Epidemiology of Pain

Occurrence

There are two main measures of disease occurrence: incidence and prevalence. The incidence rate is a measure of disease onset: the number of new cases of disease in an at-risk population within a specified period. Prevalence is a measure of disease state: the number of existing cases of disease in a population expressed as a proportion of the total population, either at a particular point in time (point prevalence) or between two specific points in time (period prevalence). Prevalence can inform us about the burden of disease and can therefore aid in planning health services and allocating resources. It is influenced by both the incidence of a disease and its persistence. In contrast, it is the study of incidence (i.e., the study of disease onset) that is most useful in determining the etiology.

In diseases of sudden onset or those that require immediate health service consultation, incidence is relatively easy to measure. However, identifying incident cases of pain is more difficult. The precise time of onset is not always easily recalled, and symptom onset does not always lead to a health care consultation. In fact, some have shown that only one in seven individuals with a new episode of low back pain (LBP) consults a family practitioner about this pain ( ). Furthermore, most people will have experienced episodes of pain from an early age. Consequently, the epidemiologist usually studies episode incidence (i.e., the onset of a new episode of pain) rather than the first-ever episode of pain (first-ever incidence). However, what constitutes an episode of pain? In the occupational setting, for example, the physician, the patient, and the employer may use different criteria to define an episode. Different definitions of an episode of LBP have been proposed for use in the population, family practice, and occupational settings ( ), and more recently, based on a Delphi consensus approach, others have proposed two definitions of LBP: a “minimal” definition covering site, time frame, and severity of pain and an “optimal” definition capturing additional information such as the duration of symptoms ( ). These definitions, however, are arbitrary and not necessarily any better (or worse) than any other. Instead, the strength of such standards lies in aiding comparability between studies.

The difficulty in measuring the onset of pain is further compounded by the fact that pain often has a natural history characterized by a pattern of relapse and remission. Consequently, identification of all incident cases of pain over a defined period in a study population can, in practice, be difficult. The researcher frequently has to instead study new prevalent cases—that is, the prevalence of pain in a population known to have been free of pain at a previous point in time.

Pain is a subjective phenomenon with no available “gold standard” clinical tools, and the researcher relies on self-reported measures of pain. Indeed, objective evidence of “abnormality” correlates poorly with pain reporting, as evidenced in the lower part of the back; most anatomical or pathological changes that have been associated with the occurrence of LBP have also been demonstrated in the symptom-free population and only poorly predict onset of pain ( ). These findings are reflected in the hip and shoulder pain literature. A consequence of relying on self-reported pain is that rather than studying the epidemiology of pain per se, the researcher is actually studying the epidemiology of the reporting of pain. However, given the subjectivity of pain and involvement of health services, it could be argued that this is indeed appropriate.

Etiology

Overall, pain may be considered an inevitable consequence of living. Indeed, in a large longitudinal study, fewer than one in five individuals consistently reported “no pain in the previous month” at each of three consecutive surveys over a 4-year period ( ). Thus, from a public health viewpoint, it is of little interest to consider all pain episodes. Instead, it is more useful to concentrate on pain that is chronic and disabling, and it is the evidence related to such conditions that will predominate in this chapter. In summarizing the epidemiological evidence, consideration has also been given to study design, the study sample, the validity of the methods used, and sample size—it is important that any study have sufficient power (i.e., be large enough) to detect an association between exposure and outcome, if one exists.

Various epidemiological study designs may be used to investigate the etiology of pain. First is an ecological study in which the occurrence of pain is compared between two or more population groups or subgroups. For example, this may involve comparing the prevalence of LBP among workers in diverse occupations. Although this type of study can give useful leads to etiology, its main drawback is the lack of information on potential confounders (i.e., known risk factors for LBP that vary between persons in different occupations). For example, some occupations will have a workforce consisting primarily of older women—such workers are known to have a high risk for LBP (irrespective of what job they do).

Observations at the population level may not hold at the individual level. For this reason it is more useful to conduct studies that collect information on the individual, and the two most common methods are case–control and cohort studies. In the former, persons with pain are compared with persons without (often, but not necessarily matched for important confounding factors) and are studied with respect to previous exposure, information on which is collected retrospectively. The important issue with regard to this methodology is to ensure comparability of information because cases and controls may recall exposures differently. In a study examining chronic widespread pain in adulthood, recall of adverse childhood exposures, such as hospitalizations or surgery, was found to differ between persons with pain (cases) and persons without (controls) ( ). Such differential recall can lead to biased study results and may indicate, erroneously, that an association exists between pain and a putative risk factor. Use of objective sources of information is desirable, if available, to help overcome this problem.

In case–control studies, the temporal relationship between risk factors and the onset of symptoms is also unclear. If persons with chronic pain are found to have higher levels of distress than control subjects, it is not possible to ascertain whether the distress preceded (and may therefore have increased the risk for) the pain or whether it was a consequence of experiencing pain over a prolonged period. In contrast, in a cohort study, rather than selecting individuals according to pain status, one selects subjects who are pain free and groups them according to their risk factor status (e.g., with and without distress). They are then monitored over time to examine whether the risk factor or factors predict subsequent disease or onset of symptoms. Thus, if any association is observed, the risk factor is known to precede disease onset and the issue then becomes whether the relationship is causal.

Studies may be undertaken on different populations from, for example, the general population, workplaces, primary care, or patients in specialist settings, and each has advantages and disadvantages. For instance, if the researcher identifies subjects from primary care consultation, it is not possible to distinguish factors related primarily to the symptoms from those related to consultation behavior. For many pain syndromes, those seeking a health service consultation may have higher levels of distress (than persons in the population without symptoms) and may be likely to have different perceptions of illness. Thus, the researcher cannot distinguish whether those with distress and with certain illness perceptions are at increased risk for the development of pain or whether those in whom pain develops, those with distress, or those with a certain perception of their symptoms are more likely to seek consultation. It is therefore important to consider the generalizability of research findings. In this respect, studies using random samples from the general population offer the greatest potential and are feasible when the condition under study is common.

Finally, one must consider the validity of the measurement of potential risk factors. Some studies measure exposures based on self-report, whereas others provide external validation of exposures from other sources, such as medical or occupational records. Poor validity resulting from random error will lead to attenuation of the relationship between a potential risk factor and symptoms and will therefore make a relationship that exists harder to identify. Alternatively, if the validity of exposure measurement differs, for example, between persons with or without symptoms, it is impossible to predict the effect on the observed association.

Introduction

A large Internet-based questionnaire study in the United States (US) collected data on 27,035 individuals ( ), 9326 of whom reported chronic pain, defined as pain of at least 6 months’ duration, for a point prevalence of 30.7% (95% confidence interval [CI], 29.8–31.7%), weighted to account for differences in a number of demographic characteristics between the sample and the U.S. standard population. The prevalence of chronic pain in men and women separately for different age groups is shown in Figure 16-1 . Chronic pain is shown clearly to increase with age, and at every age it is more common in women than in men.

In 2006, Beivik and colleagues surveyed 46,394 individuals from 15 European countries plus Israel. They found that 19% of the respondents had pain for at least 6 months and had experienced pain in the last month and several times during the last week. Despite minor differences, the prevalence, severity, and impact of pain were broadly similar across all 16 countries. The prevalence of chronic pain varied with anatomical location, as shown in Figure 16-2 , and ranged from 5% in the upper part of the back to 18% in the lower part of the back. Additionally, back pain of unspecified location was reported by 24% of the respondents. However, caution must be observed when interpreting data on specific regional pain syndromes. In addition to true underlying variations in the prevalence of regional pain by anatomical site, study design and methodology can have a large impact on the estimated prevalence. For example, estimated that differences in the case definition of migraine explained more than 36% of the variance in the prevalence estimates, and a further 30% could be explained by different age–sex strata in the study samples.

In the next part of this chapter we do not intend to provide a comprehensive review of all the literature on the epidemiology of pain. Rather, we review, individually, information on the descriptive epidemiology (occurrence) and analytical epidemiology (etiology) of the major regional pain syndromes and widespread body pain/fibromyalgia (FM) and focus, when possible, on potentially modifiable risk factors. In addition, we consider recent advances in epidemiological studies of pain and what biological and genetic parameters can add to what is known about environmental risk factors, and we consider to what extent different pain syndromes have common aspects in their etiology.