Study on blood pressure rhythm in hypertensive patients withdeficiency syndrome and a random forest model for predicting hypertension withdeficiency syndrome.

Objective: To investigate blood pressure rhythm (BPR) in deficiency syndrome of hypertension (YDSH) patients and develop a random forest model for predicting YDSH.

Methods: Our study was consistent with technical processes and specification for developing guidelines of Evidence-based Chinese medicine clinical practice (T/CACM 1032-2017). We enrolled 234 patients who had been diagnosed with primary hypertension without antihypertensive medications prior to the enrollment. All participants were divided into deficiency group (YX, 74) and non- deficiency group (NYX, 160). Participants were professionally grouped by three experienced chief Traditional Chinese Medicine (TCM) physicians according to four examinations (i.e., inspection, listening and smelling, inquiry and palpation). We collected data on 24 h ambulatory blood pressure monitoring (ABPM) and YDSH rating scale. We divided 24 h of a day into 12 two-hour periods [Chen-Shi (7:00-9:00), Si-Shi (9:00-11:00), Wu-Shi (11:00-13:00), Wei-Shi (13:00-15:00), Shen-Shi (15:00-17:00), You-Shi (17:00-19:00), Xu-Shi (19:00-21:00), Hai-Shi (21:00-23:00), Zi-Shi (23:00-1:00), Chou-Shi (1:00-3:00), Yin-Shi (3:00-5:00), Mao-Shi (5:00-7:00)] according to the theory of "midnight-midday ebb flow". We used random forest to build the diagnostic model of YDSH, with whether it was deficiency syndrome as the outcome.

Results: Compared with NYX group, YX group had more female participants with older age, lower waist circumference, body mass index (BMI), diastolic blood pressure (DBP), and smoking and drinking rate (all ＜ 0.05). The YDSH rating scores of YX group [28.5 (21.0-36.0)] were significantly higher than NYX group [13.0 (8.0-22.0)] ( < 0.001), and the typical symptoms of YX group included vexing heat in the chest, palms and soles, dizziness, dry eyes, string-like and fine pulse, soreness and weakness of lumbus and knees, palpitations, reddened cheeks, and tinnitus (all < 0.05). The ratio of non-dipper hypertension in YX group was higher than in NYX group (56.9% 44.4%, = 0.004). Compared with NYX group, 24 h DBP standard deviation (SD), nighttime DBP SD, Si-Shi DBP, Si-Shi mean arterial pressure (MAP), Hi-Shi systolic blood pressure (SBP), Hi-Shi DBP, Hi-Shi MAP, Zi-Shi SBP, Zi-Shi DBP, Zi-Shi MAP, Chou-Shi SBP SD, Chou-Shi DBP SD, Chou-Shi SBP coefficient of variation (CV) were lower in YX group (all ＜ 0.05). Binary Logistic Regression analysis showed that the diagnosis of YDSH was positively correlated with age, heart rate, YDSH rating scores, and four TCM symptoms including vexing heat in the chest, palms and soles, string-like and fine pulse, soreness and weakness of lumbus and knees, and reddened cheeks (all ＜ 0.05), but was negatively correlated with smoking (﹥0.05). In addition, the diagnosis of YDSH was positively correlated with daytime SBP SD, nighttime SBP SD, nighttime SBP CV, and Hi-Shi SBP CV, but was negatively correlated with 24 h SBP CV, daytime DBP SD, nighttime DBP SD, and Hi-Shi DBP (all ＜ 0.05). Hi-Shi SBP CV had independent and positive correlation with the diagnosis of YDSH after adjusting the variables of age, gender, course of hypertension, BMI, waist circumference, SBP, DBP, heart rate, smoking and drinking ( = 0.029). Diagnostic model of YDSH was established and verified based on the random forest. The results showed that the calculation accuracy, specificity and sensitivity were 77.3%, 77.8% and 76.9%, respectively.

Conclusion: The BPR was significantly attenuated in YDSH patients, including lower 24 h DBP SD and nighttime DBP SD, and Hi-Shi SBP CV is independently correlated with the diagnosis of YDSH. The prediction accuracy of diagnosis model of YDSH based on the random forest was good, which could be valuable for clinicians to differentiate YDSH and non- deficiency patients for more effective hypertensive treatment of TCM.