Differential relationship between colorectal cancer and diabetes in a nationally representative sample of adults

Abstract

Background:

Diabetes has been identified as a risk factor for developing colorectal cancer (CRC); however, the literature identifying groups most at risk is sparse. This study aims to understand the relationship between CRC and diabetes by age and race/ethnicity.

Methods:

This is a cross-sectional study of data from the 2001–2014 National Health and Nutrition Examination Survey (unweighted n = 37,173; weighted n = 214,363,348). Individuals were categorized as having CRC if diagnosed with colon or rectal cancer and as having diabetes if told by a doctor they had diabetes, were taking insulin, or had an HbA1c ≥ 6.5%. Covariates included gender, age, race, marital status, educational level and income as a ratio of the poverty line. Multivariable logistic regression was used to assess the relationship between CRC and diabetes overall and stratified by age and by race.

Results:

24.32% of the sample with CRC also had diabetes. After adjusting for covariates, individuals with diabetes had a 47% greater probability of having CRC (p = 0.03). While significance did not persist after stratification for those ≥65 years (OR = 1.06, p = 0.74), those <65 years with diabetes had nearly 5-times higher odds of having CRC (OR = 4.78, p < 0.001). When stratified by race, both groups had statistically higher odds of having CRC; however, the odds for non-whites (OR = 1.87, p = 0.04) were higher compared to whites (OR = 1.54, p = 0.03).

Conclusion:

Individuals younger than 65 and racial/ethnic minorities have higher odds of CRC when also diagnosed with diabetes. Targeted interventions for these populations, especially regarding screening recommendations, may result in earlier detection of CRC and improved health outcomes.

1. Introduction

In 2017, there will be an estimated 135,430 diagnoses of and 50,260 deaths attributable to colorectal cancer (CRC) in the United States.¹ Based on 2014 estimates, CRC is the third most diagnosed cancer and accounts for the third most cancer-related mortalities.¹ Racial disparities exist as Non-Hispanic Blacks (NHBs) have a 20% and 40% greater risk of developing and dying, respectively, from CRC compared to Non-Hispanic Whites (NHWs).¹ Evidence suggests these increased rates of incidence and mortality among NHBs compared to NHWs are likely due to social factors, including lower socioeconomic status, lifestyle factors such as less engagement in physical exercise, and decreased access to and use of health-care among NHBs, rather than biological factors.^{2, 3} Several studies have noted racial and ethnic disparities with respect to mortality, treatment, screening and stage of diagnosis of CRC, with racial and ethnic minorities typically being diagnosed with a later stage of CRC and experiencing poorer quality of care and health outcomes than NHWs.^1–6

It is estimated that >30 million Americans, 9.4% of the population, live with either diagnosed or undiagnosed diabetes.⁷ Giovannucci hypothesized that since similar factors contribute to both CRC and diabetes, it is possible that diabetes may be a risk factor for developing colon cancer.⁸ Hyperglycemia and hyperinsulinemia are two proposed mechanisms supported by biological and epidemiological evidence that suggest how diabetes causes CRC.^9–11 Hyperglycemia may lead to CRC either by increasing fecal bile acid concentrations and stool transit times or by providing colorectal neoplasia with the glucose necessary for tumor growth as cells switch from aerobic to anaerobic metabolism.¹⁰ Hyperinsulinemia mediates CRC development through insulin, a growth factor for both normal and cancerous colon muscosal cells, and insulin-like growth factor-1 (IGF-1), which limits the abilities of cells to undergo apoptosis, pushes the cell cycle forward and promotes angiogenesis.^{10, 11} A meta-analysis of twenty-one unique case-control and cohort studies conducted in the United States and in Europe found that having diabetes conferred a 30% greater probability of developing CRC in both women and men across all colorectal sites.¹² More recent meta-analyses that controlled for potentially confounding variables¹³ and included studies with more subjects¹⁴ similarly found that diabetes is a risk factor for both colon and rectal cancers. Furthermore, evidence suggests that different treatments for diabetes alternatively affect CRC risk. It has been shown that treatment of diabetes with insulin and sulfonylureas increases CRC risk,^15–19 especially as duration of treatment increases.²⁰ Alternatively, metformin use has been negatively associated with CRC risk.^{15, 17, 18} Given the proposed relationship between diabetes and CRC and estimates that predict an increased prevalence of diabetes in the coming years,²⁰ the incidence of CRC may also continue to increase as a result.¹²

Limited evidence has examined (1) the risk for developing CRC when diagnosed with diabetes by race and ethnicity, and (2) the relationship between colorectal adenomas, cancers or neoplasms and diabetes by age. Evidence demonstrates an increased risk for CRC when diagnosed with diabetes across all races.^{16, 21–23} Younger adults with diabetes have an increased risk for the development of colorectal adenomas, cancers or neoplasms than individuals without diabetes of comparable age.^{24, 25} The US Preventive Services Task Force (USPSTF) recommends that screening for CRC begins at age 50 for individuals with “average-risk” for CRC.²⁶ This recommendation appears to correspond to the decrease in incidence of CRC among Americans age 50 and older.^{1, 27} However, it has been shown that colorectal screening rates, particularly by colonoscopy, was higher among men and women 65 years and older than it was among men and women 50 to 64 years of age.²⁸ This is likely a result of improved access among those 65 years and older as they are beneficiaries of Medicare.²⁸ Additionally, the 22% increase in the incidence of CRC among individuals under age 50 has been attributed to factors associated with increased diabetes risk, such as obesity, unhealthy diet and a lack of physical activity.^{1, 8, 29} Evidence that CRC may begin to develop at ages younger than 50 suggests the potential need for screening to begin earlier than the current recommendations.¹ In addition to age differences in the use of colorectal screening, differences exist with respect to race and ethnicity as NHWs are more likely to be screened than minorities.^{6, 30, 31} The lack of evidence assessing the relationship between CRC and diabetes by race/ethnicity and age, coupled with the abundance of literature demonstrating differential CRC screening by race/ethnicity and age, suggests a need to further understand these relationships. Therefore, this analysis seeks to ascertain the relationship between CRC and diabetes by race/ethnicity and age.

2. Methods

2.1. Study population

The National Center for Health Statistics (NCHS) conducts the National Health and Nutrition Examination Survey (NHANES) and releases data in two year intervals. This current study makes use of seven of the two year cross-sectional surveys, beginning with the 2001–2002 survey and ending with the 2013–2014 survey. NHANES provides a representative sample of the U.S. civilian noninstitutionalized population by means of complex multistage probability sampling and oversampling certain groups such as ethnic minorities, low income, and elderly persons. Each survey consists of an interview regarding demographic, socioeconomic, dietary, and health-related information conducted at home and a physical health examination with laboratory tests completed at a mobile examination center. The Center for Disease Control (CDC) provides further details regarding NHANES design and procedures.³²

We analyzed data from the survey interview and physical examination within NHANES from 2001 to 2014. Some demographic variables were asked only for age 20 years and older, so the study population was limited to adults 20 years of age and older, who participated both in interview and examination, and responded to questions regarding diagnosis of colorectal cancer. In total, 37,173 participants were used in the analysis.

2.2. Colorectal cancer classification

Individuals were categorized as having colorectal cancer if they answered that they had colon or rectal cancer when asked “What kind of cancer was it?” as a follow-up to having answered yes to “Have you ever been told by a doctor or health professional that you had cancer or a malignancy of any kind?”

2.3. Diabetes classification

Individuals were categorized as having diabetes if they answered “yes” to any of the following three questions: “Have you ever been told by a doctor or other health professional that you have diabetes or sugar diabetes?”, “Are you now taking insulin?”, or “Are you now taking diabetic pills to lower your blood sugar?”. Additionally, to capture those that may have limited healthcare access, however, are living with undiagnosed diabetes, we included participants that answered “no” to the above questions, but had a measured blood hemoglobin HbA1c ≥ 6.5%, which is the accepted range for diagnosis of diabetes according to American Diabetes Association guidelines.³³

2.4. Covariates

All analyses were adjusted for several demographic variables, including age, gender, race/ethnicity, education, marital status, and income. In order to conserve power for the analysis, all covariates were dichotomized. Covariates included gender, age (dichotomized as <65 years and 65 years or older), race/ethnicity (dichotomized as whites and non-whites), education (dichotomized as high school or below and college or above), marital status (dichotomized as married or not married) and ratio of family income to poverty (dichotomized as 130% and less of poverty level and above 130% of poverty level).

2.5. Statistical analysis

The primary analytical method was an estimation of the relative odds for CRC using logistic models. First, the CRC status (with CRC, no CRC) were compared by demographics using Rao-Scott chi-square statistics. Second, univariate and multivariable survey logistic regressions were used to assess the relationships between CRC and diabetes, controlling for demographic variables including age; gender; race/ethnicity; education; marital status; poverty level. Third, univariate and multivariable models were run stratified by the following domains: 1) age < 65 years old and age ≥65 years old; 2) Race: White and non-White. All p values were 2-sided and p < 0.05 was considered statistically significant. NCHS recommendations were followed in the analyses to account for the complex survey design (including oversampling), survey nonresponse, and post-stratification. Data analysis was performed using SAS version 9.4 (SAS Institute). The SURVEYFREQ and SURVEYLOGISTIC procedures in SAS were used to account for the complex survey sampling design.

3. Results

The population sample included 37,173 adults 20 years and older, which represented 214,363,348 US adults 20 years and older. Table 1 provides sample demographics. The population’s demographics were 51.95% female, 82.77% younger than 65 years old, 69.16% white, 57.93% with a college level education or more, 56.25% married and 73.16% had a ratio of family income that was above 130% of the poverty level.

Table 1.

Weighted sample demographics (unweighted: 37,173; weighted: 214,363,348).

Variables	Percentage
Gender
Male	48.05%
Female	51.95%
Age group
<65 years	82.77%
65 years and older	17.23%
Race
White	69.16%
Non-white	30.84%
Education level
High school or below	41.96%
College or above	57.93%
Marital status
Married	56.25%
Not married	43.68%
Median income
130% and below poverty line	20.51%
Above 130% of poverty line	73.16%
Diabetes
No	89.39%
Yes	10.61%
Colorectal cancer
No	99.43%
Yes	0.57%

Table 2 provides information on the unadjusted relationship between colorectal cancer and diabetes. 10.53% of the sample population not diagnosed with colorectal cancer had diabetes, while 24.32% of those diagnosed with colorectal cancer had diabetes.

Table 2.

Unweighted and weighted relationship between colorectal cancer and diabetes.

Diabetes	Unweighted			Weighted
	No CRC	CRC	P-value	No CRC	CRC	P-value
No	31,573	196	<0.0001	190,688,981	925,890	<0.0001
	(85.56%)	(71.79%)	(89.47%)	(75.68%)
Yes	5327	77	22,450,926	297,550
	(14.44%)	(28.21%)	(10.53%)	(24.32%)

Table 3 presents the adjusted analyses for the overall population. After adjustment of relevant sociodemographic variables, individuals with diabetes had an adjusted odds ratio of 1.47 (95% CI: 1.04, 2.07) of having CRC.

Table 3.

Colorectal cancer multivariable logistic regression analysis.

	Adjusted odds ratio	95% CI	P-value
Diabetes			0.0301
No	Ref	Ref
Yes	1.47***	1.04–2.07
Gender			0.2564
Male	Ref	Ref
Female	1.21	0.87–1.70
Age group			<0.0001
<65 years old	Ref	Ref
≥65 years old	9.46***	6.70–13.36
Race			<0.0001
White	Ref	Ref
Non-white	0.52***	0.39–0.70
Education level			0.1430
High school or below	Ref	Ref
College or above	1.26	0.92–1.71
Marital status			0.3492
Married	Ref	Ref
Not married	0.86	0.63–1.18
Ratio of family income to poverty			0.1899
130% and less of poverty level	Ref	Ref
Above 130% of poverty level	0.82	0.60–1.11

Bold text indicates significant findings.

^*Level of significance p < 0.05,

^**Level of significance p < 0.01,

^***Level of significance p < 0.001.

Table 4 shows adjusted analyses stratified by age and race. Stratification by age revealed that individuals with diabetes who were under 65 years old had a statistically significant odds ratio of 4.78 (95% CI:2.57, 8.89) (p < 0.0001) for having CRC, whereas the odds ratio for individuals 65 years or older was not significant. When the analysis was stratified by race, whites with diabetes had an odds ratio of 1.54 (95% CI:1.04, 2.28) (p = 0.03), while non-whites with diabetes had an odds ratio of 1.87 (95% CI:1.01, 3.47) (p = 0.04) for having CRC.

Table 4.

Stratified analysis for the relationship between colorectal cancer and diabetes by age and race.

	Age stratified						Race stratified
	<65 years old			65 years old and older			White			Non-white
	Adjusted odds ratio	95% CI	P-value	Adjusted Odds Ratio	95% CI	P-value	Adjusted odds ratio	95% CI	P-value	Adjusted odds ratio	95% CI	P-value
Diabetes			p < 0.0001			p = 0.74			p = 0.03			p = 0.04
	4.78***	2.57–8.89		1.06	0.74–1.53		1.54*	1.04–2.28		1.87*	1.01–3.47
Gender			p = 0.19			p = 0.79			p = 0.11			p = 0.32
Female	1.53	0.80–2.91		1.05	0.74–1.47		1.33	0.93–1.91		0.76	0.44–1.32
Age	–	–	–	–	–	–			p < 0.0001			p < 0.0001
b65 years old	–	–	–	–	–	–	Ref			Ref
≥65 years old	–	–	–	–	–	–	8.42***	5.54–12.80		14.07***	6.98–28.37
Race			p < 0.01			p = 0.02	–	–	–	–	–	–
White	Ref			Ref			–	–	–	–	–	–
Non-white	0.36**	0.21–0.63		0.66*	0.46–0.94		–	–	–	–	–	–
Education			p = 0.48			p = 0.88			p = 0.57			p = 0.69
College or above	1.32	0.60–2.87		1.03	0.72–1.46		1.11	0.78–1.57		1.13	0.63–2.00
Marital status			p = 0.14			p = 0.38			p = 0.28			p = 0.03
Not married	0.65	0.36–1.16		1.17	0.82–1.66		0.84	0.61–1.16		1.93*	1.05–3.54

Bold text indicates significant odds ratios for the relationship between CRC and diabetes in stratified analyses.

^*Level of significance p < 0.05,

^**Level of significance p < 0.01,

^***Level of significance p < 0.001.

4. Discussion

Using a nationally representative data set, we found that those with diabetes have a 47% greater likelihood of also having CRC. Furthermore, when stratifying by age and race, those with diabetes were nearly five times more likely to also have CRC if they were younger than 65, and those with diabetes who were not white had an 87% greater risk of also having CRC. While the first finding is in agreement with the great amount of literature that exists on the relationship between CRC and diabetes, it is, to our knowledge, the first to use a nationally representative survey to do so. Additionally, the finding of a differential relationship between CRC and diabetes between adults of different ages and races/ethnicities is among the first evidence identifying increased risk for the relationship between CRC and diabetes by sociodemographics.

Previous studies have demonstrated that diabetes confers a relative risk for developing CRC that ranges between 20 and 60%.^{12, 13, 22, 34–36} In this nationally representative sample of adults, we found the risk of having CRC when diagnosed with diabetes to be 47%; however, individuals with undiagnosed diabetes are captured in this estimate by including those with a HbA1c ≥ 6.5% in our definition of diabetes. Larsson et al. discuss that because diabetes is underdiagnosed, the relationship between CRC and diabetes may be underestimated.¹² Accordingly, our estimated 47% increase in the risk between CRC and diabetes provides an estimate accounting for undiagnosed diabetes.

To our knowledge, our findings are among the first to demonstrate a differential relationship between CRC and diabetes by age. While La Vecchia et al. reported an increased risk of CRC among individuals who had diabetes for >10 years and were 60 years and older, their results were not significant. In contrast, our result suggests that those who are <65 years old are at an increased risk of having both diabetes and CRC. CRC usually develops in individuals older than 50, with the median age at diagnosis being 68 years.^{37, 38} However, the findings of Vu et al. demonstrated that individuals with diabetes between 40 and 49 are more likely to develop CRC adenomas than their peers without diabetes.³⁹ This finding suggests that diabetes may alter the timeline of CRC development in younger individuals with diabetes, making them a subpopulation of greater interest for screening.^{23, 24} Though we cannot comment on whether the screening age for those with diabetes should be <50 due to how we grouped individuals by age, our result does suggest that screening those younger than 65 with diabetes needs to be a priority due to their significantly increased risk of developing CRC. Current literature reveals that CRC screening rates are suboptimal for all age groups, but especially for those 64 years old and younger and minorities.^{1, 40, 41} This is speculated to be the case in individuals younger than 65 years because they do not qualify for Medicare reimbursement.²⁸ Future studies should continue to examine the relationship between CRC and diabetes by age and employ an alternate age dichotomy that examines clinical implications, in addition to the access implications examined by this study. If supported by further evidence, modifications to screening recommendations for individuals with diabetes and interventions that improve screening for those younger than 65 should be implemented.³⁹ The mechanism underlying the stronger relationship between CRC and diabetes in younger individuals also needs to be understood. It has been shown that the age of diabetes onset has decreased,⁴² which could play a role in the effect observed by this study. Considering that it has been shown that diabetes rates are increasing among adolescents and that it is predicted that diabetes prevalence will quadruple among adolescents and increase by 75% among those 20–44 by 2050, the increased relative risk for CRC among individuals with diabetes younger than 65 could potentially affect a great number of people.^43–45

In addition to differences by age, we also found differences in the co-occurrence of CRC and diabetes by race. Our results are supported by those of Cavicchia et al., that minorities have an increased likelihood of developing CRC if they have diabetes.¹⁶ While Cavicchia et al. found that the risk was approximately 70% higher among NHBs for developing colon or rectal cancer compared to NHWs, we found minorities to have an 87% higher odds and whites to have a 54% higher odds of having colorectcal cancer and diabetes.¹⁶ The reasons for this may be our use of national data, compared to their use of data from only individuals from South Carolina, or our grouping of all racial and ethnic minorities into one category, compared to their study of only NHBs. Cavicchia et al. suggest the need for better diabetes guidelines and interventions to mitigate the effects of diabetes on CRC among NHBs who have been shown to have higher rates of diabetes, worse glycemic control, and more complications.^{16, 46} Interventions that target disparities in the incidence and treatment of diabetes can help reduce the greater risk minorities have for developing CRC. It has already been shown that NHBs have less access to CRC screening, experience unique barriers to screening, and have suboptimal screening rates;^{3, 6, 30} therefore, solutions tailored to specific needs of different population subgroups are warranted. Considering the increased risk for CRC associated with diabetes and race/ethnicity, these solutions must take perspectives from the community and population subgroups into account.

This study has strengths and limitations that should be acknowledged. Strengths include the use of a nationally representative data set, which allows for the relationship between CRC and diabetes to be generalized to a large population. Secondly, the inclusion of undiagnosed diabetes by virtue of an individual having an HbA1c ≥ 6.5%, ensures that we do not underestimate the relationship between CRC and diabetes, especially if CRC is to be understood as a biological outcome of diabetes. Limitations of this study include the use of cross-sectional data, which removes the possibility of detecting causality or inferring a temporal relationship between the development of diabetes or colorectal cancer, and the limited number of individuals with both CRC and diabetes in the unweighted population, which limited the power of these analyses. While dichotomizing covariates conserved power, future analysis using larger numbers of individuals will be useful in supporting these findings. Finally, additional factors that could explain the relationship between CRC and diabetes, such as health behaviors, were not included in the analysis, but may be of interest in future investigations.

5. Conclusions

In this nationally representative sample of adults, having diabetes increased probability of also having CRC compared to those without diabetes. Individuals younger than 65 years of age and racial/ethnic minorities have higher relative odds of CRC if also diagnosed with diabetes. These observations of a differential relationship between CRC and diabetes among adults by age and race/ethnicity warrant further investigation into the mechanisms for these relationships and the development of interventions targeted to populations most at risk.