Association between colonic diverticular disease and colorectal cancer: a nationwide population-based study.

Association Between Colonic Diverticular

Disease and Colorectal

Cancer: A Nationwide Population-Based Study

Wen–Yen Huang,

*

,‡

Che–Chen Lin,

§

Yee–Min Jen,

*

Yen–Jung Chang,

§

Cheng–Wen Hsiao,

jj

Muh–Hwa Yang,

‡,¶

Chun–Shun Lin,

*

Fung–Chang Sung,

§,#

Ji–An

Liang,

**

,‡‡

and

Chia–Hung Kao

**

,§§

C

olonic diverticular disease and colorectal cancer (CRC) are important global health care problems. Both are extremely common, and the overall economic burden is considerable.1,2 Autopsy studies have found

colonic diverticular disease to have a prevalence rate of 5% to 52% in the West, and of 1% to 19% in Asia.3 The

sequelae arising from colonic diverticular disease are important causes of morbidity, including infammation,

obstruction, infection, bleeding, fstula formation, perforation, and death.4 CRC is the third most common cancer

in men and the second most common cancer in women worldwide, with more than 1.2 million new cancer cases and 608,700 deaths estimated to have occurred in 2008.2

Colonic diverticular disease and CRC share certain characteristics. In epidemiology, the incidence rates of them are much higher in the West than in other areas of the world and they are diseases that primarily affect the elderly.1–4 In etiology, lack of physical activity and

obesity also are associated with both diseases in prospective cohort studies and meta-analysis.5–8 Therefore,

their epidemiologic correlation is an interesting

issue. There have been some reports published to date to investigate their association. Certain reports have shown

a positive correlation between colonic diverticular disease and CRC.9–13 However, other publications have

failed to show an association14–16 or a negative correlation. 17,18 The results are discrepant and insuffciently

strong to draw frm conclusions. Most studies have focused on the population in the West. To the best of our knowledge, no study has been conducted in Asia to investigate the epidemiologic association between colonic diverticular disease and CRC using a nationwide population-based data set.

In Taiwan, CRC is the second most common cancer for both men and women, with an incidence of 53.99 per

100,000 for men and 41.43 per 100,000 for women, according to the 2008 cancer report released by the Taiwan

Department of Health.19 The Taiwan National Health Insurance

Research Database (NHIRD) is composed of deidentifed medical claims from 99% of Taiwan’s 23

million people. The database is available to researchers in Taiwan and has been used extensively in epidemiologic studies.20,21 Furthermore, the accuracy and high validity

of the data in the NHIRD also have been shown.22,23 This

study provides an opportunity to examine our hypothesis that a diagnosis of colonic diverticular disease is associated with an increased risk of subsequent CRC development in a nationwide population-based cohort study in Taiwan.

Methods

Data Source

The Taiwan NHI program became a universal insurance system in 1996, covering nearly 99% of 23 million citizens in Taiwan in 1998. The National Health Research Institute established and managed the NHIRD, which involved annual reimbursement claim data from the

Taiwan NHI program. All personal identifcation information was encoded to protect patient privacy before

being released for research. The National Health Research Institute created a secret, anonymous identifcation number to link each patient’s information, including sex,

birth date, and medical service registry such as diagnosis codes, prescription drugs, surgery code, and so forth. In this study, disease diagnosis was defned by the International Classifcation of Diseases, 9th revision, Clinical

Modifcation (ICD-9-CM) from NHIRD inpatient data. This study was approved by the Ethics Review Board at China Medical University (CMU-REC-101-012).

Study Population

This was a population-based, retrospective cohort study. The participants in both the study and

comparison cohort were extracted from the NHIRD. In the study cohort, the inclusion criteria were patients who were admitted in 2000 to 2009 with a newly diagnosed colonic diverticular disease (ICD-9-CM 562.1), and patients who were age 18 years or older. The exclusion criterion was those who had a history of any cancer. In the comparison cohort, the participants were extracted by random selection from all of NHIinsured individuals without a diagnosis of ICD-9-CM

562.1 during the same period. We matched each

diverticular disease patient to 4 nondiverticular disease patients by age, sex, and baseline year. The diagnosis of CRC (ICD-9-CM 153 and 154) was collected

from the registry of the Catastrophic Illness Patient Database, a subset of the NHIRD, which only included patients with histologic confrmation and supporting evidence. The accuracy and high validity of diagnoses identifed in the NHIRD have been shown in previous studies.22,23

We performed a subanalysis and defned the diverticulosis group as inpatients with a diagnosis of diverticulosis (ICD-9 CM-562.10 and/or 562.12). The

diverticulitis group was defned as inpatients with a diagnosis of diverticulitis (ICD-9-CM 562.11 and/or 562.13) or with both diverticulosis (ICD-9 CM-562.10 and/or 562.12) and diverticulitis (ICD-9-CM 562.11 and/or 562.13). This research excluded patients with cancer or patients younger than age 18. Subsequent CRC,

the main event of this study, was defned by ICD-9-CM 153 and 154 from the registry of the Catastrophic Illness Patient Database.

A cancer that was discovered very soon after the

diverticular disease was diagnosed actually may represent an incorrect initial diagnosis (because the symptoms are similar) or the subsequent diagnosis of cancer because a colonoscopy then was performed. To detect this problem, a sensitivity analysis was performed to assess for

ascertainment bias and we excluded the frst 12 months of follow-up evaluation after a diagnosis of colonic diverticular disease.

The follow-up evaluation was terminated when CRC occurred, when the patient lost insurance, or on

December 31, 2010. The database only provided the date of insurance withdrawal, but did not record the reason why the insurance was removed. Common reasons for insurance withdrawal include fulflling mandatory military service requirements, being expatriated to another country, imprisonment, and death.

We considered demographic factors and CRCassociated comorbidity confounding factors. The

CRC-associated comorbidities included hypertension (ICD-9-CM 401-405), diabetes (ICD-9-CM 250), and hyperlipidemia (ICD-9-CM 272). They were all risk factors for CRC and their prevalence rates were extremely high in Taiwan. We chose them as the factors that should be adjusted, with concern for their potential confounding effect. Furthermore, we also used the Charlson Comorbidity Index to adjust comorbidities. Statistical Analysis

We showed the mean and standard deviation for continuous variables, the number and percentage for category variables to describe colonic diverticular disease distribution and the comparison cohort, and we

used the t test for continuous variables and the chisquare test for category variables to test the difference

between the 2 cohorts. The total incidence, demographicspecifc incidence, and comorbidity-specifc incidence of

developing CRC was calculated per 10,000 person-years. The Kaplan–Meier method was used to estimate the cumulative CRC incidence curves for study cohorts and the

log-rank test was used to test the difference of these 2 curves. Demographic-specifc incidence and the hazard ratio (HR) of CRC stratifed by age, sex, and comorbidities also were analyzed. The Cox proportional hazards

regression model with adjusted potential confounding factors was used to estimate the HR and the confdence interval (CI) for the study cohort, compared with the comparison cohort.

Data management and analysis were performed using SAS 9.1 software (SAS Institute, Cary, NC). The competing risk analysis and the plot of the cumulative incidence curve was performed using STATA SE 11 (Stata Corp, College Station, TX). The data initially were analyzed

including the entire population. Because potential confounding factors might have exaggerated or lessened

differences, the 5 comorbidities previously noted for which information was available in the database were considered separately: age, sex, hypertension, diabetes, and hyperlipidemia. Because multiple analyses were

being performed, a P value of .01 was used as the measure of signifcance.

Results

The cohort included 41,359 patients with diverticular disease. Of these patients, 28,909 had diverticulitis and 12,450 had diverticulosis. There was a total of 165,436 individuals in the comparison cohort. Table 1

shows the demographic characteristic distribution of the cohorts. The mean age of these 2 cohorts was 56 years, and men comprised 56% of the study population. The proportion of all comorbidities in the study cohort was greater than that in the comparison cohort (all P values < .0001).

The analysis with the start of follow-up since the index date (no lag period) showed that the participants with colonic diverticular disease had a signifcantly

increased risk of subsequent CRC (adjusted HR, 4.54; 95% CI, 4.19–4.91; P < .0001) (Table 2). The cumulative subsequent CRC incidence curve for the participants with colonic diverticular disease was signifcantly higher than the curve for the comparison group (Figure 1A, P value for log-rank < .0001). The high incidence of CRC in the frst year, especially in the frst 4 months, contributed mainly to the increased risk of subsequent CRC in the follow-up period.

After excluding the frst 12 months of follow-up

evaluation after a diagnosis of colonic diverticular disease, the incidence rate of developing CRC in the study

cohort was 15.13 per 10,000 person-years (Table 3) and was not signifcantly different than the comparison cohort (15.74 per 10,000 person-years). After adjusting for age, sex, hypertension, hyperlipidemia, and the Charlson Comorbidity Index, the risk was still not signifcantly different (adjusted HR, 0.98; 95% CI,

0.85–1.13). The Kaplan–Meier method was used to estimate the cumulative subsequent CRC incidence curve for

the study cohort, which was not signifcantly different than the curve for the comparison cohort (Figure 1B, P value for log-rank ? .6102). The study cohort was

divided into 2 subcohorts: diverticulosis and diverticulitis. Relative to the comparison cohort, both the diverticulosis and diverticulitis cohort were not associated

with an increased risk of developing CRC (adjusted HR, 0.83; 95% CI, 0.67–1.04; and adjusted HR, 1.08; 95% CI, 0.91–1.28, respectively).

We also applied an analysis to measure developing CRC risk in the study population by follow-up duration

(Table 4). The analysis showed that there were no signifcant differences in CRC risks between the study and

comparison cohorts in the follow-up period at more than 1 year, more than 2 years, more than 3 years, more than 4 years, and more than 5 years.

Table 5 shows the demographic-specifc incidence

borderline higher risk for CRC

compared with the comparison cohort in the age group of patients younger than 40 years (adjusted HR, 3.38; 95% CI, 1.33–8.57; P ? .010). For both males and females, the study cohort was not associated with an

increased risk of developing CRC compared with the comparison cohort. In the subgroup analysis stratifed by comorbidities, there were no signifcant differences between the study cohort and the comparison cohort after

adjusting for potential confounding factors.

Discussion

This was a large study performed to investigate the association of colonic diverticular disease and subsequent CRC with widespread coverage of a nationwide

database and complete follow-up evaluation. We found that colonic diverticular disease is not associated with CRC and any association shown in the frst year is likely an artifact of ascertainment bias.

Although the subgroup analysis for those younger than age 40 showed a borderline greater risk for CRC, there were only 18 events in the follow-up period. Therefore, we could not draw a conclusion regarding whether the younger patients with colonic diverticular disease had an increased risk of developing subsequent CRC, compared with those patients without colonic diverticular disease. For this issue, a large-scale cohort study focusing on young adults with a long-term followup period is needed.

In this study, all the excess risk of subsequent CRC was concentrated within the frst year after the diagnosis of colonic diverticular disease. It showed that patients with colonic diverticular disease were more likely to have a subsequent diagnosis of CRC when they were followed since the date of diagnosis of colonic diverticular disease (adjusted HR, 4.54; P < .0001) (Table 2 and Figure 1A). However, a lag period of 1 year would eliminate the effect of colonic diverticular disease on CRC (adjusted HR, 0.98) (Table 3 and Figure 1B). The increased risk in the frst

year could be attributed to misclassifcation and screening effect. Colonic diverticular disease–like symptoms were reported before the diagnosis of CRC. Colonic diverticular disease and CRC have similar clinical presentations because of the considerable overlap in symptoms. Both diseases share a symptom complex with abdominal pain and gastrointestinal bleeding. Thus, it is diffcult to distinguish CRC from colonic diverticular disease at the initial intervention, especially for mild disease activity. As a result, patients with CRC might be misclassifed as having colonic diverticular disease at frst, however, later

examinations disclose CRC. A screening effect might be another cause of the increased CRC risk in the frst year

after a diagnosis of colonic diverticular disease. It generally is agreed that a follow-up colonoscopy is necessary

for patients who have acute diverticulitis and whose diseases subside from an acute stage. Participants in the

study cohort, therefore, had colorectal examinations to a greater extent than those in the comparison cohort. This could increase the chance of a diagnosis of CRC.

In the published cohort studies and in the nationwide population-based case-control studies, no consistent data indicated whether colonic diverticular disease increased the risk of CRC. Stefansson et al24 reported a populationbased

cohort study to assess CRC risk after a diagnosis of colonic diverticular disease in 1993. From 1965 to 1983, there were 7159 cases with a hospital discharge diagnosis of colonic diverticular disease who were followed

up with a mean duration of 5.95 years in the Uppsala health care region of Central Sweden. After excluding the frst 2 years of follow-up evaluation, no statistically increased risk for total CRC existed (standard incidence ratio, 1.2; 95% CI, 0.9–1.6). The same research group conducted a subsequent study to examine the extent of difference in CRC risk between patients with sigmoid diverticulitis and patients with colonic diverticulosis devoid of any clinical signs of diverticulitis.13 They found

and sigmoid diverticulitis (odds ratio, 4.2; 95% CI, 1.3–13.0), but not colonic diverticulosis.

Granlund et al16 performed a case-control study

based on a nationwide population-based design. All patients diagnosed with colon cancer as their frst cancer from 1992 to 2006 in the Sweden Cancer Register were included. Increased risk only occurred within the frst year after a diagnosis of colonic diverticular disease. They hypothesized this was caused by the screening effect and misclassifcation, and therefore concluded that

colonic diverticular disease did not increase the longterm risk of colon cancer. Their results are consistent

with our fndings.

There are variations in the prevalence and pattern of diverticulosis among Western and Eastern populations. The prevalence rates are 5% to 52% in the West and 1% to 19% in Asia. In Western populations, colonic diverticular disease predominantly occurred in the left side

colon. In Eastern populations, it often developed in the right side of the colon.25–28

We used NHIRD inpatient data and those who had

diverticulosis diagnosed as an outpatient without further hospitalization were not included in the study cohort. If outpatients with diverticulosis are not associated with the risk of CRC, this will not affect our results. If outpatients with diverticulosis have an increased risk for

CRC and they initially are misclassifed into the comparison cohort, this will underestimate the result. However,

because the number of cases was small and the

relative risk of CRC was not markedly higher than in the general population, this group will not infuence the results of our study noticeably. Patients with diverticulitis

are less likely to be excluded from the study cohort. In

clinical practice, a vast majority of the patients are indicated for admission to the hospital, which is supported

by the NHI system in Taiwan because they often are treated with bowel rest and intravenous hydration. Collectively, our results could be used for inpatients,

however, the application to outpatients should be with caution.

The large sample obtained from our nationwide

population-based data set strengthened the statistical power of our examination of associations between colonic diverticular disease and subsequent CRC. The participants in this study presented a wide range of demographic characteristics, in which we could perform stratifed analyses according to age, sex, and comorbidities. The study population was well defned and the database we used generally was accurate.22,23

Widespread coverage of the nationwide database allowed us to trace all medical service-use histories for the participants and conduct a complete follow-up evaluation.

However, limitations existed in this study. First, the data set did not contain information regarding certain health-related factors, such as smoking, diet, body mass index, or family history of CRC, and these are known risk factors for CRC. We do not know their distribution in the study and comparison cohort, thus we could not adjust for their effect on CRC risk. Furthermore, colonic diverticular disease may co-exist with some environmental and lifestyle factors, such as obesity and low physical activity,

which also are associated with CRC. The impact from these confounders on our fndings is unknown. Readers should remain aware of this limitation when assessing the fndings in the present study. Second, participants with

diverticulosis may have been included in the comparison cohort. They were clinically indolent and did not undergo colonoscopy or an imaging study, or were missed in the examinations. Third, we could not include competing risks such as death from other causes because the NHIRD database does not provide information on death. This was a study limitation for analyzing death as a competing risk. Fourth, there was a relatively small number of subjects who developed subsequent CRC (1294 cases). However, it was still true that this study included large patient samples

with complete follow-up evaluation. We believe that this did not bias our result and that our study still has strong statistical power.

In conclusion, our study does not support the hypothesis that colonic diverticular disease is associated

with an increased risk of CRC. The increased risk is only in the frst year after a diagnosis of colonic diverticular disease and this may be explained by misclassifcation and screening effect.