SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. SUBJECTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Objective

To determine whether reported widespread body pain is related to an increased incidence of cancer and/or reduced survival from cancer, since our previous population surveys have demonstrated a relationship between widespread body pain and a subsequent 2-fold increase in mortality from cancer over an 8-year period.

Methods

A total of 6,565 subjects in Northwest England participated in 2 health surveys during 1991–1992. The subjects were classified according to their reported pain status (no pain, regional pain, and widespread pain), and were subsequently followed up prospectively until December 31, 1999. During followup, information was collected on incidence of cancer and survival rates among those developing cancer. Associations between the original pain status and development of cancer and cancer survival were expressed as the incidence rate ratio (IRR) and mortality rate ratio (MRR), respectively. All analyses were adjusted for age, sex, and study location, the latter being a proxy measure of socioeconomic status.

Results

Among the study population, 6,331 had never been diagnosed with cancer at the time of participation in the survey. Of these subjects, 956 (15%) were classified as having widespread pain, 3,061 (48%) as having regional pain, and 2,314 (37%) as having no pain. There were a total of 395 first malignancies recorded during followup. In comparison with subjects reporting no pain, those with regional pain (IRR 1.19, 95% confidence interval [95% CI] 0.94–1.50) and widespread pain (IRR 1.61, 95% CI 1.21–2.13) experienced an excess incidence of cancer during the followup period. The increased incidence among subjects previously reporting widespread pain was related, most strongly, to breast cancer (IRR 3.67, 95% CI 1.39–9.68), but there were also cancers of the prostate (IRR 3.46, 95% CI 1.25–9.59), large bowel (IRR 2.35, 95% CI 0.96–5.77), and lung (IRR 2.04, 95% CI 0.96–4.34). Subjects reporting widespread pain who subsequently developed cancer, in comparison with those previously reporting no pain, had an increased risk of death (MRR 1.82, 95% CI 1.18–2.80). This decreased survival was highest among subjects with cancers of the breast and prostate, although the effects on site-specific survival were nonsignificant.

Conclusion

This study has demonstrated that widespread pain reported in population surveys is associated with a substantial subsequent increased incidence of cancer and reduced cancer survival. At present there are no satisfactory biologic explanations for this observation, although several possible leads have been identified.

Widespread body pain is the cardinal symptom of the fibromyalgia syndrome. It is commonly reported in the general population, with a 1-month period prevalence of ∼9–10% (1–4). Symptom onset is predicted by high levels of psychological distress, features of somatization, and comorbidities such as fatigue (5). Onset of pain is also strongly associated with prior aspects of illness behavior such as a propensity to seek care. Studies of clinic patients with fibromyalgia have suggested that symptoms resolve infrequently (6, 7), and long-term prospective studies in the population have similarly shown that symptoms are generally persistent (8).

The symptom of widespread pain may reflect underlying organic disease. However, studies of patients with fibromyalgia have found an organic basis for symptoms in only a very small proportion of subjects. Nevertheless, a previous study conducted by us did demonstrate an increased subsequent mortality in subjects who reported widespread pain in community surveys (9). This was almost all related to a significant excess mortality from cancer (and persisted when subjects who already had a cancer diagnosis were removed from the analyses). This result was intriguing, since we did not have an a priori hypothesis linking cancer mortality and widespread pain.

If the relationship of widespread pain were solely with mortality, through reduced survival, then possible explanations include that frequent presentation with either somatic symptoms or depression as a comorbidity may result in delays in diagnosis, or that adverse psychological factors per se may result in reduced survival. In contrast, a relationship between widespread pain and incidence of cancer would suggest an underlying biologic mechanism linking the two conditions. Therefore, in order to refine hypotheses about the mechanism linking widespread pain with cancer-related death, further data on all incident cancers were collected on the same cohort to determine whether the report of widespread body pain as part of a population-based questionnaire study was associated with increased subsequent cancer incidence and/or reduced survival among those subjects subsequently diagnosed with cancer.

SUBJECTS AND METHODS

  1. Top of page
  2. Abstract
  3. SUBJECTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Study cohort.

The design was a population-based, prospective cohort study which has been described previously (9). Subjects were individuals who had participated in 2 population surveys conducted in Northwest England between 1991 and 1992.

Study 1 was conducted during 1992 in a residential area of the city of Manchester. Study 2 was conducted during 1991 in 2 areas: a commuting suburb of the city and a semirural area of a neighboring county. These 2 study populations are distinct with respect to socioeconomic conditions, although within study areas, conditions are very similar. These individual studies have been described in detail previously (1, 10). They were conducted using very similar protocols. The sampling frames in each study were the age/sex registers of 2 local general practices. In the United Kingdom, more than 95% of the population is registered with a single general practitioner, and therefore this represents a convenient population-sampling frame. From both registers used in study 1, a simple random sample of subjects ages 18–75 years was obtained, while from both registers in study 2, an age-stratified random sample of subjects ages 18–85 years was obtained. A questionnaire was mailed to all selected subjects in both studies (with followup reminders to nonresponders), inviting them to participate in a health survey. The surveys primarily pertained to pain symptoms and gathered information on potential etiologic factors. In total, the studies involved 6,565 subjects, with participation rates in studies 1 and 2 of 65% and 75%, respectively.

In both studies, participants were asked, “During the past month, have you experienced pain lasting at least one day?” If subjects gave a positive response, they were invited to indicate the sites of pain on blank body diagrams. This allowed subjects to be classified into 3 groups according to their pain status: widespread pain, regional pain, or no pain. Widespread pain was defined according to the American College of Rheumatology (ACR) criteria for fibromyalgia; this requires the presence of pain in the axial skeleton, in addition to pain in 2 contralateral body quadrants (11). Subjects who reported pain but who did not meet this ACR definition were classified as having regional pain. In addition, study 1, which contributed 65% of all study subjects, collected information on current smoking status.

Details on the subjects were sent to the Office for National Statistics, to be identified on the National Health Service (NHS) Central Register. The NHS Central Register contains a record of all persons in the United Kingdom who are or have been registered with a general practitioner since 1991, and, using official records, contains information on incident cancers and vital status. For each subject, it was determined whether they had been diagnosed with a cancer at the time of the original survey (and were therefore excluded from the followup cohort), and for the remainder of subjects, it was determined whether they had been diagnosed with a cancer during the followup period. If a subject was registered as having died, the Office for National Statistics provided information on the date and underlying cause of death, coded according to the tenth revision of the International Classification of Diseases (12). Subjects were followed up from the date of participation in the survey until either December 31, 1999 (i.e., 9 years of followup), date of diagnosis of incident cancer, or censoring due to death or emigration from the United Kingdom.

The study gained approval from the Office for National Statistics and from the ethics committee of the University of Manchester.

Statistical analysis.

The person-years at risk of being diagnosed with cancer was calculated for each study subject, from the date of the original survey until either December 31, 1999, date of diagnosis of cancer, or date of censoring. This allowed the cancer incidence rate in each of the 3 pain groups (widespread pain, regional pain, no pain) to be determined. Thereafter, Cox proportional hazards modeling was used to account for the possible confounding effects of age (in 5-year age groups), sex, and study location, the latter being a proxy measure for socioeconomic status. The results on the incidence of all cancers, as well as cancers at specific sites, were calculated by comparing the widespread pain group with the no pain group (as reference), with values expressed as the incidence rate ratio (IRR) and 95% confidence interval (95% CI).

Among those developing cancer, the person-years at risk of dying from cancer was calculated from the date of diagnosis until December 31, 1999 or the date of death. For those subjects who died from an unrelated cause, followup was censored at the date of death. Cox proportional hazards modeling was again used, taking into account the possible confounding effects of age (in 5-year age groups), sex, and study location. The results on deaths related to cancer (all cancers as well as site-specific cancers) were calculated by comparing the widespread pain group with the no pain (reference) group, with values expressed as the mortality rate ratio (MRR) and 95% CI.

RESULTS

  1. Top of page
  2. Abstract
  3. SUBJECTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Of the 6,565 study subjects participating in the baseline surveys, 6,331 had not been diagnosed with cancer at that time, and these were the subjects who were eligible for the current study. Of these subjects, 15% (n = 956) were classified as having widespread pain, 48% (n = 3,061) as having regional pain, and 37% (n = 2,314) as having no pain (Table 1). Subjects with widespread pain (median age 55 years, 66% female) were older and more likely to be female than those with regional pain (median age 49 years, 58% female) or those with no pain (median age 42 years, 54% female).

Table 1. Characteristics of the study groups*
CharacteristicStudy 1 (n = 2,221)Study 2 (n = 4,110)
  • *

    All values are the number (%) of subjects.

Pain status  
 No pain924 (41.6)1,390 (33.8)
 Other pain929 (41.8)2,132 (51.9)
 Widespread pain368 (16.6)588 (14.3)
Age group, years  
 18–33450 (20.3)1,162 (28.3)
 34–47524 (23.6)1,040 (25.3)
 48–64645 (29.0)971 (23.6)
 65–85602 (27.1)937 (22.8)
Sex, women1,279 (57.6)2,367 (57.6)

In total, 395 malignant cancers were recorded during followup, of which 107 occurred in subjects with no pain, 198 in those who had reported regional pain, and 90 in those with widespread pain. The incidence rate was lowest in subjects who originally reported no pain (5.7 per 1,000 person-years) and increased among those reporting regional pain (8.2 per 1,000 person-years) and those reporting widespread pain (11.9 per 1,000 person-years). The incidence rates in both the regional pain group (IRR 1.19, 95% CI 0.94–1.50) and widespread pain group (IRR 1.61, 95% CI 1.21–2.13) were elevated, and the latter rate was significantly elevated, after adjustment for age group, sex, and study location (Table 2). The excess risks associated with new-onset widespread pain were evident in both the men and the women (IRRs 1.89 and 1.50, respectively), and the excess risk persisted after additional adjustment, in study 1, for smoking status (IRRs 1.42 unadjusted and 1.39 adjusted). The excess risk of developing cancer in the group with widespread pain (in comparison with those with no pain) also persisted throughout the followup period: year 1 (baseline) IRR 1.44, 95% CI 0.74–2.79, years 2–6 IRR 1.69 95% CI 1.09–2.61, and years 7–9 IRR 1.37, 95% CI 0.93–2.03.

Table 2. Pain status at baseline and subsequent incident cancer during 9-year followup
CharacteristicPerson-years of followupNumber of incident cancersIncident rate ratio (95% CI)*
  • *

    Derived from multivariate model; risks are mutually adjusted and in addition adjusted for study location. 95% CI = 95% confidence interval.

Pain status   
 No pain18,7461071
 Other pain24,2801981.19 (0.94–1.50)
 Widespread pain7,548901.61 (1.21–2.13)
Age group (5-year bands)1.41 (1.36–1.47)
Sex   
 Men21,3902011
 Women29,1841940.66 (0.54–0.80)

In addition to cancer incidence (overall), the relationship between widespread pain and common specific sites of cancer was examined. There was a doubling of the risk of cancers of the lung (incidence rate among men 1.6 per 1,000 person-years, and among women 0.7 per 1,000 person-years; IRR 2.04, 95% CI 0.96–4.34) and large bowel (incidence rate among men 0.9 per 1,000 person-years, and among women 0.6 per 1,000 person-years; IRR 2.35, 95% CI 0.96–5.77). The greatest risk was associated with the onset of female breast cancer (IRR 3.67, 95% CI 1.39–9.68) and prostate cancer (IRR 3.46, 95% CI 1.25–9.59). For all other cancers combined, there was only a small, and not statistically significant, increase in risk (Table 3).

Table 3. Widespread pain at baseline and subsequent incident cancer during 9-year followup by site-specific analysis*
Cancer siteICD-10 codesNumber of incident cancersIncident rate ratio (95% CI)
  • *

    Participants classified as having “no pain” form the reference group; results are adjusted for age, sex, and study location. ICD-10 = tenth revision of the International Classification of Diseases; 95% CI = 95% confidence interval.

All malignanciesC00–C973951.64 (1.23–2.19)
All malignancies excluding nonmelanoma skin cancerC00–C97 excluding C443101.83 (1.32–2.54)
Female breastC50413.67 (1.39–9.68)
ProstateC61293.46 (1.25–9.59)
LungC34552.04 (0.96–4.34)
Gastric and small intestineC15–C17161.59 (0.35–7.24)
Large bowelC18–C20392.35 (0.96–5.77)
All other cancersC00–C97 excluding C15–C20, C34, C44, C50, C611301.18 (0.70–2.00)

Among the 90 and 107 subjects reporting widespread pain and no pain, respectively, on the original survey who went on to develop cancer during followup, investigation was made of their survival experience (Table 4). In comparison with those originally reporting no pain, subjects with widespread pain had a significantly reduced survival when they developed cancer (MRR 1.82, 95% CI 1.18–2.80). In relation to site-specific cancers, there was a modest, but not significant, reduced survival among those subjects who developed cancer of the large bowel (MRR 1.36, 95% CI 0.42–4.37) and lung cancer (MRR 1.66, 95% CI 0.72–3.84). The greatest effect on survival, however, was once again observed among those female subjects with breast cancer (MRR 2.45, 95% CI 0.85–7.05) and those who developed prostate cancer (MRR 2.81, 95% CI 0.28–27.94), although the limited number of deaths from these cancers resulted in wide (and nonsignificant) confidence intervals. For all other cancers, there was no relationship between cancer survival and pain status as reported on the original survey.

Table 4. Widespread pain at baseline and rate of survival after development of cancer*
Cancer siteICD-10 codesNumber of deathsMortality rate ratio (95% CI)
  • *

    Participants classified as having “no pain” form the reference group; results are adjusted for age, sex, and study location. See Table 3 for definitions.

All malignanciesC00–C97861.82 (1.18–2.80)
All malignancies excluding nonmelanoma skin cancerC00–C97 excluding C44801.90 (1.21–2.96)
Female breastC50102.45 (0.85–7.05)
ProstateC6152.81 (0.28–27.94)
LungC34191.66 (0.72–3.84)
Gastric and small intestineC15–C1761.78 (0.34–9.29)
Large bowelC18–C20121.36 (0.42–4.37)
All other cancersC00–C97 excluding C15–C20, C34, C44, C50, C61280.85 (0.38–1.89)

DISCUSSION

  1. Top of page
  2. Abstract
  3. SUBJECTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

This study has shown that subjects who reported widespread pain in 2 population surveys (but had not been diagnosed with cancer) had an increased incidence of cancer over the subsequent 9 years. Among those subjects who developed cancer during the followup period, those who originally reported widespread pain had a reduced survival in comparison with those who reported no pain. The strongest relationship, with regard to both incidence and survival (and albeit with wide confidence intervals), was with hormone-related cancers, that is, prostate and female breast cancers, while there were smaller effects for gastrointestinal and lung cancers.

Nevertheless, how generalizable are the results from this population? First, there was no selection of the participating subjects apart from their decision on whether to take part in the original population pain surveys. The prevalence of regional and widespread pain in these surveys is very similar to those reported by other population surveys in the United Kingdom, elsewhere in Europe, and in the United States (2, 3, 5). Second, a comparison between cancer incidence in this study population and the whole population in Northwest England gives a standardized incidence ratio of 1.005 (comparison made using data from Northwestern England and Merseyside and Cheshire cancer registries) (13). Thus, the population participating in the surveys has an almost identical future cancer experience to that of the population of the whole region in which they are resident. It is highly improbable, therefore, that those who chose not to take part, whose overall cancer experience would have been similar to that of the participants, would exhibit a different relationship between their original pain status (when free of a cancer diagnosis) and subsequent cancer incidence and/or cancer survival.

Pain was reported at a single point in time (referring to the preceding month). In study 2, information was available on the duration of pain reported, and among those who reported widespread pain, 83% satisfied the International Association for the Study of Pain definition of “chronic pain.” Therefore, in the majority of subjects with widespread pain, these were not transient symptoms. Furthermore, a followup study of subjects in study 2 with chronic widespread pain in the community found that only 15% were free of pain 7 years later (8). Therefore, it is likely that most of the subjects in our study will have had chronic pain and have experienced pain throughout (although not necessarily continuously) the followup period. Nevertheless, some subjects will have been misclassified according to their usual pain state, and others will have changed pain state during followup. Such misclassification, however, will result in an underestimate of the strength of the association. Errors in coding of cancer and vital status on the NHS Central Register are very rare. Nevertheless, there are delays in the recording of incident cancers (and deaths) and there may have been some subjects who have been diagnosed with cancer (or died) who are not recorded as such on the Central Register. The effect of such “delays” would be random across pain groups and would result in an underestimation of any true association with onset and/or survival.

Low social class is associated both with higher risk of cancer and with reduced survival (14); however, it was unlikely to be a confounding factor in the observed relationships. Although there have been reported associations between aspects of socioeconomic status and fibromyalgia, the research in that area is limited (15, 16). Indeed, there is little evidence that pain reporting, particularly widespread pain, varies by social class. Nevertheless, there were differences in social status between the areas in which the studies were conducted, and we adjusted for study area in the analysis. Using study area as a measure of socioeconomic status is likely to reflect other measures that are primarily based on area of residence. Within each study, because the study population was sampled from persons registered with selected general practices, the variation in measures of social status (particularly if based on area of residence) would be small. Indeed, we have examined our data and are confident that, for those persons within an area who were registered with the same general practice, we have identified homogeneous groups with respect to socioeconomic status. Our method is limited in that we are unable to assess the relationship with specific aspects of socioeconomic status, e.g., household income.

In study 1, information was available on respondents' smoking habits. When adjustment was made for this in the analysis of incident cancers, there was almost no change in the observed results. The fact that there was still an excess of incident lung cancers after adjustment (for current smoking status) may, however, indicate that there was some residual confounding by smoking. If this were the case, it would not pertain to the main results of an excess risk of female breast and prostate cancers, since neither are related to smoking (17, 18).

It is important to set these results in context. Our original study of widespread pain and mortality was conducted to determine long-term outcome (9). We did not have an a priori hypothesis of an increased risk of cancer-related death. However, understanding whether subjects with widespread pain are at an increased risk of cancer onset and/or reduced cancer survival is important in trying to understand the mechanism underlying the association. It appears that the relationship is with both risk and survival. This allows us to exclude one hypothesis about the relationship with cancer mortality: it cannot be explained by “behavioral factors,” e.g., general practitioners potentially being slower to investigate new symptoms in such patients, given that the patients are likely to have a history of consultation for a variety of somatic symptoms. Indeed, one could hypothesize that such frequent consultation would result in potentially serious symptoms being detected at an earlier stage. If this were true, and in a worse-case scenario it did not result in an improved survival, there would still be an observed increase in survival due to the diagnosis of the cancer being advanced in time. However, this study has demonstrated that survival is poorer in subjects with prior widespread pain.

In contrast, biologic factors seem the most likely explanation of an association with both cancer onset and survival, but the evidence for any single such factor is weak. Subjects with widespread pain have high levels of psychological distress and mental disorder (19). However, overall, the evidence does not support an association between these factors and increased cancer onset or reduced survival (20). A recent community study of 3,000 persons who were followed up for 13 years in East Baltimore found that, overall, there was no increased risk of cancer among people with major depression. The only exception was that, among women with major depression, there was a significant increased risk of developing breast cancer (21). However, a nationwide-based study in Denmark involving 67,000 women admitted to hospital with an affective or neurotic disorder failed to confirm this association; their risk of breast cancer was similar to that of the general population (22).

As a group, patients with fibromyalgia (who have widespread body pain as the cardinal feature) have been demonstrated to have abnormalities of their hypothalamic-pituitary-adrenal axis (23). Although the precise nature of the dysfunction has not been consistent between studies, the most common finding is of an axis that is hypoactive but is hyperresponsive to challenge. The resultant changes in cortisol levels (which itself has been demonstrated to influence tumor progression [24,25]) may result in the promotion of hormone-sensitive cancers such as breast cancer. Contradicting this hypothesis, however, is the fact that depression, which is associated with hyperactivity of the axis and hypercortisolemia, is not associated with an increased incidence of cancer (not even hormone-sensitive cancers).

A further potential link between widespread pain and cancer is the influence of insulin-like growth factor 1 (IGF-1). High levels of IGF-1 and the binding protein IGF-B3 have been shown to increase the risk of cancers (female breast, prostate) (26, 27), which are in excess in this study. However, although fibromyalgia has also been consistently linked to abnormalities of IGF-1, the levels found have been lower than in subjects without fibromyalgia. Although the study subjects in our widespread pain group did not all have fibromyalgia, this would have suggested that such subjects were potentially at lower risk of developing the very cancers that have been found in excess. Despite the inconsistent pattern of IGF-1 abnormality, this may still provide a useful area for further research.

Clinic patients with fibromyalgia are characterized by high levels of disability. If this were true more generally of the subjects in our study with widespread pain, and if the disability would lead to low levels of physical activity and leisure-time exercise, this may provide a plausible mechanism linking chronic widespread pain and cancer onset. High levels of physical activity are related to a lower risk of some cancers. The evidence has recently been reviewed by the International Agency for Research on Cancer and is strongest for cancers of the colon and rectum, female breast, and endometrium, although for prostate cancer, the conclusion is that “although additional evidence is needed.… physical activity may protect against prostate cancer” (28). Despite this plausible link between chronic widespread pain and cancer, studies that have formally examined the level of physical activity in fibromyalgia patients have reported fitness levels similar to those in healthy controls (29) or higher levels of activity than those in the general population (30). Therefore, although apparently plausible, there is lack of data to support the hypothesis that physical activity may confound the relationship between widespread pain and cancer.

Finally, chronic widespread pain is one of a series of interrelated “unexplained syndromes” which include, for example, chronic fatigue syndrome, irritable bowel syndrome, and the more recent “Gulf War” syndrome. Their etiologies share common features. However, although an association with increased cancer risk for some of these syndromes has been postulated (31), strong evidence to date is absent (32).

In summary, this study has demonstrated that a previous reported association between widespread pain and increased cancer mortality is a result of both increased cancer incidence and decreased cancer survival. The associations were particularly strong for hormone-related cancers (female breast and prostate cancers). To date, there remains no convincing explanation for this observation, although several possible leads have been identified.

Acknowledgements

  1. Top of page
  2. Abstract
  3. SUBJECTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The authors thank Elaine Thomas, who identified the study population, constructed the study database, and initially communicated with the staff of the Office for National Statistics (Southport) to flag the study subjects on the NHS Central Register. We also thank Ann Papageorgiou, who coordinated the original population surveys included in this study. We are indebted to the staff and patients of the participating general practices in Wythenshawe, Handforth, and Bollington in Greater Manchester and Cheshire.

REFERENCES

  1. Top of page
  2. Abstract
  3. SUBJECTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES