| |
Periodontal disease, tooth loss, and cancer risk in male health professionals: a prospective cohort study
|
| |
| |
Lancet Oncology June 2008; 9:550-558
Dr Dominique S Michaud ScD a b , Yan Liu MS c, Mara Meyer ScM b, Prof Edward Giovannucci ScD b c d and Prof Kaumudi Joshipura ScD b e f
a. Department of Epidemiology and Public Health, Imperial College London, London, UK
b. Department of Epidemiology, Harvard School of Public Health, Boston, MA, USA
c. Department of Nutrition, Harvard School of Public Health, Boston, MA, USA
d. Channing Laboratory, Department of Medicine, Brigham and Women's Hospital and Harvard Medical School, Boston, MA, USA
e. University of Puerto Rico, Medical Sciences Campus, School of Dentistry, San Juan, Puerto Rico
f. Department of Oral Health Policy and Epidemiology, Harvard School of Dental Medicine, Boston, MA, USA
The role of oral infection, especially periodontal infection, in chronic diseases has gained strength as supportive evidence from basic to population research has accumulated...... Periodontal infection in the mouth has systemic implications: individuals with periodontal infections have increased concentrations of circulating inflammatory markers,4,5 severity of disease directly correlates with serum concentrations of inflammatory markers,6 and treatment of periodontal infection can lower markers of systemic inflammatory and endothelial dysfunction within 2-6 months.....Currently, we do not know whether systemic inflammation, pathogenic invasion into the bloodstream, or the immune response to periodontal infection might have an effect on cancer risk overall, or at various tumour sites.....we assessed whether there was an association between periodontal disease, number of teeth, and tooth loss during follow-up, and cancer incidence in a large prospective questionnaire study of male health professionals.
".....In this cohort study of health professionals, a small increase in the risk of total cancer was recorded in men who reported having periodontal disease compared with those who did not.....For individual cancer sites, periodontal disease was associated with increased risks of several solid tumours, including lung, pancreas, and kidney......A significant increase in risk of haematological cancers was recorded for men who reported periodontal disease compared with those who did not. The HRs were similar for NHL, leukaemia, and multiple myeloma, although individually, only NHL was significant. This is the first report of such an association and should be interpreted with caution. However, the consistency in the HRs across the different types of haematological cancers suggests that the associations might not be because of chance. These findings might represent a commonality in the immune function and response to inflammation, which results in susceptibility to both periodontal disease and haematological cancers. Alternatively, the systemic effect of periodontal disease on the immune system might be directly related to the risk. Because the risk of NHL has been associated with immune modulation34 and inflammatory conditions,35,36 our observation of association with lymphoma needs to be studied further.....Limitations of this study include assessment of periodontal disease by self-reports and inadequate power to study less common cancers......Given the systemic effects of periodontal disease and the potential involvement of the immune system, as a marker of susceptibility or through changes in immune surveillance, we believe that further research on the role of periodontal disease in cancer, especially haematological cancers, is warranted....."
Summary
Background
Studies suggest that tooth loss and periodontal disease might increase the risk of developing various cancers; however, smoking might have confounded the reported associations. We aimed to assess whether periodontal disease or tooth loss is associated with cancer risk.
Methods
The analysis was done in a prospective study (the Health Professionals Follow-Up Study [HPFS]), which was initiated in 1986 when US male health professionals aged 40-75 years responded to questionnaires posted by the Department of Nutrition, Harvard University School of Public Health, Boston, MA, USA. In addition to the baseline questionnaires, follow-up questionnaires were posted to all living participants every 2 years and dietary questionnaires every 4 years. At baseline, participants were asked whether they had a history of periodontal disease with bone loss. Participants also reported number of natural teeth at baseline and any tooth loss during the previous 2 years was reported on the follow-up questionnaires. Smoking status and history of smoking were obtained at baseline and in all subsequent questionnaires. Additionally at baseline, participants reported their mean frequency of food intake over the previous year on a 131-item semiquantitative food-frequency questionnaire. Participants reported any new cancer diagnosis on the follow-up questionnaires. Endpoints for this study were risk of total cancer and individual cancers with more than 100 cases. Multivariate hazard ratios (HRs) and 95% CIs were calculated by use of Cox proportional hazard models according to periodontal disease status and number of teeth at baseline.
Findings
In the main analyses, 48_375 men with median follow-up of 17·7 years (1986 to Jan 31, 2004) were eligible after excluding participants diagnosed with cancer before 1986 (other than non-melanoma skin cancer, n=2076) and those with missing data on periodontal disease (n=1078). 5720 incident cancer cases were documented (excluding non-melanoma skin cancer and non-aggressive prostate cancer). The five most common cancers were colorectal (n=1043), melanoma of the skin (n=698), lung (n=678), bladder (n=543), and advanced prostate (n=541). After adjusting for known risk factors, including detailed smoking history and dietary factors, participants with a history of periodontal disease had an increased risk of total cancer (HR 1·14 [95% CI 1·07-1·22]) compared with those with no history of periodontal disease. By cancer site, significant associations for those with a history of peridontal disease were noted for lung (1·36 [1·15-1·60]), kidney (1·49 [1·12-1·97]), pancreas (1·54 [1·16-2·04]; findings previously published), and haematological cancers (1·30 [1·11-1·53]). Fewer teeth at baseline (0-16) was associated with an increase in risk of lung cancer (1·70 [1·37-2·11]) for those with 0-16 teeth versus those with 25-32 teeth. In never-smokers, periodontal disease was associated with significant increases in total (1·21 [1·06-1·39]) and haematological cancers (1·35 [1·01-1·81]). By contrast, no association was noted for lung cancer (0·96 [0·46-1·98]).
Interpretation
Periodontal disease was associated with a small, but significant, increase in overall cancer risk, which persisted in never-smokers. The associations recorded for lung cancer are probably because of residual confounding by smoking. The increased risks noted for haematological, kidney, and pancreatic cancers need confirmation, but suggest that periodontal disease might be a marker of a susceptible immune system or might directly affect cancer risk.
Funding
National Cancer Institute, National Institutes of Health, Bethesda, MD, USA (grant number P01CA055075).
Introduction
The role of oral infection, especially periodontal infection, in chronic diseases has gained strength as supportive evidence from basic to population research has accumulated.1-3 Basic laboratory studies have shown that infections with one of the periodontal infection pathogens, Porphyromonas gingivalis, can accelerate atheroma deposition in animals.1 Periodontal infection in the mouth has systemic implications: individuals with periodontal infections have increased concentrations of circulating inflammatory markers,4,5 severity of disease directly correlates with serum concentrations of inflammatory markers,6 and treatment of periodontal infection can lower markers of systemic inflammatory and endothelial dysfunction within 2-6 months.7-10 Taken together, data from multidisciplinary studies lend support to the possibility of causal associations for diabetes, stroke, and cardiovascular disease.11 However, differing opinions remain on the relative role of confounding and bias and the causal component of these associations.12
Currently, we do not know whether systemic inflammation, pathogenic invasion into the bloodstream, or the immune response to periodontal infection might have an effect on cancer risk overall, or at various tumour sites. Interest in the effect of oral health on risk of head and neck cancers has led to many studies that assess the relation between oral health, especially tooth loss, and those cancers; these studies, all of case-control designs, have recorded strong associations for tooth loss and oral cancer even after controlling for tobacco smoking and alcohol consumption.13-17 Increased risks have also been noted for tooth loss and oesophageal,18 upper gastrointestinal,19 gastric,18,20,21 and pancreatic cancers.22 Tooth loss is most commonly caused by dental caries and periodontal disease, but the percentage contribution from each condition depends on age and other factors. Tooth loss in older age is more likely to be caused by chronic periodontal disease, whereas tooth loss at a younger age is usually caused by dental caries.23 Therefore, although tooth loss might be a marker of periodontal disease, the association between tooth loss and periodontal disease is not always strong. So far, to our knowledge, only two studies have assessed the relation between periodontal disease and any cancer risk24,25 (one of these was from the present cohort) and four studies have recorded the association between tooth loss19,26,27 or periodontal disease24 and risk of total cancer death.
To address the role of oral health on cancer, we assessed whether there was an association between periodontal disease, number of teeth, and tooth loss during follow-up, and cancer incidence in a large prospective questionnaire study of male health professionals. Also, we aimed to study whether periodontal disease with bone loss is associated with an increase in risk of cancer at specific sites. Periodontal disease develops over many years, can progress in some individuals despite treatment, and causes loss of supporting connective tissue and bone. The question on periodontal disease in this cohort study specifically asked about bone loss, and therefore, focuses on cumulative disease burden. The Health Professionals Follow-up Study (HPFS)25 population is well-suited to assess this issue because it is homogeneous in terms of socioeconomic and educational status, and includes a large group of never-smokers. To our knowledge, this the first comprehensive study on periodontal disease and cancer incidence.
Discussion
In this cohort study of health professionals, a small increase in the risk of total cancer was recorded in men who reported having periodontal disease compared with those who did not. The increase in risk persisted in never-smokers for total cancer, but not for lung cancer, suggesting that the increase in risk of malignant disease overall is not because of residual confounding by smoking. A similar increase in risk was noted for number of teeth and total cancer risk, but this association was attenuated and non-significant after adjusting for smoking, and further attenuated after adjusting for periodontal disease.
For individual cancer sites, periodontal disease was associated with increased risks of several solid tumours, including lung, pancreas, and kidney. For pancreatic and kidney cancers (but not lung cancer), the HRs from the multivariable model without smoking were not substantially different from the model that adjusted for smoking status (table 2), suggesting that smoking is not likely to account for the excess risks at those cancer sites. Furthermore, for kidney and pancreatic cancers, number of teeth was not associated with risk, as would be expected if smoking or another confounder associated with oral health was responsible for the association between periodontal disease and these cancers. We considered that the associations for kidney and pancreatic cancers might be explained by other risk factors shared by both diseases, such as diabetes and obesity. However, our findings were similar in multivariable models with and without BMI or diabetes (data not shown), or after removing diabetics from the analysis (data not shown), suggesting that these two known risk factors are not likely to be responsible for the associations noted with periodontal disease.
A significant increase in risk of haematological cancers was recorded for men who reported periodontal disease compared with those who did not. The HRs were similar for NHL, leukaemia, and multiple myeloma, although individually, only NHL was significant. This is the first report of such an association and should be interpreted with caution. However, the consistency in the HRs across the different types of haematological cancers suggests that the associations might not be because of chance. These findings might represent a commonality in the immune function and response to inflammation, which results in susceptibility to both periodontal disease and haematological cancers. Alternatively, the systemic effect of periodontal disease on the immune system might be directly related to the risk. Because the risk of NHL has been associated with immune modulation34 and inflammatory conditions,35,36 our observation of association with lymphoma needs to be studied further.
The bulk of published studies on tooth loss or periodontal disease and cancer has been on upper gastrointestinal cancers, including oral and stomach cancers. For these cancer sites, we did not note significant associations for tooth loss or periodontal disease; however, our cancer cases were low for these analyses. For lung cancer, a strong association was recorded for men with fewer teeth. Number of teeth is a marker of lifetime oral health and is highly associated with socioeconomic status and access to dental care, and therefore, residual confounding by socioeconomic status might be responsible for these findings. Alternatively, a causal mechanism might be postulated; associations have been reported between oral hygiene and chronic respiratory diseases, and aspiration of oropharyngeal bacteria might be responsible for these observations.37,38
The inverse associations noted for number of teeth and melanoma of the skin and advanced prostate cancer should be interpreted with caution. As with the data for haematological cancers, these findings might be a result of chance because we studied a large number of cancer sites. Alternatively, a risk factor not adjusted for, such as sun exposure for melanoma of the skin, which could hypothetically be related to number of teeth, and therefore be a confounding variable, might have resulted in the inverse associations we noted.
To minimise confounding by education or socioeconomic status, we undertook our study in a cohort of health professionals that is fairly homogeneous with respect to education and socioeconomic status. Although individuals with periodontal disease might change their diet as their gums become more sensitive or as they start losing teeth, controlling for several dietary factors, including intake of fruits and vegetables, and vitamins C and D, had no effect on the HRs in this study. The detailed information on smoking (including dose at each decade of life before the study and every other year during follow-up; duration calculated at baseline and time-varying over follow-up period; and time since cessation for those who quit) allowed us to rigorously control for this factor; additionally, we were able to study the periodontal associations in the large number of never-smokers in this group.
We did not collect information on treatment of periodontal disease in this study. We were interested in the association between periodontal disease with bone loss and cancer risk. Individuals with irreversible damage from chronic periodontal disease (ie, bone loss) will have experienced chronic inflammation related to periodontitis at some point in their lives. Although periodontal disease can be treated or managed to decrease further damage and grafting might be occasionally used to fill the bone loss, those who have periodontal disease cannot be cured because the effects of past periodontal disease will generally manifest as attachment loss and bone loss. Our approach is similar to other studies that relate periodontitis and systemic diseases and that focus on attachment loss or pocket depth and measure overall burden rather than specific treatments and progressions. Therefore, we do not think that the absence of treatment data is an important concern. At this point, we feel that any recommendations for prevention of cancer based on these findings are premature; patients with periodontal diseases should seek care from their dentists irrespective of the effect on cancer.
Limitations of this study include assessment of periodontal disease by self-reports and inadequate power to study less common cancers. Self-reporting of periodontal disease status could have introduced measurement error which would most probably have been non-differential and resulted in attenuation towards the null (given the binary exposure) of any underlying association. However, misclassification is unlikely to be substantial given that: the self-reported question on periodontal disease has good positive and negative predictive values when compared with radiographs in a subsample of this cohort;29,30 and this measure was previously associated with stroke in this cohort.39 Furthermore, the prevalence of periodontal disease in this study (7863 of 48_375 [16%]) is similar to the prevalence in US adults aged 30-90 years (13% for moderate or severe forms of periodontitis based on National Health and Nutrition Examination Survey III data),40 suggesting that under-reporting of periodontal disease was not likely to be high. However, some misclassification of periodontal disease probably occurred and, consequently, actual associations might in fact be stronger than those noted. Another limitation of this study was that our findings for some of the less common cancers are based on only a few cases and thus, other studies, or additional follow-up, are necessary to assess better the associations between periodontal disease and oropharyngeal, oesophageal, and subsites of gastric cancers. Also, given that this study was done in men, the findings might not be generalisable to women.
Given the systemic effects of periodontal disease and the potential involvement of the immune system, as a marker of susceptibility or through changes in immune surveillance, we believe that further research on the role of periodontal disease in cancer, especially haematological cancers, is warranted.
Results
In the main analyses, 48_375 men with median follow-up of 17·7 years (1986 to Jan 31, 2004) were eligible after excluding participants diagnosed with cancer before 1986 (other than non-melanoma skin cancer, n=2076) and those with missing data on periodontal disease (n=1078). We confirmed 5720 total incident cancers (excluding non-melanoma skin cancers and non-aggressive, organ-confined prostate cancer), which were diagnosed between the return of the baseline questionnaire and Jan 31, 2004.
Periodontal disease in this study was defined by the self-report of periodontal disease "with bone loss", which was predictive of periodontal disease assessed by use of radiographs (representing cumulative damage from periodontal disease) in our validation studies.29,30Table 1 summarises baseline characteristics for this population across categories of tooth loss and periodontal disease; this table provides some indication of which factors might be confounding variables in the main analysis. Participants with a history of periodontal disease were older, more likely to have a history of diabetes, or be current smokers than those who had no reported history of periodontal disease at baseline (table 1). Similarly, men with fewer teeth at baseline (0-16 remaining) were older, more likely to be current smokers and have a history of diabetes than men with most of their natural teeth (25-32 remaining). Additionally, men with fewer teeth were less physically active, had lower intakes of calcium and vitamins C and D, and were consuming more calories than men with more teeth (table 1).
Number of teeth at baseline represents tooth loss because of dental caries, periodontal disease, injury, and orthodontic treatment. At baseline (in participants with no missing data), 1215 of 2910 (42%) of the participants with 0-16 remaining teeth, 1722 of 5899 (29%) of those with 17-24 remaining teeth, and 5357 of 41_288 (13%) of those with 25-32 remaining teeth had periodontal disease. Therefore, although there was overlap between number of teeth and periodontal disease, the two exposures can be mutually exclusive.
Men who reported a history of periodontal disease had a slightly higher total cancer incidence compared with men who had no periodontal disease at baseline (unadjusted HR 1·28 [1·20-1·36], data not shown) and when controlling for known risk factors (HR 1·14 [95% CI 1·07-1·22], table 2). No statistically significant association was noted for men with only 0-16 teeth at baseline compared with men with 25-32 teeth in the multivariable model (HR 1·09 [0·99-1·20], table 2; unadjusted HR 1·30 [1·19-1·43], data not shown). The HR was identical for periodontal disease in a model that simultaneously adjusted for periodontal disease and teeth number, and the association remained non-significant for teeth number (HR 1·06 [0·96-1·16], data not shown).
After controlling for smoking and other risk factors, periodontal disease was significantly associated with an increased risk of lung, kidney, pancreatic, and haematological cancers (table 2). The association for oesophageal cancer was also increased, but the association was not significant after adjusting for smoking status (1·44 [0·98-2·11]). Few teeth (0-16 vs 25-32), which was similarly associated with smoking, was only associated with an increased risk of lung cancer (table 2). Significant inverse associations were noted for tooth loss and melanoma of the skin and advanced prostate cancer; no associations with periodontal disease were noted for these two cancer sites.
Associations were stronger for periodontal disease and teeth number when smoking was not included in the models of smoke-related cancers (table 2), which confirmed that smoking is a strong confounder of these associations. However, for pancreatic and kidney cancers, the associations remained strong after controlling for smoking.
In addition to the baseline questions, information on periodontal disease was collected every 2 years during follow-up. In a secondary analysis we assessed the relation between updated periodontal disease status and risk of cancer. Overall, associations were similar or slightly weaker after updating periodontal status during follow-up (data not shown), with the exception of the association with lung cancer where in the fully adjusted model the association was stronger in an updated model (1·48 [1·26-1·74], data not shown) compared with the baseline model (1·36 [1·15-1·60], table 2).
The associations for lung cancer varied by histology (after controlling for smoking and other covariates); the strongest associations for periodontal disease were recorded for small-cell carcinoma (n=75, 2·03 [1·24-3·32]) and squamous-cell carcinoma (n=104, 1·68 [1·11-2·55]). No association was noted for adenocarcinoma (n=224, 1·29 [0·96-1·74]). Number of teeth was associated with squamous-cell carcinoma (1·72 [1·03-2·85] for 0-16 teeth vs 25-32 teeth), but not with the other subtypes of lung cancer.
We further studied the potential for residual confounding by smoking by restricting the analyses to never-smokers (excludes ever cigarette, pipe, and cigar smokers). We collapsed certain cancer sites together because of small numbers of cases in never-smokers (table 3). For total cancers, a small but significant increase in risk was noted with periodontal disease in this subset of the population (1·21 [1·06-1·39; table 3). We noted no increase for lung cancer risk (0·96 [0·46-1·98]). In never-smokers, the increase in risk for haematological cancers was similar to the overall finding for this group of cancers (table 3).
All three major cancers that are defined as haematological cancers, ie, non-Hodgkin lymphoma (NHL), leukaemia, and multiple myeloma, were associated with similar increased HRs in those with periodontal disease, although the association was only significant for NHL (table 4).
We did stratified analyses to assess potential interactions of other factors with periodontal disease. For total cancer, the associations did not vary substantially by age, BMI, multivitamin use, or by number of teeth at baseline. For lung cancer, however, the associations were stronger in men with BMI under 25 kg/m2 (1·69 [1·31-2·17, 284 cases) than in those with higher BMI, and in non-users of multivitamins (1·56 [1·27-1·92], 406 cases) than in users of multivitamins. For haematological cancers, associations were stronger in men aged under 70 years (1·48 [1·19-1·83, 538 cases) than in older men, in non-users of multivitamins (1·41 [1·15-1·73], 521 cases) than in users of multivitamins, and in those with BMI over 25 kg/m2 (1·44 [1·17-1·77], 509 cases) than in those with lower BMI. The stronger associations for lung cancer in men with BMI under 25 kg/m2 and not taking multivitamins suggest that residual confounding by smoking might be involved, given that those two groups are more likely to be smokers. Some of these differences, however, could be because of the smaller number of cases in the stratified groups.
As dentists might be less prone to under-report history of periodontal disease with bone loss than other health professionals, we assessed the association between periodontal disease and risk separately by health-care profession. We noted similar associations for total and lung cancers in dentists and other health professionals (data not shown), but associations were slightly stronger in dentists for periodontal disease and haematological cancers (1·37 [1·12-1·67] in dentists, 561 cases; 1·22 [0·91-1·63] in non-dentist health professionals, 372 cases).
Any incident tooth loss during follow-up (1988-2004) compared with no tooth loss was not associated with the risk of total cancer (1·02 [0·95-1·09]), and not associated with an increase in lung cancer (1·18 [0·99-1·41]). Similarly, tooth loss within the previous 4 years was not associated with total cancer (0·99 [0·92-1·07]) compared with no tooth loss in the past 4 years, or with lung-cancer risk (1·05 [0·86-1·28]).
Methods
Participants
The Health Professionals Follow-Up Study (HPFS) was initiated in 1986 (baseline) when 51_529 mainly white US men (50_121 [97%]) aged 40-75 years responded to questionnaires posted by the Department of Nutrition, Harvard University School of Public Health, Boston, MA, USA. Participants in our current cohort were those who returned a completed questionnaire; these participants consisted of dentists (29_683 [58%]), veterinarians (10_096 [20%]), pharmacists (4185 [8%]), optometrists (3745 [7%]), osteopathic physicians (2220 [4%]), and podiatrists (1600 [3%]). Individual data on behavioural and lifestyle characteristics, including smoking history, physical activity, and diet, and also medical conditions and medications, were obtained on the baseline questionnaire. Participants were contacted by follow-up postal questionnaires every 2 years, and they would usually inform us of address changes (or we obtained new addresses from the postal service). Data on newly diagnosed medical conditions and lifestyle factors were obtained on the follow-up questionnaires, and diet was updated every 4 years by use of postal questionnaires. Deaths of most members of this cohort were reported by family members or by the postal service in response to postal questionnaires. Additionally, the National Death Index was searched every 2 years for non-respondents; this method has been shown to have a sensitivity of 98%.28 Between 1986 and 2002, follow-up rate for the HPFS cohort was greater than 96% of the total potential person-years. This study was approved by the Human Subjects Committee of the Harvard School of Public Health and conforms to the STROBE guidelines for observational studies.
Assessment of dental measures
Periodontal disease
At baseline, participants of the HPFS were asked whether they had a history of periodontal disease with bone loss. This question was validated in dentists29 and non-dentists30 in the HPFS cohort by obtaining radiographs from subsets of individuals with and without a self-reported history of periodontal disease. Radiographs for each participant were assessed by blinded examiners for bone loss in all posterior teeth present except for the third molars. Both interproximal sites, mesial and distal, were measured for the two premolars and for the first and second molars of each quadrant for a total of 32 possible sites. Bone loss assessed from the radiographs was used as the standard measure of cumulative periodontal disease. In dentists (from whom 140 radiographs were obtained), the positive predictive value was 0·76 and the negative predictive value was 0·74.29 In non-dentists (from whom 212 radiographs were obtained), the positive predictive value was 0·80 and the negative predictive value was 0·68.30
Tooth loss
Participants reported number of natural teeth at baseline, and any tooth loss during the previous 2 years was reported on the follow-up questionnaires. Self-reported number of teeth is highly correlated with the actual number of teeth on clinical assessment in a general population (r=0·97).31 Accordingly, we expected self-reported number of teeth and tooth loss during follow-up to be well reported in this cohort of educated health professionals.
Assessment of smoking history and other risk factors
Smoking status and history of smoking were obtained at baseline and in all subsequent questionnaires. Current smokers also reported amount of smoking (mean number of cigarettes smoked per day) on each questionnaire. Past-smokers reported when they last smoked, and the time since quitting was also calculated for those who quit during follow-up.
Height and current weight were reported by participants at baseline and current weight was obtained in the biennial questionnaires. We estimated body-mass index (BMI) from weight and height (ie, kg/[height in metres]2) as a measure of total adiposity.
Physical activity was estimated by use of eight different activities and a weekly physical-activity score was derived by multiplying the time spent in each activity per week by its typical energy-expenditure requirements expressed in metabolic equivalents (METs).32
Participants were also asked about history of diabetes and use of non-steroidal anti-inflammatory drugs (NSAIDs) at baseline and in all subsequent questionnaires.
Dietary assessment
A 131-item semiquantitative food-frequency questionnaire (FFQ) was posted to all participants in 1986. Participants were asked to report their mean frequency of food intake over the previous year for a specified serving size of each food. Nutrient intakes, such as calcium, were calculated by multiplication of the reported frequency of each food item by the nutrient content for the specified portion size. Additionally, on each questionnaire, participants reported the brand of breakfast cereal, duration, and frequency of use, and brand of vitamin supplements, including specific supplements. We have nutrient content from around 1400 multivitamins, which are updated every 4 years. These specific data on supplements are used to compute nutrient intakes. Food-composition data are mainly based on values obtained from the US Department of Agriculture supplemented with other data.
Vitamin D exposure was based on a score variable derived from factors that independently predicted circulation plasma 25-hydroxy-vitamin D concentrations in a subset of the HPFS cohort;33 these factors include geographical region, vitamin D intake, physical activity, BMI, and ethnic origin.
Identification of participants with cancer
Participants reported any new cancer diagnosis on the biennial postal questionnaires, and written permission to obtain the related medical records or pathology reports was then obtained from those men (or their next of kin). If previously unreported cancer was noted on a death certificate, we contacted a family member to obtain permission to retrieve medical records or to confirm the diagnosis of cancer. About 90% of cancers were confirmed by review of medical records and the remainder were confirmed with information from the participant, a family member, or by death certificate.
We excluded organ-confined prostate cancers because of their favourable prognosis and high incidence, and because their detection usually results from a prostate-specific antigen (PSA) screening test. Also, we assessed individual cancer sites for which at least 100 cases were available.
Statistical analysis
We computed person-time of follow-up for each participant from the return date of the baseline questionnaire to the date of cancer diagnosis, death from any cause, or the end of follow-up (Jan 31, 2004), whichever came first. Incidence rates of cancer were calculated by dividing the number of incident cases by the number of person-years in each category of exposure. We computed hazard ratios (HRs) for each of the exposed categories (eg, history of periodontal disease) by dividing the incidence rates in these categories by the rate in the non-exposure category.
We estimated HRs and 95% CIs by use of Cox proportional hazard models adjusting for potential confounders. A new data record was created for every questionnaire cycle at which a participant was at risk by use of covariate values at the time that the questionnaire was returned (ie, time-varying covariates were used for variables that can change over time, such as smoking status, BMI, and diabetes). For the periodontal disease analyses, baseline data on periodontal disease were used without updating because periodontal disease is chronic and generally progresses slowly. Total number of teeth identified at baseline was used in the main analyses. In a secondary analysis, we updated periodontal disease status by use of the biennial follow-up questionnaires. We also calculated cumulative incident tooth loss (ie, any tooth loss occurring during follow-up) in a secondary analysis and assessed recent tooth loss (within the past 4 years). The Cox proportional hazards models all satisfied the proportionality of hazards assumption.
All models were stratified by age (continuous in months) and calendar time. Other factors included in the models were: ethnic origin (white, Asian, black), BMI (<22, 22-24·9, 25-29·9, ≥30), physical activity (quintiles), smoking history (never, past quit [time since quit smoking] …10 years, past quit >10 years, current 1-14 cigarettes per day, 15-24 cigarettes per day, or >25 cigarettes per day), pack-years (continuous), history of diabetes (yes or no), geographical region (west, midwest, south, northeast), height (quintiles), alcohol (quartiles), vitamin D score (deciles), calcium intake (quintiles), fruit and vegetable intake (quintiles), red-meat intake (quintiles), and total calorific intake (quintiles). Most of these factors have been linked to periodontal disease and cancer, and consequently, could be confounding factors in this analysis. Individuals with missing data were excluded from the analyses. All analyses were done by use of SAS (version 9.1).
Role of the funding source
The US National Cancer Institute (Bethesda, MD, USA) funded this study (grant number P01CA055075). The funding source had no role in the study design, collection, analysis, or interpretation of the data, or in the writing of the report. All authors had full access to the raw data. The corresponding author had final responsibility for the decision to submit for publication.
|
|
| |
| |
|
|
|
