This study has added to findings about musculoskeletal pain and future general risk of cancer and mortality by demonstrating relationships between new episodes of specific regional musculoskeletal problems in older persons attending primary care and subsequent onset of individual regional cancers. The associations observed in the first year of follow-up (back problems with lung, breast and prostate cancer; hip with breast and prostate cancer; and neck with prostate cancer) generally disappeared after the first year, although remained strong but reduced for prostate cancer up to 10 years after initial consultation.
An association between widespread pain and development of cancer up to 8 years later, particularly of breast and prostate, has previously been shown.3
We have previously found that new consulters presenting with single site problems in back, neck, hip or shoulder had a general increased risk of cancer diagnosis which was not evident for other pain sites.5
Here, we have added to this picture by showing that associations between specific musculoskeletal sites and different cancers did not include colorectal cancer; that the strength of these associations diminished with time; that the associations varied in strength for the different cancers—those with prostate cancer being the strongest and longest sustained; and that the presence of widespread pain did not explain the associations.
We constructed the cohorts to exclude all persons with a record of cancer in their medical case-notes in the 2 years prior to the onset of their musculoskeletal problem (and equivalent for the comparison group). Although it is possible that persons who had cancer earlier than this were not excluded, it is more likely that some record of it would have appeared in the case-notes during the 2 year period prior to recruitment and follow-up. Furthermore, the diagnosis and recording of cancer was made at a later date than the date of onset of the baseline musculoskeletal condition which defined recruitment into the exposure group; so all cancer diagnoses occurred subsequently to the initial musculoskeletal episode. One explanation of the patterns we observed is that musculoskeletal pain is an early marker of occult or developing cancer, prior to presentation and diagnosis of the primary cancer, rather than the alternative explanation that a general component of the mechanism or experience of chronic pain is a cause of cancer. The incidence of prostate cancer in the back group in the first year of follow-up was five times higher than the comparison group who had no musculoskeletal consultation at baseline. After the first year risk of cancer was similar (years 2–5) or only slightly higher (years 6–10) in those with initial back problems compared to the comparison group. It is possible that this may be partially explained by those in the comparison group becoming “exposed,” that is, consulting for a musculoskeletal problem during follow-up and prior to being diagnosed with cancer. Regardless, this reinforces the finding that the main impact is in the first year after a musculoskeletal consultation and the likely interpretation that the musculoskeletal problems are a marker rather than cause of cancer.
The pattern of risk observed (strongest for prostate and breast, no association with colorectal cancer) does mirror the pattern of risk of skeletal metastases. The locations of musculoskeletal pain carrying the strongest risk of future cancers (spine and hip) are among the most frequent sites of skeletal metastatic spread from prostate, breast and lung.15
Pain due to bony metastases can be the first presentation of prostate or breast cancer but usually only months before diagnosis and are likely to be due to the mechanical effect of the rapidly enlarging metastases in the bone.6
Micrometastases of prostate cancer in the spine may occur over 10 years prior to progression, whilst Ross reported the importance of both disseminated and circulating tumor cells to the prognosis in breast cancer.19,20
A review of micrometastases suggested they occur early in the life of a solid tumor and there may be a tendency for disseminated tumour cells to lodge in bone marrow.7,21
Two reviews concerning the relevance of disseminated tumour cells to progress of cancer concluded that tumour cells can lay dormant in bone marrow and for prostate cancer this may be for 10 years.7,19
It has not been reported that these micrometastases are related to pain.8,9,22
Metastases to bone in prostate cancer are common in several areas of the skeleton.23
Prostate cancer itself may lay dormant for many years, the trigger to why there is growth in some cancers but not others is uncertain.24–26
The importance of the micro-environment of the skeletal bone tissue in providing a stimulus for the seeding and growth of metastatic cancer has been highlighted in recent literature and may provide another explanation of musculoskeletal pain as a herald of future cancer.27
Certain rheumatic diseases are associated with the development of malignancy at sites in the body remote from the cancer (paraneoplastic syndromes).8–10
These paraneoplastic rheumatic disorders can precede the cancer by over 2 years and some bear similarities to premalignant conditions.8
Their aetiology is not clear but various mechanisms have been implicated including autoantibodies linked to cell death, inducing humoral and cell mediated immunity and release of cytokines.8,9
It is possible that some of the cancer diagnoses in this study might have fitted this category but there was no evidence from the available primary care records to confirm this.
There are other explanations of the observations, notably confounding by causal factors common to both musculoskeletal problems and cancer. We adjusted for potential confounders such as BMI and smoking which are routinely recorded in the medical records. It is possible that other confounders, such as physical activity, which were not included because they are not routinely recorded, could explain the findings, although the specific associations observed make this less likely.
The study has some potential weaknesses. It is possible that symptoms of cancer may be recorded elsewhere than under the Neoplasms Chapter. However, our study is based on definitive diagnoses of malignant and premalignant neoplasms. Furthermore, based on national cancer incidence figures for the United Kingdom, 17 new lung cancer cases per 10,000 population were expected in our comparison group in the first year. The actual figure of 16 per 10,000 suggests a comparable cancer rate in this group to the United Kingdom general population. We arbitrarily selected 2 years free of consultation for musculoskeletal problems as our definition of an incident case. Not all repeat consultations for chronic conditions in the GPRD need to be recorded.28
However, on-going consultations for the same problem should be recorded at least once during a 2 year period as the GPRD requires recording of morbidity if treatment changes. Morbidity codes are required to be entered at first diagnosis and hence incident cases should be comprehensively recorded.
This study has investigated problems in body locations rather than specific diagnoses. This is more relevant to the patient who will present initially with pain in a specific location, and may not obtain a clear diagnosis initially. Our assessment of widespread problems was based on primary care records and is not a stringent definition of widespread pain. There was some loss to follow-up but there is no reason to believe these patients were substantially different to those with longer follow-up. The groups had similar lengths of follow-up. The estimation of comorbidity ignores comorbidity occurring within the same Read Code Chapter, but reflects the range of comorbidities with which patients present.
With respect to the relevance of our findings for clinical practice, our previous study suggested that in a group of 1,000 new consulters to primary care aged 50 plus with back pain, there will be approximately 12 extra cases of cancer than would be expected in the next 12 months. Our new analysis suggests that these extra cases will be in the lung, prostate or breast. Prostate cancer is the most common but the absolute risk is small. Our study suggests prostate cancer will occur in the first year after presentation of back pain in one out of every 120 male new back pain consulters, and one out of every 75 male new hip pain consulters. Stratifying on age, prostate cancer will occur within the first year after presentation of back pain in one out of every 200 new male back pain consulters aged under 75 years and one in 40 aged 75 years or over. The equivalent figures for hip pain are one in every 156 new consultations (aged under 75) and one in 32 (aged 75 and over).
The absolute excess risk of cancer in patients presenting with common musculoskeletal problems is therefore low, and general practitioners and their patients can in general be reassured by these estimates from a large and representative national database of primary care consultations that new onset musculoskeletal pain is only rarely a harbinger of cancer. However, there is a need to ensure that primary care physicians, most of whom will know of such occasional rare possibilities and of patients' concerns about them, are also aware of the estimated size and nature of the increased risk for their discussions with older patients presenting with new back, hip or neck problems and to inform future consultations with those who return with other problems. Consultations for new back pain in the elderly are considered as red flag symptoms but this study shows that prostate cancer related to new hip pain is as common. Although the link between certain musculoskeletal conditions, such as hypertrophic osteoarthropathy, and cancer are recognised, most common musculoskeletal conditions seen in primary care are not closely related to cancer.8
As, we have discussed above, the reduction in risk over time suggests musculoskeletal problems are not a cause of cancer.
Regional musculoskeletal problems in the back, hip and neck are related to onset of new prostate, lung, and breast cancer in the elderly, particularly in those aged over 75 years, and in the first year after the initial musculoskeletal consultation, and one explanation is that these associations represent an early manifestation of the cancer. For prostate cancer, the link appears to extend for up to 10 years. However, the likelihood of cancer overall in older persons presenting in primary care with symptoms in these musculoskeletal sites is low, and the clear clinical message is that general practitioners and their patients do not, on this basis, need to search for cancer in the presence of uncomplicated back, hip or neck pain. Our study has provided specific estimates that can support discussion and reassurance in the consultation, and inform clinical practice if patients continue to consult with troublesome or additional symptoms. From the perspective of the science and biology of pain and cancer, further research is warranted into the mechanisms of these associations, including the possibility that they indicate early micrometastases to bone or an early marker of the potential for metastasis. Our study has found no evidence to support the hypothesis that musculoskeletal pain itself is a contributory cause of future cancer.