Crohn's Disease in Olmsted County, Minnesota, 1940-1993:
Incidence, Prevalence, and Survival

Edward V. Loftus, Jr.,¹ Marc D. Silverstein,^2,4,5, William J. Sandborn,¹
William J. Tremaine,¹ W. Scott Harmsen³, and Alan R. Zinsmeister³

¹Inflammatory Bowel Disease Clinic, Division of Gastroenterology, ²Division of Area General Internal Medicine, and Sections of ³Biostatistics and ⁴Epidemiology, Mayo Clinic, Rochester, Minnesota; and ⁵Center for Health Care Research, Medical University of South Carolina, Charleston South Carolina

Background & Aims: Many centers worldwide have reported an increased incidence of Crohn's disease, but population-based data in North America are sparse. We studied the incidence and prevalence of Crohn's disease in Olmsted County, Minnesota, and examined temporal trends in incidence and survival. Methods: Residents diagnosed with Crohn's disease between 1970 and 1993 were incidence cases, and residents with Crohn's disease who were alive on January 1, 1991, were prevalence cases. Cases from previous studies were reconfirmed. Rates were adjusted using 1990 U.S. Census figures for whites. Incidence trends were evaluated with a Poisson regression model. Survival from diagnosis was compared with that expected for U.S. north-central whites. Results: Between 1940 and 1993, 225 incidence cases were identified, for an adjusted incidence rate of 5.8 per 100,000 person-years. On January 1, 1991, there were 145 residents with Crohn's disease, an adjusted prevalence rate of 133 per 100,000, 46% higher than that seen in 1980. Incidence rates before 1964 were significantly lower than those of 1989-1993. Observed survival was less than expected (P = 0.007). Conclusions: The incidence of Crohn's disease has stabilized since the 1970s at a rate higher than that seen previously. Prevalence has increased by 46% since 1980. Overall survival is slightly decreased.

Introduction

The cause of Crohn's disease, a chronic inflammatory disease of the gut that results in substantial morbidity, remains unknown.¹ Environmental, genetic, and immunologic causes have been suggested. Numerous epidemiological studies have been undertaken at centers worldwide in an effort to elucidate the causal mechanism of this disease.^2-26 Many centers, especially those in northern latitudes, have described an increase in the incidence and prevalence of Crohn's disease.^{3-6,7,9-12,17-20,22} This illness may be associated with a small but significant decrease in overall survival,^27,28 but data are conflicting.²⁹

Incidence and prevalence data from North America have also suggested trends of increasing incidence and prevalence,^14-26 but many such studies have been limited by incomplete case ascertainment. The nature of the health care delivery systems in the United States makes it difficult to identify all cases of a particular disease within a defined region, especially if some patients with the disease never require hospitalization. However, in Olmsted County, Minnesota, population-based studies disease are possible because the small number of institutions that provide health care share a linked diagnosis index.^30,31 Population-based epidemiological studies may reflect the true spectrum of illness, in contrast to studies based on patients seen at referral centers.^30,31

The incidence and prevalence rates of Crohn's disease in Olmsted County have been reported previously.^15,19 Sedlack et al. noted a higher incidence in urban residents, a nearly fourfold increase in incidence between 1935 and 1975 survival rate only slightly diminish compared with that expected, resulting in one of the highest prevalence rates reported at the time, 106 per 100,000 persons.¹⁵ Gollop et al. extended the study until 1982 and noted that the incidence rate of Crohn's disease was greater in women, was higher in the urban part of the county, and increased between 1943 and 1982.¹⁹ In the latter study, it appeared that the incidence rate had plateaued at a high but stable rate.¹⁹ Overall survival was only slightly less than expected. ¹⁹ Whether the rise in incidence of Crohn's disease represented improved diagnostic techniques or a worsening environmental risk factor profile is unclear.

Since publication of these studies, some centers reported a continued increase in the incidence of Crohn’s disease,^5,10,12 whereas others have noted that rates are stable.^7,8,11,22 Continued study of the trends in incidence and prevalence rates and mortality of Crohn's disease may shed light on its causes. Moreover, an updated, population-based inception cohort may serve as a platform for further detailed epidemiological study of Crohn's disease. In the current study, the inception cohort of Olmsted County residents with Crohn's disease was updated to 1993, previously identified cases were reviewed to assure consistent diagnostic criteria, and the prevalence of Crohn's disease in Olmsted County on January 1, 1991, was determined. Temporal trends in the incidence of Crohn's disease over a 54-year period were examined. Finally, overall survival was estimated and potential risk factors for death were examined.

Patients and Methods

Setting

Olmsted County, situated in southeastern Minnesota, comprised approximately 106,000 people in the 1990 U.S. Census. The county's urban center, Rochester, comprised approximately 71,000 people in 1990. The remainder of the county is rural. In 1990, 96% of the population was white. Although 25% of the residents are employed in health care services (vs. 8% nationwide) and the level of education is higher (30% have completed college vs. 21% nationwide), the residents of Olmsted County are socioeconomically similar to the U.S. white population.³¹

Rochester Epidemiology Project

The resources of the Rochester Epidemiology Project were used to identify permanent residents of Olmsted County who were diagnosed with Crohn's disease between 1970 and 1993. This unique medical records linkage system exploits the fact that virtually all the health care of the residents of the county is provided by two organizations: the Mayo Medical Center, consisting of the Mayo Clinic and its two affiliated hospitals, and Olmsted Medical Center, consisting of a smaller multispecialty group and its affiliated hospital.^30,31 In any 3-year period, 90% of county residents are examined at Mayo Clinic,³¹ and the percentage is almost certainly higher if visits to either one of the two health care systems are included. Diagnoses generated from all outpatient visits, emergency room visits, hospitalizations, nursing home visits, surgical procedures, autopsy examinations, and death certificates are recorded in a central diagnostic index. Thus, it is possible to identify all cases of a disease for which patients sought medical attention.^30,31

Case Ascertainment

The study was approved by the institutional review boards of Mayo Medical Center and Olmsted Medical Center. The complete (inpatient and outpatient) medical records of all potential cases identified through the diagnostic index were reviewed, and a diagnosis of Crohn's disease was made if a potential case met at least two of the following criteria: (1) clinical history of abdominal pain, weight loss, malaise, diarrhea, and/or rectal bleeding; (2) endoscopic findings of mucosal cobblestoning, linear ulceration, skip areas, or perianal disease; (3) radiological findings of stricture, fistula, mucosal cobblestoning, or ulceration; (4) macroscopic appearance of bowel wall induration, mesenteric lymphadenopathy, and "creeping fat" at laparotomy; or (5) pathological findings of transmural inflammation and/or epithehoid granulomas. Only patients who met these criteria for at least 2 months were included as definite cases. Patients who had been residents of Olmsted County for <1 year before diagnosis were excluded. (This resulted in the exclusion of 10 patients who met the diagnostic criteria.) These criteria were identical to those used in a previous study of Crohn's disease in Olmsted County.¹⁹

The medical records of all patients in previous studies of Crohn's disease^15,19 or ulcerative colitis^32,33 in Olmsted County (1940-1982) were re-reviewed to confirm the diagnosis of Crohn's disease. We performed a concurrent study of county residents with ulcerative colitis.³⁴ Potential cases in both studies were categorized as having Crohn's disease or ulcerative colitis based on all available clinical, radiographic, endoscopic, and histological evidence. By definition, an individual patient could not be counted in both studies.

The date of onset of symptoms, date of diagnosis, residency on January 1, 1991, last follow-up date, and vital status at last follow-up were recorded. Death certificates were obtained for all patients who had died. All cases from previous studies were also re-reviewed for the above data. Follow-up was complete to death or January 1, 1994, in 182 patients (81%).

Incidence and Prevalence Calculations

The entire population of Olmsted County from 1940 to 1993 was considered at risk for Crohn's disease. The denominator age- and gender-specific person-years were derived from decennial U.S. Census figures. When applicable, rates were age- and gender-adjusted to the 1990 U.S. white population. All residents of Olmsted County with Crohn's disease on January 1, 1991, were included in the prevalence calculation (including those who had moved to Olmsted County after diagnosis or <1 year before diagnosis). Ninety-five percent confidence intervals (95% CIs) of incidence and prevalence rates were estimated, assuming a Poisson distribution of cases. Trends in the incidence of Crohn's disease by age at diagnosis, gender, and calendar year of diagnosis (5-year intervals) were evaluated with a Poisson regression model.

Survival

Overall survival for the entire Crohn's disease inception cohort (date of diagnosis to date of last follow-up or death) was estimated using the product-limit life table method. Observed survival was compared with expected survival using age- and gender-specific expected death rates for Minnesota whites in 1990 using the log rank test. A Cox proportional hazards regression model examined the association of age at diagnosis, gender, and calendar decade of diagnosis with survival.

Results

Demographic Characteristics

Between 1940 and 1993 a new diagnosis of Crohn's disease was made in 225 Olmsted County residents. One hundred twenty-two patients were female (54%), resulting in a female to male ratio of 1.2:1. The mean age at diagnosis was 33.4 years. The median age at diagnosis was 29.5 years (range, 4.2-83.8 years). The diagnosis was most commonly made in the third and fourth decades of life (Figure 1). There was significant difference between males and females in the overall age distribution (Figure 1). The median interval between onset of symptoms and actual diagnosis was 90 days (range, asymptomatic to 10.4 years). Ileocolonic Crohn's disease was noted in 101 patients (45%). Seventy-one patients had Crohn's colitis (32%). Disease confined to the terminal ileum was noted in 49 patients (22%). Perianal involvement was detected in 58 patients(26%).

Incidence

The crude incidence rate of Crohn's disease in Olmsted County between 1940 and 1993 was 5.7 cases per 100,000 person-years. The age- and gender-adjusted incidence rate was 5.8 cases per 100,000 person-years (95% CI, 5.0-6.5). The adjusted incidence rate in the city of Rochester was 6.4 cases per 100,000 person-years (95% CI, 5.4-7.4), compared with an adjusted rate of 4.6 cases per 100,000 person-years in rural Olmsted County (95% CI, 3.4-5.8).

Temporal Trends

The female to male ratio of incidence cases increased from 1.0:1 in 1940-1968 to 1.3:1 in 1969-1993 (P = 0.37, _X²). The median interval between onset of symptoms and actual diagnosis was 26 days (range, 9 days to 8.3 years) in patients diagnosed between 1940 and 1953. This interval lengthened in subsequent decades. The longest median interval was 101 days, in both the 1954-1963 and 1974-1983 time periods.

The incidence of Crohn's disease increased markedly during the study period in both males and females (Figure 2). When analyzed by 10-year intervals, the adjusted incidence rate rose from 1.0 cases per 100,000 person-years in 1940-1943 to 6.9 cases per 100,000 person-years in 1984-1993, with a peak rate of 7.8 cases per 100,000 person-years in 1964-1973 (Table 1). For males, the adjusted incidence increased from 1.2 cases per 100,000 person-years in 1940-1943 to 7.6 cases per 100,000 person-years in 1974-1983, and apparently stabilized thereafter. In women, the adjusted incidence rose from 0.9 cases per 100,000 person-years in 1940-1943 to a maximum of 9.3 cases per 100,000 person years in 1964-1973.

The age distribution at onset of Crohn's disease changed over the course of the study (Figure 3). The median age at diagnosis decreased from 43.7 years in 1940-1953 to 29.8 years in 1984-1993. Between 1940 and 1963, the distribution of age at diagnosis was wide and no particular age group was afflicted to a greater degree. After 1964 there was a striking increase incidence among residents aged 20-29 years (Table 1 and Figure 3). Before 1964, only 36% of all patients wet diagnosed with Crohn's disease in the third and fourth decades of life. After 1964, 60% of patients were diagnosed in this age group. No pediatric cases of Crohn’s disease were diagnosed before 1954, but 17% of patients diagnosed in the last two decades of the study period were under the age of 20 years.

A Poisson regression model was used to examine effects of age at diagnosis, gender, and calendar year of diagnosis (5-year intervals) on incidence rates. Using the 1989-1993 incidence rate as the reference, the incidence rates before 1964 were significantly lower (P = 0.0001).

Using the overall incidence rate in patients at least 70 years old at diagnosis as the reference, the incidence rate in patients between 20 and 29 years old at diagnosis was significantly higher (P = 0.0005). The incidence rate in patients between 30 and 39 years old at diagnosis appeared to be higher, but this finding was of borderline statistical significance (P = 0.09). There was no association between gender and incidence rates (P = 0.7).

Prevalence

On January 1, 1991, there were 145 Olmsted County residents alive with a diagnosis of Crohn's disease. The crude prevalence rate on this date was 137 cases per 100,000 persons. The age- and gender-specific prevalence rates are shown in Table 2. The age- and gender-adjusted prevalence rate was 133 cases per 100,000 persons (95% CI, 111-155). Only 8 prevalence cases (5%) were not incidence cases. The age- and gender-adjusted prevalence rate in 1991 was 46% higher than that measured in 1980 (91 cases per 100,000 persons; 95% CI, 71-111).¹⁹

Survival

The inception cohort of Crohn's disease patients was observed for a total of 3407 person-years. The median duration of follow-up was 13.3 years (range, 0.1-51.3 years). Among those alive at last follow-up, the median duration was 13.4 years (range 0.1-51.3 years). Forty-three patients died (19%). Median follow-up for those who died was 11.2 years (range, 0.1-41.6 years). The causes of death are listed in Table 3. The deaths of 7 patients were attributed directly to Crohn's disease (perforation in 3, "toxemia"/sepsis in 2, and malnutrition/short-bowel syndrome in 2). Five of these deaths occurred before 1970, 1 occurred in 1972, and the last occurred in 1994. Gastrointestinal cancer was responsible for 7 deaths (4 colorectal carcinomas, 1 small bowel lymphoma, 1 ileal leiomyosarcoma, and 1 abdominal carcinomatosis of unknown primary). Additionally, one patient died of bile duct cancer. Thus, 15 deaths (35%) may have been related to Crohn's disease.

Overall survival was slightly less than that expected (P = 0.067, log rank test; Figure 4). For example, 20-year survival was 79% (vs. 86% expected) and 30-year survival was 73% (vs. 74% expected). When overall survival was stratified by decade of diagnosis, only survival for patients diagnosed between 1963 and 1974 was significantly less than expected (P = 0.021, log rank test). In this group of 60 patients, the 20-year survival was 83% (vs. 88% expected). Survival in the other decade groups was similar to that expected.

A Cox proportional hazards regression model examined the association of age at diagnosis, gender, and calendar year of diagnosis with time to death. There was no effect of calendar year of diagnosis (P > 0.5) or gender (P > 0.2) on survival. As expected, older age at diagnosis was associated with diminished survival (P = 0.0001).

Discussion

The incidence of Crohn's disease in Olmsted County, Minnesota, increased dramatically during the 1960s and early 1970s, and has since stabilized at a rate of roughly 7 cases per 100,000 person-years. The prevalence rate increased by 46% between 1980 and 1991. The "classic" age distribution of onset of Crohn's disease, with the majority of cases diagnosed in the third and fourth decades of life, did not become evident in our county until the mid-1960s. The overall survival of this inception cohort was slightly less than that expected for Minnesota whites of the same age and gender. One third of all deaths could be attributed to Crohn's disease or gastrointestinal cancer.

The main strength of the study is that ascertainment of cases within a defined geographic region approached 100%. Complete medical records (outpatient and inpatient) for all potential cases were available for review, and stringent but consistent diagnostic criteria were applied. Follow-up to 1994 was available in more than 80% of the patients. The population of the county is relatively stable as evidenced by the fact that 95% of the prevalence cases had been incidence cases. Moreover, we were able to build on previous reports of Crohn's disease in Olmsted County and extend the period of study to 54 years, allowing analysis of temporal trends in incidence and survival. Thus, the study provides true population-based information on the incidence and prevalence of Crohn's disease in a small region of the United States.

Our study is limited by the small size of the geographic region and therefore the relatively small number of cases. However, even with the small number of cases, the changes in incidence rates over time and by age of diagnosis were significant in a Poisson regression analysis. Second, only limited follow-up was available for recent years (vital status after January 1, 1994, was unavailable in almost 20% of patients). The majority of patients lost to follow-up appear to have moved out of the county. Mortality may have been underestimated, but only if patients lost to follow-up were somehow at increased risk of death. Third, the results of our study may not be generalizable to the United States as a whole because of racial and ethnic differences; as of 1990, only 4% of the Olmsted County population was nonwhite, and a substantial portion of the whites are of northern European heritage.

The crude incidence rate of Crohn's disease in the last decade of our study (1984-1993) was 6.9 cases per person-years. This rate, although historically very high, is not the highest reported, even for North America (Table 4). Pinchbeck et al. reported an incidence rate of 10 cases per 100,000 person-years in northern Alberta in 1981.²⁰ A recent preliminary report from Manitoba suggested an incidence rate as high as 15 cases per 100,000 person-years between 1989 and 1993²⁶; however, this was based on an administrative database and case ascertainment may have been imprecise and incomplete. When compared with crude rates obtained in 1991-1993 under the auspices of the European collaborative study of inflammatory bowel disease (EC-IBD),¹³ the incidence rate in Olmsted County ranks among the highest seen. Only rates in Reykjavik, Maastricht, Oslo, and Amiens were higher. Age- and sex-adjusted rates in the EC-IBD cannot be directly compared with adjusted rates in our study because different standards for adjustment were used. However, it appears that the rates in Olmsted County (north-central United States), Alberta, and Manitoba would be compatible with the hypothesis that a north-south gradient of Crohn's disease incidence exists. Although the magnitude of the north-south gradient in the EC-IBD study was less than that previously reported, a definite trend was still observed.¹³ The high incidence rate in the present study is also compatible with previous hospital-based observations of a north-south gradient within the United States.^23,24

The precipitous increase in incidence of Crohn's disease in Olmsted County between the 1950s and 1970s, especially in selected age groups, is striking. Had the rise been merely an artifact of the introduction of better diagnostic techniques, one might have expected a peak in incidence followed by a decline and stabilization at a rate lower than the peak. The two major modalities used to diagnose Crohn's disease are barium radiography and endoscopy. Barium radiography of the gastrointestinal tract and proctoscopy were in widespread use in Olmsted County before the dramatic increase in incidence, whereas colonoscopy was introduced after the increase had taken place. One might expect a shorter interval between symptom onset and diagnosis if the rise in incidence were caused by increased health care seeking or better diagnostic techniques. in fact, the shortest median interval between symptoms and diagnosis was seen in the first 14 years of the study and the interval remained stable thereafter. The reclassification of granulomatous colitis as a subtype of Crohn's disease rather than ulcerative colitis cannot explain the increase, because all cases of ulcerative colitis were concurrently studied and those with features of Crohn's colitis diagnosed before recognition of this entity were reclassified. The fact that most of the increase in incidence took place among those aged 20-39 years and not among all age groups suggests that it was not caused merely by better diagnostic techniques. The rising incidence among younger patients has also been noted in other regions (Scotland; Cardiff, Wales; Rochester, NY).^5,11,22 The rise remains unexplained and warrants further study.

The increase in incidence of Crohn's disease followed by a stabilization in the incidence rate has been noted in some but not all longitudinal population-based studies. ^3,4,6,7,11 Northeastern Scotland may be an exception in that the upward trend in the incidence rate curve had shown no sign of leveling off as of 1988 (9.8 cases per 10⁵ person-years). ¹⁰ On the other hand, incidence rates in Uppsala, Sweden, have remained stable.⁸ It is interesting to note that some regions, such as Florence, that had previously reported very low incidence rates are now seeing a rise in incidence rather than a stabilization.¹² This phenomenon has occurred in areas with long histories of advanced medical care, again indirectly suggesting that the rise in incidence cannot be explained solely on the basis of increased awareness and diagnostic techniques.

The adjusted prevalence of Crohn's disease in 1991 (133 per 100,000) is one of the highest reported and has increased by 46% over 11 years. The rising prevalence likely reflects the rise in incidence combined with the relatively prolonged survival of patients with Crohn's disease. If the results of the study are extrapolated to the U.S. white population of 1990 (approximately 209 million), there may have been as many as 278,000 patients with Crohn's disease among U.S. whites in the early 1990s, and each year another 14,600 patients are diagnosed. This calculation most likely underestimates all cases in the entire U.S. population, because black Americans, who numbered almost 31 million in the 1990 census, may have an incidence approaching that of U.S. whites.^21,25 The prevalence rates in younger Olmsted County residents are strikingly high (Table 2); if these younger patients experience a relatively normal, or even slightly diminished, life expectancy, the prevalence will continue to rise. If one assumes a stable incidence rate of 7 cases per 100,000 person-years and a median survival of 36 years after diagnosis (conservative estimate based on our survival curve), then the prevalence in our county should eventually stabilize at roughly 250 cases per 100,000 persons, almost double the prevalence in 1991. If this estimate is extrapolated to the projected U.S. white population in 2005 (232 million), there may be as many as 580,000 cases of Crohn's disease among U.S. whites in the middle of the next decade. This projection does not include cases among black Americans and other minority groups, in whom the prevalence of Crohn's disease may be higher than previously believed.²⁵

The small but statistically significant decreased survival in our cohort has been noted in most^27,28 but not all²⁹ recent population-based inception cohorts. The relative (actual/expected) survival in our cohort at 20 years was 92%, roughly comparable to a relative survival of 96% at 20 years in Uppsala, Sweden,²⁷ and 94% at 15 years in Stockholm.²⁸ Mortality directly related to Crohn's disease (e.g., perforation, sepsis, inanition) appears to have decreased, because only of 7 such deaths occurred after 1970, whereas most of the deaths overall occurred after 1970. This might imply that IBD-related mortality may have decreased as medical and surgical treatments have improved. When the observed vs. expected survival comparison was stratified by decade of diagnosis, there was an association between decade of diagnosis 1964-1973 and decreased overall survival; however, the absolute difference between actual and expected survival at 20 years was small (5%). Moreover, formal analysis with the Cox proportional hazards regression model could not detect an association between survival and calendar year of diagnosis.

The finding of 7 deaths caused by gastrointestinal cancer, including 4 colorectal cancers and 1 small bowel lymphoma, suggests that mortality from intestinal cancer may be increased. Although a previous study in Olmsted County detected a twofold increased risk of colorectal cancer in Crohn's disease, it was not statistically significant.¹⁹ Other population-based studies have not found an increased mortality caused by colorectal cancer when all patients with Crohn's disease are studied.^27-29 The risk of any colorectal cancer (fatal or nonfatal) may be increased for patients with Crohn's disease, especially for those with Crohn's colitis,³⁵ although data are conflicting in this regard.^28,29 Further study of the risk of gastrointestinal cancer in this cohort is warranted.

Figures and Tables

Figure 1. Crude incidence of Crohn's disease in Olmsted County, 1940-1993, by age of diagnosis

Figure 2. Temporal trends in age- and sex-adjusted incidence of Crohn's disease in Olmsted County, 1944-1993, with 95% CIs

Figure 3. Crude incidence of Crohn's disease in Olmsted County, 1944-1993, by age and year of diagnosis.

Figure 4. Overall survival of Crohn's disease inception cohort Olmsted County, 1940-1993 (solid line), and expected survival of sex- and age-matched Minnesota whites in 1990 (dashed line; P = 0.007, log rank test).

References