Results

EDUCATION AND THE INCIDENCE, RECURRENCE AND PREVALENCE OF BACK PAIN

Table 1 summarises 19 major studies that have examined the association between education and measures of back pain frequency in 23 instances. In 19 instances reported in 16 studies, low educational status was significantly associated with increased prevalence of back pain.4 11 18 29 36 50-60 Only four major studies found no significant association of formal education with the frequency of back pain.58 61-63 There was no apparent pattern of similarity among the negative studies by design, country nor outcome measure. However, three of the four negative studies had sample sizes among the four smallest of all major studies (n=238 to 481 compared with 1135 to 84 572).61-63 The other negative report was that of Hurwitz and Morgenstern (1997), who, with the largest sample size of all major studies (n=84 572), found a significant association between low education and two week period prevalence of disabling back pain but no association with the two week prevalence of non-disabling back pain (no back related restricted activity days).58Only one study with a small sample size (n=154) found an association between low education and a higher frequency of back pain, but it was significant only among women.18 These observations have two implications: (1) given that the association of education with back pain frequency is often observed in large population surveys, one might argue that it is an artefact created by the very high statistical power of these studies. Although this could be true in some instances, this position is challenged by the fact that one small study found the association to be significant and that a quite large survey did not; (2) there seems to be some important differences according to the outcome measures used, although back pain is defined so heterogeneously across studies that it is very difficult to draw consistent lines. As there might be very important differences between various forms of back pain beyond the strict duration of symptoms,64 it is possible that education is associated only with the most severe forms of back pain, but that revealing this relation requires higher statistical power. Thus, it seems possible that heterogeneity in results across studies comes in part from a combination of differences in outcome measures (more or less stringent definitions) and variation in statistical power.

In the only major incidence study, a report on 271 subjects without any prior history of back pain from a probability sample of a large Washington State HMO's enrollees, the odds ratios comparing back pain incidence among subjects with some college and college graduates and subjects with high school or less were 0.59 and 1.01 respectively, both statistically non-significant.61

In positive major studies, although the strength of the association or the differences are sometimes relatively small, the point estimates are generally adjusted for several variables that often include other measures of SES like occupation and income and other variables that could be intermediate between education and outcomes. Such overadjustment would usually bias results toward the null (that is, toward finding no association),72 as would also do random misclassification of education, as it has been demonstrated that most people tend to exaggerate the amount of schooling they have obtained.73 The association between low education and the frequency of back pain observed in most major studies thus seems to be quite robust.

The majority of studies reviewed were cross sectional and examined self reported point, period or lifetime prevalence of back pain. As amount of formal schooling rarely changes after age 20–25 and disabling back pain typically begins after that age, there is probably no important problems with regard to temporal sequence in making cross sectional comparisons. Relying on self reported prevalence of back pain leads to two main concerns: firstly, the association of education with back pain frequency is potentially confounded with differences in recall and reporting of back symptoms across education groups. Prior research using physician validated outcomes suggests this is not an important source of bias. For instance, in the study of Cunningham and Kelsey (1984) low education was found to be associated with higher point prevalence of signs and symptoms of back pain, in agreement with most other prevalence studies that used self report of back pain.4 Secondly, as for chronic episodic diseases prevalence is approximately (incidence) × (average episode duration) × (average number of episodes),74 the association of formal education with back pain prevalence could reflect an effect of education on back pain onset rates, duration, or number of back pain episodes or some combination of them. Using period prevalence of back pain constitutes a particular problem, because this measure fails to distinguish between incident and prevalent cases. No major studies have looked specifically at the association of education with the recurrence of back pain.

Overall, the current limited available evidence points indirectly to a stronger effect of education on the duration and/or recurrence of back pain than to an association with onset.

EDUCATION AND THE OUTCOMES OF BACK PAIN EPISODES

Studies of education as a predictor of the outcomes of back pain episodes are summarised in table 2. In 20 instances reported in the 11 major studies, worse outcomes were associated significantly with low education.4 58 75-83 Negative results were found in only five instances, reported in two studies.76 77 No major studies reported worse outcomes among better educated subjects. The validity of comparing study results is impaired by the heterogeneity of the methods used, including the choice of subjects and the definition of back pain and outcome measures. The following examples illustrate this diversity and the many methodological issues to consider in interpreting their findings.

Badley and Ibañez (1994) used data from the 1986 Canada Health and Activity Limitation Survey (HALS), which included 132 337 non-institutionalised persons aged 16 years and older.81More than half of this sample (54%) had some limitation of activity, 8228 from back pain (in neck or back). The authors found an independent association of low education with activity limitation attributable to back pain (≤ grade 8 versus > secondary education: OR=1.25, 95% confidence intervals: 1.11, 1.40). This study used a crude index of disability (a “yes/no” answer to a single question). It had an extremely large sample size that might have caused most associations to be statistically significant.

In a three month longitudinal study of 179 patients presenting with low back pain (pain below T12) to the outpatient clinic of a hospital,76 seeking compensation decreased and self rated pain improvement increased with increasing education. No association was found with days of pain, improvement in disability, employment status and the number of visits made to a physician. Education was, however, also strongly associated with the language used by the study subjects (English versus Spanish), which predicted disability in multivariate analyses. Although some findings of this study were replicated using NHANES II survey data, the population under study was largely indigent and Hispanic and not representative of the US population, which may explain why some of the dependent variables were not significantly associated with education.

In a 22 year longitudinal study of 391 male employees of a Swedish company, the longest longitudinal study by far, Åstrand and Isacsson (1988) did not find education to be predictive of early retirement with a back diagnosis (disorders in thoracic or lumbar areas). However, they identified a protective effect of education manifested by a relative risk of early retirement for all diagnoses of 0.8 (p=0.004) for each additional year of education. There were few cases of early retirement for back pain in this study, which also had the disadvantage of including only blue collar workers with very similar educational backgrounds, making it difficult to contrast differences between education groups.77

In one study that aimed specifically at explaining the association between education and the consequences of back pain, 1213 HMO enrollees who consulted a primary care physician for back pain in 1989–1990, were followed up for two years. Low education was associated cross sectionally and longitudinally with back related functional limitations measured with a modified version of the Roland-Morris scale. The statistical associations were explained mostly by combinations of psychological (symptoms of somatisation and depression, patient's expectation of continuous pain), behavioural (cigarette smoking, body mass index) and occupational (handling, kneeling and job strength) factors.82 These results have not been replicated yet.

The many reports of a relation of low education with adverse consequences of back pain, even after multivariate adjustment for several variables, suggest that the course of a back pain episode is less favourable among persons with low educational attainment.

EDUCATION AND THE OUTCOMES OF INTERVENTIONS AMONG BACK PAIN PATIENTS

Synthesising the results of studies on the third role of education is much less straightforward. Only two studies among 17 (table 3) met our definition of major studies and both concluded to no association between formal education and outcomes.92 93 Other studies found very heterogeneous results. As the study objectives, populations, interventions, outcome measures, statistical techniques, length of follow up and the way education was considered are very different from one study to another, it is particularly difficult to draw firm conclusions from these results.

At this stage of the natural history of the disease, a reduction in the variability of educational attainment could be advanced as a possible difficulty in detecting a difference on outcomes between groups. If less educated people get more episodes of back pain and are also more likely to shift towards chronicity than people with higher level of education, theoretically there must be relatively few people in the highest groups of educational attainment among severe cases, making less easy to contrast differences between groups of education. This seems to be true in several “other” studies where the baseline educational status of subjects was sufficiently described.94 97 100 101 103-105

Thus, although the limited evidence for an association of educational status with the outcomes of surgical and rehabilitative interventions among back pain patients is negative, this role of education could not be denied yet and needs more research efforts to be understood.

POSSIBLE UNDERLYING MECHANISMS

The fact that many health related events are linked to formal education suggests that something about low SES or other specific traits linked to education increases susceptibility or impairs adaptation to illness. Five hypotheses that could explain this association are outlined below and explored in the context of back pain. These hypotheses are not mutually exclusive.

Behavioural and environmental risk factors

Persons in lower socioeconomic groups are more likely to live in a toxic, hazardous and non-hygienic environment, resulting in a broad array of disease concerns.1 14 73 Besides benefiting from better living conditions, people in the highest groups of education are also more likely to be aware of risks and to adjust their behaviours accordingly.37 Strong evidence exists for the impact on health of cigarette smoking, obesity, alcohol consumption, substance misuse, life changes, chronic stressful events, social support and dietary habits, among others. These factors are often linked with education,9 20 30 37 111-114 and there is some evidence for their adverse effects on the back.19 53 54 87 115-118 For instance, cigarette smoking has been postulated to act on back pain by way of impaired fibrinolysis and reduced nutrition to intervertebral discs, induction of osteoporosis, and mechanical stress to the spine from increased abdominal pressure provoked by coughing.116 119 Obesity is considered to affect the spine by conferring mechanical disadvantages.120 Given the current state of knowledge, however, it is difficult to propose any particular combination of these risk factors as a sufficient explanation for the association of education with back pain.

The mechanisms linking psychological factors to back pain are still obscure; anxious or depressed patients may have greater awareness of pain or they may have more difficulty coping with it and consult health care providers earlier. Life stress could also be a precursor of back pain and act directly by an unknown mechanism.115 117 121 The work of Pincus and Callahan, who noticed that low formal education was consistently associated with increased morbidity and mortality in rheumatoid arthritis, has lead to identify other important psychological variables (helplessness and hostility) that might become the focus of future interventions.6 7 12 122-126

Occupational factors

It is widely believed that a hazardous work environment and physically demanding jobs play an important part in the aetiology of back pain.65 84 127 128 People with low education and low paying jobs are more likely to work in physically demanding jobs and tasks involving stresses on the spine. After an episode of back pain, they may have to wait longer to resume a demanding job than persons with sedentary occupations.4 8 20 41 49 75 84 129 130 They may be less satisfied with their jobs and less motivated to continue working or to work as hard after an episode of back pain.5 76Conversely, persons with low education may have poorer sick leave benefits or fear loss of their jobs because of back pain, and be more prone to return to a physically demanding job before the back injury has resolved, risking reinjury or prolongation of symptoms.51 This explanation was not sufficient, however, to account for differences in outcomes across educational groups in several studies.1 18 19 60 75 100 131 For instance, Leigh and Sheetz (1989) found a significant association between back pain period prevalence and low education even after adjustment for other risk factors, including occupation, physical demands and repetitiveness of work.54 Although these results suggest that occupational factors alone are not sufficient to explain the association, the findings may have resulted from imperfect classification of occupation and physical demands of the job, as these are difficult to measure and classify.132

Compromised “health stock”

In the perspective of Grossman's application of consumer theory to health care, a person's health is viewed as capital (“health stock”) that is inherited and reduced by depreciation (biological aging) and finally by death, when the capital falls below a certain minimal level.133 There is some evidence that socioeconomic circumstances during childhood may influence adult health by compromising this “health stock” at the beginning of adult life.20 This hypothesis is compatible with the “pain prone disorder” suggested in the literature on back pain.134 In a study on premorbid social adjustment of chronic back pain sufferers, the subjects were found to have limited formal education.135 “Most had experienced unmet dependency needs...and many had familial models for pain and disability present in their early life experience”. The study of Porter and Oakshot,136 who observed a higher prevalence of small adult vertebral canal among subjects with lower education (as a marker of early impairment of growth) also supports this hypothesis.

Health services access and utilisation

Differential access to or use of health services across groups of education, or differences in the benefits obtained from the use of health care are other possible explanations to account for the relation between education and health.1 4 12 14 24 People of lower SES do less well in the health care system, although the reasons for this are unknown.24 One hypothesis is that low education would impair communication and trust with health professionals, and make it more difficult to use services effectively.24 For back pain, it is possible that patients with lower education have less access to some specialised interventions, wait longer before consulting or have lower compliance with health professionals' recommendations. However, several reports found that people with back pain in low education groups more often consulted a physician and were more frequently hospitalised.1 9 20 37 50 In a national US survey, the relation between education and duration of back related disability was unexplained by medical care utilisation. There was a significant trend toward more frequent hospitalisation among those with less education. The proportion of subjects who had sought medical care or had undergone back surgery was similar in each educational group.75These results are in agreement with the conclusion of a recent paper by Pincus et al (1998) on access to care as a determinant of health, in which the authors demonstrate, on the basis of many international studies (several of them from countries with longstanding universal access to medical care), that widening disparities in health according to SES are not explained mainly by access to care.137 The association between education and back pain thus seems unlikely to be explained by differences in the use of health services.1 11 14 19 20 81 122

Adaptation to stressful events

There is evidence for the joint effect of stressful events and an altered social environment as an explanation for the relation between SES and health.2 4 14 19 84 138 Such an effect would be explained as an opportunity for pathogenic agents (stressful events) to affect health when host resistance is lowered by way of endocrine or immunological changes. There are two major circumstances in which the social environment might impair host resistance: (1) when the person is unprepared to function in their social setting and fails to receive adequate feedback that their problem solving efforts are effective and (2) when there is a deficiency in the social support provided by the primary group. This mechanism is supported by studies on animals.139 In humans, this hypothesis may explain the increased risk of mortality from myocardial infarction among men with low educational achievement. Men who are socially isolated and experience important changes in their life are more likely to die than others.34 35 In view of the current dominant theory of pain140 such an effect may operate on back pain by influencing the central mechanisms that modulate the transmission of nociceptive messages to the brain, facilitating pain that persists after the original injury, leading to chronicity. Obesity, cigarette smoking, alcoholism, type A behaviour and neurotic diseases may all be considered as responses to stress.1 This hypothesis, tested in only one back pain study to date, has not been supported.82

Recently, Kubzansky et al (1999) found a strong association between education and allostatic load, a measure of biological risk across several systems, that was explained by a measure of “hostility”, a broad concept associated with “a tendency to devalue the worth and motives of others, to expect that others are likely to be doing wrong, to view one's self as being in opposite toward others, and to desire to inflict harm or see others harmed”. Hostility would develop as a response pattern to difficult life circumstances.141 The authors concluded that lower levels of education and greater hostility are associated with greater “wear and tear” on the body and that the relation between education and allostatic load may be mediated by hostility. This mechanism needs to be studied further.

SUGGESTIONS FOR FUTURE RESEARCH

Frequent methodological problems that affected the results of studies reviewed for this paper include: reliance on cross sectional study design and prevalence of outcomes; inconsistency in defining “back pain”, education, interventions, outcome measures and length of follow up; selection bias; limited variability in educational status, making it difficult to find differences between groups based on educational status; and low statistical power because of a small number of outcome events. The diversity of the populations studied may also have contributed to variation in results. Many studies limited their analyses to bivariate relations, while others used multivariate models.

As back pain investigators are still looking for strong and modifiable risk factors, many studies examine as many variables as possible. Although such “fishing expeditions” are necessary as a first step toward the identification of major risk factors, hypothesis testing research must focus on a limited number of variables and develop further knowledge of their impact on the disease. Education is a good candidate for this purpose. Studies on back pain should include it as a risk factor on its own. Measuring education as a continuous variable (years of schooling) should be preferred to categorical classifications, and a clear description of the way it is considered in the analyses should be provided. Multivariate analysis methods allowing adjustment for confounding factors should be used to extend bivariate and correlation analyses. Reporting statistical power or confidence limits in studies that did not find statistically significant associations would also be helpful. Determining if education is related to both incidence/recurrence of back pain (or back related disability) and duration of episodes would certainly be an important contribution to the field. Clarifying the role of sex as an effect modifier would also be valuable, as the literature is very inconsistent on this point.

Few reports among the 64 reviewed for this paper were developed specifically to examine the relation of sociodemographic variables with the outcomes. More research efforts geared toward testing specific hypotheses are needed. Prospective studies on subjects with no prior history of back pain seems to be the most promising study design to this end, and a sine qua non condition to answer some questions, particularly those dealing with the directionality of associations.

Another key question concerns exposure measurement; many suspected risk factors are complicated constructs that require care in their measurement, a need more likely to be met in prospective studies. This is certainly the case for psychological factors and job requirements. Studies of the effect of interactions between stress and coping strategies are also needed. Heterogeneity in the definition of study subjects, back pain, interventions, outcomes and length of follow up was a constant concern during our review. Although it is difficult to reach a consensus to define these parameters, the field would certainly benefit from a restriction to some commonly used definitions.

The association between education and back pain has an important immediate implication: adjusting for educational status in multivariate analyses could lead to underestimation of the association between some independent variable education is associated with (for instance, cigarette smoking) and back pain outcomes. It thus seems advisable to compare results obtained with and without adjustment on educational status in aetiological research on back pain.

Conclusion

Scientific evidence supports the hypothesis that less well educated people are more likely to be affected by disabling back pain. However, the evidence is stronger for an effect of education on the duration and/or recurrence of back pain episodes than for an effect of education on the incidence of back pain. Educational status may be a “marker” for other factors involved in the aetiology or natural history of the disease, acting as a proxy for SES. Back pain is still an enormous public health problem in all Western societies.44 48 Despite extensive research work, our understanding of the aetiology and natural history of this condition is still very limited. While the contribution of some risk factors has been highlighted, the search toward the development of a model that accounts for occupational, sociodemographic, biological and psychological characteristics must be pursued. Few traits have demonstrated as a strong and consistent association with back pain as education. It may well be a pathway to a better understanding of this illness and certainly deserves scientific attention.