https://orcid.org/0000-0003-4325-4492
Abstract
The relationship between obesity and back pain in older populations is poorly understood. This study aimed to examine (a) the impacts of changes in obesity status on back pain risk and (b) the heterogeneity in the influence of changes in obesity status according to muscle strength.
We analyzed 6 868 participants in waves 4 (2008–2009), 6 (2012–2013), and 7 (2014–2015) of the English Longitudinal Study of Ageing. The exposure comprised continuous body mass index (BMI) values, whereas the outcome comprised self-reported moderate or severe back pain. The targeted minimum loss-based estimator was used to estimate the impacts of hypothetical changes in BMI in each wave under 10 scenarios encompassing a 5%−25% hypothetical reduction/increase in BMI. We also performed stratified analysis using handgrip strength at the baseline.
For the hypothetical reduction scenarios, a 10% (relative risk [RR]; 95% confidence interval [CI] = 0.82 [0.73–0.92], p = .001) reduction in BMI estimated a significantly lower back pain risk compared to the observed data. For the hypothetical increase scenarios, a 5% (RR [95% CI] = 1.11 [1.04–1.19], p = .002) increase in BMI estimated a significantly higher back pain risk. Increased BMI had a higher risk of back pain among those with weak strength when stratified by handgrip strength but not among those with strong strength.
Our study confirmed that weight gain leads to a greater risk of back pain as well as heterogeneity in the influence of changes in obesity status according to the handgrip strength.
A number of prior studies have linked obesity to noncommunicable diseases, such as cancer and cardiovascular disease, as well as musculoskeletal problems, including back pain (1–5). The prevalence of obesity is on the rise worldwide, resulting in an obesity epidemic (1). Thus, prevention and reduction strategies are needed to address obesity at the population level from a public health perspective.
Back pain is one of the leading causes of disability (6,7), and several previous epidemiological studies have examined risk factors for back pain across various age groups (5,8). Accordingly, 2 systematic reviews revealed that obesity was associated with an increased risk of back pain in occupational or general populations among the identified risk factors (5,8). However, approximately half of the study results were not significantly associated with back pain after reviewing each longitudinal study included in the aforementioned systematic reviews (5,8). This result possibly indicates that heterogeneity in the influence of obesity status on back pain might exist.
One possible factor for heterogeneity in the risk of back pain is muscle strength. Although body mass index (BMI) is a common metric used for defining obesity status, BMI is also moderately correlated with muscle mass in older people (9). A population-based cohort study conducted in the United States reported that individuals with coexisting obesity and weak handgrip strength had the highest risk of mortality but preserving handgrip strength had a lower risk of mortality, even for those with obesity (10). Generally, weight gain increases the load on the spinal components, possibly leading to back pain (11). Previous studies have reported that spine and trunk muscle masses and activities were weaker in older individuals compared with younger individuals (12,13). Besides, greater intervertebral disk deterioration at all lumbar levels was observed in older individuals compared with younger (14). Thus, the magnitude of obesity might be greater in those individuals with weak handgrip strength given that handgrip strength is a proxy for overall muscle strength (15–17).
Thus far, evidence on the magnitude of the effect of weight gain/loss on the risk of back pain in the older population is scarce. This might be owing to methodological challenges to treat BMI as an exposure variable. Recently, many modern methods for causal inference using observational data have been developed and these methods use a potential outcome framework that calculates the contrast between outcome distributions in various hypothetical counterfactual scenarios (eg, the difference in mean outcome between “what if all study participants were exposed” and “what if all study participants were not exposed”). However, the above approach is not realistic in the case of BMI, a continuous exposure, because it is hard (or a zero change) to make everyone underweight or obese (ie, positivity violation) (18,19). Moreover, there assumes a differential effect of an increase in BMI for those individuals between underweight and another status, that is, an increase in BMI is a treatment for underweight but increases the risk of adverse health outcomes for those normal/overweight. The longitudinal modified treatment policies approach enables us to deal with continuous exposure variables using a targeted minimum loss-based estimator (TMLE) to estimate statistical parameters (18). This method also enables us for estimating the expected change in mean outcome for a given change in the continuous exposure (eg, 10% increase in BMI only among those with normal weight) without the need for relying upon strict parametric assumptions (eg, linearity) (18).
Using this framework, we hypothesized that (a) weight gain increases the risk of back pain prevalence while weight loss reduces it and (b) individuals with muscle weakness are more prone to suffer from back pain due to weight gain than those without muscle weakness. The current study, therefore, aimed to use longitudinal data from England to examine the impact of changes in obesity status on the risk of back pain and examine the heterogeneity in the influence of changes in obesity status according to muscle strength.
Materials and Methods
Study Population
This study was conducted in accordance with the Declaration of Helsinki. The National Research and Ethics Committee provided the ethics approval for all English Longitudinal Study of Ageing (ELSA) waves. All participants provided informed consent to participate in this study.
We constructed 3 waves of panel data using waves 4 (2008–2009), 6 (2012–2013), and 7 (2014–2015) of the ELSA, a nationally representative survey in England. The ELSA survey, which involves independently living people over the age of 50 in England, is conducted every 2 years, with physical examinations also being performed every 4 years (20). To be eligible for the ELSA survey, at least 1 household member must have consented to undergo follow-up by participating in the Health Survey for England, an annual health survey of the general population; must have been born before March 1, 1952; and must have resided in a household in England (20). Therefore, samples included in the ELSA survey are representative of the noninstitutionalized UK general population aged 50 years or older. This study used waves 4, 6, and 7 because the latest data on BMI was measured in wave 6; moreover, conducting an analysis using more up-to-date data for clinical implications is considered favorable.
A total of 8,643 participants eligible to be surveyed in all 3 waves were included in the analysis. After excluding 1 043 participants who had missing information on the baseline survey variables, we further excluded 732 participants who responded to all waves but had missing variables for waves 6 and 7. Consequently, 6 868 participants (mean age at baseline: 65.0 ± 9.3 years) were included in our primary analysis (Supplementary Figure 1).
Outcome: Back Pain Severity
The outcome of this study was moderate or severe back pain measured in wave 7 to avoid violating temporality (back pain measured in waves 4 and 6 was used as covariates) (21). Participants were asked the following question: “In which parts of the body do you feel pain?” Responses that included “back” were taken to indicate back pain at the time of the survey. Participants were then asked to rate the level of pain on a scale of 0–10, with 0 indicating no pain at all and 10 indicating the worst pain imaginable. Back pain graded 5 and more (moderate/severe) was characterized as having pain in line with earlier studies (22–24).
Exposure: Body Mass Index
Body weight (kg) and height (m) measured by trained nurses were used for BMI calculation: body weight divided by the square of height. Given that physical examination data were assessed every 4 years, data from waves 4 and 6 were used in this study. Continuous BMI values were used in this study.
Covariates
This study used both the time-variant (eg, age and marital status) and time-invariant covariates (eg, sex and race). The following variables, which were assessed in wave 4, were used as covariates: time-invariant = sex (binary; male vs female) (23,25–27), race (binary; White individuals vs other races) (23,28), duration of education (continuous, 9–14 years) (26,29) and time-variant = age (continuous) (23,25–27), equalized household income (continuous) (26,30), longstanding illness (binary; no vs yes) (22,23,31), arthritis (binary; no vs yes) (23,26,32), marital status (binary; married/cohabiting vs other) (26,33), engagement in mild/moderate/vigorous physical activity (categorical; more than once per week, once per week, one to three times per month, or hardly ever) (23,24,34), depressive symptoms (continuous; the Center for Epidemiologic Studies Depression Scale) (26,35), handgrip strength (continuous) (9,36,37), and back pain (binary). Handgrip strength in our study was assessed three times per person (in the dominant hand), using the Smedley handheld dynamometer (Stoelting Co., Chicago, IL). Moreover, we calculated the average value and used this as a covariate. We also employed time-variant covariates in wave 6, which included the equalized household income, longstanding illness, arthritis, marital status, engagement in light/moderate/vigorous physical activity, depressive symptoms, handgrip strength, and back pain.
Statistical Analysis
Longitudinal modified treatment policies with TMLE were used to estimate the impacts of changes in BMI in each wave (ie, waves 4 and 6) under several hypothetical scenarios on the risk of back pain (18). A weighted combination of multiple machine learning models (generalized linear models, gradient boosting models, and neural net) were used within the TMLE (18). TMLE is a doubly robust estimator in which the G-computation and inverse probability of treatment weighting approaches are combined, wherein, unbiased estimates can be obtained for the counterfactual hypothetical outcome and its contrast if the estimates of either approach are constant (18). Thus, this method enables the treatment effect calculation using counterfactual modeling, which differs from the traditional approaches that examine “association.” Moreover, in contrast to the conventional models (eg, logistic regression), the covariates were used for the calculation of weighted average probability to estimate the expected outcome in the TMLE approach; thus, the calculation of each estimate against the outcome was not required. Additionally, TMLE does not rely on parametric assumptions; thus, a weighted combination (using the SuperLearner algorithm) of previously mentioned machine learning algorithms were used to estimate the G-computation and propensity scores for inverse probability calculation (18).
The following 10 scenarios were investigated: 5%, 10%, 15%, 20%, and 25% hypothetical reduced/increased BMI in each wave (waves 4 and 6). The relative risks (RRs) and their 95% confidence intervals (CIs) were obtained by contrasting each estimate with the originally observed data. An increased or decreased hypothetical scenario was determined by the obesity status in each wave. More specifically, the ceiling value was 40 kg/m2 in the increased BMI scenarios (ie, none of the hypothetical interventions increased the BMI to >40 kg/m2) for participants with a BMI of ≥18.5 kg/m2. For participants with a BMI of <18.5 kg/m2, BMI increase was not considered (ie, there was no change in BMI). Additionally, observational data were used, considering the differential effect of an increased BMI between individuals with underweight and those with another status (38). BMI increase is a treatment for those who are underweight, but it increases the risk of adverse health outcomes for those with normal weight or overweight (38). However, the floor value was 25 kg/m2 for participants with a BMI of >25 kg/m2 in the decreased BMI scenario. For participants with normal weight and underweight (BMI of <25.0 kg/m2), BMI decrease was not considered (ie, no change in BMI), and the observed data were used. The earlier-mentioned floor and ceiling values were determined according to the BMI classification (39).
Additionally, the E value was estimated to check the robustness of our results to residual confounders (40). This value quantifies the minimum strength of association between any unmeasured confounder, BMI, and back pain.
We also performed stratified analysis using handgrip strength at wave 4. The 50th percentile of the normative values for handgrip strength according to age, sex, height, and hand side in the United Kingdom was used as the cutoff value, with the same analyses being performed on the participants (41). Another stratified analysis by age (middle age, ≤64 years; older age, ≥65 years) at wave 4 was also conducted, given that handgrip strength tends to decline with age (42).
Participants who responded to the baseline survey but did not participate in either wave 6 or 7 surveys (n = 2 334; Supplementary Figure 1) were included as censoring events to control for attrition bias in all TMLE models. All statistical analyses were conducted using R software (version 4.0.5 for Windows).
Results
Table 1 shows the baseline characteristics according to back pain at the follow-up survey (wave 7). Overall, participants who reported back pain at the follow-up survey were older; predominantly female; had a lower duration of education and income levels; had more comorbid conditions; had poorer physical activity levels and were more obese compared to those who did not report back pain. Table 2 shows changes in obesity status in waves 4 and 6. Among participants categorized as having normal weight (18.5–24.9), 16.0% transitioned to overweight. For participants whose BMI was categorized as overweight, approximately 80% remained overweight, and 10% transitioned to obesity. Moreover, among obese participants in the wave 4 survey, 84% remained obese, and 15% transitioned to overweight. Supplementary Table 1 shows the results of the comparison of baseline characteristics between the analytic sample and censored sample, indicating that attrition was associated with older age, lower socioeconomic status, poorer baseline health-related conditions, and weaker handgrip strength.
Baseline Characteristics of Study Participants Who Responded to All Three Waves, Stratified According to Back Pain at Follow-up; England; 2008–2012–2014
| Characteristic | Did Not Report Back Pain at Follow-up n = 4 048 | Reported Back Pain at Follow-up n = 486 | p Values |
|---|---|---|---|
| Age (years), mean ± SD | 64.0 ± 8.0 | 65.3 ± 8.0 | <.001* |
| Female | 2 144 (53.0) | 337 (69.3) | <.001* |
| White individuals | 3 973 (98.2) | 467 (96.1) | .003* |
| Married/cohabiting | 2 951 (72.9) | 318 (65.4) | <.001* |
| Duration of education (years), mean ± SD | 11.5 ± 1.0 | 10.8 ± 1.4 | <.001† |
| Equalized household income (£), mean ± SD | 367.8 ± 263.2 | 280.5 ± 192.2 | <.001† |
| Existing longstanding illness | 1 878 (46.4) | 382 (78.6) | <.001* |
| Existing arthritis | 775 (19.2) | 221 (45.5) | <.001* |
| Reported back pain | 234 (5.8) | 223 (45.9) | <.001* |
| No light physical activity at all | 227 (5.6) | 48 (9.9) | <.001* |
| No moderate physical activity at all | 350 (8.7) | 117 (24.1) | <.001* |
| No vigorous physical activity at all | 2 072 (51.2) | 367 (75.5) | <.001* |
| Handgrip strength (kg), mean ± SD (range, 1–66) | 31.1 ± 10.9 | 26.0 ± 10.6 | <.001* |
| Depressive symptom (CES-D), mean ± SD (range, 0–8) | 1.0 ± 1.6 | 2.2 ± 2.3 | <.001† |
| Body mass index (kg/m2), (range, 15.8–63.2) | 27.9 ± 4.8 | 30.5 ± 5.7 | <.001† |
| Characteristic | Did Not Report Back Pain at Follow-up n = 4 048 | Reported Back Pain at Follow-up n = 486 | p Values |
|---|---|---|---|
| Age (years), mean ± SD | 64.0 ± 8.0 | 65.3 ± 8.0 | <.001* |
| Female | 2 144 (53.0) | 337 (69.3) | <.001* |
| White individuals | 3 973 (98.2) | 467 (96.1) | .003* |
| Married/cohabiting | 2 951 (72.9) | 318 (65.4) | <.001* |
| Duration of education (years), mean ± SD | 11.5 ± 1.0 | 10.8 ± 1.4 | <.001† |
| Equalized household income (£), mean ± SD | 367.8 ± 263.2 | 280.5 ± 192.2 | <.001† |
| Existing longstanding illness | 1 878 (46.4) | 382 (78.6) | <.001* |
| Existing arthritis | 775 (19.2) | 221 (45.5) | <.001* |
| Reported back pain | 234 (5.8) | 223 (45.9) | <.001* |
| No light physical activity at all | 227 (5.6) | 48 (9.9) | <.001* |
| No moderate physical activity at all | 350 (8.7) | 117 (24.1) | <.001* |
| No vigorous physical activity at all | 2 072 (51.2) | 367 (75.5) | <.001* |
| Handgrip strength (kg), mean ± SD (range, 1–66) | 31.1 ± 10.9 | 26.0 ± 10.6 | <.001* |
| Depressive symptom (CES-D), mean ± SD (range, 0–8) | 1.0 ± 1.6 | 2.2 ± 2.3 | <.001† |
| Body mass index (kg/m2), (range, 15.8–63.2) | 27.9 ± 4.8 | 30.5 ± 5.7 | <.001† |
Notes: Values are presented as numbers and percentages unless otherwise noted. CES-D = Center for Epidemiologic Studies Depression Scale; SD = standard deviation.
*Chi-squared test was performed.
†t-test was performed.
Baseline Characteristics of Study Participants Who Responded to All Three Waves, Stratified According to Back Pain at Follow-up; England; 2008–2012–2014
| Characteristic | Did Not Report Back Pain at Follow-up n = 4 048 | Reported Back Pain at Follow-up n = 486 | p Values |
|---|---|---|---|
| Age (years), mean ± SD | 64.0 ± 8.0 | 65.3 ± 8.0 | <.001* |
| Female | 2 144 (53.0) | 337 (69.3) | <.001* |
| White individuals | 3 973 (98.2) | 467 (96.1) | .003* |
| Married/cohabiting | 2 951 (72.9) | 318 (65.4) | <.001* |
| Duration of education (years), mean ± SD | 11.5 ± 1.0 | 10.8 ± 1.4 | <.001† |
| Equalized household income (£), mean ± SD | 367.8 ± 263.2 | 280.5 ± 192.2 | <.001† |
| Existing longstanding illness | 1 878 (46.4) | 382 (78.6) | <.001* |
| Existing arthritis | 775 (19.2) | 221 (45.5) | <.001* |
| Reported back pain | 234 (5.8) | 223 (45.9) | <.001* |
| No light physical activity at all | 227 (5.6) | 48 (9.9) | <.001* |
| No moderate physical activity at all | 350 (8.7) | 117 (24.1) | <.001* |
| No vigorous physical activity at all | 2 072 (51.2) | 367 (75.5) | <.001* |
| Handgrip strength (kg), mean ± SD (range, 1–66) | 31.1 ± 10.9 | 26.0 ± 10.6 | <.001* |
| Depressive symptom (CES-D), mean ± SD (range, 0–8) | 1.0 ± 1.6 | 2.2 ± 2.3 | <.001† |
| Body mass index (kg/m2), (range, 15.8–63.2) | 27.9 ± 4.8 | 30.5 ± 5.7 | <.001† |
| Characteristic | Did Not Report Back Pain at Follow-up n = 4 048 | Reported Back Pain at Follow-up n = 486 | p Values |
|---|---|---|---|
| Age (years), mean ± SD | 64.0 ± 8.0 | 65.3 ± 8.0 | <.001* |
| Female | 2 144 (53.0) | 337 (69.3) | <.001* |
| White individuals | 3 973 (98.2) | 467 (96.1) | .003* |
| Married/cohabiting | 2 951 (72.9) | 318 (65.4) | <.001* |
| Duration of education (years), mean ± SD | 11.5 ± 1.0 | 10.8 ± 1.4 | <.001† |
| Equalized household income (£), mean ± SD | 367.8 ± 263.2 | 280.5 ± 192.2 | <.001† |
| Existing longstanding illness | 1 878 (46.4) | 382 (78.6) | <.001* |
| Existing arthritis | 775 (19.2) | 221 (45.5) | <.001* |
| Reported back pain | 234 (5.8) | 223 (45.9) | <.001* |
| No light physical activity at all | 227 (5.6) | 48 (9.9) | <.001* |
| No moderate physical activity at all | 350 (8.7) | 117 (24.1) | <.001* |
| No vigorous physical activity at all | 2 072 (51.2) | 367 (75.5) | <.001* |
| Handgrip strength (kg), mean ± SD (range, 1–66) | 31.1 ± 10.9 | 26.0 ± 10.6 | <.001* |
| Depressive symptom (CES-D), mean ± SD (range, 0–8) | 1.0 ± 1.6 | 2.2 ± 2.3 | <.001† |
| Body mass index (kg/m2), (range, 15.8–63.2) | 27.9 ± 4.8 | 30.5 ± 5.7 | <.001† |
Notes: Values are presented as numbers and percentages unless otherwise noted. CES-D = Center for Epidemiologic Studies Depression Scale; SD = standard deviation.
*Chi-squared test was performed.
†t-test was performed.
Changes in Obesity Status Over 4 Years; England; 2008–2012
| Wave 6 | |||||
|---|---|---|---|---|---|
| Obesity Status | Normal (BMI, 18.5–24.9) | Under (BMI, <18.5) | Over (BMI, 25.0–29.9) | Obese (BMI, >30) | |
| Wave 4 | Normal (BMI, 18.5–24.9) | 951 (82.1) | 20 (1.7) | 185 (16.0) | 2 (0.2) |
| Under (BMI, <18.5) | 9 (33.3) | 18 (66.7) | 0 (0.0) | 0 (0.0) | |
| Over (BMI, 25.0–29.9) | 207 (10.6) | 0 (0.0) | 1 551 (79.4) | 195 (10.0) | |
| Obese (BMI, >30) | 8 (0.6) | 0 (0.0) | 211 (15.1) | 1 177 (84.3) |
| Wave 6 | |||||
|---|---|---|---|---|---|
| Obesity Status | Normal (BMI, 18.5–24.9) | Under (BMI, <18.5) | Over (BMI, 25.0–29.9) | Obese (BMI, >30) | |
| Wave 4 | Normal (BMI, 18.5–24.9) | 951 (82.1) | 20 (1.7) | 185 (16.0) | 2 (0.2) |
| Under (BMI, <18.5) | 9 (33.3) | 18 (66.7) | 0 (0.0) | 0 (0.0) | |
| Over (BMI, 25.0–29.9) | 207 (10.6) | 0 (0.0) | 1 551 (79.4) | 195 (10.0) | |
| Obese (BMI, >30) | 8 (0.6) | 0 (0.0) | 211 (15.1) | 1 177 (84.3) |
Notes: Wave 4, 2008–2009; wave 6, 2012–2013. Values are presented as numbers and proportions. BMI = body mass index.
Changes in Obesity Status Over 4 Years; England; 2008–2012
| Wave 6 | |||||
|---|---|---|---|---|---|
| Obesity Status | Normal (BMI, 18.5–24.9) | Under (BMI, <18.5) | Over (BMI, 25.0–29.9) | Obese (BMI, >30) | |
| Wave 4 | Normal (BMI, 18.5–24.9) | 951 (82.1) | 20 (1.7) | 185 (16.0) | 2 (0.2) |
| Under (BMI, <18.5) | 9 (33.3) | 18 (66.7) | 0 (0.0) | 0 (0.0) | |
| Over (BMI, 25.0–29.9) | 207 (10.6) | 0 (0.0) | 1 551 (79.4) | 195 (10.0) | |
| Obese (BMI, >30) | 8 (0.6) | 0 (0.0) | 211 (15.1) | 1 177 (84.3) |
| Wave 6 | |||||
|---|---|---|---|---|---|
| Obesity Status | Normal (BMI, 18.5–24.9) | Under (BMI, <18.5) | Over (BMI, 25.0–29.9) | Obese (BMI, >30) | |
| Wave 4 | Normal (BMI, 18.5–24.9) | 951 (82.1) | 20 (1.7) | 185 (16.0) | 2 (0.2) |
| Under (BMI, <18.5) | 9 (33.3) | 18 (66.7) | 0 (0.0) | 0 (0.0) | |
| Over (BMI, 25.0–29.9) | 207 (10.6) | 0 (0.0) | 1 551 (79.4) | 195 (10.0) | |
| Obese (BMI, >30) | 8 (0.6) | 0 (0.0) | 211 (15.1) | 1 177 (84.3) |
Notes: Wave 4, 2008–2009; wave 6, 2012–2013. Values are presented as numbers and proportions. BMI = body mass index.
Figure 1 shows the results of our TMLE models, which compared the RRs and their 95% CIs between the 10 counterfactual scenarios for changes in BMI and the original (observed) data. For the hypothetical reduction scenarios, a BMI reduction of 10% (RR [95% CI] = 0.82 [0.73–0.92], p = .001), 15% (RR [95% CI] = 0.80 [0.68–0.94), p = .01), 20% (RR [95% CI] = 0.77 [0.64–0.94], p = .01), and 25% (RR [95% CI] = 0.75 [0.62–0.91], p = .003) in waves 4 and 6 estimated significantly lower risk for back pain compared to the observed data, whereas a 5% reduction in BMI in waves 4 and 6 estimated no significant reduction in risk for back pain (RR [95% CI] = 0.94 [0.87–1.02], p = .12). For the hypothetical increase scenarios, a BMI increase of 5% (RR [95% CI] = 1.11 [1.04–1.19], p = .002), 10% (RR [95% CI] = 1.21 [1.08–1.37], p = .002), 15% (RR [95% CI] = 1.23 [1.05–1.45], p = .01), 20% (RR [95% CI] = 1.33 [1.07–1.64], p = .01), 25% (RR [95% CI] = 1.49 [1.20–1.84], p < .001) in waves 4 and 6 estimated significantly higher risk for back pain compared to the observed data.
Relative risk for back pain in wave 7 according to a hypothetically increased and decreased body mass index in waves 4 and 6. All models were adjusted for baseline (ie, wave 4) covariates, including age, sex, race, education, equalized household income, marital status, longstanding illness, arthritis, depressive symptoms, handgrip strength, light/moderate/vigorous physical activities, and back pain. All models in wave 6 were also adjusted for all the earlier-mentioned covariates except sex and race. RR = relative risk; CI = confidence interval.
Figures 2 and 3 show the results of our stratified analyses according to handgrip strength at the baseline survey. For the hypothetical reduction scenarios among participants with weak handgrip strength at the baseline survey, a 20% and 25% reduction in BMI in waves 4 and 6 estimated a significantly lower risk for back pain compared to the observed data (RR [95% CIs) of 0.78 [0.61–0.99], p = .04 and 0.77 [0.64–0.92], p = .004, respectively). Furthermore, for the hypothetical increase scenarios, a BMI increase of 5% (RR [95% CI] = 1.17 [1.05–1.31], p = .004), 10% (RR [95% CI] = 1.36 [1.12–1.65], p = .002), 15% (RR [95% CI] = 1.42 [1.03–1.95], p = .03), 20% (RR [95% CI] = 1.46 [1.08–1.98], p = .01) in waves 4 and 6 estimated a significant higher risk for back pain compared to the observed data. Similar results were observed for a hypothetical 25% increase in BMI in waves 4 and 6, although statistical significance was not reached (RR [95% CI] = 1.57 [0.97–2.55], p = .07).
Results of the stratified analysis according to handgrip strength at baseline survey—participants with weak handgrip strength (n = 3 371). All models were adjusted for baseline (ie, wave 4) covariates, including age, sex, race, education, equalized household income, marital status, longstanding illness, arthritis, depressive symptoms, handgrip strength, light/moderate/vigorous physical activities, and back pain. All models in wave 6 were also adjusted for all the earlier-mentioned covariates except sex and race. RR = relative risk; CI = confidence interval.
Results of the stratified analysis according to handgrip strength at baseline survey—participants with strong handgrip strength (n = 3 497). All models were adjusted for baseline (ie, wave 4) covariates, including age, sex, race, education, equalized household income, marital status, longstanding illness, arthritis, depressive symptoms, handgrip strength, light/moderate/vigorous physical activities, and back pain. All models in wave 6 were also adjusted for all the earlier-mentioned covariates except sex and race. RR = relative risk; CI = confidence interval.
In contrast, for the hypothetical reduction scenarios among participants with a strong handgrip strength at the baseline survey, a 20% and 25% reduction in BMI in waves 4 and 6 estimated a significantly lower risk for back pain compared to the observed data (RR [95% CIs] of 0.66 [0.49–0.89], p = .01 and 0.62 [0.53–0.73], p < .001, respectively). Meanwhile, for the hypothetical increase scenarios, no significant differences were observed.
The calculated E values suggested that the observed associations were moderately robust to an unmeasured confounder (Supplementary Table 2). For example, an unmeasured confounder would necessarily be associated with both a 10% increased BMI and back pain by RR of >1.71 and above the adjusted covariates, which could be explained by the observed associations.
Supplementary Figures 2 and 3 show the results of other types of stratified analysis by age. Although a reduced BMI in waves 4 and 6 demonstrated a significantly lower risk of back pain than that observed in both middle- and older-aged individuals; the risk was higher among older adults. Hypothetical increase scenarios revealed a more apparent difference between these age groups. Among middle-aged adults, 5%−15% increases in BMI had a lower risk of back pain. In contrast, a 10% increase in BMI had a higher risk of low back pain among older adults, while other hypothetical increases were not observed a significant difference.
Discussion
The current study used nationally representative data in England to determine whether changes in BMI were associated with reducing or increasing the risk for back pain. Our findings showed that a 10%–25% reduction in BMI by over 2 survey periods (4 years) significantly reduced the risk for back pain at the 6-year follow-up. The current study also revealed that a 5%–25% increase in BMI over 2 survey periods (4 years) significantly increased the risk for back pain at the 6-year follow-up. Moreover, an increase in BMI among participants with weak handgrip strength increased the risk for back pain compared to that among participants with strong handgrip strength. In contrast, the magnitude of the impact of a reduction in BMI on lowering the risk for back pain was greater in participants with strong handgrip strength.
Overall, the current study found that a hypothetical increase in BMI in each survey period (ie, waves 4 and 6) estimated a higher risk for back pain at the follow-up survey (wave 7) with the observed data, which were partially consistent with previous studies (5,8,43). The current study found that a 5% increase in BMI in each survey wave increased the risk for back pain and that the risk of back pain increased with each additional 5% increase in BMI (Figure 1). Thus, a holistic approach toward tackling the obesity epidemic is essential for reducing the burden associated with adverse health outcomes resulting from back pain (6). The possible mechanism underlying the relationships between obesity and back pain might be complex. As mentioned earlier, weight gain increases the load on the spinal components, such as the intervertebral discs, leading to excessive wear and degeneration (11). Individuals with obesity have lower maximum muscular strength than nonobese individuals (44). Besides, individuals with back pain had lower trunk muscle strength (45).
Obesity is a state of low-grade inflammation (46). Also, a positive association was observed between elevated inflammatory biomarkers and back pain (47). Thus, inflammation resulting from obesity status might be a potential cause of back pain. Psychosocial factors might link to the back pain risk. For example, a cross-sectional study in Japan reported that lower education, income level, and depressive symptoms were associated with back pain in older population (26). Besides, obesity is also reported to link to lower socioeconomic status (29,30) and depressive symptoms (35). Therefore, it is highly likely that the heterogeneity in psychosocial factors might exist between BMI and back pain. Further studies are warranted to understand the complex relationships between obesity and back pain risk.
Our results also found that a 10% and more hypothetical reduction in BMI and maintenance for individuals with normal weight in each survey period estimated a lower risk for back pain compared to the observed data. In contrast to the evidence reported suggesting obesity to be a risk factor for back pain, we found only two studies linking weight loss to the lower risk for back pain in both the overweight (48) and obese populations (49). Thus, the current study adds new evidence suggesting that a reduction in BMI among individuals with overweight and obesity and maintenance of BMI among normal weight might reduce back pain risk. In contrast to the hypothetical increase in BMI scenarios, an additional impact of a 5% each additional reduction in BMI on back pain risk was modest (Figure 1). Hence, a 10% weight loss for overweight and obese individuals and weight maintenance for individuals with normal weight can be considered important target values for the prevention and treatment of back pain.
Our study observed that hypothetical increases in BMI had different impacts on back pain risk when stratified according to handgrip strength at the baseline survey. Notably, no clear associations of a hypothetical increase in BMI were observed among participants with strong (above 50% of the British norm) handgrip strength, whereas an apparent additional association of a 5% each additional increase in BMI on back pain risk was observed between a hypothetical increase in BMI and greater back pain risk among those with weak handgrip strength. Our findings are consistent with a previous study reporting that poor muscle mass was associated with a greater risk for musculoskeletal pains in older populations (50). By contrast, the associations of a hypothetical reduction in BMI on the risk of back pain were relatively greater in participants with strong than weak handgrip strength. Integrating previous findings with our own, intervention for weight control is important with considering muscle strength which is one of the components of assessing frailty and sarcopenia.
Our stratified analysis by age (Supplementary Figures 2 and 3) revealed that BMI reduction had a significantly lower risk of back pain than that observed in both middle- and older-aged individuals, and the risk was higher among older adults. Notably, the difference between the age groups was more apparent in the hypothetical BMI increase scenarios. Among middle-aged adults, 5%−15% increases in BMI had a lower risk of back pain. In contrast, a 10% increase in BMI had a higher risk of low back pain among older adults, while other hypothetical increases were not observed a significant difference. Therefore, obesity is a more important risk factor for the older population than for the younger population.
The present study has several limitations worth noting. The first limitation is that a high number of participants were dropped from the follow-up surveys due to attrition, which could have resulted in selection bias (4 534 individuals responded to all three survey waves, whereas 2 334 individuals did not respond to any of the three survey waves). However, participants with attrition were considered in our estimation modeling. Second, given that this was not a randomized control trial but rather a longitudinal observational study, our findings may have been influenced by unknown confounders, despite adjusting for them. However, we assessed the E values to check the robustness of the observed associations for unmeasured confounders and revealed moderate evidence for causality. Third, our results might not be generalized to other younger populations because we primarily analyzed older populations. Fourth, we used handgrip strength and age to assess the heterogeneity in the influence of obesity on back pain. Thus, the heterogeneity in physical functions (eg, walking speed) and other psychosocial factors (eg, socioeconomic status) were not assessed. Therefore, future studies are warranted to assess the influence of obesity status on health outcomes in the older population according to various health conditions, including physical functions. Fifth, there are alternative methods for assessing the impact of back pain (eg, disabling pain), which might be a crucial outcome in this population. However, we did not use that as an outcome due to a lack of information. Thus, future studies where disabling pain is used as an outcome are warranted.
In conclusion, we examined the impacts of changes in BMI on the risk of back pain using a nationally representative survey conducted in England. This study revealed that an increase in BMI had a greater risk for back pain, whereas a reduction in BMI had protective associations. Moreover, the impacts of changes in BMI on back pain risk differed according to handgrip strength. Our findings suggest the importance of aggressive weight management interventions according to physical function for back pain.
Funding
The study was supported by a grant by the Grants-in-Aid for Scientific Research (19K19818, 22K17648) from the Japan Society for the Promotion of Science. The sponsor was not involved in the study design; in the collection, analysis, and interpretation of the data; in the writing of the report; or in the decision to submit the article for publication.
Conflict of Interest
None declared.
Author Contributions
Concept and design: T.I. and U.C.; Acquisition, analysis, or interpretation of data: all authors; Drafting of the manuscript: T.I.; Critical revision of the manuscript for important intellectual content: all authors; Statistical analysis: T.I. and U.C.; Obtained funding: T.I.
