- Top of page
- MATERIALS AND METHODS
- AUTHOR CONTRIBUTIONS
Arthritis is a leading cause of pain and long-term disability, with estimated direct and indirect costs of $128 billion annually (1). Osteoarthritis (OA), the most prevalent type of arthritis, was estimated to affect 27 million adults in the US in 2005 (1). OA is associated with chronic pain, functional disability, and lower quality of life (1, 2). To reduce the personal, social, and economic burden of arthritis, it is of paramount importance that its risk factors be recognized and addressed.
One understudied topic that warrants attention is the potential relationship between childhood abuse and arthritis. Childhood abuse is a significant stressor in early life that precipitates deleterious consequences in adulthood relating to both mental (3–6) and physical health. A wide range of adult health conditions has been linked to childhood abuse, including heart disease (3, 7, 8), cancer (9), functional somatic syndromes (10–12), asthma, bronchitis/emphysema, and ulcers (5).
Previous research has identified an association between childhood abuse and arthritis in adulthood (5, 7, 13, 14). Abuse can include physical, sexual, or emotional types, and/or neglect. The evidence of a relationship between arthritis and childhood physical abuse is the strongest and most consistent (5, 7). Goodwin and Stein (7) did not find an association between childhood neglect and arthritis when other forms of abuse, including physical abuse, were controlled for. The evidence of a relationship between childhood sexual abuse and arthritis is conflicting; one study found an association (14) and another did not (7). Finally, to our knowledge, no quantitative study has investigated the association between emotional abuse and arthritis.
The literature suggests that there are at least 3 types of factors that potentially link childhood physical abuse to arthritis: 1) childhood stressors, 2) adult health behaviors, and 3) depression.
Childhood abuse often co-occurs with other childhood risk factors such as parental addictions (5), parental conflict (5, 15), household poverty (16), and parental unemployment (15, 16). Kopec and Sayre (13) found that a cluster of childhood stressors, including parental drug or alcohol abuse, divorce, and long-term unemployment, was associated with a moderately elevated risk of arthritis, with a trend toward statistical significance.
It has been hypothesized that adult survivors of childhood physical abuse adopt unhealthy behaviors to cope with their adverse early experiences. These behaviors are thought to elevate the risk for chronic diseases (3). Childhood physical abuse has been associated with obesity and physical inactivity (3), smoking (3, 17–19), and alcohol use in adulthood (3, 17–19).
Obesity is a widely established risk factor for OA, demonstrated by self-report and radiographic evidence in the knee and hip among both sexes (20–25). Moderate recreational exercise is found to be associated with reduced rates of OA of the knee (26). Adults with doctor-diagnosed arthritis, however, are less likely to participate in leisure time physical activity as compared with their peers without arthritis (27). Smoking has recently been associated with femoral cartilage loss among both sexes (28) and with higher rates of self-reported arthritic conditions (29). Previous studies, however, have not found an association between smoking and radiographic OA of the knee, hand, foot, and cervical spine (30), and yet others have found smoking to be a protective factor for knee OA in women (31) and hip OA in men (22, 24). The evidence of the role of alcohol consumption as a risk factor for OA is also unclear. A recent international study using pooled data from several countries found slightly higher odds of alcohol dependence among those with self-reported arthritis (32). This finding is supported by a previous study in the US that found a positive association between alcohol consumption and knee and hip OA among younger men (33); however, another US study did not find any association between OA and alcohol consumption (22).
Childhood physical abuse is a widely recognized risk factor for depression in adulthood (3–5). In turn, population-based research has demonstrated a positive association between depression and self-reported arthritis (32). This finding has also been confirmed in prospective research (34). Few studies have focused on depression prevalence specifically among those with OA; however, recent clinical research suggests that depression is elevated among those with OA as compared with general population estimates (35).
Previous research has found consistent patterns of associations between older age (20, 21, 31, 36), female sex (22, 36, 37), and higher rates of OA. Race has also been associated with significant variations in rates of OA (29, 36, 38). Additionally, low socioeconomic status (SES) measured by education level and income has been associated with higher rates of self-reported arthritic conditions (29). In particular, an inverse relationship between education level and rates of knee OA has been found (39, 40).
This study investigated the relationship between self-reported childhood physical abuse and a self-reported health professional diagnosis of OA based on a large Canadian regionally representative community sample. In order to study this relationship, the control variables age, sex, race, and SES (education and income) were adjusted for in addition to the following 3 types of risk factors: 1) childhood stressors, 2) adult health behaviors, and 3) mood disorders. Unfortunately, certain risk factors for OA could not be controlled for due to lack of data availability. Such factors included joint vulnerability characteristics such as acute joint injury (20, 22, 41) or a high frequency or intensity of repeated loading, for example, due to a history of heavy lifting (20, 42), or high levels of physical activity or sports participation (22, 23, 33, 43); bone mineral density (20); estrogen deficiency (20); family history; and genetic factors relating to OA (20, 44, 45).
Based on our previous research (9), several hypotheses were made as follows: those reporting childhood physical abuse were hypothesized to be more likely to also report OA when adjusting for the control variables (age, race, sex, and SES) only. Adjusting for the control variables in addition to any one of the 3 groups of factors alone was hypothesized to reduce the association between childhood physical abuse and OA, but not below significance. Health behavioral factors were hypothesized to reduce the association the greatest (3). Finally, adjusting for control variables in addition to all of the 3 groups of factors together was hypothesized to reduce the association between childhood physical abuse and OA more than any one of the groups alone; however, it was anticipated that the association would remain significant.
- Top of page
- MATERIALS AND METHODS
- AUTHOR CONTRIBUTIONS
Based on the full sample of 13,093 respondents, 9.6% (95% CI 8.4–10.8) reported OA diagnosed by a health professional. The results reported hereafter were based on the restricted sample (n = 11,108), which excluded individuals with missing data on the arthritis questions and those with arthritis types other than OA. Based on this subsample, 10.1% (95% CI 8.9–11.3) reported OA diagnosed by a health professional and 6.9% (95% CI 5.9–7.9) reported a history of childhood physical abuse. A description of the sample comparing those with OA (n = 1,452) with those without any arthritis (n = 9,656) is shown in Table 1.
Table 1. Description of the sample of those with osteoarthritis and those without any arthritic condition from the Canadian provinces of Manitoba and Saskatchewan*
| ||No arthritis (n = 9,656)||Osteoarthritis (n = 1,452)||Total (n = 11,108)|
|Key variable of interest|| || || |
| Childhood physical abuse|| || || |
| No||8,951 (93.5)||1,303 (89.8)||10,254 (93.1)|
| Yes||705 (6.5)||149 (10.2)||854 (6.9)|
|Control variables|| || || |
| Sex|| || || |
| Male||4,540 (50.6)||374 (30.7)||4,914 (48.6)|
| Female||5,116 (49.4)||1,078 (69.3)||6,194 (51.4)|
| Age, years|| || || |
| 12–39||4,069 (46.9)||–†||4,121 (42.5)|
| 40–49||1,725 (22.0)||121 (14.2)||1,846 (21.2)|
| 50–59||1,525 (15.6)||338 (29.4)||1,863 (17.0)|
| 60–69||1,054 (7.9)||349 (22.2)||1,403 (9.3)|
| 70–79||777 (5.1)||356 (19.0)||1,133 (6.5)|
| ≥80||506 (2.5)||236 (11.8)||742 (3.5)|
| Ethnicity|| || || |
| White||8,340 (85.1)||1,341 (91.5)||9,681 (85.8)|
| Nonwhite||1,210 (14.9)||102 (8.5)‡||1,312 (14.2)|
| Education|| || || |
| Less than high school||2,074 (16.1)||507 (28.9)||2,581 (17.4)|
| High school/some postsecondary||2,750 (31.1)||276 (21.8)||3,026 (30.2)|
| Postsecondary||4,729 (52.8)||659 (49.3)||5,388 (52.4)|
| Household income|| || || |
| <$15,000||804 (4.7)||189 (9.7)‡||933 (5.2)|
| $15,000–29,999||1,434 (10.2)||362 (19.1)||1,796 (11.1)|
| $30,000–49,999||1,929 (18.4)||316 (21.6)||2,245 (18.8)|
| ≥$50,000||4,161 (53.1)||373 (35.6)||4,534 (51.4)|
| Missing||1,328 (13.5)||212 (14.0)||1,540 (13.5)|
|Childhood stressors|| || || |
| Parents unemployed for a long time|| || || |
| No||8,743 (90.9)||1,314 (90.4)||10,057 (90.9)|
| Yes||816 (9.1)||121 (9.6)‡||937 (9.1)|
| Parents addicted to alcohol/drugs|| || || |
| No||8,245 (85.9)||1,232 (82.4)||9,477 (85.6)|
| Yes||1,386 (14.1)||213 (17.6)||1,599 (14.4)|
| Parents divorced|| || || |
| No||8,524 (86.9)||1,357 (91.1)||9,881 (87.4)|
| Yes||1,121 (13.1)||90 (8.1)‡||1,211 (12.6)|
|Adult health behaviors|| || || |
| Body mass index category, kg/m2|| || || |
| ≤25||4,135 (45.3)||493 (34.2)||4,628 (44.1)|
| 26–29.99||3,323 (34.6)||503 (35.5)||3,826 (34.7)|
| ≥30||1,809 (16.8)||414 (27.6)||2,223 (17.9)|
| Missing||389 (3.3)||–†||431 (3.3)|
| Physical activity level|| || || |
| Active||2,317 (24.1)||259 (18.5)||2,576 (23.5)|
| Moderate||2,380 (25.2)||340 (24.8)||2,720 (25.1)|
| Inactive||4,949 (50.8)||849 (11.1)||5,798 (51.3)|
| Smoker|| || || |
| Current or former||6,532 (65.3)||1,002 (72.2)||7,534 (66.0)|
| Never||3,114 (34.7)||449 (27.8)||3,563 (34.0)|
| Alcohol use|| || || |
| Abstainer/very light drinker||6,874 (69.8)||1,192 (79.7)||8,066 (70.8)|
| Low consumption||2,099 (24.8)||204 (18.0)||2,303 (24.1)|
| Hazardous/harmful drinking||508 (5.4)||–†||544 (5.1)|
|Depression|| || || |
| Have a mood disorder|| || || |
| No||9,158 (95.0)||1,330 (92.8)||10,488 (94.8)|
| Yes||493 (5.0)||120 (7.2)‡||613 (5.2)|
Compared with individuals who did not report abuse, those who reported childhood physical abuse had 1.99 (95% CI 1.57–2.52) higher odds of having a diagnosis of OA when adjusting for age, sex, race, and SES only (Figure 1).
Figure 1. Odds ratios and 95% confidence intervals for osteoarthritis among individuals reporting childhood physical abuse versus individuals not reporting abuse. All of the data are adjusted for age, sex, race, and socioeconomic status (SES). Sample sizes vary from 10,965 in the first model to 10,629 in the fully adjusted model. Source: representative, regional sample of the Canadian Community Health Survey, 2005.
Download figure to PowerPoint
Adjustment for each of the 3 groups of risk factors separately acted to reduce the magnitude of the association, but not below significance (childhood stressors: OR 1.73, 95% CI 1.35–2.22; adult health behaviors: OR 1.84, 95% CI 1.45–2.34; mood disorders: OR 1.89, 95% CI 1.50–2.40). Variance inflation factor analysis of the independent variables yielded a maximum average variance inflation factor score of 1.17, thus indicating little evidence of multicolinearity.
Each of the 3 preceding models was associated with a 5% or greater change in the magnitude of the OR for childhood physical abuse compared with the base model; therefore, all 3 groups were retained in the final model. In contrast, the effect modification analysis demonstrated that sex did not act as a significant modifier in the relationship between childhood physical abuse and OA; thus, the interaction was dropped from further analyses.
In the final model adjusting for control variables and all 3 groups of risk factors concomitantly, the odds of reporting a diagnosis of OA remained significant (OR 1.56, 95% CI 1.21–2.00), although reduced in magnitude compared with the base model. The fully adjusted logistic regression model is shown in Table 2.
Table 2. Fully adjusted logistic regression model of individuals with osteoarthritis as compared with those without any arthritic conditions based on a representative sample of Canadians from Manitoba and Saskatchewan*
| ||Final model (n = 11,108), OR (95% CI)|
|Independent variable|| |
| Exposure Childhood physical abuse|| |
| Yes||1.56 (1.21–2.00)†|
| No||1.00 (referent)|
|Control variables|| |
| Sex|| |
| Male||1.00 (referent)|
| Female||2.30 (1.97–2.68)†|
| Race|| |
| White||1.00 (referent)|
| Visible minority||0.91 (0.71–1.16)|
| Age, years|| |
| 12–39||1.00 (referent)|
| 40–49||8.67 (6.07–12.38)†|
| 50–59||25.10 (17.81–35.38)†|
| 60–69||36.98 (25.88–52.86)†|
| 70–79||49.07 (33.85–71.14)†|
| ≥80||63.67 (42.48–95.43)†|
| Education|| |
| Less than high school||0.91 (0.76–1.09)|
| High school/some postsecondary||0.85 (0.71–1.01)|
| Postsecondary||1.00 (referent)|
| Household income|| |
| <$15,000||1.53 (1.16–2.02)†|
| $15,000–29,999||1.47 (1.18–1.83)†|
| $30,000–49,999||1.35 (1.12–1.63)†|
| ≥$50,000||1.00 (referent)|
| Missing||1.05 (0.84–1.33)|
| Childhood stressors|| |
| Parental divorce|| |
| Yes||0.94 (0.72–1.23)|
| No||1.00 (referent)|
| Parental addictions|| |
| Yes||1.49 (1.22–1.83)†|
| No||1.00 (referent)|
| Parental unemployment|| |
| Yes||1.29 (1.01–1.65)†|
| No||1.00 (referent)|
| Adult health behaviors|| |
| BMI category, kg/m2|| |
| ≤25||1.00 (referent)|
| 26–29.99||1.29 (1.10–1.52)†|
| ≥30||2.08 (1.73–2.49)†|
| Missing||1.39 (0.90–2.15)|
| Physical activity level|| |
| Active||1.00 (referent)|
| Moderate||0.98 (0.80–1.20)|
| Inactive||0.88 (0.73–1.06)|
| Smoking|| |
| Current/former||1.32 (1.13–1.54)†|
| Never||1.00 (referent)|
| Alcohol use|| |
| Abstainer/very light drinker||1.21 (0.78–1.86)|
| Low consumption||1.12 (0.71–1.76)|
| Hazardous/harmful drinking||1.00 (referent)|
| Depression|| |
| Mood disorder|| |
| No||1.00 (referent)|
| Yes||1.55 (1.17–2.06)†|
- Top of page
- MATERIALS AND METHODS
- AUTHOR CONTRIBUTIONS
Individuals reporting a diagnosis of OA had significantly higher odds of reporting childhood physical abuse. This association remained significant even when controlling for demographic and SES characteristics, in addition to the 3 groups of factors associated with childhood physical abuse and arthritis, namely childhood stressors, adult health behaviors, and mood disorders.
These findings support previous US research demonstrating an increased risk of arthritis in adults who reported childhood physical abuse (5, 7). Goodwin and Stein (7) and Springer et al (5) found ORs for childhood physical abuse of 1.7 (95% CI 1.2–2.3) and 1.34 (95% CI 1.05–1.72), respectively, among those reporting arthritis. These estimates are comparable to our OA-specific findings (OR 1.56, 95% CI 1.21–2.00) in the final model. Both of the studies controlled for several of the same factors controlled for in this study (Goodwin and Stein: demographics, depression, and alcohol use ; Springer et al: demographics, childhood stressors, and depression ); however, our study included a wider range of risk factors. Unique variables to this study included adult obesity, physical activity level, and smoking, the former two of which have most clearly been associated with OA (20–26). Furthermore, the outcome of this study was OA specific, whereas the aforementioned studies used a broader outcome comprising any self-reported arthritic conditions.
The Canadian prospective cohort study by Kopec and Sayre (13) found an elevated but nonsignificant risk of arthritis (hazard risk ratio 1.14, 95% CI 0.79–1.66) among those with a self-reported history of childhood physical abuse who were free of arthritis 4 years earlier. This study is the strongest to date, due to their inclusion of a large range of potential confounders and use of a longitudinal research design. However, Kopec and Sayre used a broad outcome of any self-reported arthritic condition. Future prospective studies would benefit from larger samples, longer followup periods, and a specific focus on several different arthritic conditions.
The prevalence of self-reported OA diagnosed by a health professional based on the full sample of all those in Manitoba and Saskatchewan (9.6%; 95% CI 8.4–10.8) is consistent with previous Canadian estimates of OA comprising approximately 10% of the population (48). The estimate is slightly lower than representative American estimates of clinically defined OA on the basis of symptoms and physical examination of 12.1% for adults ages 25–74 years (36).
The prevalence of childhood physical abuse reported in this study (6.9%) is based only on 2 Canadian provinces, due to the fact that the CCHS respondents from the rest of the country were not asked childhood abuse questions. Unfortunately, therefore, comparisons of the reported childhood physical abuse rate cannot be readily made using the CCHS data. The ability to make Canadian comparisons is further complicated by the lack of national data on the actual number of children in the care of the child welfare system (49). The US nationally representative estimate of childhood physical abuse reported by Afifi et al in 2006 (50) is higher than the rate of childhood physical abuse reported here (16.5% versus 6.9%). However, in contrast to the CCHS, the population survey used by Afifi et al (50) contained a more objective assessment of childhood physical abuse, which may have resulted in higher rates of reporting abuse.
The association between childhood physical abuse and OA varied only slightly when a range of potentially mediating risk factors was adjusted for. Among the risk factors explored, the role of adult health behaviors as a potential mediator between childhood physical abuse and adult health conditions has received the most attention (3). However, adjusting for health behaviors resulted in a very similar OR (1.84, 95% CI 1.45–2.34) compared with the base model (1.99, 95% CI 1.57–2.52). Interestingly, adjustment for childhood stressors reduced the magnitude of the association between childhood physical abuse and OA the most (OR 1.73, 95% CI 1.35–2.22), while adjustment for mood disorders reduced it the least (OR 1.89, 95% CI 1.50–2.40) as compared with the base model (OR 1.99, 95% CI 1.57–2.52). Speculation about the link between childhood physical abuse and OA is further complicated by the fact that the analysis could not contain all of the known risk factors for OA, such as joint vulnerability (20, 22, 41, 42) and genetic factors (20, 44, 45).
Recent interest in the “biological embedding” of health risks during childhood and the role of the hypothalamus-pituitary-adrenal axis (HPA axis) in the development of chronic disease may point to an interesting line of future research. According to the “biological embedding” hypothesis, early exposure to prolonged or highly traumatic events, such as childhood physical abuse, can result in the increased risk of health problems in adulthood, independent of health behaviors (51, 52). Researchers speculate that exposure to high levels of stress at an early age precipitates irregular functioning of the HPA axis, the system that regulates the human stress response through the secretion of cortisol (51), and the heightened sensitivity of the HPA axis to stress throughout the lifespan. Related to this is the concept of allostatic load (53). Allostatic load represents the cumulative physiologic strain placed on organ and tissue systems resulting from chronic exposure to the stress response that, over time, can lead to the onset of disease. Early experiences of childhood abuse and neglect in particular are thought to lead to allostatic load later in life (53).
Clinical research focusing on adult survivors of childhood abuse has demonstrated the presence of irregular patterns of HPA axis functioning in response to stress-inducing events (54–56). In turn, a recent study including a review of the literature of rheumatic diseases and an analysis of primary data on patients with hand, knee, spine, and primary and secondary hip OA concluded that there was evidence to suggest that HPA axis abnormalities were associated with the clinical symptoms of the various sites of OA investigated (57). Such a convergence of evidence that links HPA axis dysfunction both to early experiences of abuse and the clinical symptoms of adult chronic diseases, including OA, suggests a need for future research to explore whether the HPA axis can help explain the association between childhood physical abuse and OA.
There are several important limitations to this study that must be considered when interpreting the results, including the potential for measurement bias, selection bias, reverse causality, and confounding.
The outcome variable was determined by self-report of a health professional's diagnosis of OA that could not be validated. Although reasonable concordance (87%  and 81% ) has been found between self-report of non–joint-specific OA (such as the question in the CCHS) and a physician's diagnosis, future research would benefit from the use of medical examinations or chart reviews. This is particularly relevant, considering that individuals may have radiographic evidence of OA without experiencing any symptoms, and as such, may not independently seek medical care for a diagnosis (60).
Childhood physical abuse was based on retrospective, self-reported data, and could not be validated. Previous research has demonstrated that retrospective reports of childhood abuse are more likely to be underreported as opposed to overreported (61). If misclassifications in abuse status were made, then they were likely false-negatives, thus rendering the results more conservative.
Finally, the only method available for ascertaining the presence of mood disorders as diagnosed by a medical professional was a single-item, self-report question. Self-report of a diagnosis of depression, however, has been shown to be reasonably accurate. A recent cohort study reported a verified diagnosis in 74% of self-reported depression cases (62).
The cross-sectional nature of the data used in this study precludes casual inferences about the direction of association between the variables. Respondents' current health and emotional status may alter their likelihood of disclosing childhood abuse (3, 5). Of note is the finding by McBeth et al (63) of differential recall of adverse childhood events among those with and without current pain. Because OA is characterized by pain, it is possible that some of the relationship between childhood physical abuse and OA can be explained by a discrepancy in abuse reporting.
Furthermore, it has been speculated that the association between childhood physical abuse and arthritis may be due to a common association with pain, based on the finding that childhood abuse is associated with a number of pain-related functional somatic syndromes such as fibromyalgia (10) and irritable bowel syndrome (12). It is possible that those who have experienced abuse are less resilient to pain due to depressive symptoms (64), decreased use of positive coping mechanisms such as social support (65), and lowered self-efficacy (66). The increased perception of pain may in turn escalate the probability of receiving a diagnosis of OA and other pain-related disorders. However, we conducted a supplementary analysis of inflammatory bowel disease (IBD), which is a disease that is also associated with high levels of pain. We did not find a significant association between childhood physical abuse and IBD (data not shown). Future research should determine if an association between abuse and radiographic-diagnosed OA can be found among those without symptoms of pain.
As mentioned earlier, several potential risk factors for OA could not be controlled for in the analysis due to limitations in the CCHS data set and thus may have confounded the results. Factors include joint vulnerability (20, 22, 41, 42), bone mineral density (20), estrogen deficiency (20), and genetic factors (20, 44, 45). Finally, forms of abuse such as sexual and emotional abuse and neglect were not available in the 2005 cycle of the CCHS, but should be controlled for in future research. Although previous research has not found a relationship between childhood neglect and arthritis when controlling for other forms of abuse (7), there are conflicting findings regarding the relationship with childhood sexual abuse (7, 14). As far as we know, the relationship between childhood emotional abuse and OA has yet to be examined in the quantitative research literature.
The strengths of this study include the use of a large, regionally representative sample, with an excellent response rate of 84%. The use of the CCHS allowed for the control of a relatively broad range of risk factors and a large enough sample size to specifically focus on OA.
Future research would benefit from a focus on resilience. In this regard, it is important to highlight the 705 of 854 individuals who, despite experiencing childhood physical abuse, did not report a diagnosis of OA. Resilience among those experiencing childhood abuse has been associated with the perception of parental care, adolescent peer relationships, quality of adult romantic relationships, and personality styles (67). If the relationship between childhood physical abuse and OA is verified in future prospective studies, it may be worth investigating if interventions can be developed that help foster resilience though the development of positive relationships in adolescence and adulthood.
In conclusion, the positive association between childhood physical abuse and OA is unexpectedly robust in spite of controlling for a wide range of risk factors that may mediate the relationship. Further research would benefit from a prospective design that could help identify the potential pathways that link childhood physical abuse to OA. In particular, research on pathways that incorporate psychophysiologic factors such as HPA axis dysfunction may be an interesting line of future study.