Consideration of body mass index (BMI) in the association between hand grip strength and hypertension: Korean Longitudinal Study of Ageing (KLoSA)

Objective The purpose of this study was to investigate the association between grip strength and hypertension in the Korean population aged 65 years or older. Furthermore, individual differences in BMI were taken into account to examine whether grip strength or a relative grip strength predicted hypertension better. Methods Data from the Korean Longitudinal Study of Ageing from 2006 to 2016 were assessed, and a total of 3,383 participants were analyzed in our study (Male: 1,527, Female: 1,856). Using the generalized estimating equation model, the association between grip strength and hypertension, assessed by the response to the question ‘have you ever been diagnosed with hypertension from your doctor?’, over the follow-up period was analyzed. The relative grip strength, calculated by dividing the mean grip strength by BMI, was also analyzed in association of hypertension. Results Both grip strength and relative grip strength were significantly associated with hypertension in our sample. However, the results were more significant in the total sample when relative grip strength was used. In terms of grip strength, as the High group as reference: Low (Odds Ratio (OR): 1.238, 95% Confidence Interval (CI): 1.096, 1.397), Middle Low (OR: 1.104, 95% CI: 0.990, 1.231), and Middle high (OR: 1.024, 95% CI: 0.934, 1.122). In the analysis using relative grip strength, as High group as reference: Low (OR: 1.393, 95% CI: 1.234, 1.573), Middle low (OR: 1.232, 95% CI: 1.104, 1.374), and Middle high (OR:1.104, 95% CI: 1.009, 1.209). Furthermore, the lower QIC measure in the model with relative grip strength (QIC: 25,251) compared with the one using grip strength (QIC: 25,266) indicated a better model fit in the former. Conclusions The results of the current study strengthen the previous findings in regards to hand grip strength and health. Furthermore, the results of our study shines light on the necessity of considering individual differences in BMI, when using a physical measure as a study variable.

Introduction purpose of preparing for the aged society in terms of system reform and policy decision. The data is composed of 7 categories such as population, family, health, employment, income, wealth, subjective expectation and life expectation. This biennial survey involves multistage stratified sampling based on geographical areas and housing types across Korea. Participants were selected randomly using a multistage, stratified probability sampling design to create a nationally representative sample of community-dwelling Koreans 45 years of age and older. Participant selection was performed by the Korea Labor Institute for these rapidly growing populations, including individuals from both urban and rural areas. In case of refusal to participate, another subject was selected from an additional, similar sample from the same district. From the original 10,254 participants, those aged 65 years and older were included in our analysis. In our final analysis, 3,383 participants (Male: 1,527, Female: 1,856) without missing values on the variables of interest (e.g., grip strength, BMI, heart disease, Mini Mental State Examination (MMSE)), were included. Out of the public data in Korea, KLoSA was considered as the most suitable data for the analysis involved in the current study.

Dependent variables
The presence of circulatory related diseases. The dependent variable was the diagnosis of hypertension by a doctor. The participants were asked the following question, 'Have you ever been diagnosed with hypertension from your doctor?' The response to the question was dichotomized as either 'yes' or 'no'. This response was collected from all waves so that the possible changes in the presence of hypertension can be accounted for in the final analysis.

Independent variables
Grip strength and relative grip strength. The independent variable was grip strength. Grip strength was measured by a handgrip dynamometer (Model number: NO6103, Manufacturer: TANITA, Japan). The test was performed in a sitting position with the elbow flexed at 90˚on both the right and the left sides. The mean strength was calculated from grip strengths on both sides [15]. Grip strength in each year was divided into four groups: Q1, Q2, Q3, Q4 using SAS Rank function. For relative grip strength, grip strength was divided by BMI, which was calculated from the reported height and weight (kg/m 2 ). Grip strength at all waves were used in the final analysis to account for the possible changes in the strength of participants.
Control variables. This study used educational level (elementary school or less, middle school, high school, and college or more), gender (male or female), age (65-69 years old, 70-74, and 85 years or more)residential region (metropolitan (e.g. Seoul), urban (e.g. Daejeon, Daegu, Busan, Incheon, Kwangju, or Ulsan), or rural (not classified as administrative of a city), national health insurance (health insurance, medical aid), Mini Mental State Examination (MMSE) (dementia (0-17), cognitive decline (18)(19)(20)(21)(22)(23), normal (24-30)), smoking status (smoker, former smoker, never), alcohol use (never, former drinker, drinker), labor (yes or no), BMI (thin (0-18.4), moderate (18.5-22.9), overweight (23-24.9), obese (>25)), heart disease (yes or no), Year (2006Year ( , 2008Year ( , 2010Year ( , 2012Year ( , 2014Year ( , 2016 as covariates. Analytical approach and statistics. Chi-square test, and generalized estimating equation (GEE) regression model with a binary distribution which controls for characteristics that change over time, such as confounding variables, were used to investigate the association between degree of grip strength and hypertension. The GEE model is a useful analytical tool for longitudinal studies, because it offers a way to handle unbalanced and missing data. For example, the GEE is able to control for the change in the presence of hypertension over time. In GEE, proc genmod was used, with link logit, distribution normal. For all analyses, SAS statistical software package, version 9.4 (SAS Institute, Inc., Cary, NC, USA) was used. All statistical tests were two-tailed, with the null hypothesis of no difference being rejected if p < 0.05. Table 1 shows the general characteristics of the participants. The participants had a mean age of 73.354 (Standard Deviation (SD): 6.217), mean grip strength of 21.934 (SD: 7.768), mean BMI of 23.114 (SD: 6.779), and mean relative grip strength of 0.968 (SD: 0.356). Of the 3,383 participants selected for the study, 1,335 (39.5%) people had hypertension, and both grip strength and relative grip strength showed a significant chi-square value (p-value: < .001). In terms of grip strength, people with hypertension in the Low group, ranging between 1.250 and 16.475, were 391 (45.1%), Middle low, ranging between 16.50 and 21.00, 382 (41.7%), Middle high, ranging between 21.025 and 28.225, 280 (36.9%) and High, ranging between 28.250 and 84.300, 282 (33.6%). In terms of relative grip strength, people in the Low group were 419 (47.5%), Middle low 376 (45.1%), Middle high 305 (35.8%) and 235 (28.8%) in the High relative grip strength group. More female participants (N: 808) than male participants (N: 527) experienced hypertension, and there were significant differences in hypertension (p-value: < .001) between age groups. In terms of other control variables, national health insurance status (p-value: .015), smoking status (p-value: < .001), labor status (p-value: < .001), and the presence of heart disease (p-value: < .001) differed significantly in terms of hypertension, but the rest did not.

Adjusted association between grip strength and hypertension
The results of the fully adjusted model are shown in Table 2. In the total sample, the association between grip strength and hypertension was only statistically significant in the Low group (OR: 1.238, 95% CI: 1.096, 1.397) as the High group as reference. However, the results showed slightly different trends when gender was taken into account. The results did not differ much in the male sample, but in terms of the female sample, the associations between grip strength and hypertension were statistically significant in all grip strength groups: Low (OR: 1.684, 95% CI: 1.252, 2.265), Middle low (OR: 1.584, 95% CI: 1.180, 2.126), and Middle high (OR 1.482, 95% CI: 1.099, 1.999) with High as reference. Furthermore, there was a gradient increase in the risk of hypertension in accordance with weaker grip strength, providing indication to the importance of taking gender into account.

Adjusted association between relative grip strength and hypertension
Considering relative grip strength in association with hypertension (Table 3), the results were as follows: Low (OR: 1.393, 95% CI: 1.234, 1.573), Middle low (OR: 1.232, 95% CI: 1.104, 1.374), and Middle High (OR: 1.104, 95% CI: 1.009, 1.209) with High as reference. When relative grip strength was used, the OR for all categories of grip strength were statistically significant and this was also true in the case of males. Contrarily, relative grip strength was only statistically significantly associated with hypertension for females in the Low group (OR: 1.356, 95%: 1.356, 1.777). Furthermore, the Quasi-information criterion (QIC) measures for the total sample as well as gender stratified samples using relative grip strength were lower than that using grip strength, indicating a better model fit with the use of relative grip strength as a measure of physical fitness. We compared the difference between the relative grip strength and grip strength further by creating the receiver operating characteristic (ROC) curves for both. The bigger area under the curve (AUC) in the case of using the relative grip strength in association with hypertension provides additional support for considering BMI in the association between HGS and hypertension.

Discussion
In this study, we were able to utilize a nationally representative data of Korea to find better ways to predict hypertension in the elderly population. Similar to previous studies [13,16], the results from our study indicated a significant association between HGS and hypertension. Moreover, considering previous studies [17,18], which presented significant association between relative grip strength and biomarkers (e.g., blood pressure), we included BMI in this association and performed a separate analysis using a relative grip strength. The results of our study presented a more significant association between relative grip strength and hypertension compared with grip strength and hypertension. Also, using relative grip strength in the analysis improved the model fit of the analysis compared with when grip strength was used.
Although studies considering BMI in the association between HGS and hypertension are scarce, studies in this field of research have provided evidence to the importance of considering body composition [19,20]. Therefore, it is plausible to use the relative grip strength in future studies, and interpret the results accordingly.
In the past, Gale et al. [21] included BMI and HGS as separate variables in association with cardiovascular mortality, but contrary to our study, BMI remained significant in women, and not in men. Given the fact that the outcomes of the two studies as well as the method, in which, BMI was used as a variable differed, it is worth noting the differing effect of BMI between genders in similar circumstances. Also, the QIC measure provided in our study strengthens the plausibility of our model including BMI. Furthermore, Dong et al. [22] found that higher HGS was associated with low blood pressure, but an inverse association when BMI was included in the analysis. The discrepancy between their study findings and ours may be due to the age difference in sample groups. Whereas the participants in our study were older adults 65 years and above, the participants in Dong et al.'s study were teenagers. Accordingly, comparing varying age groups in regards to the association between BMI adjusted HGS and hypertension could provide valuable information for better predictive ability.
There are a few possible mechanisms relating HGS to hypertension. First, decreased physical activity has shown to be associated with hypertension [23], and that people with poor

PLOS ONE
Consideration of body mass index (BMI) in the association between hand grip strength and hypertension strength reported more difficulty exercising [24]. Because HGS is a widely recommended measure of muscle strength [25], it is possible to consider that people with low HGS are less physically active, consequently increasing their risk of hypertension. Therefore, HGS could be used as a useful predictive indicator of hypertension in a clinical setting, as well as a reason to prescribe more exercise to those who exhibit low HGS. Another mechanism that could associate HGS to hypertension is the arterial structure. A previous study reported an improved structure in the brachial artery due to isometric handgrip exercise [26]. Accordingly, lower HGS could be associated with poor status of the brachial artery, which in term could lead to hypertension. Subsequently, a type of exercise training could be recommended to those with hypertension, as well as those at-risk of obtaining hypertension. The findings of this study provide important implications for the elderly population of Korea, as well as present various strengths. First, the data used in this study were based on a nationally representative sample of 65 years and older. Therefore, the generalizability of the study is viable. Also, the residential regions of the study subjects were controlled for in our analysis in order to reduce the possibly biased results from over sampling in a certain region of Korea. Second, this was a prospective cohort data with 10 years of follow-up and a good follow-up rate (78.8%). This was important for our analysis, because we only included a specific age group of the entire KLoSA sample and the ones with no missing values for the variables of interest. Acquiring a good number of final sample enabled us to reach a good statistical power to show valuable findings. Furthermore, the longitudinal design of our analysis enabled us to infer a causal relationship between HGS and hypertension.
Despite the strengths of our study, the limitations need to be considered as well. Even though the results are generalizable to the Korean elderly population, it is unable to represent more specific populations, such as people who are hospitalized. Therefore, research in this line of work should continue considering more variety of population. Also, regardless of the good follow-up rate, there were people who dropped out of the study, and could have biased our results. Furthermore, a possibility of selection bias due to differences in characteristics between included and excluded participants, and a misclassification bias due participants falsely reporting the presence of hypertension, need to be taken into account. Lastly, the height and the  weight variables were based on reported values instead of physical measurements, possibly reducing the accuracy of the BMI value used in our study. In the future, considering more various age groups will provide additional information regarding HGS and hypertension. Also, researchers and clinicians should try to develop interventions to prevent hypertension, as well as ameliorate the severity of those suffering from it.