Periodontal disease, tooth loss, and colorectal cancer risk: results from the Nurses’ Health Study

Abstract

Periodontal diseases including tooth loss might increase systemic inflammation, lead to immune dysregulation, and alter gut microbiota, thereby possibly influencing colorectal carcinogenesis. Few epidemiological studies have examined the association between periodontal diseases and colorectal cancer (CRC) risk. We collected information on the periodontal disease (defined as history of periodontal bone loss) and number of natural teeth in the Nurses’ Health Study. A total of 77,443 women were followed since 1992. We used Cox proportional hazard models to calculate multivariable hazard ratios (HRs) and 95% confidence intervals (95% CIs) after adjustment for smoking and other known risk factors for CRC. We documented 1,165 incident CRC through 2010. Compared to women with 25–32 teeth, the multivariable HR (95% CI) for CRC for women with < 17 teeth was 1.20 (1.04–1.39). With regard to tumor site, the HRs (95% CIs) for the same comparison were 1.23 (1.01–1.51) for proximal colon cancer, 1.03 (0.76–1.38) for distal colon cancer, and 1.48 (1.07–2.05) for rectal cancer. Additionally, compared to those without periodontal disease, HRs for CRC were 0.91 (95% CI 0.74–1.12) for periodontal disease, and 1.22 (95% CI 0.91–1.63) when limited to moderate to severe periodontal disease. The results were not modified by smoking status, body mass index, or alcohol consumption. Women with fewer teeth, possibly moderate or severe periodontal disease, might be at a modest increased risk of developing CRC, suggesting a potential role of oral health in colorectal carcinogenesis.

INTRODUCTION

Colorectal cancer is the third most common cancer in the US and often arises within the context of increased inflammation and altered immunity as well as changes in the gut microbiota^1–3. Previous studies have shown that aspirin, a nonsteroidal anti-inflammatory drug (NSAID), is associated with a lower risk of colorectal cancer⁴. In addition, plasma levels of certain inflammatory biomarkers or proinflammatory cytokines, such as soluble tumor necrosis factor receptor 2 (TNFRSF1B, or sTNFR-2), interleukin-8 (IL8), and interleukin-6 (IL6) were associated with an increased risk of colorectal cancer in some but not all studies^5–8.

Poor oral health might increase systemic inflammation, lead to local overly aggressive immune response, and potentially alter gut microbiota by pathogens that cause periodontal disease, and thus could have important implications for colorectal cancer development^{9, 10}. Two commonly used measures of oral health are periodontal disease and tooth loss^{11, 12}. Periodontal disease is a chronic inflammatory disease of the supporting tissues of the teeth¹³, and in older populations, tooth loss is the main manifestation of persistent periodontal disease. This process develops over many years through the spread of bacterial infections and inflammation of the gums to the ligaments and bone that support the teeth^14–16.

Previous studies have reported positive associations between periodontal diseases and risk of cancers at various sites including oral cancer, lung cancer, pancreas cancer, breast cancer, and upper gastrointestinal cancer^17–24. In addition, some studies have reported that tooth loss was independently associated with an increased risk of head and neck cancer, esophageal cancer, and lung cancer^25–27. However, the association between periodontal disease, tooth loss, and colorectal cancer remains much less studied. Among the four studies identified, null associations were reported for history of periodontal disease with colorectal cancer risk^{20, 26, 28, 29}. In contrast, another study observed that periodontal disease was associated with excess colorectal cancer mortality³⁰. Hence, although oral health might play an important role in colorectal carcinogenesis, related epidemiological studies are limited and have not resulted in definitive conclusions.

Herein, we conducted this study and hypothesized that individuals with a history of periodontal disease and tooth loss might be at a greater risk for development of colorectal cancer, independent of smoking, an important contributor to tooth loss. We tested this hypothesis in the Nurses’ Health Study (NHS).

METHODS

Study population

The NHS has been described in detail elsewhere^{31, 32}. Briefly, the NHS is an ongoing prospective cohort study of 121,700 female nurses aged 30–55 years in 1976 living in 11 states in the United States. Since 1976 participants have completed questionnaires regarding information on medical history and lifestyle factors and newly diagnosed diseases every two years. The follow-up rate has been greater than 90%. The study was approved by the Brigham and Women’s Hospital institutional review board, and return of the questionnaires was considered to imply informed consent.

Assessment of periodontal disease and tooth loss

Information on periodontal disease, defined as a history of periodontal bone loss, was first assessed in 1998 and then again in 2000. If periodontal bone loss was reported, participants indicated the severity of the bone loss (none, mild, moderate/severe). Although self-reported history of periodontal disease was not validated in the NHS cohort, it has been validated by medical/dental records or radiograph review in the Health Professionals Follow-up Study (HPFS), a comparable cohort of male health professionals^{26, 33, 34}. The validation of this question in the male health professionals was done by obtaining radiographs from subset of individuals with and without self-reported history of periodontal disease. Interproximal bone loss was evaluated for each participant in all posterior teeth (premolars and molars) except third molars by a clinician who was masked to the self-report. Interproximal bone loss is an indicator of cumulative periodontal disease. Among non-dentists male health professionals in that study, the positive predictive value of the question on periodontal bone loss was 0.80 and the negative predictive value was 0.68³⁴.

On the 1992 questionnaire we also collected information on the number of natural teeth (none,1–10, 11–16, 17–24, 25–32). The question does not consider the replaced teeth (implant or bridge) as natural teeth. Again, this question was not validated in NHS, however, the correlation between self-reported number of teeth and the number of teeth from a clinical assessment in a general population sample was 0.97³⁵. In addition, women were also asked the number of tooth lost in the prior 2 years (on the 1992, 1996, and 2000 questionnaire).

Assessment of other colorectal cancer risk factors

Information on height was obtained in 1976. Alcohol consumption and other dietary factors, such as intakes of folate, calcium, and vitamin D, were first assessed in 1980 using a validated semi quantitative food frequency questionnaire, and every nearly 4 years thereafter³⁶. The baseline and biennial questionnaires collected information on other colorectal cancer risk factors such as adult body weight, physical activity (METs-hrs/wk), cigarette smoking, history of sigmoidoscopy/colonoscopy screening, family history of colorectal cancer, aspirin use, and postmenopausal hormone use (PMH).

Ascertainment of cases of colorectal cancer

The ascertainment of colorectal cancer cases has been described in detail elsewhere³⁷. Briefly, participants were asked on biennial questionnaires to report a diagnosis of colorectal cancer and other diseases. When a diagnosis was reported, medical and pathology records were obtained and the incidence of colorectal cancer was confirmed. The physicians assessing medical records were blinded to exposure information. Colorectal cancer was defined according to the International Classification of Diseases, Ninth Revision (ICD-9). Deaths were identified through the registry of National Death Index³⁸ or by family report.

Statistical analysis

We used 1992 as baseline for number of teeth analysis and 1998 as baseline for periodontal disease (i.e., self-reported periodontal bone loss) because the information was first assessed in those years. We excluded participants with a history of cancer at or before baseline (except non-melanoma skin cancer), death before baseline, with ulcerative colitis or Crohn diseases, and with missing information on the exposure of interest. The analytic cohort included 77,449 women with 1,291,462 person-years through 2010 for the analysis of number of teeth, and 69,656 women with 779,439 person-years through 2010 for the periodontal bone loss analysis. Person-years were calculated from baseline questionnaire return date to the date of colorectal cancer diagnosis, death or the end of follow-up (May 31, 2010), whichever occurred first. Women with other cancers were censored at their date of diagnosis.

We used Cox proportional hazards models³⁹ to estimate the age- and multivariable-adjusted hazard ratios (HR) and 95% confidence intervals (95% CI) for tooth loss or periodontal bone loss. Consistent with the categories used in a previous study²⁶, we analyzed the data on number of teeth using 25–32 as the reference group (other two groups were: 17–25 and 0–16). For the analysis on periodontal disease, we first categorized periodontal bone as a binary variable (no vs. mild/moderate/severe), and further evaluated the association with severity (no vs. mild vs. moderate/severe). As a sensitivity analysis, we examined the association with combined categories of number of teeth and periodontal bone loss. We also evaluated these associations by anatomic subsites (i.e., colon, proximal colon, distal colon and rectal cancers). Lastly, we evaluated whether the associations varied by smoking status (never, ever), body mass index (BMI; < 25, ≥ 25 kg/m²), and alcohol consumption (< 5, ≥ 5 g/day). We used a Wald test to examine whether the coefficients of the cross-product terms between these variables and each of periodontal disease measures were statistically significant.

All models were stratified by age (continuous in months) and calendar time. We observed no violation of the proportional hazard assumption based on the likelihood ratio test. We adjusted for known colorectal cancer risk factors and accounted for covariates that can change over time, such as weight and physical activity (see table footnote for these variables and their categorizations). All analyses were conducted using Statistical Analysis Software version 9.3 (SAS) (Cary, North Carolina).

RESULTS

We documented 1,165 incident colorectal cancer cases over 18 years of follow-up for number of teeth analysis, and 739 cases for the periodontal bone loss analysis. Women with fewer teeth (0–16) at baseline were older and more likely to be past or current smokers and have type 2 diabetes, but less likely to use PMH, and had lower intakes of alcohol, calcium and vitamin D (Table 1). A greater proportion of women with history of periodontal bone loss had ever smoked compared to those with no history of periodontal disease. Alcohol consumption was slightly higher in women with history of periodontal bone loss.

Table 1.

Baseline age-adjusted characteristics of participants by number of teeth (1992) and periodontal disease with bone loss (1998) in the Nurses’ Health Study

	Number of natural teeth			History of periodontal disease with bone loss
	25–32 (n=45,252)	17–24 (n=14,932)	0–16 (n=12,564)	No (n=59,931)	Yes (n=9,725)
Age, years*	57.0(7.0)	59.8(6.9)	61.9(6.2)	64.3(7.2)	64.5(6.9)
White, %	97.9	96.9	96.7	97.5	97.4
Body mass indexa, kg/m2	24.5(4.1)	25.3(4.5)	25.8(4.8)	25.3(4.4)	24.8(4.1)
Activity, METsb -hrs/week	17.3(18.1)	16.0(18.1)	14.9(17.2)	17.5(16.8)	17.4(16.3)
Family history of colorectal cancer, %	13.5	13.6	13.4	14.5	14.5
Regular aspirin use (2 or more tablets/wk), %	37.1	38.9	39.7	42.6	42.0
Past smoking, %	42.0	45.6	46.0	43.4	52.7
Current smoking, %	7.5	12.3	19.4	9.6	17.5
Type 2 diabetes, %	2.9	4.3	5.7	5.8	5.5
History of sigmoidoscopy/colonoscopy, %	18.8	16.0	13.9	31.6	32.7
Postmenopausal status, %	96.6	97.4	98.1	96.6	97.3
PMH use among menopausal women, %	68.7	64.0	57.7	67.6	70.5
Multivitamin use, %	43.5	41.6	40.3	59.0	60.1
Dietary intakesc
Alcohol, g/day	6.3(9.2)	6.0(9.3)	5.6(9.3)	5.6(8.4)	6.8(9.3)
Total calcium intake, mg/d	933(339)	888(327)	856(333)	1015(364)	1036(372)
Total folate intake, ug/d	404(185)	386(178)	373(182)	458(178)	456(177)
Total vitamin D, IU/d	338(203)	324(196)	315(200)	363(190)	368(193)
Red meat, servings/wk	2.1(1.3)	2.1(1.3)	2.1(1.3)	1.9(1.1)	1.9(1.1)
Processed meat, servings/wk	0.9(1.1)	1.0(1.2)	1.1(1.2)	0.9(1.0)	0.9(1.0)

Values are means(SD) or percentages and are standardized to the age distribution of the study population.

^aBody mass index was calculated as weight in kilograms divided by the square of height in meters.

^bHours of metabolic equivalent tasks.

^cDietary intakes were estimated with food-frequency questionnaire (1992 used for number of natural teeth; 1998 used for periodontal disease with bone loss).

^*Value is not age adjusted

Among women with information on both number of teeth and periodontal disease in 1998, 3,079 of 12,245 (17%) of the participants with <17 remaining teeth, 2,652 of 15,150 (18%) of those with 17–24 remaining teeth, and 5,638 of 46,457 (12%) of those with over 24 remaining teeth had periodontal disease. Thus, tooth loss and periodontal disease were partially correlated and provide different information on oral health.

Compared with women with 25–32 teeth, women with <17 teeth were at a higher risk of incident colorectal cancer (HR: 1.20, 95% CI: 1.04–1.39) (Table 2). Positive associations were observed for proximal colon cancer (HR: 1.23, 95% CI: 1.01–1.51) and rectal cancer (HR: 1.48, 95% CI: 1.07–2.05), but not with distal colon cancer (HR: 1.03, 95% CI: 0.76–1.38), although none of the heterogeneity test by subsite was statistically significant (all p-values > 0.05). In addition, any incident tooth loss (measured 1992–2000) during follow-up was not associated with colorectal cancer risk (HR: 0.89; 95% CI: 0.76–1.03).

Table 2.

Age and multivariable^a adjusted hazard ratios for colorectal cancer according to number of natural teeth in the Nurses’ Health Study

		Number of natural teeth
		25–32	17–24	0–16
Colorectal cancer
No. cases (n=1,165)		618	245	302
Age-adjusted		1 (reference)	1.03 (0.89–1.20)	1.44 (1.25–1.66)
Multivariablea		1 (reference)	0.94 (0.80–1.09)	1.20 (1.04–1.39)b
Colon cancer
No. cases (n=926)		499	189	238
Multivariablea		1 (reference)	0.88 (0.74–1.04)	1.14 (0.97–1.35)
Proximal colon cancer
No. cases (n=600)		317	121	162
Multivariablea		1 (reference)	0.89(0.72–1.10)	1.23 (1.01–1.51)c
Distal colon cancer
No. cases (n=305)		169	64	72
Multivariablea		1 (reference)	0.86(0.64–1.16)	1.03 (0.76–1.38)c
Rectal cancer
No. cases (n=239)		119	56	64
Multivariablea		1 (reference)	1.20(0.88–1.67)	1.48 (1.07–2.05)c

^a: Multivariable hazard ratios were adjusted for age (in month), race (Caucasian or not), smoking before age 30 (0, 1–4, 5–< 10, or ≥10 pack-years), history of colorectal cancer in a parent or sibling (yes or no), history of sigmoidoscopy / colonoscopy (yes or no), current physical activity (< 3, 3 –< 27, ≥ 27 METs-hrs/wk), regular aspirin use (yes or no), multivitamin use (yes or no), type 2 diabetes (yes or no), alcohol consumption (< 5, 5 –<15, or ≥ 15 g/d), adult BMI (< 25, 25 –< 27.5, 27.5– < 30, ≥ 30 kg/m²), energy-adjusted intake of total calcium, vitamin D, folate, red meat and processed meat (all in tertiles) and postmenopausal hormone use (premenopausal, never, past, or current user).

^b: HR was 1.24 (95% CI: 0.92–1.66) among never smokers.

^c: p-values for heterogeneity were 0.33 comparing proximal colon cancer and distal colon cancer, 0.35 comparing proximal colon cancer and rectal cancer, and 0.11 comparing distal colon cancer and rectal cancer.

Compared to women with no reported history of bone loss, self-reported history of periodontal bone loss was not associated with colorectal cancer risk overall (HR: 0.89, 95% CI: 0.72–1.10) or by subsite (Table 3). Interestingly, when considering severity, women with moderate or severe periodontal disease were at a suggestively higher risk of colorectal cancer (HR: 1.22, 95% CI: 0.91–1.63) compared to women without history of bone loss.

Table 3.

Age and multivariable adjusted hazard ratios for colorectal cancer according to self-reported periodontal disease with bone loss in the Nurses’ Health Study.

		Periodontal disease with bone loss			Periodontal disease with bone loss severity
		No	Yes	No	Mild	Moderate-severe
Colorectal cancer
No. cases (n=739)		641	98	641	48	50
Age-adjusted		1 (reference)	0.91 (0.74–1.12)	1 (ref)	0.72 (0.54–0.97)	1.30 (0.97–1.74)
Multivariablea		1 (reference)	0.89 (0.72–1.10)	1 (ref)	0.73 (0.55–0.98)	1.22 (0.91–1.63)
Colon cancer
No. cases (n=596)		518	78	518	40	38
Multivariablea		1 (reference)	0.88 (0.70–1.12)	1 (ref)	0.76 (0.55–1.04)	1.13 (0.81–1.58)
Proximal colon cancer
No. cases (n=395)		341	54	341	29	25
Multivariablea		1 (reference)	0.89(0.67–1.18)	1 (ref)	0.81 (0.55–1.19)	1.12 (0.74–1.70)
Distal colon cancer
No. cases (n=185)		162	23	162	10	13
Multivariablea		1 (reference)	0.89(0.58–1.36)	1 (ref)	0.64 (0.34–1.21)	1.27 (0.71–2.27)
Rectal cancer
No. cases (n=143)		123	20	123	8	12
Multivariablea		1 (reference)	0.92(0.57–1.49)	1 (ref)	0.65 (0.31–1.33)	1.60 (0.87–2.95)

^a: Multivariable hazard ratios were adjusted for age (in month), race (Caucasian or not), smoking before age 30 (0, 1–4, 5–< 10, or ≥10 pack-years), history of colorectal cancer in a parent or sibling (yes or no), history of sigmoidoscopy / colonoscopy (yes or no), current physical activity (< 3, 3 –< 27, ≥ 27 METs-hrs/wk), regular aspirin use (yes or no), multivitamin use (yes or no), type 2 diabetes (yes or no), alcohol consumption (< 5, 5 –<15, or ≥ 15 g/d), adult BMI (< 25, 25 –< 27.5, 27.5– < 30, ≥ 30 kg/m²), energy-adjusted intake of total calcium, vitamin D, folate, red meat and processed meat (all in tertiles) and postmenopausal hormone use (premenopausal, never, past, or current user).

When we jointly considered the number of teeth and history of periodontal disease, none of results reached statistical significance except that individuals who had both history of periodontal bone loss and <17 teeth had a significantly higher risk of rectal cancer (HR: 2.20, 95% CI: 1.05–4.63, based on eight cases) compared to those with neither (Supplementary Tables 1, 2). Lastly, the observed results for number of teeth or periodontal diseases were not modified by smoking status, BMI, or alcohol consumption (all p-value for interaction > 0.10, data not shown).

DISCUSSION

In this prospective cohort study, women with <17 teeth were at a modest 20% higher risk for developing colorectal cancer compared to women with over 24 teeth, independent of smoking and other known colorectal cancer risk factors. Significant positive association was seen for proximal colon cancer but not distal colon cancer, and consistently appeared to be stronger for rectal cancer, although the difference by tumor subsite was not statistically significant. History of self-reported periodontal disease with bone loss was not associated with colorectal cancer risk, although a suggestive positive association was seen for moderate to severe periodontal disease bone loss.

The majority of published studies on tooth loss and cancer have been on upper gastrointestinal cancers, including oral and gastric cancers¹⁵. We identified two studies that have examined the association between tooth loss and colorectal cancer and both did not observe statistically significant associations for colorectal cancer although tooth loss was positively associated with total cancer in one study²⁶ and with orodigestive cancer in the other²⁹. In contrast, we observed, for the first time, that women with fewer teeth were at a modestly increased risk of developing colorectal cancer. Although chance or residual confounding cannot be ruled out, this finding may be biologically plausible. Fusobacteria, a well-known periodontal pathogen, is found at increased abundance in stool samples of colorectal cancer patients, and are enriched in colorectal cancer tumors relative to normal colon tissue^{40, 41}. In addition, immune cells are key components of tumors. Fusobacteria has been shown to accelerate colorectal carcinogenesis by modulating the tumor immune microenvironment⁴¹. In the Apc^Min/+ mouse model of colonic tumors, an increase in intratumoral immune cells (i.e., the myeloid) was observed in the tumors with Fusobacterium-treated but not in control group⁴¹. Moreover, introduction of Fusobacteria to Apc^Min/+ mice resulted in activation of NF-κB proinflammatory pathway, a central link between inflammation and cancer⁴². Furthermore, early oral dysbiosis that contributes to tooth loss may also promote colorectal specific dysbiosis that also promote colorectal cancer^{43, 44}. Other oral related bacteria such as P.gingivalis was reported to be associated with risk of orodigestive and pancreatic cancers^{30, 45}. Genus level studies showed that the microbial composition was significantly different in cancerous areas compared to the noncancerous lesions. Periodontal associated genera such as Fusobacterium, Prevotella, and Bacteroides were more enriched in the cancerous tissue than adjacent noncancerous lesions⁴⁰. Although test of difference by subsite was not statistically significant, we observed a significant positive association for proximal colon cancer, the site most strongly associated with distinct feature of microbial organization⁴⁶. Presence of structured invasive polymicrobial bacterial biofilms was reported at 89% on proximal tumors in contrast to 12% of distal colon tumors⁴⁶. Lastly, the consistently observed stronger association of tooth loss with rectal cancer might be due to chance (limited number of cases) or partly because Prevotella was reported to be among the most abundant bacterial species in the rectum^{40, 47}. Take together, epidemiological data on oral health and colorectal cancer is limited, and clearly, more study is needed to confirm our findings.

The overall null results of periodontal disease with bone loss and colorectal cancer risk were consistent with four other studies^{20, 26, 28, 29} but not with another study that examined colorectal cancer mortality³⁰. The lack of association with periodontal disease in this study might be due to more under-report of periodontal disease than for number of teeth¹⁸. Interestingly, some studies reported positive associations of other cancers with periodontal disease but not with tooth loss, suggesting that oral bacteria profiles might be different with inflammation (caused by periodontal disease) and with tooth loss^{15, 48}. Nonetheless, although extent of tooth loss may not be the ideal surrogate for periodontal disease and capture other measures of oral health, it may provide new insights into the overall role of oral health in relation to development of cancer.

Although the present study has several strengths, including a large sample size, prospective nature of the study design, and extensive information on covariates, it has some limitations. First, the main exposures of number of teeth and history of periodontal bone loss were self-reported which was subjected to error. However, these self-reported measures were reasonably validated against objective measures in the comparable HPFS cohort, and these measures were previously associated with overall cancer risk^{34, 49}. Second, although we have adjusted for known colorectal cancer risk factors, unmeasured or residual confounding may still exist, because tooth loss may be related to childhood social economic status or early life exposures. Lastly, generalizability of our findings to other racial/ethnic groups is limited because women in our study population are mainly of European ancestry and relatively homogeneous with similar social economic status. Studies on African-Americans and Hispanic US population are clearly warranted because periodontal disease is more prevalent in these populations⁵⁰.

In summary, women with fewer teeth (0–16) and possibly those who have moderate or severe periodontal bone loss might be at a modest increased risk of developing colorectal cancer. Given limited research on this topic, more prospective studies with objective measures of periodontal disease such as clinical attachment loss, radiographic bone loss, histological and serological markers of periodontal disease, and different dimensions of oral health are warranted to confirm this finding and consider underlying mechanisms.

Impact and Novelty.

Data on oral health including periodontal diseases and colorectal cancer risk are scarce. We conducted a prospective study among 77,443 women during 18 years of follow-up in the Nurses’ Health Study. We found that independent of smoking and other risk factors, women with fewer teeth were at a modest increased risk of developing colorectal cancer, suggesting a potential role of oral health in colorectal carcinogenesis.