Handgrip strength conditional tolerance regions suggest the need for personalized sarcopenia definition: an analysis of the American NHANES database

Handgrip strength (HGS) is a well-established clinical biomarker that assesses functional capacity in older populations. In addition, HGS is a diagnostic tool that forecasts aging health conditions, such as sarcopenia. This paper provides HGS statistical tolerance regions and presents the need to establish HGS reference values according to patients’ characteristics. For this purpose, we used a conditional tolerance algorithm for HGS, and we observed the tolerances regions in different age strata and sex of non-sarcopenic individuals from the National Health and Nutrition Examination Survey (NHANES, wave 2011–2012). Our results have critical implications for sarcopenia, since conventional and available HGS cut-offs do not consider age range. This paper offers new perspectives on the evolution of traditional definitions of sarcopenia according to the principles of precision medicine.


Introduction
In older adults, sarcopenia is one of the main factors leading to the loss of independence, with a hazard ratio of 1.6 in men and 1.7 in women [1]. In addition, sarcopenia increases the chance of premature mortality and affects older adults [2]. Because the older population is increasing in many regions of the world, it is projected that sarcopenia prevalence is going to rise from 10 to 27% in the next 3 decades [2]. Consequently, the economic burden related to sarcopenia will rise by at least 34%, primarily because of the hospitalizations of older adults [3].
Historically, sarcopenia was considered as a geriatric syndrome characterized by reduced muscle mass, and muscle strength was an elective variable for sarcopenia diagnosis. Although sarcopenia is currently classified as a disease, there is still a lack of unanimous definition in the scientific literature about it [4,5], and its identification requires measurements of muscle strength [majorly handgrip strength (HGS)] and physical performance (majorly gait speed). In terms of muscle strength, HGS has become mandatory in Marcos Matabuena and Pedro Pugliesi Abdalla have contributed equally to this study.
Although HGS has well-established reference values to set thresholds for low HGS across all the consensuses mentioned above and consortia, these reference values do not consider individual characteristics. Additionally, HGS reference values are exclusively for older adults and are not age-specific despite the strong association between HGS and age [9]. For instance, recent studies have reported sarcopenia in children with specific clinical conditions, regardless of the absence of children-HGS reference values. Therefore, considering specific reference regions for each sex and age group would help in identifying the natural limits of sarcopenia definitions in other age groups and not exclusively in older adults.
This study aimed, for the first time, to examine how the conditional bivariate tolerance regions for HGS and HGS/ weight vary according to sex and age in a population of nonsarcopenic adults. This will help to overcome the limitation of current HGS cut-off points to identify sarcopenia. In addition, the new reference value methods should provide greater clinical sensibility to incorporate the effect of two biomarkers (e.g., absolute HGS and HGS/weight) in an integrated and holistic way into the analysis.

Study design and participants
We used data from the National Health and Nutrition Examination Survey (NHANES) waves 2011-12 from participants who had accelerometry data [10]. The study design follows a multi-stage complex survey design. The NHANES provides a broad range of descriptive health and nutrition statistics for the U.S. civilian non-institutionalized population. Data collection consisted of an interview and an examination. The interview gathered person-level demographic, health, and nutrition information; the examination included physical measurements, such as blood pressure, accelerometer devices, dental examination, and the collection of blood and urine specimens for laboratory testing. A total of 3500 participants were considered in the analysis (inclusion criteria: age between 18 and 80 years, non-sarcopenic, and with highlevel accelerometry data disponible). Further details about NHANES 2011-12 can be found elsewhere.
Participants provided written informed consent, and study procedures were approved by the National Center for Health Statistics (NCHS) research ethics review board.

Handgrip strength (HGS)
The Takei digital dynamometer (model T.K.K.5401) measured the maximum HGS in kilograms (range of 5.0-100 kg). Calibration checks guarantee quality control procedures. The entire protocol of HGS can be found elsewhere. Briefly, participants squeeze as hard they can three times on each hand with 60 s of pause between the trials. Best values of the dominant hand (or higher, for ambidextrous), expressed in kilograms, were considered in the analysis.

Sarcopenia identification
We removed participants with sarcopenia from the sample. Sarcopenia was defined following SDOC recommendations with the presence of low muscle strength [for women: HGS < 20 kg, HGS/weight < 0.337 kg/kg, HGS/ BMI < 0.79 kg/kg/m2, and for men: HGS < 35 kg, HGS/ weight < 0.45 kg/kg, HGS/BMI < 1.05 kg/kg/m2], and low physical performance, derivate based on daily total activity time (TAC). Specifically, we determined low physical performance based on the TAC variable distribution in the sample. In addition, we defined that a patient presents with low physical performance if his/her TAC value is lower than the fifth percentile distribution in the sample analyzed.

Statistical procedures
We calculate the β-content tolerance regions to the HGS, and HGS/weight (95% CI), using the conditional β-content tolerance algorithm described in [11], conditioned to age (as a continuous biomarker into the model and the graphical region was restricted to every 10 years) and sex. The software R version 4.1.3 and the R package "refreg" were used to estimate the conditional β-content tolerance regions. The conditional β-content tolerance region method estimated the distribution function OF HGS and HGS/weight conditioned to sex and age. It determined a multivariate quantile that identifies the 5% of lower value of HGS and HGS/weight in a bivariate region. Here, we illustrate with a general sample of non-sarcopenic individuals. Furthermore, introducing the HGS and HGS/ weight correlation structure provides greater power to the analysis, incorporating the relationship between the two clinical biomarkers. To interpret the results and the outputs, one must focus on the bivariate regions instead of observing the data points outside the bivariate β-content tolerance region areas, which are estimated by the algorithm.

Results
Initially, total sample size was 7719 (n = 3859 women; n = 3860 men). After removing participants without highlevel accelerometry data, remained 3931. From those, we finally excluded 243 sarcopenic (6.4%). Descriptive statistics comparing sarcopenic and non-sarcopenic are depicted in the Table 1 of the Supplemental Material.
The final non-sarcopenic sample size used to propose the expected conditional bivariate tolerance regions was 3500 participants (i = 1708 women). Figure 1 (A for men; B for women) shows the expected conditional bivariate tolerance regions with confidence levels of 95% in the age range of 18-80 years.
In a simple way, the algorithm returns bivariate regions in 95% of individuals with values in the normal range of the two bidimensional biomarkers. In addition, we show evident variations between sex and age in the expected conditional tolerance. These results indicate that HGS cut-off points cannot be similar in individuals of different ages and sex. Specific definitions for each group are required to obtain the recommended range of the biomarker. For example, for 80-year-old men, the expected values in HGS and HGS/ weights are around 35-50 and 0.25 and 0.6, respectively. However, for individuals aged 20 years, the range varies from 40 to 60, and 0.4 and 0.75 for HGS and HGS/weight, respectively.

Discussion
Our study investigates conditional tolerance regions according to age and sex. The results demonstrate the importance of discrepancies between the age decades and sex groups. Therefore, HGS reference values must consider at least sex and age variability [9]. A critical issue in this subject is obtaining optimal conditional reference regions from a clinical point of view. This analysis can drive the development of new sarcopenia definitions in terms of precision medicine principles to overcome the inherent current limitations of the available sarcopenic screening tools. An individual has low muscle mass if his or her HGS and HGS/weight values are not within the tolerance region linked to his or her clinical characteristics.
In clinical practice, the diagnosis and treatment of many health conditions lie in quantifiable results obtained from patients' biomarkers and compared against healthy populations. However, in the case of sarcopenia, patients' age and sex must be considered in all interpretations. For instance, a study with 49,964 participants (26,687 female) from Great Britain described HGS changes per decade and highlighted lifelong changes in a univariate analysis [9]. In early adult life, HGS has expected an increase to a peak, and in midlife, HGS is maintained. However, from midlife onwards, a decline occurs [9]. Similarly, men and women are biologically different, resulting in a considerable sex variation in Fig. 1 Expected conditional bivariate tolerance regions with confidence levels of 95% for HGS and HGS/weight according to sex and age groups. Note: HGS = handgrip strength. Conditional tolerance regions by sex [men (A); women (B)] and age for the response variables HGS and HGS/weight. The observation out of the regions corresponding to each group strata is considered an outlier muscular strength capacity [9]. Taking this biological difference into consideration (sex and age) is undoubtedly relevant in the classification of normal or abnormal results [9]. Importantly, nowadays in the case of sarcopenia, the cut-off points recommended for all consensus are sex-specific but not age-specific [6][7][8][9]12], which might incur an incorrect disease diagnosis.
To interpret and extrapolate the clinical results in terms of sarcopenia disease, we selected the non-sarcopenic target population. Still, other inclusion criteria for clinical sarcopenia outcomes can be used, such as physical performance or skeletal muscle quantity/quality. One of the potential advantages of the present study proposed here is the NHANES population-based data precedence, from the American population, which provides a unique opportunity to monitor the habits of American individuals in a broader and more reliable sense.
Another advantage is the recent methodology of conditional tolerance algorithm used, which, together with the advantages of the study design, creates specific reference regions that overcome the limitations of using the standard cut-off or reference values. This new method introduces the individuals' particular characteristics (such as age, sex, and other desired characteristic) and incorporates the multivariate nature with simultaneous analysis of multiple biomarkers, such as HGS and HGS/weight. This is important, because we added the correlation structure into the analysis and optimized the statistical efficiency of the estimators.
One limitation of our study was that we derived mobility variables from the accelerometer measures, and not from usual gait speed, typically used in the older mobility reduction contexts. In NHANES, there are limited mobility data, obliging us to derivate this information from TAC. Therefore, our criteria are conservative because according to other reference values papers of the literature the 5% of individuals of populations with greater values than the cut of 0.8 m/s. These conservative rules guarantee that we are included in the sample of healthy individuals in the analysis. According to the results of the reference study about gait movement [13], the cut-off of 0.8 m/s cannot be equal between women and men. Consequently, although the definition of mobility problems will be conservative, it guarantees and supports our study in a general population with non-sarcopenic individuals. Here, we are introducing the primary characteristics (age and sex) that influence the changes in strength production, which can be interpreted well with the statistical models introduced [11]. However, there are many other characteristics that influence strength (for instance, the presence of comorbidity, physical activity level, and nutritional status), but if we introduce more predictors, interpreting the results will be more difficult. Finally, further analysis for such a purpose requires a longitudinal analysis of prognosis variables that would help to refine and confirm the optimal regions. Unfortunately, the needed longitudinal information is not entirely available in the NHANES cohort.
Other variables can be used to normalize HGS, such as BMI, as proposed by SDOC [8]. However, as the conditional tolerance algorithm is a bivariate analysis, future studies should explore other possibilities (HGS and HGS/ BMI; HGS/weight and HGS/BMI. In addition, the utilization of mortality as outcome may contribute for predicting the future health of the elders and confirm the value of personalized definitions [3] in standard clinical practice.

Conclusion
In this study, we examine the variations in grip strength according to sex and age with multivariate analysis and conditional tolerance regions in a random sample of nonsarcopenic individuals. Sarcopenia definitions must evolve based on the principles of precision medicine [14], which considers individual variability environment, genes, and/or lifestyle for each person.