| Objective:We aim to explore a convenience new model of 24-hour urinary sodium and potassium excretion by spot urine suitable for the stroke patients and compare the accuracy of the new model and the three developed formulas to facilitate the standardized management and treatment of stroke patients in China.Methods:The study participants with cardiovascular and cerebrovascular diseases were from Gaoping and Changzhi cities of Shanxi province,China.In detail,6 villages from each city and 20 subjects from each village were randomly selected for subject enrollment for questionnaire survey,physical examination,spot urine samples and 24-hour urine samples collection.After 5 trials,970 spot urine samples and 970 24-hour urine samples were assessed in the study,and the samples were formed according to the 7:3 ratio,which is the requirement of the sample size of form the new model and the sample size of the verification.70%of spot urine samples and 24-hour urine samples were randomly divided into the training cohort to develop the new model and 30%samples were randomly divided into the testing cohort for validation and comparison of the new model with the developed Kawasaki formula The INTERSALT formula and the Tanaka formula comparing the measured values of 24-hour urine sodium and potassium.A sex-specific equation for 24UNa V and 24UKV was established using the training data by stepwise multivariate linear regression.The equation for estimating 24UNa V in-cluded age,body mass index（BMI）,spot urinary sodium concentration（spot Na）,and spot urinary sodium-to-spot creatinine ratio（spot Na/spot Cr）.The equation for estimating24UKV included age,BMI,spot urinary potassium concentration（spot K）,and the natural log（ln）-transformed Ln[spot urinary potassium-to-spot creatinine ratio（spot K/spot Cr）]and Ln（spot Na/spot Cr）.We calculated the means and standard deviations of the measured 24UNa V and24UKV and the estimated 24UNa V and 24UKV using published formulas and the newly developed formula.The measured and estimated 24UNa V and 24UKV were compared using the paired samples T-test.Meanwhile,the mean bias,the Pearson correlation coefficient（r）,and the intraclass correlation coefficient（ICC）of the 24UNa V and 24UKV,between measured and estimated values.We calculated P30（the proportion of the estimated value within 30%of the measured sodium excretion）and Bland–Altman plots to verify and compare the accuracy of the new model with the developed formula for estimating the values of 24UNa V and 24UKV.Results:1.Demographic characteristicsAcross the follow-up period,A total of 1287 participants were initially included.Excluding unqualified urine samples,the data of 970（75.4%）stroke patients participated in the development and verification of the new model for estimating 24UNa V and 24UKV by spot urine samples.Among them,urine sample data from 689 participants were used for model development and urine sample data from 281 participants were used for validation of the new model and 3 different formulas.Among the training cohort for model development,there were 389 men（56.46%）and300 women（43.54%）,with an average age of 64.43±7.92 years.Among the testing cohort for model validation and comparison of the new model and the developed formulas with measured values,142（50.53%）were men and 1,39（49.47%）were women The average age was 64.76±75 years.There were no statistically significant differences in age,sex,education level,BMI,Spot Urine Specimen,and 24-hour urine electrolyte excretion levels（P>0.05）.2.Urine sample collectionA total of 2896 urine samples were collected,including spot urine samples,1606 and24-hour urine samples,1290.233 spot urine samples without they pair urine（24-hour urine samples）were excluded;403 unqualified spot urine samples and 320 unqualified 24-hour urine samples were excluded.1940 pair urine samples（spot urine and 24-hour urine samples）were finally included in the study and participated in the construction and validation of the new model.3.Development of new modelThe correlation coefficient of 24UNa V was 0.42 for males and 0.33 for females.The correlation coefficient of 24UKV was 0.71 for both males and females.The obtained new model were:（1）Y Estimated 24UNa V（mmol/day）=-0.191×Age+4.349×BMI+0.229×Spot Na+1.744×Spot Na/Spot Cr+41.492（for male）;-1.030×Age+2.011×BMI+0.143×Spot Na+1.035×Spot Na/Spot Cr+147.159（for female）;and（2）Y Estimated 24UKV（mmol/day）=-0.052×Age+0.410×BMI+0.031×Spot K+33.280×Ln(Spot K/spot Cr)-5.789×Ln(Spot Na/spot Cr)-1.035（for male）;-0.235×Age+0.530×BMI+0.040×Spot K+30.990×Ln（spot K/spot Cr）-7.837×Ln(Spot Na/Spot Cr)+4.318（for female）.3.Validation of the new modelThe comparison between the three previously published equations and the new model for estimating 24UNa V and 24UKV in stroke patients showed that these published formulas might be inferior to the new model.Compared with the measured values of 24UNa V and 24UKV,the mean bias of the new model was the lowest（24-hour urine sodium excretion:5.17 mmol/day;24-hour urine potassium excretion:0.85 mmol/day）,the difference was not statistically significant（P>0.05）,and the different ratio of measured and estimated values of sodium and potassium excretion Within 30%(P30)of the new model has the highest estimated proportion（24UNa V:59.43%;24UKV:70.11%）.For the excretion of 24UNa V,the average deviations of the developed 3 formulas included Kawasaki formula INTERSALT formula Tanaka formula are respectively 38.02mmol/day,37.80 mmol/Day and11.3mmol/day;For the excretion of 24UKV,The average deviations of Kawasaki Formula and Tanaka formula are respectively 13.66 mmol/day and2.21mmol/day and Compared with the measured values of 24UNa V and 24UKV,the differences between the estimates of the new model were statistically significant（P<0.05）.The result of Pearson correlation coefficient（r）and ICC show that these published formulas might be inferior to the new model.The Pearson correlation coefficient and ICC between the estimate values of the new model and measured values of 24UNa V and24UKV are the largest,respectively 0.45,0.71 and 0.35、0.69.The Pearson correlation coefficient and ICC between the estimate values of the developed formulas and the measured values of 24UNa V respectively 0.37,0.40,0.35 and 0.32,0.30,0.31.The Pearson correlation coefficient and ICC between the estimate values of the Kawasaki formula,the Tanaka formula and the measured values of 24UKV respectively 0.64,0.62 and 0.46,0.49.There was a strong correlation between measured 24UNa V and 24UKV,the estimate values by the new model and the values estimating by the developed three formulas（P<0.05）.Through the Bland–Altman plot,we can found that all measured values on the left side of the abscissa of the Bland–Altman plot is overestimated,the value on the right is underestimated,there is a certain systematic deviation,the linear trend is obvious,and the result suggests:There was a tendency to underestimate by the different verification formulas when the urinary sodium and potassium excretion was high and overestimate by the different verification formulas when the urinary sodium and potassium excretion was at the low level.Comparing 24UNa V and 24UKV estimated by the new model with the measured values,most of the individual estimates of the Bland–Altman plot were distributed in a narrower interval of±1.96 standard deviations.The effect is relatively good.The Bland–Altman plot of 24-hour urine sodium estimation shows that the linear trend of the new model is more pronounced in people with high dietary sodium intake,and underestimation trends are more likely.Conclusion:The study explores the new model for estimating diary sodium and potassium intake suitable for stroke patients in China to provide convenient means for the clinical treatment and daily management of stroke patients. |
PDF Full Text Request