Cancer risks in ulcerative colitis patients
Article first published online: 17 JUN 2008
Copyright © 2008 Wiley-Liss, Inc.
International Journal of Cancer
Volume 123, Issue 6, pages 1417–1421, 15 September 2008
How to Cite
Hemminki, K., Li, X., Sundquist, J. and Sundquist, K. (2008), Cancer risks in ulcerative colitis patients. Int. J. Cancer, 123: 1417–1421. doi: 10.1002/ijc.23666
- Issue published online: 16 JUL 2008
- Article first published online: 17 JUN 2008
- Manuscript Accepted: 8 APR 2008
- Manuscript Received: 19 FEB 2008
- EU. Grant Number: LSHC-CT-2004-503465
- Deutsche Krebshilfe
- NGFN Plus
- Swedish Cancer Society
- The Swedish Council for Working Life and Social Research
- cancer risk;
- inflammatory bowel disease;
- age at hospitalization
Patients diagnosed with ulcerative colitis (UC) are known to be at an increased risk of colorectal and liver cancers and leukemia. UC is an autoimmune disease, which may present a wider spectrum of cancers. We wanted to examine the risk of cancer in a large population of UC patients in order to reach high statistical power. A UC research database was constructed by identifying UC patients from the Swedish Hospital Discharge Register and cancer patients from the Cancer Registry. Follow-up of 27,606 UC patients hospitalized for the first time during the years 1964–2004 identified 2,058 patients with cancer. Standardized incidence ratios were calculated for cancer in UC patients by comparing to subjects without hospitalization for UC. The novel tumor sites in UC patients included small intestinal (carcinoid), pancreatic, breast and prostate cancers, nonthyroid endocrine gland tumors, non-Hodgkin lymphoma and multiple myeloma. A total of 11 sites showed an increased risk, which remained at 6 sites when tumors diagnosed in the year of UC hospitalization were excluded; even chronic myeloid leukemia was in excess. Cancer risks depended on the age at first hospitalization for UC. The SIRs for colon, rectal, liver and pancreatic cancers declined by age at hospitalization for UC, while for endocrine tumors the older patients were at higher risk. Our large study identified novel subsequent cancers in UC patients. However, some of these, including small intestinal carcinoids, prostate cancers and nonthyroid endocrine tumors, may be in excess because of intensified medical surveillance of the patients. © 2008 Wiley-Liss, Inc.
Ulcerative colitis (UC), together with Crohn disease, is a common type of inflammatory bowel disease, which is characterized by a chronic dysregulated immune response against the mucosal linings.1 The sequence of events is thought to be initiated by nonpathogenic, commensal intestinal bacteria. The disease is mainly manifested in the colon in UC, whereas in Crohn disease the small intestine may also be affected. A differential diagnosis between UC and Crohn disease is made through pathognomonic clinical manifestations and endoscopic appearance. Family history is a risk factor of UC, according to family and twin studies.2–4 The familial risk for UC is lower than for Crohn disease and a family history for 1 is also a risk factor for the other. Accordingly, the recently described susceptibility genes for UC, i.e., IL23R and NPSR1 also predispose to Crohn disease,5, 6 whereas the TLR4 variant only predisposes to Crohn disease.7 However, a family history of inflammatory bowel disease does not predispose other family members to cancer.8
UC patients are at an increased risk of colorectal, hepatobiliary, nonmelanoma skin cancers and acute myeloid leukemia.9–12 However, considering that UC is an autoimmune disease, the spectrum of established cancers appears quite limited compared to other autoimmune diseases, such as rheumatoid arthritis and immunosuppressed patients, both of which present a wide spectrum of cancers.13–18 The traditional medication for UC with 5-aminosalicylates is believed to alleviate colorectal cancer risks in patients19 in contrast to the immunosuppressive agents, such as methotrexate and azathioprine, which are carcinogenic.20, 21 In the end of the 1990s tumor necrosis factor (TNF) antagonists were taken into clinical use, particularly for patients with a severe disease. Emerging data suggest an association between TNF antagonists and lymphoma in Crohn disease patients.20–23
We identified a total of 27,656 hospitalized UC patients throughout Sweden and followed their cancers from the National Cancer Registry. Together with a previous Swedish study, which focused on lymphohematopoietic tumors,11 our study is by far the largest published on UC, providing a statistical power to observe possible increases in rare cancers and to identify new associations.
Material and methods
The UC research database, used for our study, was constructed by linking several national Swedish registers. Statistics Sweden, the Swedish government-owned statistics bureau, provided the Multigeneration Register where persons born in Sweden in 1932 and later (second generation) are linked to their parents (first generation), registered shortly after birth; this made the base population for our study. Linkages were carried out to national census data, to obtain individual occupational status. Individual data on UC were obtained from the Swedish Hospital Discharge Register that records data on all discharges with dates of hospitalization and diagnoses since the 1960s with a complete nation-wide coverage since 1987. Cancers were retrieved from the nationwide Swedish Cancer Registry from the period 1958 to 2004. A 4-digit diagnostic code according to the 7th revision of the International Classification of Diseases (ICD-7) was used. All linkages were performed by the use of an individual national identification number that is assigned to each person in Sweden for their lifetime. This number was replaced by a serial number for each person in order to provide anonymity. The serial number was used to check that each individual was only entered once, for his or her first appearance with a UC diagnosis. Over 11.5 million individuals were included in this database. The Cancer Registry records new cases of cancer and practically 100% of the cases are histologically or cytologically confirmed at a full national coverage.24 Reports are requested on all malignant tumors, and additionally, as relevant to our study, on all carcinoid tumors, central nervous system and endocrine gland tumors. For example, 10–20% of liver cancers are recorded as “not primary.”24
Outcome and adjustment variables
UC patients were retrieved from hospital discharges reported according to the 7th (1964–1968), 8th (1969–1986), 9th (1987–1996) and 10th (1997–present) ICD revision. The codes for UC were as follows: ICD-7: 572.20 and 572.21, ICD-8: 563.10 and 569.02, ICD-9: 556, ICD-10: K51. A total of 28,080 UC patients were identified. Using codes for operations, we were able to identify 424 patients who had undergone colectomy and these patients were removed from the analysis. For diagnostic accuracy, we checked that among the 27,656 UC patients only 50 (0.18%) of them were subsequently discharged with a diagnosis for Crohn disease; these were removed from the analysis. More than one-half (54.6%) of the UC patients were discharged 2 or more times with the same diagnosis and 25.3% of the patients had 4 or more discharges with the diagnosis UC.
For cancers we had information on the year of diagnosis but not on the exact date. The follow-up of cancer was started on January 1 in the year of UC diagnosis and it was ended at the diagnosis of cancer, death, emigration or closing date on December 31, 2004. The individual variables controlled for in the analysis included gender, age at diagnosis (categorized as <30, 30–39, 40–49, 50–59, and 60–69 years), occupational status [6 groups: (1) farmers, (2) unskilled/skilled workers, (3) white collar workers (4), professionals (5), self-employed and (6) all others] and region [3 groups: (1) large cities, Stockholm, Gothenburg and Malmo, (2) Southern Sweden and (3) Northern Sweden]. Region allowed adjustment for possible regional differences in hospitalization rates.
Person-years were calculated from start of follow-up at diagnosis of UC until diagnosis of cancer, death, emigration, or closing date on December 31, 2004. Standardized incidence ratios (SIRs) were calculated as the ratio of observed (O) to expected (E) number of cases. The expected number of cases was calculated for age (5-year groups), sex, period (5-year groups), region and socioeconomic status-specific standard incidence rates for persons without hospitalization for UC.
Among the 27,606 UC patients who were hospitalized for the first time during the years 1964–2004, 2,058 patients developed cancer, giving an overall SIR of 1.46 (“All”) and a SIR of 1.29 for cancer diagnosed later than 1 year after the hospitalization for UC (1+), as shown in Table I (for cancer, only the year of diagnosis was available; thus the follow-up for 1+ started on January 1 in the year following the first hospitalization for UC). Only sites with more than 10 cases for the whole period were included; none of the rarer sites showed a significant difference. We show separately data for “All” and “All 1+” because for some cancers the risks were very high during the first year (i.e., the calendar year when both UC and cancer were diagnosed), probably because of a concomitant diagnosis of UC and cancer. Thus, 11 sites showed an increased risk for “All” but only 6 remained for the “All 1+” category, even though cancers diagnosed during the first year constituted no more than 14.3% of all cancers. Diagnosis of cancer at the time of first hospitalization for UC may be due to lead time bias. Yet, we prefer to discuss the results for “All” because even the cancers diagnosed within the first year were true diagnoses, verified as any other cancer in the Cancer Registry. Leaving out the cancers diagnosed during the first year would cause deficits to the subsequent years. The highest SIRs for “All” were noted for liver (SIR 4.30), colon (3.60), small intestinal (2.42) and rectal (2.26) cancers. Common sites, such as the breast (1.25) and the prostate (1.14), were also in excess, whereas the lung was not (1.00). Among commonly considered markers of depressed immune function, e.g. non-Hodgkin lymphoma, the SIR was increased (1.52), whereas squamous cell skin cancer was not (1.07).
|Cancer site||Follow-up interval (years)|
|O||SIR||95% CI||O||SIR||95% CI||O||SIR||95% CI||O||SIR||95% CI||O||SIR||95% CI||O||SIR||95% CI|
|Upper aerodigestive tract||1||1.02||0.00||5.85||5||0.74||0.23||1.74||11||1.51||0.75||2.72||26||1.56||1.02||2.29||43||1.36||0.98||1.83||42||1.37||0.98||1.85|
|Skin, squamous cell||4||2.44||0.63||6.31||16||1.45||0.82||2.35||12||0.99||0.51||1.73||27||0.90||0.59||1.30||59||1.07||0.82||1.38||55||1.03||0.78||1.34|
|Other endocrine glands||5||6.10||1.92||14.34||14||2.40||1.31||4.03||14||2.19||1.20||3.69||20||1.40||0.85||2.16||53||1.94||1.45||2.53||48||1.81||1.33||2.40|
In some more detailed analyses (data not shown), the increased risk for small intestinal cancer was due to carcinoid tumors (“All” N = 10, SIR 3.04, 95% CIs 1.45–5.61; “All 1+” N = 7, SIR 2.20, 95% CIs 0.87–4.56). The increased risk for nonthyroid endocrine gland tumors was entirely explained by parathyroid adenomas (“All” N = 42, SIR 2.51, 95% CIs 1.81–3.39; “All 1+” N = 37, SIR 2.28, 95% CIs 1.60–3.14). Among the leukemia subtypes, only chronic myeloid leukemia was in excess (“All” N = 13, SIR 3.40, 95% CIs 1.80–5.84; “All 1+” N = 11, SIR 2.97, 95% CIs 1.48–5.34).
The length of the follow-up period was also considered in Table I. When the first year after the diagnosis was not considered, the risks for colon and rectal cancers did not depend on the follow-up time; small intestinal and endocrine tumors and non-Hodgkin lymphoma were only increased in the early follow-up periods, whereas liver and pancreatic cancers were increased when the follow-up periods were longer.
The effect of age at first hospitalization for UC is shown in Table II. The data are shown for patients whose cancer follow-up was started the year following UC diagnosis (1+). The data for the whole patient group were essentially similar. Colon, rectal, liver and pancreatic cancers were increased mostly in those hospitalized before the age of 25 years and the SIRs were particularly high for liver (35.09) and colon cancers (20.41). Endocrine tumors showed the highest SIRs in those hospitalized at ages over 64 years.
|Cancer site||Age at hospitalization for ulcerative colitis (years)|
|O||SIR||95% CI||O||SIR||95% CI||O||SIR||95% CI||O||SIR||95% CI|
|Upper aerodigestive tract||4||3.01||0.78||7.78||7||0.71||0.28||1.48||21||1.54||0.95||2.36||10||1.67||0.80||3.09|
|Skin, squamous cell||4||3.25||0.85||8.41||10||1.10||0.52||2.03||21||0.98||0.60||1.49||20||0.93||0.57||1.44|
|Other endocrine glands||3||1.21||0.23||3.60||19||1.74||1.05||2.72||16||1.66||0.95||2.70||10||2.83||1.35||5.23|
Our study is the largest so far published, including 27,656 patients hospitalized for UC. Using data on operations, we were able to remove the patients with colectomies from the study population. The diagnostic accuracy for hospitalized cases of UC has previously been estimated in Sweden by comparing the hospital discharge data with regional registers including individually reviewed patient data.11 The concordance for the specific types of inflammatory bowel disease was 96%. According to our study, only 0.18% of the UC patients were diagnosed as Crohn disease patients in a subsequent discharge. Moreover, more than one-half of the patients had been discharged multiple times with the UC diagnosis. Another indicator for a good diagnostic distinction between UC and Crohn in the Hospital Discharge Register is that the SIR for colon cancer was higher than for small intestinal cancer (the SIR for small intestinal cancer was not significant for “All 1+”), whereas in Crohn disease the relative risk of small intestinal cancer is vastly higher than that of colon cancer (28.37 vs. 2.59 according to previous research25). According to the aforementioned Swedish study, 65.9% of the patients in the regional registers with apparently complete catchment of UC patients were found in the Hospital Discharge Register.11 However, this figure is an underestimate because the regional registers were started a decade before the Hospital Discharge Register, and 3 decades before, it reached a nation-wide coverage. Nevertheless, for studies of the present kind, diagnostic accuracy rather than coverage of all cases is crucial, yet with the caveat that severe cases are most likely to be hospitalized.
We hypothesized that many cancers would be in excess, considering that UC is an autoimmune disease and considering the statistical power of our study. Indeed, 11 sites showed an increased risk for “All” and 6 remained for the “All 1+” periods. The spectrum of cancers that were increased after UC included many sites that have not been reported before: small intestine, pancreas, breast, prostate, nonthyroid endocrine glands, non-Hodgkin lymphoma and multiple myeloma. However, some of these cancers were probably in excess because of the increased medical surveillance of UC patients. Thus small intestinal and prostate cancer were increased in “All” and not in “All 1+” periods, and the increase for small intestinal cancers was mainly explained by indolent carcinoid tumors. Similarly, other endocrine tumors, most of which are relatively indolent adenomas, showed an unusually high risk in the oldest age group, opposite to other tumors that were in excess in the younger UC patients. The previously established cancers, i.e. colon, rectal and liver cancers and chronic myeloid leukemia were confirmed.9–12 They showed the highest SIRs (2.26 for rectal cancer ranging to 4.30 for liver cancer). An evidence of the statistical power in our study is that a SIR of 1.25 for breast cancer was found significant. Somewhat surprisingly for an autoimmune disease, there was no increase for squamous cell skin cancer. These tumors have been in large excess in rheumatoid arthritis patients according to one of our recent studies.26 Nonmelanoma skin cancer was in excess in a previous study on UC patients.9 Opposite to an earlier Swedish study of 1,547 patients, there was no decreased risk of lung cancer.10
It is likely that UC often is diagnosed before the first hospitalization. In a previous Swedish study on inflammatory bowel disease patients diagnosed until 1990, the median delay time for hospitalization was 0.5 years (mean 4.4 years).11 Thus the data in Table II refer to the maximal age at diagnosis. Cancer risks depended on the age at first hospitalization for UC but the trends varied somewhat by cancer type. The SIRs for colon, rectal, liver and pancreatic cancers declined by age at hospitalization. The SIRs for colorectal and liver cancers were very high when the first hospitalization took place before the age 45 years. When the first hospitalization took place before the age 25 years, the SIR were particularly high, i.e., 20.41 for colon, 11.11 for rectal and 35.09 for liver cancers; the risk for pancreatic cancer was also high, 7.14. It should be recommendable that patients diagnosed before the age 45 years should undergo medical surveillance for early detection of preneoplastic lesions or locally defined tumors. In contrast, endocrine tumors showed the highest significant SIRs in those hospitalized at ages over 63 years, which may be indicative of fortuitous diagnosis of these indolent tumors, of which most are adenomas. The sites most severely affected by fulminate inflammation showed the highest risk in those hospitalized at an early age. As pointed out earlier, tumors diagnosed in patients hospitalized at a higher age may be incidental findings, such as parathyroid adenomas that are frequently diagnosed in nonsymptomatic individuals through serum calcium determination.
In summary, the follow-up of UC patients showed the highest relative risks for colon, rectal and liver cancers, particularly when the first hospitalization was before 45 years. These patients would benefit from medical surveillance. Our study also detected increased relative risks for many subsequent cancers that have not been reported earlier, due to its high statistical power.
The database used was created by linking registers maintained at the Statistics Sweden and the Swedish Cancer Registry.
- 2Risk of ulcerative colitis and Crohn's disease among offspring of patients with chronic inflammatory bowel disease. Am J Gastroenterol 1999; 94: 3236–8., , .Direct Link:
- 24Centre for Epidemiology. Cancer incidence in Sweden 2005. Stockholm: The National Board of Health and Welfare, 2007.
- 26Cancer risk in hospitalized rheumatoid arthritis patients. Rheumatology (Oxford), in press., , , .