- Top of page
- Materials and Methods
- Supporting Information
Colorectal cancer is the third most common malignancy worldwide and the disease has a serious prognosis. Synthetic glucocorticoids are widely used in the treatment of chronic and acute inflammatory diseases due to their potent immunosuppressive effects. In Denmark, 3.3% of the population was treated with systemic glucocorticoids in 2010. Concerns have been raised that prolonged therapy may increase cancer risk.[3-7] Use of glucocorticoids could possibly affect the risk of colorectal cancer through several mechanisms. Glucocorticoids are involved in the regulation of metabolism, cell growth, proliferation, apoptosis and immune function, which all play major roles in the prevention of cancer development and spread.[9, 10] As glucocorticoids are potent inhibitors of immunosurveillance, they may facilitate carcinogenesis and rapid progression of colorectal cancer. Moreover, insulin resistance, a well-known side effect of glucocorticoids, has been suggested to increase colorectal cancer risk. In contrast, glucocorticoids exert antiproliferative and proapoptotic effects, and could thereby even facilitate chemoprevention. In addition, glucocorticoids belong to the same steroid superfamily as oestrogen and progesterone, which appear inversely associated with risk of colorectal cancer.
Although glucocorticoids are commonly prescribed in clinical practice, an association with colorectal cancer development is unclear and existing epidemiological data are few. Previous studies evaluating glucocorticoid therapy and colorectal cancer risk all were restricted to patients with inflammatory bowel disease, who a priori have an increased risk of cancer. Moreover, glucocorticoid use was not the main exposure of interest in any of the studies, and their results were conflicting. We therefore conducted a large prospective population-based case–control study examining the association between glucocorticoid use and colorectal cancer risk, both overall and according to cancer stage.
- Top of page
- Materials and Methods
- Supporting Information
We identified 14 158 colorectal cancer cases and 141 580 population controls during the study period, of which 782 (5.5%) and 8434 (6.0%), respectively, were frequent users of systemic glucocorticoids (Table 1). More men (52.5%) than women (47.5%) had colorectal cancer, and most patients were diagnosed between the ages of 70 and 79 years (32.6%). The distribution of potential confounders was nearly the same for cases and controls.
Table 1. Characteristics of colorectal cancer cases and matched population controls, Northern Denmark, 1991–2010
|Characteristics||Cases N (%)||Controls N (%)|
|Never/rare systemic use||12 122 (85.6)||121 271 (85.7)|
|Frequent systemic use||782 (5.5)||8434 (6.0)|
|Combined usea||1254 (8.9)||11 875 (8.4)|
|Female||6727 (47.5)||67 270 (47.5)|
|Male||7431 (52.5)||74 310 (52.5)|
|Age at diagnosis, years|
|<50||749 (5.3)||7505 (5.3)|
|50–59||1910 (13.5)||19 260 (13.6)|
|60–69||3668 (25.9)||36 404 (25.7)|
|70–79||4596 (32.5)||46 206 (32.6)|
|80+||3235 (22.9)||32 205 (22.8)|
|1991–1995||1391 (9.8)||13 910 (9.8)|
|1996–2000||2672 (18.9)||26 720 (18.9)|
|2001–2005||4777 (33.7)||47 770 (33.7)|
|2006–2010||5318 (37.6)||53 180 (37.6)|
|Diagnoses or related medication before index date|
|Diabetesc||1203 (8.5)||10 176 (7.2)|
|Obesity||352 (2.5)||3040 (2.2)|
|Alcoholismd||366 (2.6)||3398 (2.4)|
|Pulmonary diseasese||2191 (15.5)||20 868 (14.7)|
|Inflammatory bowel diseases||107 (0.8)||1036 (0.7)|
|Rheumatoid arthritis||154 (1.1)||1593 (1.1)|
|Medication before index date|
|NSAIDs||8230 (58.1)||83 257 (58.9)|
|Low-dose aspirinf||4038 (28.5)||40 180 (28.4)|
|High-dose aspirinb||68 (0.5)||942 (0.7)|
|Immunosuppressants||176 (1.2)||1727 (1.2)|
Among frequent users of systemic glucocorticoids, each subject filled 11 prescriptions on average, ranging from 3 to 311, during a mean period of 4.4 years. The mean cumulative prednisolone-equivalent dose prescribed was 4295 mg, ranging from 75 to 87 550 mg. Grouped according to quartiles, low, medium and high doses were 75–350 mg, 350–5500 mg, and more than 5500 mg, respectively.
We found no association between frequent use of systemic glucocorticoids and overall risk of colorectal cancer [aOR = 0.93 (95% CI: 0.85–1.00)]. Recent vs. former use did not affect this overall OR (data not shown). Table 2 outlines the ORs according to duration of use and dose. The results are virtually identical to the overall OR, although short-term high-dose systemic glucocorticoid use was associated with a slightly lower aOR of 0.74 (95% CI: 0.59–0.94). In the analysis by colorectal cancer stage, associations between long-term use of medium-dose [aOR = 1.16 (95% CI: 0.89–1.53)] or high-dose [aOR = 1.12 (95% CI: 0.81–1.55)] systemic glucocorticoids and localised cancer were near the null (Table 3). Corresponding associations for metastatic cancer were also almost null [aOR = 0.79 (95% CI: 0.59–1.05) and aOR = 0.82 (95% CI: 0.59–1.14)].
Table 2. Associations between systemic glucocorticoid use and overall risk of colorectal cancer
| ||Cases||Controls|| || |
|Systemic glucocorticoids||N (%)||N (%)||OR (95% CI)||aORa (95% CI)|
|Never/rare use||12 122 (85.6)||121 271 (85.7)||1.0 (referent)||1.0 (referent)|
|Combined useb||1254 (8.9)||11 875 (8.4)||1.06 (0.99–1.12)||1.02 (0.95–1.10)|
|Low dosec||142 (1.0)||1553 (1.1)||0.91 (0.77–1.09)||0.92 (0.78–1.10)|
|Medium dosed||276 (2.0)||2835 (2.0)||0.97 (0.86–1.10)||0.97 (0.85–1.10)|
|High dosee||79 (0.6)||1044 (0.7)||0.76 (0.60–0.95)||0.74 (0.59–0.94)|
|Low dose||54 (0.4)||666 (0.5)||0.81 (0.61–1.07)||0.81 (0.62–1.08)|
|Medium dose||133 (0.9)||1318 (0.9)||1.01 (0.84–1.21)||1.01 (0.84–1.21)|
|High dose||98 (0.7)||1018 (0.7)||0.96 (0.78–1.19)||0.95 (0.76–1.17)|
Table 3. Associations between systemic glucocorticoid use and risk of colorectal cancer by stagea
|Systemic glucocorticoids||Localised colorectal cancer||Metastatic colorectal cancer|
|N||OR (95% CI)||aORb (95% CI)|| N ||OR (95% CI)||aOR (95% CI)|
|Never/rare use||4635||1.0 (referent)||1.0 (referent)||6136||1.0 (referent)||1.0 (referent)|
|Combined usec||542||1.19 (1.09–1.31)||1.17 (1.04–1.31)||527||0.88 (0.80–0.96)||0.88 (0.79–0.99)|
|Low dose||45||0.76 (0.56–1.02)||0.77 (0.57–1.03)||80||1.02 (0.81–1.28)||1.05 (0.84–1.32)|
|Medium dose||116||1.07 (0.89–1.29)||1.09 (0.90–1.31)||116||0.81 (0.67–0.98)||0.84 (0.70–1.02)|
|High dose||30||0.75 (0.52–1.08)||0.76 (0.53–1.10)||39||0.74 (0.54–1.02)||0.77 (0.56–1.07)|
|Low dose||17||0.67 (0.41–1.08)||0.70 (0.43–1.14)||31||0.92 (0.64–1.32)||0.96 (0.67–1.37)|
|Medium dose||55||1.09 (0.83–1.43)||1.16 (0.89–1.53)||50||0.75 (0.56–1.00)||0.79 (0.59–1.05)|
|High dose||41||1.05 (0.77–1.44)||1.12 (0.81–1.55)||40||0.78 (0.57–1.07)||0.82 (0.59–1.14)|
Subanalyses across strata of comorbidities and drug use did not change the association between frequent use of systemic glucocorticoids and overall colorectal cancer risk, except for use of NSAIDs [aOR = 0.89 (95% CI: 0.81–0.97)]. The sensitivity analysis stratified by time period (1991–2002 and 2003–2010) also yielded results near the null (data not shown). Results of our sensitivity analyses, including glucocorticoid prescriptions filled within 1 year before the index date, were not substantially different from the results of the main analysis (data not shown).
- Top of page
- Materials and Methods
- Supporting Information
In this large population-based case–control study, we found no evidence of an overall association between frequent use of systemic glucocorticoids and colorectal cancer risk. Although results were close to null, long-term use of medium- or high-dose systemic glucocorticoids slightly increased risk of localised colorectal cancer, and decreased risk of metastatic cancer. We cannot exclude the possibility that this association was causal; however, it seems likely that increased surveillance influenced our findings. Systemic glucocorticoids are used for the treatment of various medical conditions and only by prescription. Thus, patients using these drugs may have frequent contacts with physicians. Moreover, glucocorticoids can cause serious side effects, including peptic ulcers, upper gastrointestinal bleeding, and anaemia, which may lead to blood tests, diagnostic endoscopies and early detection of colorectal neoplasia. In this context, we would expect fewer metastatic colorectal cancers at the time of diagnosis. Our study findings supported this assumption.
To our knowledge, this is the first study to investigate the association between systemic glucocorticoid use and risk of colorectal cancer in the general population. Previous studies suggest that immune suppression by glucocorticoids increases the risk of non-Hodgkin lymphomas, and cancers of the skin, bladder and prostate.[3-7] Still, no association was found between glucocorticoid therapy and risk of breast cancer in a Danish population-based case–control study, and an inverse association was reported for inhaled corticosteroids and lung cancer among patients with chronic obstructive pulmonary disease. Differences in study results may indicate that associations between glucocorticoids and cancer risk depend on where the cancer is located.
Major strengths of our study include its population-based design, encompassing all incident colorectal cancers in the study period. Our study population was identified in registries with complete follow-up.[16, 19] In addition, accurate registry data on prescriptions and diagnoses eliminated recall bias and permitted adjustment for potential confounders.[20, 23]
Our study also has potential weaknesses. Misclassification of both exposure and covariates is a major concern. First, filled prescriptions for medications were used to estimate actual intake, as we lacked information on adherence. However, a recent Danish study reported complete correspondence between corticosteroid treatment reported by general practitioners and time of prescription dispensation within 3 months of a set index date. Second, we lacked data on glucocorticoids dispensed to hospital inpatients, thereby potentially underestimating the actual use of glucocorticoids. However, adjustment and stratification based on hospitalisations and corresponding diagnoses did not change the overall risk estimates. Third, the prescription database does not include over-the-counter NSAIDs, which account for 14% of total NSAID use in Denmark. Also, low-dose aspirin is available without prescription. These drugs are inversely associated with colorectal cancer risk and may also be related to glucocorticoid exposure. We assumed that these drugs were likely to be prescribed when used to treat chronic conditions, to allow patients to get reimbursed. Fourth, left truncation of prescription data could lead to misclassification of users as non-users. However, potential bias due to under-ascertainment of glucocorticoid exposure before establishment of the prescription database cannot explain our findings.
We had no information on the indication for prescription of systemic glucocorticoids. However, we adjusted for relevant diagnoses before the index date and included glucocorticoid dose and duration of use, which are surrogate markers of disease severity. Moreover, our definition of systemic glucocorticoids minimised confounding by the indication for concurrent use of either inhaled glucocorticoids or glucocorticoids acting locally in the intestines (i.e. related to severe underlying conditions).
Finally, unmeasured factors such as smoking, physical activity, and diet could affect colorectal cancer risk. However, confounding by lifestyle factors would have caused an overestimation of the ORs and thus cannot explain a null result or an inverse association.
In conclusion, despite immunological and metabolic effects of frequent use of systemic glucocorticoids, which would be expected to increase colorectal cancer risk, we found no overall association between the two. Associations between frequent use of systemic glucocorticoids and colorectal cancer stage are probably influenced by detection bias, due to increased medical surveillance among these patients.