High fragmented physical activity as an early risk indicator of frailty and 2 mortality in adults aged 50 years and over

1 Tittle: High fragmented physical activity as an early risk indicator of frailty and 1 mortality in adults aged 50 years and over. 2 Short Tittle: Physical Activity Fragmentation and Frailty 3 4 Associate Professor Jesús del Pozo-Cruz, PhD*1,2; Associate Professor Borja del Pozo Cruz, 5 PhD2,5; Miguel Ángel Perez-Sousa, PhD2,4; Rosa M Alfonso-Rosa, PhD2,3 6 1.Departament of Physical Education and Sport, Faculty of Education, University of Seville 7 Seville, Spain 8 2.Epidemiology of Physical Activity and Fitness Across the Lifespan Research Group 9 (EPAFit), Seville, Spain. 10 3. Departament of Human Motricity and Sport Performance, University of Seville, Faculty of 11 Education, University of Seville Seville, Spain 12 4. Faculty of Sport Sciences, University of Extremadura, 10003 Cáceres, Spain. 13 5. Centre for Active and Healthy Ageing, Department of Sports Science and Clinical 14 Biomechanics, University of Southern Denmark, Odense, Denmark. 15 Keywords 16 Pattern of Activity, Physical Activity fragmentation, Adults, Active-to-Sedentary Transition 17 Probability, Frailty 18 Word count abstract/word count main/references/tables/Figures: 296/1963/20/3/2 19 20 **Corresponding author 21 Dr. Jesus del Pozo Cruz. Associate Professor 22 Department of Physical Education and Sport 23 University of Seville 24 Street: Pirotecnia s/n 25 Postal: 41013 26 Office: +34 955420468; Mobile: +34 657661443 27 Email: [email protected] 28 2 Abstract 29 Introduction. This study aimed to explore the associations of activity fragmentation with 30 frailty status and all-cause mortality in a representative US sample of people 50 years and 31 over. 32 Methods 33 This prospective study used data from the 2003-2006 waves of the National Health and 34 Nutrition Examination Survey (NHANES). Participants 50 years or over were included in the 35 study (n = 2,586). Frailty status was assessed using a valid modification of the Fried criteria. 36 Linked data from the National Death Index registry were used to ascertain mortality. Physical 37 activity fragmentation was measured by accelerometry. To calculate activity fragmentation, 38 an active-to-sedentary transition probability was calculated as the number of physical activity 39 bouts divided by the total sum of minutes spent in physical activity. Age, gender, ethnicity, 40 education, mobility issues, drinking status, smoking status, bmi, and self-reported chronic 41 diseases were reported in the NHANES study. 42 Results 43 An increment of 1SD in activity fragmentation was associated with an increased likelihood of 44 frailty (OR [95%CI] = 1.36 [1.13 to 1.664]). Compared with participants in the high activity 45 fragmentation/low physical activity category, participants in the low activity fragmentation/low 46 physical activity and low activity fragmentation/high physical activity categories were 47 associated with a lower likelihood of frailty. We found a non-linear association between 48 activity fragmentation and all-cause mortality. Compared with participants in the high activity 49 fragmentation/low physical activity category, participants in the low activity fragmentation/low 50 physical activity, low activity fragmentation/high physical activity, and high activity 51 fragmentation/high physical activity category categories were associated with a lower mortality 52 risk. Participants with a low fragmented activity pattern may also overcome some of the 53 detrimental effects associated with sedentary behavior. 54 55 Conclusions 56 Our results suggest that a high fragmented physical activity pattern is associated with frailty 57 and risk of mortality in adults and older adults. This association was independent of total 58 volume of physical activity and time spent sedentary. 59 60 61 3 Introduction 62 Frailty is a condition of increased vulnerability [1,2] associated with aging, leading to several 63 adverse health outcomes, including disability, falls, hospitalization, and death [3]. Due to the 64 ageing population, the prevalence of frailty is rising rapidly [3], and the individual burden and 65 costs associated with this condition are sizable [4]. Nonetheless, frailty is dynamic (i.e., 66 individuals move across several states of frailty) and could be prevented to some extent. 67 Therefore, finding strategies to prevent and slow the progression of frailty are of greatest 68 importance [5]. 69 Age, sex, body mass index, education, ethnicity, and the presence of chronic conditions are 70 well-known risk factors for frailty and mortality [6,7]. Lifestyle risk factors such as alcohol or 71 smoking may also contribute to a greater risk of frailty[8]. Insufficient physical activity (i.e., 72 not meeting the physical activity guidelines) and excessive time spent in sedentary activities 73 (i.e., sitting) are well-documented major risk factors for a number of adverse health outcomes 74 [9], including frailty [10]. The patterns of accumulation of physical activity and sedentary 75 behavior may also be relevant [10–12]. Activity fragmentation, defined as the frequency with 76 which a person transitions into a sedentary state from an active state throughout the day, may 77 also contribute to frailty. It is likely that more frail individuals depict a more fragmented pattern 78 of activity, possibly due to an increased fatigue and declines in physical functioning, which 79 may result in the inability to sustain longer bouts of physical activity, and individuals adapting 80 their mobility patterns accordingly [13]. A fragmented pattern of activity has been previously 81 associated with cancer [14], fatigue [15], and subjective memory complaints[16]. Recent 82 evidence has also suggested that more fragmented patterns of physical activity may also be 83 associated with frailty. These previous studies highlight the value of accelerometers to assess 84 physical activity patterns (including activity fragmentation), and to identify adults at risk of 85 becoming frail, even before other clinical manifestations are present. These novel 86 accelerometer metrics may also reveal important insights for the design of tailored 87 interventions to prevent frailty among those at risk. As previously noted, fragmentation in 88 adults may precede declines in functional capability and overall physical activity that typically 89 indicate impeding mortality [17]. However, the dose-response between activity fragmentation 90 and mortality risk remains unexplored. This information can also be useful to identify relevant 91 thresholds of activity fragmentation that may be able to identify plausible impeding health 92 outcomes, previously unknown. 93 4 Therefore, this study aimed to explore the associations of activity fragmentation with frailty 94 status and the risk of mortality in a US-based population sample of people 50 years and over. 95 In doing so, we leveraged the potential of accelerometers and, through continuous assessment 96 of minute-by-minute activity as well as sedentary cycles throughout the day, derived a measure 97 of activity fragmentation (i.e., frequency with which a person transitions into a sedentary state 98 from an active state throughout the day) [14]. We hypothesized that a higher activity 99 fragmentation will be associated with greater odds of frailty status and an increased risk of 100 mortality in the study population. We also hypothesized that participants with a low total 101 volume of physical activity and a high fragmented pattern of physical activity will display 102 higher odds of frailty and an increased risk of mortality compared with other patterns of 103 activity. 104 Methods 105 Study design and participants 106 We used data from the waves 2003-2006 of the National Health and Nutrition Examination 107 Survey (NHANES). The NHANES is a series of cross-sectional surveys delivered every two 108 years and is designed to gather health and nutrition information in a representative sample of 109 the civilian, non-institutionalized US population. The original study was approved by the 110 Centers for Disease Control and Prevention Ethics Committee, and all participants gave 111 informed consent. Participants 50 years or over [18] with at least 3 days of valid accelerometry 112 data and available frailty data were included in the study (n = 2,586). Figure 1 shows the flow 113 diagram of participants in the study. 114 Measures 115 Frailty status 116 We defined frailty status based on a modification of the Fried criteria [1], validated for 117 application to NHANES data [19]. The following criteria were used to classify participants’ 118 frailty status [18]: exhaustion, defined by ‘‘some difficulty’’, ‘‘much difficulty’’, or ‘‘unable 119 to do’’ when asked how much difficulty they have ‘‘walking from one room to the other on the 120 same level’’. Low physical activity, defined as ‘‘less active’’ when asked ‘‘Compared with 121 most (men/women) your age, would you say that you are more active, less active, or about the 122 same?’’. Weakness, defined by ‘‘some difficulty”, ‘‘much difficulty’’, or ‘‘unable to do’’ when 123 5 asked how much difficulty they have ‘‘lifting or carrying something as heavy as 10 pounds 124 [like a sack of potatoes or rice]’’. Low body weight, defined by BMI 18.5 kg/m2 or lower. 125 Robust individuals were those with no criteria present. Participants with 1 or 2 criteria were 126 considered pre-frail, and participants with 3 or 4 were considered frail. Because of the low 127 proportion of frail individuals within our sample (4%), the pre-frail and frail groups were 128 merged into a unique category. Participants who did not present any of the aforementioned 129 criteria were classified as robust. Individuals with missing data on any of the criteria were 130 excluded from the study. 131 Mortality 132 A subset of the individuals included is linked to death records from the National Death Index 133 until the 31st of December of 2011, which provides information on whether or not the 134 participant was alive at follow-up. If deceased, the length of time (in months) between the 135 NHANES examination and the subject's death is provided, as well as the cause of death. 136 Accelerometer variables 137 Physical activity and activity fragmentation were measured by accelerometry. Participants 138 wore accelerometers (AM-7164, ActiGraph, LLC, Fort Walton Beach, Florida) on their waist 139 according to the protocol for 7-consecutive days during waking hours. Accelerometer data 140 reduction in the current study followed validated procedures [20]. Accelerometers were set to 141 record accelerometer counts in one-minute epochs. Non-wear time was defined as 60 142 consecutive minutes or longer of zero intensity counts, with no more than two minutes of counts 143 between zero and 100. The study included data only from participants with at least 3 valid days 144 with 10 or more hours per day of wear time [14]. Valid accelerometer counts were classified 145 into sedentary (<100 counts/min) or otherwise active (100+ counts/min). Two different 146 continuous measures were then derived. First, to calculate total physical activity volume, 147 activity counts were summed across all minutes for each valid day and averaged across all valid 148 days for each participant. Second, to calculate activity fragmentation, an active-to-sedentary 149 transition probability was calculated as the number of active physical activity bouts divided by 150 the total sum of minutes spent in active physical activity (i.e., consecutive minutes registering 151 100+ counts/min). We used the sample-based median value of the total volume of physical 152 activity and activity fragmentation to derive high and low physical activity and activity 153 fragmentation groups. Based on these values, we classified participants into four mutually 154 6 exclusive physical activity pattern categories: low physical activity fragmentation/high 155 physical activity, low physical activity fragmentation/low physical activity, high physical 156 activity fragmentation/high physical activity, and high physical activity fragmentation/low 157 physical activity [14] . We then used the sample-based median value of sedentary behavior and 158 activity fragmentation to derive high and low sedentary behavior and activity fragmentation 159 groups. Based on these values, we created four mutually exclusive sedentary pattern categories: 160 low physical activity fragmentation/high sedentary behavior, low physical activity 161 fragmentation/low sedentary behavior, high physical activity fragmentation/high sedentary 162 behavior, and high physical activity fragmentation/low sedentary behavior. 163 Covariates 164 Based on previous literature [6–8]and data availability, the following covariates were selected: 165 A computer-assisted personal interviewing methodology was used to collect in-home 166 information regarding age, gender, ethnicity, education, mobility issues, drinking status 167 (former-never drinker, heavy drinker, moderate drinker, non-drinker), smoking status (never, 168 former, current), and self-reported chronic diseases (Diabetes, Coronary heart disease, 169 Congestive heart failure, Stroke, Cancer) in the NHANES study. Body mass index (BMI) was 170 calculated as weight in kilograms divided by height in meters squared. 171 Statistical analysis 172 We first described the sample characteristics using mean (SD) or frequency (percentage) for 173 continuous and categorical variables respectively. 174 Frailty 175 A multivariable ordinal logistic regression model was conducted to examine the cross-sectional 176 associations between total physical activity volume, activity fragmentation, and joint physical 177 activity pattern category (reference, high physical activity fragmentation/low physical activity 178 group) with the probability of frailty. We also tested these associations for joint sedentary 179 behavior patterns (reference, high physical activity fragmentation/high sedentary behavior). 180 The results are reported as odds ratio (OR) and associated 95% Confidence Interval (CI). The 181 magnitude of the association is illustrated through the average marginal effect (AME) and 182 associated 95% CI. For additional interpretability of the results, we standardized the activity 183 fragmentation index by subtracting the population-level mean and dividing by the population- 184 7 level standard deviation, resulting in a model coefficient that corresponds to one standard 185 deviation change. A locally weighted scatterplot-smoothing curve was used to represent the 186 adjusted estimated probability of frailty associated with total volume of physical activity and 187 activity fragmentation. 188 Mortality 189 We used a cox proportional hazard model to estimate the prospective associations between 190 baseline physical activity (and sedentary behavior) patterns and the risk of mortality. Results 191 are displayed as hazard ratios (HRs) of mortality during follow-up, and time-on-study in 192 months was used as the timescale. We assessed the dose-response associations of activity 193 fragmentation (modelled as a continuous exposure) and all-cause mortality using a restricted 194 cubic spline model to allow for potential non-linearity. For the purpose of this analysis, we 195 trimmed observations less than 5% and greater than 95% of the distribution. Then, we pre- 196 specified knots placed at the 10th, 50th (reference), and 90th percentiles of the exposure 197 distribution. We assumed linearity for values below the 10th percentile and for values above 198 the 90th percentile. Departure from linearity was assessed by a Wald test examining the null 199 hypothesis that the coefficient of the second spline was equal to zero. A Cox proportional 200 hazard model was also used to estimate the HRs of mortality associated with the different joint 201 categories of physical activity patterns (reference, high physical activity fragmentation/low 202 physical activity group). Results are reported as HRs with 95% CIs and levels of significance 203 were set at p <0.05. 204 All models were adjusted for age, sex, ethnicity, education level, smoking status (never, former, 205 current), drinking status (former-never drinker, heavy drinker, moderate drinker, non-drinker), 206 BMI (kg/m2), mobility issues (yes, no), self-reported diagnosis of diabetes, coronary heart 207 disease, congestive heart failure, stroke, and cancer as well as sedentary time and total wear 208 time. Sample weights were used to account for the complexity of the NHANES design. All 209 analyses were conducted with R software (version 3.5.1). The alpha level was set at 0.05, two- 210 tailed. 211 Results 212 Table 1 shows the characteristics of the study participants. Robust individuals were slightly 213 older, mostly male, and had a lower BMI. They were also more educated, healthier, non- 214 8 smokers or drinkers, and had less mobility issues. Finally, robust individuals were also more 215 active, and had a less fragmented activity pattern. 216 Frailty 217 An increment of 30 min/day of physical activity was associated with a decreased likelihood of 218 frailty (OR [95%CI] = 0.91 [0.87 to 0.94]; AME [95%CI] = –0.014 [–0.019 to –0.009]). In 219 contrast, an increment of 1SD in activity fragmentation was associated with an increased 220 likelihood of frailty (OR [95%CI] = 1.36 [1.13 to 1.664]; AME [95%CI] = 0.048 [0.019 to 221 0.077]) in the population of the study (Table 2; Figure 2). 222 Table 3 shows the association between joint physical activity pattern category and frailty. 223 Compared with participants in the high activity fragmentation/low physical activity category, 224 participants in the low activity fragmentation/low physical activity and low activity 225 fragmentation/high physical activity categories were associated with a lower likelihood of 226 frailty. There was no detectable association between the high activity fragmentation/high 227 physical activity category and frailty when compared to the high activity fragmentation/low 228 physical activity category. Compared with participants in the high activity fragmentation/high 229 sedentary behavior category, other participants in the rest of the joint sedentary behavior 230 pattern groups depicted a lower likelihood of frailty (Table 3) 231 Mortality 232 The dose-response analysis revealed a non-linear (p-value from second spline 0.040) 233 association between activity fragmentation and all causes of mortality (minimal dose, 23%) 234 (Figure 3). Compared with participants in the high activity fragmentation/low physical activity 235 category, participants in the low activity fragmentation/low physical activity, low activity 236 fragmentation/high physical activity, and high activity fragmentation/high physical activity 237 categories were associated with a lower mortality risk (Table 4). A lower mortality risk was 238 also observed for participants in the high activity fragmentation/low sedentary behavior, low 239 activity fragmentation/low sedentary behavior, and low activity fragmentation/high sedentary 240 behavior groups when compared with those participants classified in the high activity 241 fragmentation/high sedentary behavior (Table 4). 242 Discussion 243 The current study was designed to examine the association between physical activity 244 9 fragmentation and frailty in a population of adults from the NHANES survey. Activity 245 fragmentation has already been associated with several clinical outcomes. For example, 246 Palmberg et al. [15] and Wanigatunga et al. [17] found that a higher physical activity 247 fragmentation was associated with an increased risk of mental fatigability and mortality, 248 respectively. Similarly, del Pozo Cruz et al. found that a higher activity fragmentation was 249 associated with subjective memory complaints in older adults [16]. Recent evidence suggested 250 that fragmented patters of physical activity may also be associated with frailty[21]. Our results 251 complement and extend those from previous studies and suggest that adults and older with a 252 fragmented physical activity pattern could be more prone to frailty, over and above total time 253 of physical activity. Similarly, we also found that a more fragmented patter of activity was 254 associated with an increased risk of mortality. Another relevant finding of this study was that 255 a low fragmented activity may be beneficial even in the presence of high volumes of sedentary 256 behavior. 257 Previous studies have indicated the major role total volume of physical activity and sedentary 258 behavior have on frailty among adults and older adults [22–24]. Others have indicated that 259 patterns of physical activity and sedentary behavior accumulation may also be relevant [10,16]. 260 In agreement with a previous smaller study[21], we found that an increased probability of 261 transitioning from an active state to a sedentary state (i.e., more fragmented patterns of physical 262 activity) may also be associated with an increased risk of frailty. Biologically, these findings 263 could be partly explained by the functional deterioration associated with age, and an increase 264 in time spent in sedentary behaviours [24,25]. This is consistent with previous evidence 265 suggesting that the inability to sustain prolonged bouts of physical activity is associated with 266 fatigability and loss of physical function among older adults, which could ultimately lead to 267 sarcopenia and frailty[26]. As a novelty, we found that even in the presence of high levels of 268 total time of physical activity, participants with a higher fragmentation index had an increased 269 likelihood of being frail. In contrast, lower activity fragmentation may reduce the odds of frailty 270 even in the presence of high volumes of sedentary behavior. Together, our findings and those 271 from previous reports underscore the relevance of assessing the patterns of physical activity 272 accumulation as well as total time of physical activity and sedentary behavior to identify people 273 at risk of frailty and plan interventions accordingly. 274 Our study also supports previous evidence from well-functioning older adults[17] and 275 confirms, in a representative sample, that more fragmented activity patterns are associated with 276