Introduction

Obesity is a pandemic and growing public health concern [1]. It is recognized as the main public health problem in industrialized countries [2] and is linked to morbidity and high mortality rates [1]. Obesity has also been found to be associated with various musculoskeletal disorders, including low back pain (LBP) [3].

LBP is common with the 1-month prevalence being estimated as 23.2 % (95 % CI 20.3–26.1). It is the highest contributor to the number of years that people live with disability in the world [4]. LBP is more common in women than in men [5] and little is understood about its etiology [6]. There is a consensus in the field that research efforts need to be allocated to the investigation of causes and risk factors for LBP, as this understanding is crucial for effective prevention [7]. From a public health perspective, it would be important to know if lifestyle factors, such as excessive body weight, are contributors to LBP incidence especially when considering that obesity is a potential modifiable risk factor.

Although it is generally assumed that obesity and LBP are associated, the actual path between these conditions remains controversial [6, 8]. For instance, whereas some studies have shown that obesity increases LBP prevalence [9, 10], others have failed to observe any association between the two [11, 12]. Moreover, in a study that used a twin sample [10], the positive association between body mass index (BMI) and LBP found in the total sample analysis disappeared in monozygotic twins dissimilar in body weight, suggesting that genetics possibly influence this relationship.

One of the limitations of the studies investigating the relationship between obesity and LBP is how obesity has been assessed. The frequent measure used to classify obesity in previous studies was BMI [2, 8, 13] which does not account for the distribution of fat in the body. Although there is evidence that body fat distribution rather than absolute total fat is associated with increases in the risk of diseases such as diabetes and coronary artery disease [14], most studies [2, 8, 13] have not investigated it. To our knowledge, only two studies have looked at the relationship between body fat distribution (where the excessive adipose tissue is stored) and LBP and found higher prevalence of LBP in women with a predominant central obesity, measured by waist circumference and waist–hip ratio [15, 16]. Although familial factors (genetics and early environment) were not investigated in these studies, the findings indicate that body fat distribution, in addition to total fat, should be considered when analyzing the obesity–LBP relationship.

In this current study, we aim to investigate the relationship between chronic LBP and different measures of obesity that account for body fat distribution in female twins. Previous evidence has indicated that women are more likely to report back pain [5, 17], seek medical care more frequently [18], and suffer from pain for longer periods [17] than men. It is likely that women represent a specific subgroup of patients with LBP and deserves special attention particularly when investigating factors such as body fat distribution. By performing a within-pair twin case–control analysis (where one twin has chronic LBP while the co-twin does not), we are able to control for possible genetic and early shared environmental effects on the obesity–LBP relationship.

Methods

Design

Cross-sectional observational study with a within-pair twin case–control design.

Study sample and data collection

All twins included in this study were registered in the Murcia Twin Registry (MTR), a population-based twin registry of adult multiples born between 1940 and 1966 in the region of Murcia, Southeast Spain. Information regarding the MTR characteristics and recruitment procedures can be found elsewhere [19]. All registry and data collection procedures involved in this study were approved by the Murcia University Ethics Committee, and informed consent was obtained from all twins.

Assessment of chronic LBP

The main outcome of this study was lifetime prevalence of chronic LBP with participants being asked the following question: “Have you ever suffered from chronic low back pain?”, based on the corresponding item from the Spanish National Health Survey. The Survey defines ‘chronic’ as a health problem lasting for at least 6 months to screen and eliminate isolated acute processes. This includes seasonal or recurrent episodes. Participants answering “yes” to this question were categorized as having a history of chronic LBP.

Measures of obesity-related measures

Self-reported measures of weight and height were obtained for 38 % of the sample (430 participants). For the other 62 % of the sample (698 participants), standardized anthropometric measurements were obtained on participants by a blinded research assistant for weight, height, waist circumference and percent body fat. BMI was calculated by dividing the individuals’ body weight in kilograms by the square of their height in meters. Percentage of body fat was measured by bioelectrical impedance using TANITA BC-420 MA (Tanita Corporation of America, USA) equipment. A single new and calibrated device was used during the whole study. Subjects were instructed to fast and not practise physical exercise during the previous 4 h, refrain from drinking alcoholic beverages during the last 24 h, and urinate closely prior to the appointment. Waist circumference was measured at the narrowest torso circumference or, alternatively, at the midpoint between the inferior border of the ribcage and the superior aspect of the iliac crest using an inelastic measuring tape. Hip circumference was measured at the widest point or, alternatively, over the buttocks”. Waist–hip ratio (WHR) was calculated as the ratio between their respective components.

Statistical analysis

The analysis was conducted in three stages: (i) total sample analysis; (ii) within-pair case–control analysis for monozygotic (MZ) and dizygotic (DZ) twins together; (iii) within-pair case–control analysis separated by DZ (iii.a) and MZ (iii.b) (Fig. 1). BMI, percent body fat, waist circumference and waist-hip ratio were classified in four categories according to percentile distributions of the data (i.e., category one ≤25th lowest percentile; 25th percentile < category two ≤ 50th percentile; 50th < category three ≤ 75th percentile; category four > 75th percentile). The specific cut-off points used to define the quartiles for obesity-related measures are defined in Table 1.

Fig. 1
figure 1

Statistical analysis schema. LBP low back pain, MZ monozygotic twins, DZ dizygotic twins

Full size image
Table 1 Cut-off points used for the obesity-related measures
Full size table

Potential confounders for the total sample analysis included age, engagement in leisure physical activity, engagement in daily physical activity (work and domestic related) and smoking. The same confounders were investigated for the within-pair twin case–control analyses, except age. Leisure physical activity was dichotomised into low/no physical activity engagement in recreational physical activity (mainly sedentary) or moderate/vigorous physical activity engagement (regular physical activity or training several times a week/month, ex: jogging, swimming, cycling). Daily physical activity was dichotomised into low/no engagement in work-related physical activity (mainly sitting or light physical efforts) or moderate/vigorous physical activity engagement (doing tasks that require a strong physical effort). Smoking was dichotomized as ex/never smoker or current smoker. We included confounders in the multivariate logistic regression models if p values for associations in univariate models were <0.2.

Total sample analysis

For the total sample analysis, we investigated the association between obesity-related measures (BMI, percent body fat, waist circumference and waist–hip ratio) and lifetime prevalence of chronic LBP using separate multivariate unconditional regression models for each obesity measure. All participants were included and twins were analyzed as individuals rather than pairs.

Within-pair twin case–control analyses

To control for the effect of genetics and early shared environment on a possible association between obesity-related measures and the lifetime prevalence of chronic LBP, we performed a subsequent within-pair twin case–control analysis on all complete and discordant pairs for LBP status (one twin reported chronic LBP while the other did not) using conditional logistic regression. In addition, separated analyses were conducted for DZ and MZ twin pairs. Theoretically, when the magnitude of the association between two variables (i.e., BMI and LBP) increases sequentially from the total sample analysis (no adjustment for genetics or early shared environment) to a DZ within-pair case–control analysis (adjustment for early shared environment and approximately 50 % of genetics occurs) and then to a MZ within-pair case–control analysis (adjustment for early shared environment and approximately 100 % of genetics), the relationship between the two variables is more direct and possibly more consistent with a direct causal path [20].

We set p < 0.05 as our level of significance for the estimates of association in the multivariate models and presented estimates as odds ratios (OR) and 95 % confidence intervals (CI). OR represents the odds of having chronic LBP per quartile step. Data analyses were performed using STATA statistical software (version 12.0).

Results

Sample characteristics

Data on lifetime prevalence of chronic LBP for the total sample of 1128 females was estimated as 41.3 % (95 % CI 38.4–44.2) with the prevalence for MZ and DZ estimated as 43.1 % (95 % CI 38.3–47.9) and 40.3 % (95 % CI 36.7–43.9), respectively. Among all twins, the mean age was 54 years with 64 % of the twins being DZ (Table 2). Results for the obesity-related measures for twins with and without chronic LBP are described in Table 3.

Table 2 Study sample characteristics of anthropometric data and lifestyle factors
Full size table
Table 3 Study sample characteristics of obesity-related measures for the total sample and cases and controls within a twin pair
Full size table

BMI

BMI (OR 1.13; 95 % CI 1.02–1.26) was weakly associated with lifetime prevalence of chronic LBP in the total sample analysis but no association was found between BMI and chronic LBP in any of the within-pair twin case–control analyses (Table 4).

Table 4 Total sample analysis and within-pair twin case–control analysis for chronic low back pain
Full size table

Percent body fat

Percent body fat (OR 1.15; 95 % CI 1.01–1.32; p = 0.05) was weakly associated with lifetime prevalence of chronic LBP but no association was identified in any of the within-pair case–control analyses.

Waist circumference

No association was found between waist circumference and lifetime prevalence of chronic LBP for the total sample or the within-pair case–control analyses.

Waist-hip ratio

No association was found between waist–hip ratio and lifetime prevalence of chronic LBP for the total sample analysis. Waist–hip ratio was associated with chronic LBP in the within-pair case–control analysis with MZ and DZ twins included (OR 0.67; 95 % CI 0.47–0.94). When the analyses were performed separately for zygosity, a stronger association was found for DZ twins (OR 0.59; 95 % CI 0.35–0.98) and the association disappeared when the analysis was conducted for MZ twins (OR 0.77; 95 % CI 0.48–1.25).

Discussion

To our knowledge, this is the first study investigating the obesity–LBP relationship in females that considered not only traditional measures of obesity such as BMI but also measures of body fat distribution and explored the effects of genetics and early shared environment. We found that lifetime prevalence of chronic LBP was weakly associated with measures of obesity. However, the association was no longer present after the full adjustment for genetics and early shared environmental factors in MZ twins dissimilar for LBP status. These results suggest that a causal direct link between obesity and chronic LBP is unlikely.

BMI and percent body fat

Our results demonstrated that lifetime prevalence of chronic LBP was associated with BMI and percent body fat when the total sample of twins (with no adjustment for genetics or early shared environment among twins) was used. The association was small (OR 1.1 and OR 1.2 for BMI and percent body fat, respectively), but in agreement with previous cross-sectional studies [21–23].

Obesity and LBP are complex traits resulting from multiple interactions between genetic and environmental factors. For example, 35–60 % of the body fat [24], and 67 % of LBP [25] variances can be accounted for by the transmission of genetic and familial environmental factors. Consequently, the true extent of the effect of obesity on LBP is difficult to estimate, and for a clear and more direct identification of obesity–LBP relationship other factors, including familial factors, should be considered.

We found that after the adjustment for familial factors, using the within-pair twin case–control design, the significant association between chronic LBP and both BMI and percent body fat did not persist. This pattern of attenuated associations after controlling for familial factors is in agreement with the previous twin studies that investigated the obesity and LBP relationship using a within-pair twin case–control design [10, 21]. It suggests that genetics and common shared environment facts play an important role when measures such as BMI and percent body fat are investigated as part of the obesity–LBP relationship.

Waist circumference and waist–hip ratio

Our results did not reveal any association between waist circumference and lifetime prevalence of chronic LBP for the total sample analysis or for any of the within-pair case–control analyses. In addition, no association was found between waist–hip ratio and lifetime prevalence of chronic LBP for the total sample analysis. However, the within-pair twin case–control analysis, with both MZ and DZ twins included, showed a significant inverse relationship between waist–hip ratio and chronic LBP (OR 0.7; 95 % CI 0.5–0.9), meaning that women with greater waist–hip ratios had lower estimates of prevalence of chronic LBP.

When we sequentially separated the analysis for DZ and then MZ twins pairs, even though the pattern of association remained the same (inverse relationship between waist–hip ratio and chronic LBP), the association was significant only in DZ twins (OR 0.6; 95 % CI 0.4–1.0), disappearing after the full adjustment for genetic factors in the MZ twins (OR 0.8; 95 % CI 0.5–1.3). The pattern of a significant relationship being observed in the within-pair twin case–control analysis and in the within-pair twin DZ only case–control analysis but not in the total sample analysis indicates that when genetics and early shared environment among twins are considered, a relationship between waist–hip ratio and chronic LBP is stronger and possibly more direct. This association is attenuated and was not statistically significant when only MZ twins are analyzed (OR 0.8; early shared environment and genetic component are fully controlled for) as opposed to DZ twins only (OR 0.6; early shared environment fully controlled for but genetic component only partially controlled for). Although there is not a clear explanation for this effect, this finding could point to the possibility that when a full adjustment for genetics and early environment is implemented, the significant relationship initially observed in DZ twins disappears. Thus, genetics could be responsible for the possible relationship between waist–hip ratio and chronic LBP.

The direction of the relationship between waist–hip ratio and chronic LBP were somewhat unexpected. Our findings suggest that chronic LBP prevalence is smaller in women with a higher waist–hip ratio. Thus, those individuals with greater hip circumferences (accumulation of fat and corresponding weight around the hip bones) were more likely to have chronic LBP. Our results are in agreement with a population-based cross-sectional study in middle-aged women, which showed that even after the adjustment of many possible confounder factors such as work-related and physical activity, high waist-to-hip ratio was still inversely associated with the risk of severe LBP [26]. However, this study did not control for genetics factors and early shared environment. These findings are in contrast to other earlier studies that used samples of women from the general population where greater levels of central obesity were associated with LBP [15, 16]. From a biomechanical perspective, it is plausible that greater fat mass around the waist area loads the spine through gravity. However, it is also mechanically plausible that greater hip to waist mass could potentially unbalance the forces around the spine, leading to lumbar–pelvic instability and LBP [27]. It is important to note that the assessment of waist–hip ratio not only incorporates the distribution of fat but also is a reflection of bony anatomical features such as the shape of the pelvis. However, it is possible that the results found in the intermediate analytical steps are a result of genetics confounding the association between waist–hip ratio and chronic LBP. It is important to note that this finding was only present in the intermediate analytical steps (MZ/DZ analysis and DZ analysis) where DZ pairs are included, and consequently 50 % of genetics are not controlled for.

Limitations

We acknowledge several limitations in our study design. Firstly this was a cross-sectional analysis, which limits possible insights on a causation path between the variables of interest. Secondly, the measure of chronic LBP used in this study was somewhat simple and did not include additional assessments of the severity, frequency, as well as disability levels associated with LBP. This assessment of LBP might have influenced the results as patients’ understanding of what constitutes chronic LBP and degrees or patterns of chronicity may vary. Thirdly, the accuracy of BMI data could have been affected by the combination of self-reported with a direct measure of weight and height used to calculate BMI given that self-reported assessment methods seem to underestimate weight and overestimate height values [28]. In our sample, the difference between the subgroups for BMI data, measured (27.53 kg/m2) and self-reported (26.58 kg/m2), was 0.96 kg/m2. We have tested the association between LBP and both subgroups for BMI for all analytical steps, and the difference in OR found were very small (≠OR < 0.2) and clinically not significant. Therefore, we believe that the combination of self-reported with a direct measure of weight and height have little or no effect in our results.

Also, in spite of being a practical and widely used assessment method, uncertainty has been raised regarding the validity and reliability of bioelectrical impedance measurements for estimation of body fat [29]. However, we should take into account the homogeneous character of our sample and the measurement conditions, and the fact that the objective of this study is not to determine the exact value of body fat percentage but analyzing its possible association with LBP. Consequently, we believe that the method of body fat measurement does not have a relevant effect on our results and conclusions. Lastly, we recognize that the smaller sample size in the case–control analyses reduced the power required to identify a relationship between obesity-related measures and chronic LBP, if in fact, the relationship exists. Therefore, we cannot exclude that a much larger twin sample would show a significant result. Although smaller samples and larger confidence intervals observed in the case–control analyses could add uncertainty to the results of our study, the imprecision of the data is unlikely to be the explanation for this finding. Firstly, the average values between LBP and non LBP groups are very small for all variables in all analysis steps. Furthermore, the magnitude of the ORs for the significant predictors in the total sample (BMI and percent body fat) is still similar to the all within-pair twin case–control analyses, which points to the fact that the association between obesity-related measures and chronic LBP in fact was weak and non-reliable when twins were considered as individuals. Secondly, according to a systematic review recently published, the reduced association between LBP and obesity seems to be consistent across studies when genetics and the environment factors are considered [30]. Pooled results of two MZ case–control studies with greater sample sizes than ours, 442 [10] and 413 [21] pairs, has shown no association between obesity and LBP (OR 1.4; 95 % CI 0.8–2.3) [30]. Notwithstanding, this study represents a step forward in the investigation of the relationship between obesity and chronic LBP relationship because we used a comprehensive assessment of obesity that accounts for body fat distribution and employed a within-pair case–control design to allow for more direct and precise estimates of obesity–LBP relationship.

In summary, BMI, percent of fat mass and greater depositions of fat and mass around the hips are associated with increases in chronic LBP prevalence in women. However, these associations are small and disappear with the full adjustment for genetics and early shared environment effects. Therefore, our results do not support a causal direct relationship between obesity and chronic LBP. We advocate that the results observed in this study should be tested in the future in a longitudinal twin research design.