were diagnosed as OSA and included in the study.
Then, the HC, AC, NC, BMI, height, and age were measured. Furthermore, using Mallampati classification (observations of the patient's throat), patients were classified into four classes. Class I: soft palate was completely visible; Class II: uvula was visible; Class III: only base of uvula was visible; and Class IV: only hard palate was visible.
Finally, collected data were analyzed with SPSS (version 22, Chicago, IL, USA). The qualitative data and quantitative data were expressed as frequencies (percentage) and mean ± standard deviation and given the result of Kolmogorov–Smirnov test (normality of data distribution), and we used independent sample t-test and Pearson correlation coefficient. Furthermore, to evaluate and achieve a decision model for predicting the severity of OSA, we applied decision tree (C5.0) algorithm using input variables and the target variable. To establish decision tree models, input variables including gender, NC, age, BMI, AC, HC, and Mallampati index were used as the predictive variables, and two categorical variables of AHI (AHI <15 or AHI ≥15; AHI<30 or AHI ≥30) were applied as the target variables. Severity of OSA was considered as “dependent variable.” P < 0.05 was considered statistically significant.
| Results|| |
The current study included 205 OSA patients (mean = 111 [54.1%]; mean age = 55.40 ± 14.16 years; mean AHI = 38.87 ± 24.23/h). Among anthropometric indices, the values of height and NC were significantly greater in men than women versus higher values of BMI, AC, and HC in women than men (P < 0.05) [Table 1].
Assessing the correlation between anthropometric indices and AHI showed that age, Mallampati index, and all anthropometric indices (except height) of patients were correlated with AHI (P < 0.05) but taking into account the gender; among female patients, there was no significant correlation between AHI and indices such as Mallampati index, height, and AC/HC ratio (P > 0.05).
Except AC/HC, there were significant differences between all other anthropometric and Mallampati indices and OSA severity [Table 2]. Given the presence of significant correlation of anthropometric indices, age, and gender with AHI of patients, using decision tree (C5.0), a decision model was provided to compare patients with AHI ≥ 15/h versus patients with AHI <15/h (Model 1) and to compare patients with AHI ≥ 30/h versus patients with AHI <30/h (Model 2). The results suggested that the likelihood of AHI ≥15 in patients with Mallampati >2, age > 51 years, and NC >36 cm was estimated to be 94.16%. In Model 2, the analysis showed that gender plays a significant role. According to this decision model, the likelihood of AHI ≥30/h in female patients with BMI >35.8 kg/m2 and age >44 years was 89.98% and the likelihood of AHI ≥30 in male patients with Mallampati ≥2 and HC >112 was estimated to be 90.32% [Table 3]. Finally, the assessment of sensitivity and specificity of both models showed a higher sensitivity of Model 1 (93.5%) than Model 2 but a higher specificity of Model 2 (89.66%) than Model 1. Thereby, to predict the severity of OSA, Model 1 benefits from a higher sensitivity to diagnose patients with AHI ≥15, whereas Model 2 is more specific to predict AHI ≥30 in OSA patients [Table 4].
|Table 2: Anthropometric and Mallampati indices with respect to Apnea–Hypopnea Index|
Click here to view
|Table 4: The comparison of diagnostic values of two models obtained from decision tree (C=5.0)|
Click here to view
| Discussion|| |
This study represented a predictive model for the OSA severity in the Iranian adult population. Hence, accessible and easy-to-obtain indices (including anthropometric indices, age, and gender of patients) were used, and to achieve a proper model, decision tree (C.5) was applied. At the first step, a significant positive correlation was found between anthropometric indices (NC, AC, and age) and Mallampati index with AHI. Afterward, these parameters were inserted into the decision model for predictors of OSA severity. Therefore, two decision models were represented using these variables.
Model 1 represented a decision model for the identification and anticipation of patients with AHI ≥15. The results indicated that the likelihood of AHI ≥15 in patients with Mallampati >2, age >51 years, and NC >36 cm was approximately 94.16%. The sensitivity of the model was very high and its positive predictive value was acceptable. Thus, the model can be identified as a predictive model to distinguish patients based on AHI <15 versus AHI ≥15.
Model 2 represented a decision model for the identification and anticipation of patients with AHI ≥30. In Model 2, gender was an important factor, and the predictive model was different for both genders. The likelihood of AHI ≥30 in female patients with BMI >35.8 kg/m2 and age >44 years was approximately 89.98%, and the likelihood of AHI ≥ 30 in male patients with Mallampati ≥2 and HC >112 cm was approximately 90.32%. The sensitivity and specificity of the model in predicting AHI <30 versus AHI ≥ 30 were 56.78% and 89.66%, respectively. Although the sensitivity of this model is lower than Model 1, its negative predictive value is higher.
In line with this study, Kang et al. in South Korean population suggested obesity as a major risk factor for OSA and a significant positive correlation between all anthropometric indices and the OSA severity. In addition, many studies have shown a significant correlation between indices including age, height, weight, BMI, neck/waist circumference (WC), HC and AC/HC ratio, and AHI. This correlation was significant among OSA patients compared to controls.,, Moreover, the division based on gender by investigators showed a higher prevalence of OSA in men than women. Furthermore, there is a stronger correlation between AC/HC ratio and the severity of OSA in men than women whereas the presence of stronger correlation between BMI and the severity of OSA in women than men.,,
Liu et al. classified OSA patients based on AHI scores of 15 and 30 and attempted to predict the OSA severity based on support vector machine as the model which is called “body profile model.” They found that three parameters of WC, BMI, and NC as well as age can predict OSA severity. They suggested that it could be accurate to distinguish between patients with AHI ≥30 and AHI < 30 among young females in Asia, but the result was slightly worse for males. Their model could be a reasonable instrument to screen patients and rule out moderate-to-severe OSA.
Few studies have pointed out the importance of Mallampati index to diagnose and predict OSA severity. For example, Davidson and Patel have shown that Mallampati index could be effective in evaluating fat deposition in the tongue and measuring obesity. They found the correlation between Mallampati index and AHI in male patients.
Nashi et al. found the relationship between increased tongue volume and a higher Mallampati index. Therefore, it was associated with a high outbreak and severity of sleep disorders. Pinto et al. also applied the Mallampati index as an anthropometric measurement and found that it is significantly associated with PSG parameters such as the AHI.
In line with this study, previous literature has reported that male OSA patients even with similar BMI and WC had greater upper body (chest and abdomen) obesity. It means that fat distribution can play an important role in OSA development. Furthermore, more neck fat in male than female OSA patients has been reported.
Another study had applied logistic regression to identify effective anthropometric indices in OSA severity. The study showed that after adjusting for age and gender, indices including NC, WC, and BMI were effective in OSA, and using receiver operating characteristic analysis, the authors suggested different cutoff values to predict OSA in females and males. The cutoff values for females were NC of 34.5 cm, HC of 76.5 cm, and BMI of 23.05 kg/m2, whereas the cutoff values for males were NC of 38.75 cm, HC of 88.5 cm, and BMI of 24.95 kg/m2.
To predict the presence of OSA, Cizza et al. have reported that an NC ≥ 38 cm had a sensitivity of 58% and a specificity of 79%. Soylu et al. have noted a greater value of NC than WC with respect to an association with OSA in Turkish adults. Furthermore, they stated that to predict OSA, the optimal cutoff values of NC were 35.5 cm in females and 39 cm in males. Zhou et al. have mentioned NC values of 33 cm for females and 37 cm for males as the optimal cutoff values for metabolic syndrome in Chinese adults with OSA.
Martins et al. in Brazil have suggested a similar trend in the literature. The study determined a cutoff value of NC = 40 cm for OSA risk in males compared to the related cutoff value of 36 cm in females. Furthermore, Davidson et al. in the USA have reported that the presence of OSA could be determined by related values of 43 cm for males and 38 cm for females.
Overall, the results of the current study and other studies indicated that anthropometric indices, as well as age and gender, could be effective in predicting the presence and the severity of OSA. While a small sample size of this study could be mentioned as one of the limitations, larger sample size can increase the accuracy of provided model to identify the target variables as well as the possibility of generalization. Thus, it seemingly requires more studies with larger sample sizes and in various populations of different countries because it is likely that the anthropometric conditions of people in different societies follow a variety of patterns, and we need further studies in this field to achieve a unified predictive model.
| Conclusion|| |
According to the results, anthropometric indices are positively correlated with AHI. Based on decision model (C 5.0), the likelihood of AHI ≥15 in patients with Mallampati score >2, age >51 years, and NC >36 cm was estimated to be 94.16% (Model 1) and the likelihood of AHI ≥30 in female patients with BMI >35.8 kg/m2 and age >44 years was estimated to be 89.98%, whereas the likelihood of AHI ≥30 in male patients with Mallampati score ≥2 and AC >112 cm was estimated to be 90.32% (Model 2). In general, Model 1 showed a higher accuracy, sensitivity, and predictive value than Model 2.
Financial support and sponsorship
Conflicts of interest
There are no conflicts of interest.
| References|| |
Liu WT, Wu HT, Juang JN, Wisniewski A, Lee HC, Wu D, et al.
Prediction of the severity of obstructive sleep apnea by anthropometric features via support vector machine. PLoS One 2017;12:e0176991.
Soltaninejad F, Sadeghi A, Amra B. Compliance with continuous positive airway pressure in Persian patients with obstructive sleep apnea. J Res Med Sci 2017;22:114.
] [Full text]
Saddki N, Mohamad H, Mohd Yusof NI, Mohamad D, Mokhtar N, Wan Bakar WZ. Validity and reliability of the Malay version of sleep apnea quality of life index – Preliminary results. Health Qual Life Outcomes 2013;11:100.
Hersi AS. Obstructive sleep apnea and cardiac arrhythmias. Ann Thorac Med 2010;5:10-7.
] [Full text]
Yilmaz A, Akcaalan M. What can anthropometric measurements tell us about obstructive sleep apnoea? Folia Morphol (Warsz) 2017;76:301-6.
Young T, Palta M, Dempsey J, Peppard PE, Nieto FJ, Hla KM. Burden of sleep apnea: Rationale, design, and major findings of the Wisconsin sleep cohort study. WMJ 2009;108:246-9.
Leger D, Bayon V, Laaban JP, Philip P. Impact of sleep apnea on economics. Sleep Med Rev 2012;16:455-62.
Oǧretmenoǧlu O, Süslü AE, Yücel OT, Onerci TM, Sahin A. Body fat composition: A predictive factor for obstructive sleep apnea. Laryngoscope 2005;115:1493-8.
Silva GE, Vana KD, Goodwin JL, Sherrill DL, Quan SF. Identification of patients with sleep disordered breathing: Comparing the four-variable screening tool, STOP, STOP-bang, and Epworth sleepiness scales. J Clin Sleep Med 2011;7:467-72.
Mirrakhimov AE, Sooronbaev T, Mirrakhimov EM. Prevalence of obstructive sleep apnea in Asian adults: A systematic review of the literature. BMC Pulm Med 2013;13:10.
Li KK, Kushida C, Powell NB, Riley RW, Guilleminault C. Obstructive sleep apnea syndrome: A comparison between Far-East Asian and white men. Laryngoscope 2000;110:1689-93.
Dobrowolska-Zarzycka M, Dunin-Wilczyńska I, Szymańska J. Craniofacial structure in patients with obstructive sleep apnoea. Folia Morphol (Warsz) 2016;75:311-5.
Gulati A, Chate RA, Howes TQ. Can a single cephalometric measurement predict obstructive sleep apnea severity? J Clin Sleep Med 2010;6:64-8.
Wysocki J, Charuta A, Kowalcze K, Ptaszyńska-Sarosiek I. Anthropometric and physiologic assessment in sleep apnoea patients regarding body fat distribution. Folia Morphol (Warsz) 2016;75:393-9.
Katz SL, Vaccani JP, Barrowman N, Momoli F, Bradbury CL, Murto K. Does neck-to-waist ratio predict obstructive sleep apnea in children? J Clin Sleep Med 2014;10:1303-8.
Ruehland WR, Rochford PD, O'Donoghue FJ, Pierce RJ, Singh P, Thornton AT. The new AASM criteria for scoring hypopneas: Impact on the apnea hypopnea index. Sleep 2009;32:150-7.
Manoochehri Z, Salari N, Rezaei M, Khazaie H, Manoochehri S, Pavah BK. Comparison of support vector machine based on genetic algorithm with logistic regression to diagnose obstructive sleep apnea. J Res Med Sci 2018;23:65.
] [Full text]
Berry RB, Gamaldo CE, Harding SM, Brooks R, Lloyd RM, Vaughn BV, et al.
AASM scoring manual version 2.2 updates: New chapters for scoring infant sleep staging and home sleep apnea testing. J Clin Sleep Med 2015;11:1253-4.
Han J, Kamber M, Pei J. Data Mining: Concepts and Techniques. 3rd
ed. Massachusetts, US: Morgan Kaufmann; 2012.
Kang HH, Kang JY, Ha JH, Lee J, Kim SK, Moon HS, et al.
The associations between anthropometric indices and obstructive sleep apnea in a Korean population. PLoS One 2014;9:e114463.
Pinto JA, Godoy LB, Marquis VW, Sonego TB, Leal Cde F, Artico MS. Anthropometric data as predictors of obstructive sleep apnea severity. Braz J Otorhinolaryngol 2011;77:516-21.
Yildirim Y, Yilmaz S, Güven M, Kılınç F, Kara AV, Yilmaz Z, et al.
Evaluation of anthropometric and metabolic parameters in obstructive sleep apnea. Pulm Med 2015;2015:189761.
Subramanian S, Jayaraman G, Majid H, Aguilar R, Surani S. Influence of gender and anthropometric measures on severity of obstructive sleep apnea. Sleep Breath 2012;16:1091-5.
Lim YH, Choi J, Kim KR, Shin J, Hwang KG, Ryu S, et al.
Sex-specific characteristics of anthropometry in patients with obstructive sleep apnea: Neck circumference and waist-hip ratio. Ann Otol Rhinol Laryngol 2014;123:517-23.
Peppard PE, Young T, Barnet JH, Palta M, Hagen EW, Hla KM. Increased prevalence of sleep-disordered breathing in adults. Am J Epidemiol 2013;177:1006-14.
Nashi N, Kang S, Barkdull GC, Lucas J, Davidson TM. Lingual fat at autopsy. Laryngoscope 2007;117:1467-73.
Millman RP, Carlisle CC, McGarvey ST, Eveloff SE, Levinson PD. Body fat distribution and sleep apnea severity in women. Chest 1995;107:362-6.
Whittle AT, Marshall I, Mortimore IL, Wraith PK, Sellar RJ, Douglas NJ. Neck soft tissue and fat distribution: Comparison between normal men and women by magnetic resonance imaging. Thorax 1999;54:323-8.
Cizza G, de Jonge L, Piaggi P, Mattingly M, Zhao X, Lucassen E, et al.
Neck circumference is a predictor of metabolic syndrome and obstructive sleep apnea in short-sleeping obese men and women. Metab Syndr Relat Disord 2014;12:231-41.
Soylu AC, Levent E, Sarıman N, Yurtlu S, Alparslan S, Saygı A. Obstructive sleep apnea syndrome and anthropometric obesity indexes. Sleep Breath 2012;16:1151-8.
Zhou JY, Ge H, Zhu MF, Wang LJ, Chen L, Tan YZ, et al.
Neck circumference as an independent predictive contributor to cardio-metabolic syndrome. Cardiovasc Diabetol 2013;12:76.
Martins AB, Tufik S, Moura SM. Physiopathology of obstructive sleep apnea-hypopnea syndrome. J Bras Pneumol 2007;33:93-100.
Davidson TM, Patel MR. Waist circumference and sleep disordered breathing. Laryngoscopy 2008;118:339-47.
[Table 1], [Table 2], [Table 3], [Table 4]