A risk prediction nomogram for resistant hypertension in patients with obstructive sleep apnea

Patients with obstructive sleep apnea (OSA) are liable to have resistant hypertension (RH) associated with unfavorable cardiovascular events. It is of necessity to predict OSA patients who are susceptible to resistant hypertension. Hence, we conducted a retrospective study based on the clinical records of OSA patients admitted to Yixing Hospital Affiliated to Jiangsu University from January 2018 to December 2022. According to different time periods, patients diagnosed between January 2018 and December 2021 were included in the training set (n = 539) for modeling, and those diagnosed between January 2022 and December 2022 were enrolled into the validation set (n = 259) for further assessment. The incidence of RH in the training set and external validation set was comparable (P = 0.396). The related clinical data of patients enrolled were collected and analyzed through univariate analysis and least absolute shrinkage and selection operator (LASSO) logistic regression analysis to identify independent risk factors and construct a nomogram. Finally, five variables were confirmed as independent risk factors for OSA patients with RH, including smoking, heart disease, neck circumference, AHI and T90. The nomogram established on the basis of variables above was shown to have good discrimination and calibration in both the training set and validation set. Decision curve analysis indicated that the nomogram was useful for a majority of OSA patients. Therefore, our nomogram might be useful to identify OSA patients at high risk of developing RH and facilitate the individualized management of OSA patients in clinical practice.


Risk factors
Baseline information of the enrolled patients was recorded, including gender, age, body mass index (BMI), neck circumference, waist circumference, smoking, alcohol drinking, family history of hypertension, and comorbidities composed of RH, diabetes and heart disease (coronary heart disease, atrial fibrillation and myocardial infraction).The Epworth sleepiness scale (ESS) was also filled to assess daytime sleepiness.All patients included in this study underwent overnight PSG monitoring using SOMNOlab 2 model machine (Wanman, Germany), and the collected indicators included total sleep time, apnea-hypopnea index (AHI), oxygen desaturation index (ODI), percentage of total time with oxygen saturation level < 90% (T90), lowest arterial oxygen saturation (LSaO2) and mean arterial oxygen saturation (MSaO2).In addition, fasting venous blood (FBG) of all subjects was collected within 24 h before discharge using SIEMENS ADVIA 1200 automatic biochemical analyzer (SIEMENS, Germany) to obtain data on total cholesterol, triglyceride, high density lipoprotein and low density lipoprotein.These variables were considered as potential risk factors for further selection of predictors.

Statistical analysis
Statistical analysis was conducted using IBM SPSS Statistics (version 27.0.1,Chicago, USA) and R software 4.2.2 (R foundation for Statistical Computing, Vienna, Austria).Data were described as mean (standard deviation [SD]) or median (interquartile range [IQR]) for continuous variables and number (percentage) for categorical variables as appropriate.In order to compare the differences in baseline clinical characteristics between the training set and the validation set, and between patients with RH and without RH in training set, bivariate analysis was performed by chi-squared test or Fisher's exact test for categorical variables, t test for normal distribution continuous variables, and Mann-Whitney U test for non-normal distribution variables.The variables with P < 0.05 in the univariate analysis of the training set were substituted into LASSO analysis for further variable screening.According to the results of LASSO analysis, multivariate logistic regression analysis was implemented to further clarify independent risk factors of RH in OSA patients.The results were reported as odds ratios (ORs) and 95% confidence intervals (95%CIs).We then constructed a nomogram based on independent risk factors to predict the risk of developing RH in patients with OSA using package 'rms' in R. Patients in the external validation set were used for assessing the discrimination and calibration of the predictive nomogram.The discriminative performance was evaluated using the area under receiver operating characteristic (ROC) curve (AUC), known as the c-statistic.Calibration of the model was assessed by comparison of the predicted and observed probability of RH in patients with OSA.The fitting degree of the scoring model was measured by Hosmer-Lemeshow goodness of fit test.Decision curve analysis was performed to evaluate the clinical utility of the nomogram, using the package 'dcurves' in R. Unless otherwise stated, a two-tailed P value < 0.05 was considered statistically significant.All methods were carried out in accordance with relevant guidelines and regulations.

Nomogram construction
According to the results of initial screening using univariate analysis, fourteen variables had significant differences between OSA patients with RH and OSA patients without RH in the training set, including male gender, BMI, smoking, family history of hypertension, comorbidities of heart disease and diabetes, neck circumference, waist circumference, ESS, AHI, T90, ODI, MSaO2 and FBG (Table 2).These variables were further analyzed through LASSO regression analysis, and six variables were selected for multivariate logistic regression analysis (Fig. 2a, b).In the multivariate logistic regression analysis, smoking, heart disease, neck circumference, AHI, T90 were demonstrated as significant independent risk factors for RH in OSA patients.The detailed OR values and 95%CI in the multivariate analysis were listed in Table S1.We developed a risk prediction model based on the above five parameters through logistic regression analysis (Table 3).In order to facilitate the assessment of the individual risk of RH in OSA patients, a nomogram was constructed according to the contribution weight of each factor in the risk model (Fig. 3).In the nomogram, the values of each factor have corresponding points on the scale axis.The total score for each patient can be easily calculated by adding up single points of each factor.By projecting the total score to the bottom of the total subscale, the risk of developing RH in each patient with OSA could be estimated.

Nomogram assessment
The c-statistic was 0.806 (95%CI: 0.761-0.851) in the training set (Fig. 4a) and 0.800 (95%CI: 0.732-0.867) in the validation set (Fig. 4b).The calibration curves demonstrated that the probability predicted by the nomogram was relatively consistent with the actual observed probability in the training set (Fig. 5a) and validation set (Fig. 5b).The P values of Hosmer-Lemeshow test in the training set and validation set were both more than 0.05, which were 0.618 and 0.525, respectively, indicating a favorable calibration of the risk prediction model.In conclusion, the nomogram for RH had a decent discriminative and calibrating performance in both the training set and validation set.
Decision curve analysis was conducted to evaluate the practicality and clinical application value of the nomogram, which showed that the majority of the threshold probabilities had great net benefit in both the training set (Fig. 6a) and validation set (Fig. 6b).In the decision curve analysis, the nomogram achieved a positive net benefit www.nature.com/scientificreports/ in the training set and validation set when predicted risk thresholds was between 19% and 81%, suggesting that the nomogram established could be alternative tool for clinical practice.

Discussion
Nomogram is a fast and cost-effective method that is easy to implement in clinical practice.This retrospective study is the first study to develop and validated a risk prediction nomogram to identify OSA patients at high risk of RH.The nomogram performed a good discrimination and calibration in the training set, which was confirmed in the validation set.Decision curve analysis supported that the nomogram could provide an ideal reference value for clinical practice.Our nomogram consisted of five clinical characteristics, including smoking, heart disease, neck circumference, AHI and T90.In the nomogram, the individual risk score of RH could be calculated manually by collecting related medical history and routine parameters tested in the laboratory.OSA is the most common comorbidity and independent risk factor of RH.Epidemiological data reported that the prevalence of OSA in patients with RH was as high as 70-83% 21,22 .Hou et al. 23 found a significant association between OSA and RH with a pooled OR value of 2.842 (95%CI: 1.703-3.980),indicating that the risk of RH in OSA patients was 1.842 times higher than that in non-OSA individuals.However, the prevalence of RH in OSA patients has not been directly investigated.The prevalence of RH was 14.7% in 3.2 million patients with hypertension 24 .In the present study, the prevalence of RH in patients with OSA was as high as 21.5% in the training set and 18.9% in the validation set.The reason for the relatively high prevalence of RH may be that all the subjects included in this study were OSA patients, resulting in a higher proportion of RH in the enrolled population than in the general population.
In comparison with patients with controlled blood pressure, associated risk factors of RH are older age, obesity, diabetes, and silent target organ damage (left ventricular hypertrophy or chronic kidney disease) 25,26 .However, there are few studies on the risk factors of RH in patients with OSA.In the present study, the LASSOmultivariate logistic regression analysis demonstrated that smoking, heart disease, neck circumference, AHI and T90 were independent risk factors of RH in OSA patients.Ahmed et al. 22 found that smoking and obesity were associated with increased risk for resistant hypertension in OSA patients.Smoking plays an important role in increasing blood pressure via activating sympathetic nerve system and interfering with the effects induced by certain antihypertensive drugs 27 .Neck circumference is commonly used to measure the degree of obesity in www.nature.com/scientificreports/individuals, which has been demonstrated as independent risk factor for OSA with hypertension 28 .Increased neck circumference means raised fact accumulation around the upper respiratory tract, contributing to upper respiratory tract stenosis and intermittent hypoxia, subsequently activating the renin-angiotensin aldosterone system, thereby leading to water and sodium retention and making blood pressure difficult to control 29,30 .In this study, heart diseases consisted of coronary heart disease, atrial fibrillation and myocardial infraction.OSA is quite common in patients with heart disease and interact with heart disease to lead to unfavorable cardiovascular events 31 .Zhu et al. 32 determined that the prevalence of resistant hypertension was relatively high in patients with heart disease.AHI and T90 are important indicators for evaluating intermittent hypoxia in OSA patients.Shi et al. 33 investigated that T90 was a risk factor for hypertension related with OSA.Mechanically, intermittent hypoxia in patients with OSA can also cause excessive activation of the renin-angiotensin aldosterone system and vascular endothelial dysfunction, resulting in elevated blood pressure without control and a significant increase in the risk of RH 34 .
The consensus on resistant hypertension has identified the causal relationship between OSA and RH as relevant 35 .OSA is characterized by brain microarousal, intrathoracic pressure changes and intermittent hypoxia, which in turn triggers adverse biological processes such as oxidative stress, sympathetic activation, inflammation, hypercoagulability, endothelial dysfunction and metabolic disorders, making OSA patients susceptible to hypertension 36 .Prabhakar et al. 34 found that intermittent hypoxia increased the levels of reactive oxygen species (ROS) by up-regulating hypoxia-inducible factor (HIF) -1α and reducing HIF-2α protein levels, which activated chemoreflex and inhibiting baroreflex, thereby resulting in hypertension-related sympathetic activation.In addition, patients with resistant hypertension have high aldosterone levels, and fluid retention associated with elevated aldosterone aggravates OSA by promoting fluid movement to the surrounding tissues, especially the neck, leading to secondary pharyngeal edema and an increased tendency of airway collapse during sleep 37 .
Hypertension is one of the most important risk factors for cardiovascular disease related to death, especially stroke, ischemic heart disease, congestive heart failure, aortic aneurysm and peripheral arterial disease 38,39 .RH patients suffer a higher risk of end-organ damage contributing to death in comparison to those with controlled blood pressure, including carotid stenosis, retinopathy, left ventricular hypertrophy and heart failure, myocardial infraction, stroke, impaired renal function 40,41 .In addition to its clinical importance, RH also bears a considerable public health, economic and social burden due to the cost of treatment and related disability and premature death.Hence, early detection of RH in OSA patients and timely intervention are of great significance to avoid adverse consequence.For the issue how to treat with OSA patients at high risk of developing RH, several studies suggested that continuous positive airway pressure (CPAP) treatment which is the gold standard treatment for OSA played an important role in lowering blood pressure [42][43][44][45][46] .Labarca et al. 7 conducted a systematic review and meta-analysis and proved that CPAP benefited patients with RH and OSA by reducing blood pressure, especially nocturnal blood pressure.In a randomized controlled trail, CPAP treatment for 12 weeks reduced 24-h mean blood pressure and diastolic blood pressure and improved nocturnal blood pressure pattern in OSA patients with RH 43 .Bocoum et al. 45 indicated that long-term CPAP for more than 4 h/night could reduce the all-cause mortality of RH patients.Overall, CPAP treatment seems to be an effective supplement to antihypertensive drugs for OSA patients with RH.Further studies on whether CPAP treatment can help prevent the occurrence of RH in patients with OSA are vital.

Conclusion
Currently, most studies focus on the relationship between OSA and RH, but few studies pay attention to the risk factors of RH in OSA patients.In this study, we identified five independent risk factors for RH in patients with OSA and established a corresponding prediction nomogram for the first time.The nomogram is considered to be a reliable tool for predicting individual risk of RH in OSA patients, which could avail to early detection and precise intervention for susceptible patients.The follow-up multi-center prospective study are needed to validate the accuracy of the nomogram in clinical practice.

Figure 2 .
Figure 2. LASSO regression analysis.(a) LASSO coefficient profiles of the 14 variables.As the value of λ decreased, the later the coefficient was compressed, the more important the variable was.(b) Cross-validation results.The value between the two dashed lines is the range of positive and negative standard deviations of log(λ).The dashed line on the right reveals that when the fitting effect of the model is good, the number of variables contained is the least.Finally, six variables were selected when log(λ) = 0.045.

Figure 3 .
Figure 3.The nomogram of predicting the risk of RH for individual patient with OSA.RH, resistant hypertension; OSA, obstructive sleep apnea; HD, heart disease; NC, neck circumference; AHI, apnea-hypopnea index; T90, percentage of total time with oxygen saturation level < 90%.

Figure 4 .
Figure 4. Receiver operating characteristic (ROC) curve analysis.(a) ROC curve of the prediction nomogram in the training set.(b) ROC curve of the prediction nomogram in the validation set.AUC , area under the curve.

Figure 5 .
Figure 5. (a) Calibration curve of the nomogram in the training set.(b) Calibration curve of the nomogram in the validation set.

Figure 6 .
Figure 6.(a) Decision curve analysis of the nomogram in the training set.(b) Decision curve analysis of the model in the validation set.The red line (All) is the net benefit of RH treatment for all OSA patients.The green line (None) represents the net benefit of not providing RH treatment for OSA patients.The blue curve indicates the net benefit provided by the nomogram.

Table 2 .
Comparison of baseline characteristics between OSA patients with and without RH in the training set.RH, resistant hypertension; BMI, body mass index; ESS, Epworth sleepiness scale; AHI, apnea-hypopnea index; T90, percentage of total time with oxygen saturation level < 90%; ODI, oxygen desaturation index; LSaO2, lowest arterial oxygen saturation; MSaO2, mean arterial oxygen saturation; TC, total cholesterol; HDL, high density lipoprotein; LDL, low density lipoprotein; FBG, fasting blood glucose.

Table 3 .
Multivariate logistic regression for RH in the training set.AHI, apnea-hypopnea index; T90, percentage of total time with oxygen saturation level < 90%; SE, standard error; OR, odds ratio; CI, confidence interval.