Dietary risk factors for the development of inflammatory polyarthritis: Evidence for a role of high level of red meat consumption
Abstract
Objective
To investigate the association of red meat and other specific dietary components in predicting the development of inflammatory polyarthritis.
Methods
This nested case–control study was conducted within a prospective population-based study of cancer incidence (European Prospective Investigation of Cancer in Norfolk [EPIC-Norfolk]). EPIC-Norfolk recruited 25,630 subjects ages 45–75 years between 1993 and 1997. Dietary intake was assessed at baseline using a 7-day food diary, and the information was analyzed using dietary analysis software. Patients with new cases of inflammatory polyarthritis were identified by linkage with the Norfolk Arthritis Register, a primary care–based inception study of inflammatory polyarthritis, and were matched for age and sex to 2 controls from EPIC-Norfolk. The risk for development of inflammatory polyarthritis was compared between subjects in the highest and lowest tertiles of dietary intake using conditional logistic regression and was expressed as odds ratios (ORs) with 95% confidence intervals (95% CIs).
Results
Between 1993 and 2002, 88 new patients with inflammatory polyarthritis were identified and matched with 176 controls. Among patients, the level of red meat intake was higher (P = 0.04) and that of vitamin C was lower (P = 0.03) compared with intake among controls, but no difference in total energy intake was observed. Patients were more likely to be smokers. After adjusting for total energy intake, smoking, and other possible dietary confounders, subjects with the highest level of consumption of red meat (OR 1.9, 95% CI 0.9–4.0), meat and meat products combined (OR 2.3, 95% CI 1.1–4.9), and total protein (OR 2.9, 95% CI 1.1–7.5) were at an increased risk for inflammatory polyarthritis.
Conclusion
A high level of red meat consumption may represent a novel risk factor for inflammatory arthritis or may act as a marker for a group of persons with an increased risk from other lifestyle causes.
Nongenetic factors, such as aspects of lifestyle, may explain ∼40% of the risk of rheumatoid arthritis (RA) (1). Cigarette smoking has consistently been found to be a risk factor for the development of inflammatory polyarthritis and its subset RA (2-5). There is much less evidence of a role for nutritional factors, and the existing evidence is somewhat variable. Results of 2 case–control studies have suggested a possible protective role for fish consumption, particularly fish high in omega-3 fatty acids (6, 7). Greater consumption of coffee has been associated with an increased risk of RA (8, 9), and a higher level of alcohol intake might reduce the risk of RA in women (3, 10), but such associations lack consistency (11, 12).
The most interesting data relate to fruits and vegetables. A higher level of intake of cooked, but not raw, vegetables (13), cruciferous vegetables, and β-cryptoxanthin (14), a carotenoid found in fruit and vegetables (15), has been reported to protect against the onset of RA. Furthermore, we recently demonstrated that a lower level of fruit intake and especially a lower level of vitamin C intake are associated with a >3-fold increase in the risk of new-onset inflammatory polyarthritis (16). Although the exact mechanisms remain uncertain, these associations might be attributable to the antioxidant activity of vitamin C and other antioxidant micronutrients, such as carotenoids.
Alternatively, these findings could be explained by dietary confounding, e.g., the association with vitamin C may actually be attributable to associations with other dietary variables. Indeed, although some evidence suggests that red meat could be a candidate confounder, this food group has not been widely investigated. An ecologic study demonstrated a positive correlation between the national prevalence of RA and per capita consumption of red meat in 16 countries (17). Red meat provides a dietary source of arachidonic acid, a fatty acid involved in the production of 2-series eicosanoids, several of which have potent proinflammatory activity (18). Consequently, Western diets, which are characterized by high levels of red meat consumption, are higher in arachidonic acid compared with, for example, the typical Mediterranean-type diet, in which white meats and fish are more commonly eaten.
Studies of the role of diet in the etiology of inflammatory polyarthritis and RA are, however, difficult to perform. Retrospective studies are subject to both recall errors and recall bias when attempting to capture premorbid dietary patterns (19). We used the opportunity afforded by the prospective assessment and followup of a large population sample participating in an intensive diet and general health survey to address the role of specific foods and nutrients in the onset of inflammatory arthritis. Specifically, we tested the hypothesis, based on our previous findings, that the risk associated with a low level of intake of fruit and vitamin C might be explained, at least in part, by an increased risk from other potentially “hazardous” dietary habits (e.g., high level of intake of red meat, meat fat, or protein). We also investigated the association between intake of coffee, tea, alcohol, dairy products, and vitamin D and the risk for development of inflammatory polyarthritis in light of the most recently published findings. In this particular population, 3 different methods of dietary assessment were used. We report our findings using the 7-day food diary, which has been shown to be subject to less measurement error than are the food frequency questionnaire (FFQ) or 24-hour recall methods (20, 21).
SUBJECTS AND METHODS
Study population.
This was a population-based, prospective, nested case–control study. New cases of inflammatory polyarthritis were ascertained from within a population sample; subjects in this sample had previously participated in the Norfolk arm of the European Prospective Investigation of Cancer (EPIC-Norfolk), which included diet and lifestyle assessments, details of which have been published elsewhere (22). Briefly, between 1993 and 1997, individuals ages 45–74 years living in Norfolk, UK, were selected from 35 general practice age–sex registers in that area. In all, 25,630 men and women were recruited and underwent a clinical examination that included the measurement of current weight (in kilograms) and height (in meters), using standardized procedures, and provided a blood sample to be stored for future analysis (22).
Dietary assessment.
Subjects were asked to complete a detailed 7-day “estimated” food diary (20) at baseline. They were encouraged to be as specific as possible in describing their intake and were instructed on the use of household measures and food portion photographs that were printed in the diary. Dietary intake was recorded over 7 consecutive days and never more than once on any 1 day of the week. Completed diaries were stored for subsequent coding and analysis in future nested case–control studies (such as the current investigation). One of us (DJP) performed the subsequent diary coding. The rate of missing and ambiguous data was surprisingly low, although some assumptions had to be made about food type and portion size when specific information was not provided. In addition to the food diaries (and of relevance to the current study), the EPIC assessment also included a measurement of both past and current cigarette consumption (22) from which pack-years of smoking could be calculated. Participants were assigned a Townsend deprivation score (an index of social deprivation), calculated for the enumeration district (the smallest unit of census geography) of residence for each subject (23). The scores were allocated to quintiles using reference data for England and Wales based on their residential postal code and used as a proxy for socioeconomic status in the analysis.
Ascertainment of cases of inflammatory polyarthritis.
New cases of inflammatory polyarthritis in the EPIC population were ascertained by linkage of the EPIC database to the Norfolk Arthritis Register (NOAR), which covered the general practices from which subjects for EPIC-Norfolk were recruited. NOAR patients participating in EPIC were identified by date of birth, National Health Service number, and surname. Linkage is repeated annually, with blinding to the previous year's search, to capture any new NOAR cases. Linkage in our previous study was undertaken until March 2001 (16), but for this current analysis we were able to extend the time point to October 2002 to include additional new cases of inflammatory polyarthritis. Details of NOAR have also been published elsewhere (24). Briefly, all patients presenting to a general practitioner and being newly diagnosed with inflammatory polyarthritis, defined as inflammation affecting ≥2 peripheral joints and persisting for >4 weeks, were sought. A research nurse saw patients within 2 weeks from the time at which the general practitioner notified NOAR of a patient's eligibility. Patients were interviewed and examined using a structured approach and had blood drawn for analysis of rheumatoid factor (24). Subjects who fulfilled the above entry criteria and were not given another diagnosis by a consultant (other than RA, psoriatic arthritis, or postviral arthritis) were registered by NOAR. Ethical approval for both studies was obtained from the Norwich District Health Authority Research Ethics Committee, and all participants gave informed consent.
Statistical analysis.
Subjects in whom inflammatory polyarthritis developed while they were in the EPIC cohort, and for whom a date of recalled symptom onset (defined as the date on which they first noticed joint swelling) was available after they had completed their EPIC dietary assessment, were designated as cases for this analysis. Each patient was then matched for age (±3 years) and sex and to within 3 months of their baseline dietary assessment with 2 controls from the EPIC database who had remained free of inflammatory polyarthritis during the followup period. The diet diaries of the patients and controls were coded by one of us (DJP), with blinding to subject status, using the dietary analysis software DINER (Data Into Nutrients for Epidemiological Research) (25). The DINER program is based on standard UK food composition tables; by incorporating the estimated weights of each item consumed, based on the description on the portion sizes, each individual's daily intake of all major food groups and nutrients is derived. Absolute intake of red meat, meat products (e.g., sausages, ham), red meat combined with meat products, coffee, tea, alcohol, and total energy were compared between the matched patients and controls using Wilcoxon's signed rank test. We also compared intake of the major nutrients in meat, i.e., protein, iron, total fat, and polyunsaturated fatty acids. Last, we compared intake of dairy products (milk, cheese, and yogurt combined) and vitamin D. The intake of these dietary variables was then stratified by tertiles to assess the association between exposure and the risk for development of inflammatory polyarthritis, using the lowest tertile as the referent group.
Associations between the upper tertiles of intake and the onset of inflammatory polyarthritis were expressed as odds ratios (ORs) with their 95% confidence intervals (95% CIs), using conditional logistic regression to retain the matched sets. Analyses were undertaken univariately for each diet variable and again after adjustment for total energy intake and then pack-years of smoking, which were entered into the model as continuous variables. The Townsend deprivation score and obesity were also considered as potentially confounding factors but were not retained in these analyses. Finally, for each item of interest, analyses were undertaken after adjustment for food groups or nutrients, as appropriate, that demonstrated any trend toward an association, to examine the role of confounding in the multivariate setting. Although patients were significantly taller than controls, height was not considered as a potential confounder, because differences in energy intake have been corrected for by adjusting for total energy intake, and height is not known to be associated with the onset of RA.
RESULTS
The EPIC baseline survey was conducted between 1993 and 1997. By October 2002, a total of 88 new cases of inflammatory polyarthritis were identified in NOAR patients who had previously participated in EPIC and had completed the 7-day diet diary. The mean (±SD) time from initial dietary assessment to symptom onset was 2.5 ± 1.8 years (range 0.02–7.4 years). Nearly 40% of patients satisfied the American College of Rheumatology (ACR; formerly, the American Rheumatism Association) criteria for RA (26) at baseline. In those who had completed 5 years of followup by the end of 2002 (n = 63), the proportion satisfying ACR criteria (applied cumulatively) rose to 60%. Given the design, patients and controls were closely matched for age (mean ± SD age 60.7 ± 8.9 years) and sex (69% women).
There were few nondietary differences between patients and controls, and these are shown in Table 1. Although weights were similar, patients were taller, and the groups were well balanced for body mass index. Patients were more likely to be current smokers, and among those who smoked, patients had significantly higher lifetime cigarette consumption than controls.
| Characteristic | Patients (n = 88) | Controls (n = 176) | P |
|---|---|---|---|
| Height, median (IQR) meters | 164.3 (159.8–170.4) | 162.2 (157.9–169.5) | 0.02† |
| Weight, median (IQR) kg | 71.5 (65.6–81.6) | 71.0 (63.8–78.9) | 0.4† |
| BMI, median (IQR) kg/m2 | 26.2 (24.4–29.1) | 26.3 (24.3–29.3) | 0.7† |
| BMI category, no. (%) of patients | |||
| <25 kg/m2 | 28 (31.8) | 55 (31.2) | |
| 25–30 kg/m2 | 42 (47.7) | 88 (50.0) | 0.9‡ |
| >30 kg/m2 | 18 (20.5) | 33 (18.8) | |
| Smoking status, no. (%) of patients | |||
| Never | 35 (40.7) | 85 (48.9) | |
| Former | 33 (38.4) | 67 (38.5) | 0.2‡ |
| Current | 18 (20.9) | 22 (12.6) | |
| Ever smoked | 51 (59.3) | 89 (51.1) | |
| Pack-years of smoking, median (IQR) | 2.7 (0.0–25.1) | 0.5 (0.0–13.0) | 0.04† |
| Pack-years of smoking in smokers, median (IQR) | 22.5 (9.0–34.0) | 14.7 (5.9–27.8) | 0.15† |
| Townsend deprivation score, no. (%)§ | |||
| Quintile 1 (≤−3.11) | 31 (35.2) | 69 (39.2) | |
| Quintile 2 (−3.10 to −1.56) | 27 (30.7) | 50 (28.4) | |
| Quintile 3 (−1.55 to 0.38) | 16 (18.2) | 32 (18.2) | 0.9‡ |
| Quintile 4 (0.39 to 3.13) | 10 (11.4) | 20 (11.4) | |
| Quintile 5 (>3.13) | 4 (4.5) | 5 (2.8) | |
- * IQR = interquartile range; BMI = body mass index.
- † Patients versus controls, by Wilcoxon's signed rank test.
- ‡ Patients versus controls, by chi-square test.
- § Quintile 1 is the lowest level of social deprivation, and quintile 5 is the highest level.
Data on the difference in median daily consumption of the major food groups and nutrients are shown in Tables 2 and 3. Total energy intake (in kilocalories) was similar in the 2 groups. Patients had a higher median intake of red meat (P = 0.04) and red meat combined with meat products (P = 0.02) than their matched controls, although there was no difference in intake of coffee, tea, alcohol, or dairy products. Intake of fat and the different fatty acids was virtually identical between the 2 groups, and there were no striking differences in the intake of protein, iron, or vitamin D. Associations between dietary intake of these foods and nutrients were explored further using conditional logistic regression analysis.
| Food group/beverage | Patients (n = 88) | Controls (n = 176) | P† |
|---|---|---|---|
| Red meat, gm | 48.0 (22.1–87.2) | 41.3 (14.3–65.3) | 0.04 |
| Meat products, gm | 18.1 (11.4–35.6) | 19.6 (8.1–30.6) | 0.3 |
| Red meat + meat products, gm | 78.1 (44.3–118.8) | 63.9 (37.2–92.0) | 0.02 |
| Fruit, gm | 98.4 (33.9–212.8) | 132.0 (74.8–215.6) | 0.01 |
| Vegetables, gm | 86.4 (47.4–121.9) | 85.9 (51.8–122.9) | 0.7 |
| Fruit + vegetables, gm | 188.0 (110.8–297.0) | 222.8 (157.4–320.8) | 0.07 |
| Dairy products, gm | 210.5 (137.5–290.1) | 203.0 (126.0–280.9) | 0.6 |
| Caffeinated coffee, ml | 129.0 (13.5–436.5) | 153.7 (24.6–406.1) | 0.6 |
| Tea, ml | 755.6 (384.9–1111.9) | 724.9 (442.3–909.1) | 0.1 |
| Alcohol, gm | 4.3 (0.0–12.7) | 3.0 (0.0–12.9) | 0.6 |
- * Values are the median (interquartile range).
- † Patients versus controls, by Wilcoxon's signed rank test.
| Patients (n = 88) | Controls (n = 176) | P† | |
|---|---|---|---|
| Total energy, kcal | 1,762.9 (1,447.6–2,168.1) | 1,799.0 (1,529.0–2,092.7) | 0.9 |
| Protein, gm | 71.1 (61.0–80.1) | 69.2 (58.3–78.5) | 0.3 |
| Total fat, gm | 67.3 (53.8–84.8) | 67.5 (56.2–84.5) | 0.6 |
| Saturated fat, gm | 21.9 (16.4–29.0) | 22.1 (16.9–29.2) | 1.0 |
| Polyunsaturated fat, gm | 9.4 (7.6–12.6) | 10.4 (7.5–12.9) | 0.6 |
| Monounsaturated fat, gm | 19.3 (14.5–23.1) | 19.1 (15.6–23.7) | 0.7 |
| Iron, mg | 10.8 (8.8–13.3) | 10.6 (8.5–13.2) | 0.5 |
| Vitamin D, μg | 3.0 (1.7–4.4) | 2.6 (1.9–3.9) | 0.5 |
| Vitamin C, mg | 63.7 (44.0–95.9) | 77.7 (50.7–107.2) | 0.03 |
- * Values are the median (interquartile range).
- † Patients versus controls, by Wilcoxon's signed rank test.
In a previous analysis, we showed that patients with inflammatory polyarthritis had a lower level of intake of fruit and dietary vitamin C (16), and the same was observed for this larger cohort (Tables 2 and 3). Therefore, several models were generated, including a model adjusted only for energy intake (thus allowing for age and sex), a model adjusted for smoking, and, finally, a model adjusted for the energy intake, smoking, and fruit intake. When nutrient intake was investigated, we adjusted for dietary vitamin C intake and other nutrients found to be associated with the development of RA. The results (Table 4) showed that persons in the tertile with the highest level of red meat consumption had a >2-fold increased risk for the development of inflammatory polyarthritis that remained essentially unaltered after the adjustments noted above (adjusted OR 1.9, 95% CI 0.9–4.0). A similar analysis of the composite meat variable showed a stronger association for those in the tertile with the highest level of intake (adjusted OR 2.3, 95% CI 1.1–4.9).
| Daily intake | Patients (n = 88) | Controls (n = 176) | Energy-adjusted OR (95% CI) | Fully adjusted OR (95% CI)† | P for trend |
|---|---|---|---|---|---|
| Red meat, gm | |||||
| <25.5 (referent) | 25 (28.4) | 63 (35.8) | 1.0 | 1.0 | |
| 25.5–58.0 | 26 (29.6) | 62 (35.2) | 1.2 (0.6–2.2) | 1.1 (0.6–2.2) | |
| >58.0 | 37 (42.0) | 51 (29.0) | 2.1 (1.1–4.3) | 1.9 (0.9–4.0) | 0.08 |
| Meat products, gm | |||||
| <12.2 (referent) | 27 (30.7) | 61 (34.7) | 1.0 | 1.0 | |
| 12.2–27.8 | 28 (31.8) | 60 (34.1) | 1.0 (0.5–2.0) | 0.9 (0.5–1.8) | |
| >27.8 | 33 (37.5) | 55 (31.2) | 1.4 (0.8–2.7) | 1.4 (0.7–2.7) | 0.4 |
| Total meat, gm | |||||
| <49.0 (referent) | 25 (28.4) | 63 (35.8) | 1.0 | 1.0 | |
| 49.0–87.8 | 24 (27.3) | 64 (36.4) | 1.1 (0.6–2.2) | 1.0 (0.5–2.1) | |
| >87.8 | 39 (44.3) | 49 (27.8) | 2.5 (1.2–5.2) | 2.3 (1.1–4.9) | 0.03 |
| Dairy products, gm | |||||
| <153 | 26 (29.6) | 62 (35.2) | 1.0 | 1.0 | |
| 153–260 | 31 (35.2) | 57 (32.4) | 1.4 (0.7–2.8) | 1.6 (0.8–3.5) | |
| >260 | 31 (35.2) | 57 (32.4) | 1.4 (0.7–2.9) | 1.9 (0.9–4.2) | 0.09 |
| Fruit, gm | |||||
| <76.0 | 41 (46.6) | 47 (26.7) | 2.3 (1.2–4.3) | 2.1 (1.1–4.2) | 0.03 |
| 76.0–171.2 | 21 (23.9) | 67 (38.1) | 0.8 (0.4–1.5) | 0.7 (0.4–1.3) | |
| >171.2 (referent) | 26 (29.5) | 62 (35.2) | 1.0 | 1.0 | |
| Caffeinated coffee, ml | |||||
| <49.4 (referent) | 33 (37.5) | 60 (34.1) | 1.0 | 1.0 | |
| 49.4–295 | 22 (25.0) | 63 (35.8) | 0.7 (0.4–1.3) | 0.7 (0.4–1.3) | |
| >295 | 33 (37.5) | 53 (30.1) | 1.1 (0.6–2.2) | 1.1 (0.6–2.2) | 0.8 |
| 4 cups/day (vs. <4 cups/day) | 8 (9.1) | 12 (6.8) | 1.5 (0.5–4.0) | 1.4 (0.5–4.1) | |
| Tea, ml | |||||
| <570 (referent) | 30 (34.1) | 60 (34.1) | 1.0 | 1.0 | |
| 570–885 | 21 (23.9) | 68 (38.6) | 0.7 (0.4–1.3) | 0.6 (0.3–1.2) | |
| >885 | 37 (42.0) | 48 (27.3) | 1.6 (0.8–2.9) | 1.5 (0.8–2.8) | 0.2 |
| 3 cups/day (vs. <3 cups/day) | 58 (65.9) | 116 (65.9) | 1.0 (0.6–1.7) | 0.9 (0.5–1.7) | 0.8 |
| Alcohol, gm | |||||
| 0.0 (referent) | 28 (31.8) | 62 (35.2) | 1.0 | 1.0 | |
| ≤8.9 | 31 (35.2) | 55 (31.3) | 1.3 (0.7–2.3) | 1.2 (0.6–2.2) | |
| >8.9 | 29 (33.0) | 59 (33.5) | 1.1 (0.6–2.2) | 1.0 (0.5–2.0) | 0.9 |
| Ever (vs. never) | 60 (68.2) | 114 (64.8) | 1.2 (0.7–2.1) | 1.1 (0.6–2.0) |
- * Values are the number (%). OR = odds ratio; 95% CI = 95% confidence interval.
- † Adjusted for total energy intake, pack-years of smoking, total meat intake, and fruit intake, when these were not the variables being investigated.
We further postulated that increased iron consumption might be an explanation for the increased risk from a high level of red meat intake, but analysis showed no support for this hypothesis (Table 5). There was a modest, but not significant, association between greater intake of dairy products and the risk for development of inflammatory polyarthritis, whereas no appreciable associations were observed between caffeinated coffee, tea, or alcohol consumption and the risk of inflammatory polyarthritis. There were very few patients (n = 8) with matched controls who drank decaffeinated coffee; hence, an investigation of an association with decaffeinated coffee was not justified.
| Daily intake | Patients (n = 88) | Controls (n = 176) | Energy-adjusted OR (95% CI) | Fully adjusted OR (95% CI)† | P for trend |
|---|---|---|---|---|---|
| Energy, MJ | |||||
| <1,604 (referent) | 31 (35.2) | 57 (32.4) | 1.0 | ||
| 1,604–1,970 | 27 (30.7) | 61 (34.6) | 0.7 (0.4–1.4) | ||
| >1,970 | 30 (34.1) | 58 (33.0) | 0.7 (0.3–1.9) | 0.5 | |
| Total fat, gm | |||||
| <59.0 (referent) | 32 (36.4) | 56 (31.8) | 1.0 | 1.0 | |
| 59.0–77.7 | 29 (33.0) | 59 (33.5) | 0.8 (0.4–1.6) | 0.7 (0.4–1.5) | |
| >77.7 | 27 (30.7) | 61 (34.7) | 0.7 (0.2–1.9) | 0.6 (0.2–1.9) | 0.4 |
| Saturated fatty acid, gm | |||||
| <18.9 (referent) | 29 (33.0) | 59 (33.5) | 1.0 | 1.0 | |
| 18.9–26.2 | 28 (31.8) | 60 (34.1) | 1.0 (0.5–2.0) | 1.0 (0.5–2.0) | |
| >26.2 | 31 (35.2) | 57 (32.4) | 1.3 (0.6–3.0) | 1.4 (0.6–3.2) | 0.5 |
| Polyunsaturated fats, gm | |||||
| <8.3 (referent) | 33 (37.5) | 55 (31.3) | 1.0 | 1.0 | |
| 8.3–11.9 | 29 (33.0) | 59 (33.5) | 0.8 (0.4–1.5) | 0.8 (0.4–1.5) | |
| >11.9 | 26 (29.5) | 62 (35.2) | 0.6 (0.3–1.4) | 0.6 (0.3–1.3) | 0.2 |
| Monounsaturated fats, gm | |||||
| <16.7 (referent) | 31 (35.2) | 57 (32.4) | 1.0 | 1.0 | |
| 16.7–22.1 | 30 (34.1) | 58 (33.0) | 0.9 (0.5–1.8) | 0.9 (0.5–1.8) | |
| >22.1 | 27 (30.7) | 61 (34.6) | 0.8 (0.3–1.9) | 0.7 (0.3–1.7) | 0.4 |
| Protein, gm | |||||
| <62.4 (referent) | 26 (29.6) | 62 (35.2) | 1.0 | 1.0 | |
| 62.4–75.3 | 28 (31.8) | 60 (34.1) | 1.5 (0.7–2.9) | 1.6 (0.8–3.3) | |
| >75.3 | 34 (38.6) | 54 (30.7) | 2.8 (1.1–7.3) | 2.9 (1.1–7.5) | 0.04 |
| Iron, mg | |||||
| <9.2 (referent) | 29 (33.0) | 59 (33.5) | 1.0 | 1.0 | |
| 9.2–12.5 | 29 (33.0) | 59 (33.5) | 1.1 (0.5–2.1) | 1.0 (0.5–2.1) | |
| >12.5 | 30 (34.1) | 58 (33.0) | 1.2 (0.6–2.4) | 1.1 (0.5–2.3) | 0.9 |
| Vitamin C, mg | |||||
| <55.9 | 38 (43.2) | 50 (28.4) | 2.2 (1.1–4.4) | 2.8 (1.3–5.8) | <0.01 |
| 55.9–94.0 | 25 (28.4) | 63 (35.8) | 1.0 (0.5–2.0) | 0.9 (0.4–1.9) | |
| >94.0 | 25 (28.4) | 63 (35.8) | 1.0 | 1.0 | |
| Vitamin D, μg | |||||
| <2.15 | 28 (31.8) | 60 (34.1) | 1.0 | 1.0 | |
| 2.15–3.54 | 24 (27.3) | 64 (36.4) | 0.8 (0.4–1.7) | 0.8 (0.4–1.6) | |
| >3.54 | 36 (40.9) | 52 (29.5) | 1.6 (0.8–3.2) | 1.5 (0.7–2.9) | 0.2 |
- * Values are the number (%). OR = odds ratio; 95% CI = 95% confidence interval.
- † Adjusted for total energy intake, pack-years of smoking, protein intake, and vitamin C intake, when these were not the variables being investigated.
Interestingly, we observed that a higher level of intake of total protein (highest versus lowest tertile) increased the risk of inflammatory polyarthritis by almost 3-fold. A higher level of intake of polyunsaturated and monounsaturated fats was weakly protective, whereas saturated fat was associated with a small increased risk of inflammatory polyarthritis (Table 5). Vitamin D intake was positively but weakly associated with the risk of inflammatory polyarthritis. In our previous study, a 3-fold risk for the development of inflammatory polyarthritis was observed for subjects in the tertile with the lowest level of vitamin C intake compared with those in the upper tertiles (16), whereas in this study the strength of associations was slightly weaker (Table 5). These analyses did not include vitamin C intake from dietary supplements. However, supplements were used by 21% of controls and 15% of patients, thus supporting the possible protective effect of higher levels of vitamin C intake.
Although the associations for vitamin C, red meat, and the onset of inflammatory polyarthritis seem to be independent of each other, when we tested for a statistical interaction between 1) a low level of fruit intake and a high level of red meat intake, and 2) a low level of vitamin C intake and a high level of total protein intake, in neither instance was any evidence of an interaction observed. However, it is likely that a study of this size lacks the power needed to truly detect such statistical interaction.
DISCUSSION
In this detailed analysis, based on an accurate and realistic assessment of premorbid dietary intake, we observed that a high level of consumption of red meat and total meat (red meat combined with meat products) presents additional, independent risk factors for the development of inflammatory polyarthritis.
Several methodologic issues are related to the interpretation of both the positive and negative findings. First, residual confounders, both dietary and nondietary, may explain the associations observed, although it is difficult to know what these might be. In this study, we adjusted for both cigarette smoking and dietary vitamin C, which we previously found to be associated (16), and other dietary factors associated with the risk of inflammatory polyarthritis, but we cannot exclude the possibility of other factors explaining the link.
We have studied inflammatory polyarthritis and not only the subgroup of patients who satisfy criteria for RA (26). We deliberately chose not to restrict our inclusion criteria to subjects who satisfied the criteria for RA at the time of disease onset. We have argued elsewhere (27) that the criteria for RA do not perform well in the setting of early disease; thus, it is not appropriate to classify patients with inflammatory polyarthritis at disease onset as being positive or negative for RA, because RA evolves over time. Data from NOAR have shown that the number of patients who can be classified as RA positive after 5 years increases to >60% (28). Given this inherent uncertainty in defining early RA, it is possible that, for example, some patients were included who had other causes for joint inflammation, such as those with the inflammatory phase of osteoarthritis. However, the presence of substantial etiologic heterogeneity within the patients recruited would have attenuated any underlying true association, and hence is unlikely to explain the results we obtained.
It was therefore of interest to consider whether our findings were restricted to patients with definite RA, but stratification by cumulative RA status did not reveal any difference in the strength of the association with meat consumption. The relatively small number of patients limits the power to detect weaker associations than those ascertained. This might explain the inverse findings from this study, e.g., regarding coffee and tea, which differ from results of 2 other substantially larger studies (8, 9) but are in agreement with one other (11). Methodologic differences are also the most likely explanation of why we observed a weakly positive association between higher level of intakes of dairy products and vitamin D compared with the protective effect reported from the Iowa Women's Health Study (29).
The advantages of this study are those afforded by the more detailed nature of the dietary inquiry and the case assessment used, because the intensity of such approaches is frequently prohibited in larger studies. Although the diet diary is considered a powerful method for nutrient inquiry in epidemiologic studies, it is not without problems (30). It clearly measures only immediate diet, but even so, when compared with biomarkers in blood and urine measured over 1 year, intakes of protein and potassium from a 7-day diary were more closely associated with their respective biomarkers than was intake calculated from an FFQ in this population (31). Additionally, no significant differences in reported dietary intake were observed between 7-day diaries administered on 2 separate occasions over a 9-month period, thus reflecting the longer-term stability of the 7-day diary (32). This pattern of association was also observed for nutrients such as the carotenoids, which may be consumed intermittently. However, for foods eaten very sporadically (e.g., less than once weekly), the FFQ might detect dietary intake that is not detected by the diary method.
The EPIC-Norfolk study used an estimated diet diary approach based on simple descriptions of portion size as opposed to the more rigorous weighed food intake method, which may have contributed to the high rates of participation (93%) that were achieved (20). The estimated diet diary has been extensively calibrated against other diet assessment methods and urine and serum biomarkers of nutrient intake (31), and a further comparative evaluation of the construct validity of these methods has clearly demonstrated the superiority of the 7-day diary to the FFQ in the context of an investigation into the role of fat in breast cancer (33). Dietary misclassification in this prospective study is likely to be random and again would not explain the findings observed, and the rate of misclassification is still likely to be substantially lower than that if the FFQ were used.
Finally, we have undertaken an investigation of several concurrent hypotheses relating to diet, and the possibility of a false-positive result occurring by chance should be discussed. Given the number of dietary components that we considered, it could be argued that we should have adjusted the P value accordingly by presenting, for example, 99% confidence intervals. However, in epidemiologic studies it has been suggested that adjusting for multiple testing is not appropriate if there is a cogent a priori hypothesis (34). In this regard, several lines of evidence suggested that consumption of red meat was a plausible a priori risk factor. An ecologic study with all the weaknesses inherent in that approach showed that per capita meat consumption correlates strongly with the prevalence of RA at a national level (17). A study of Seventh-Day Adventists in California also suggested that meat consumption was related to self-reported arthritis, although the majority of such cases were unlikely to have been RA (35). Data are fairly limited, but the occurrence of RA is also lower in Mediterranean countries, where the level of consumption of red meat is typically lower than that in most Western diets. Additional health benefits of eating Mediterranean-type and meat-free diets are the improvement of both subjective and objective measures of disease activity in patients with RA (36-39). A survey of patients with rheumatic diseases in Scandinavia found that almost 50% of patients with RA reported that their symptoms were aggravated following the consumption of meat (40). On balance, therefore, we believe that it was unnecessary to adjust the significance levels, but that the associations need verifying in an independent data set.
Although the size of the effect is moderate, the biologic plausibility of the main finding, an association between higher red meat consumption and an increased risk for the development of inflammatory polyarthritis, needs to be considered. Possible links are roles for iron and fat. Red meat is a rich source of iron, which has been shown to accumulate in rheumatoid synovial membrane, causing tissue damage (41), and, despite having a tendency toward anemia, patients with RA often do not tolerate iron supplementation well. Iron-catalyzed oxidative reactions have been shown to be the causative factors in the exacerbation of synovial inflammation following infusions of iron (42, 43). However, we found the effect of meat to be independent of iron intake, and there were no appreciable associations between different levels of fat exposure or types of fat and the risk of inflammatory polyarthritis. The effect was also independent of other confounding factors such as cigarette smoking and intake of fruit and vitamin C (an antioxidant). A higher level of protein intake from all dietary sources was associated with an increased risk of inflammatory polyarthritis, which may have some relevance. It may be that the high collagen content of meat leads to collagen sensitization and consequent production of anticollagen antibodies, most likely in a subgroup of susceptible individuals. Meat consumption may be linked to either additives or even infectious agents, but, again, there is no evidence as to what might be important in relation to RA.
In summary, the results of this study demonstrated that a high level of red meat consumption is an independent risk factor for the development of inflammatory polyarthritis, although it is unclear whether the association is a causative one.
Acknowledgements
We acknowledge the active collaboration of the rheumatologists in Norfolk, led by Professor David Scott, and the major contribution by the NOAR research nurses, led by Diane Bunn.
