Whole grain consumption and risk of radiographic knee osteoarthritis: a prospective study from the Osteoarthritis Initiative

Abstract

Objectives

To assess the association of whole grain consumption with the risk of incident knee OA.

Material and methods

We followed 2846 participants in the Osteoarthritis Initiative ages 45–79 years. Participants were free from radiographic knee OA (Kellgren–Lawrence grade <2) in at least one knee at baseline. Dietary data from baseline were obtained using the Block Brief Food Frequency Questionnaire. We defined radiographic knee OA incidence as a Kellgren–Lawrence grade ≥2 during the subsequent 96 months. Cox proportional hazards models were used to assess the association between whole grain food intake and the risk of incident knee OA.

Results

During the 96 month follow-up, 518 participants (691 knees) developed incident radiographic knee OA. Higher total whole grain consumption was significantly associated with a lower knee OA risk [hazard ratio (HR)_{quartile 4vs1} = 0.66 (95% CI 0.52, 0.84), P for trend < 0.01] after adjusting for demographic and socio-economic factors, clinical factors and other dietary factors related to OA. Consistently, a significant inverse association of dark bread consumption with knee OA risk was observed [HR_{quartile 4vs1} = 0.68 (95% CI 0.53, 0.87), P for trend < 0.01). In addition, we observed a significant inverse association between higher cereal fibre intake and reduced knee OA risk [HR_{quartile 4vs1} = 0.61 (95% CI 0.46, 0.81), P for trend < 0.01).

Conclusions

Our findings revealed a significant inverse association of whole grain consumption with knee OA risk. These findings provide evidence that eating a diet rich in whole grains may be a potential nutritional strategy to prevent knee OA.

Rheumatology key messages.

• This 96 month follow-up study revealed a significant inverse association between whole grain consumption and radiographic knee OA risk.

• The observed inverse association was independent of demographic and socio-economic factors.

Introduction

Whole grains are widely recommended as healthy foods because they are rich sources of dietary fibre, resistant starch, oligosaccharides and antioxidants, including trace minerals and phenolic compounds [1]. Compared with refined grains that only contain the endosperm, whole grains have three principal anatomical components (the starchy endosperm, germ, and bran) that must be present in the same relative proportions as they exist in the natural state [2]. Previous studies have consistently revealed the potential health benefits of whole grain consumption against the risk of major chronic diseases such as cardiovascular diseases [3], type 2 diabetes [4], some cancers [5, 6] and metabolic syndrome [7, 8]. Moreover, human trials and laboratory data also suggest that whole grain intake may alter gut microbiota composition [9], decrease systemic inflammation [10, 11], reduce low-density lipoprotein (LDL) cholesterol and lipid peroxidation and improve insulin sensitivity [12, 13].

Previous studies have suggested that some individual foods and dietary patterns may be associated with the incidence and progression of knee OA [14–17]. No study has yet examined the relationship between whole grain consumption and the risk of knee OA. Globally, as the most common form of arthritis, OA is associated with a great disease burden and substantial medical costs [18]. Knee OA is characterized by degenerative joint deformities, patients with knee OA were usually treated at severe stages, for instance, patients already with progressive damage of articular [19]. Currently there is a call for a shift from palliative care towards preventative care, which would allow OA patients to benefit from self-management through better nutrition and lifestyle [20–22]. In the present study we prospectively examined the association between whole grain consumption and knee OA risk using publicly available data from the Osteoarthritis Initiative (OAI).

Patients and methods

OAI

The OAI is a multicentre, longitudinal archive launched by the National Institutes of Health that includes 4796 participants ages 45–79 years with either established symptomatic knee OA or significant risk factors for the development of knee OA at four clinical centres (Columbus, OH; Baltimore, MD; Pawtucket, RI; Pittsburgh, PA). Participants’ clinical features, physical assessments and knee radiographs were collected at baseline and follow-up visits. The detailed protocol of the OAI can be found elsewhere [23]. Ethical approval was not required for this study. The OAI was approved by institutional review boards at the participating sites and written informed consent was obtained from each participant.

Study population

At baseline, we included participants with at least one knee free of radiographic OA [Kellgren–Lawrence (KL) grade of 0 or 1]. According to the exclusion criteria, daily total energy intake (800–4200 kcal for men, 500–3500 kcal for women) and an individual without any follow-up visits, we followed 2846 participants (4590 knees) for up to 96 months (Fig. 1). More than 83% of individuals were followed up for at least 4 years. The rate of loss to follow-up was similar among different quartile groups of whole grain consumption.

Figure 1.

Study population and exclusions in the OAI. After baseline, participants were lost to follow-up due to death, total knee replacement or nonresponse: 120 in year 2, 109 in year 3, 146 in year 4 and 469 in year 6

Assessment of dietary consumption

Whole grains and other food intake were evaluated using the Block Brief 2000 Food Frequency Questionnaire (Block FFQ) at baseline [24]. Detailed information for 70 food items (with standard units or portion size) was recorded by using nine categories of food frequency categorized as ‘never or less’, ‘a few times per year’, ‘once per month’, ‘5–6 times per week’ and ‘every day’, on average, during the previous year [25]. Based on previous studies [22, 26], participants were asked to indicate how often, on average, they had consumed various whole grain foods during the past year, including dark bread, cooked cereals (oatmeal, wheat semolina), cold breakfast cereals (Corn Flakes/Cheerios) and tortillas. Whole grain food intakes were calculated from questionnaire responses using estimated food serving sizes. We also estimated the consumption of total refined grains from rice, white bread, muffins, bagels and spaghetti. The nutrient intakes, such as total energy and dietary fibre in cereals, beans, fruits and vegetables, were derived by multiplying each food frequency and portion size by the nutrient content and summing the nutrients according to US Department of Agriculture food composition data (www.nutritionquest.com).

Ascertainment of radiographic onset of knee OA

For all the participants, bilateral weight-bearing, fixed-flexion posterior-anterior radiographs at baseline and the 12, 24, 36, 48, 72 and 96 months follow-up visits were recorded for KL grade (0–4). The interrater reliability consistency for these readings was good (weighted κ = 0.70–0.80). These KL data were publicly available [files: kXR_SQ_BU##_SAS (versions 0.6, 1.6, 3.5, 5.5, 6.3, 8.2, and 10.2)] [27]. We defined radiographic knee OA onset as any KL grade ≥2 during follow-up visits for a specific knee.

Assessment of covariates

Demographic and socio-economic factors were collected at baseline, including age, sex, race/ethnicity (characterized as African American, White or other), education level (high school or less, college and above college) and annual household income. Lifestyle and clinical factors included smoking status (never, past, current), alcohol consumption (g/day; continuous), total energy intake (kcal/day; continuous), baseline KL grade (0 or 1), history of traumatic knee injury or knee surgery, depression (defined as the 20-item Center for Epidemiological Studies–Depression scale score >16) [28], Physical Activity Scale for the Elderly (PASE) score [29] and body mass index (BMI; continuous).

Statistical analyses

We calculated person-years of follow-up from baseline to knee OA onset, death, loss to follow-up or the end of study at 96 months, whichever occurred first. Baseline characteristics were summarized as mean (s.d.) for continuous variables and percentage for categorical variables according to quartiles of whole grain consumption. Analysis of variance, chi-squared test or nonparametric tests were used for comparisons when applicable. Less than 1% of participants had missing values for BMI or PASE score, which were replaced with the sex-specific sample median. We adjusted total energy intake for whole grain and dietary fibre consumption using residual methods [16, 26].

Cox proportional hazards models were used to assess the association between whole grain consumption and knee OA risk. We applied the discrete likelihood method to handle ties of the failure times in the models. Robust sandwich covariance estimators were used to assess the intracluster dependence between two knees for the same individual [30]. In the primary multivariable model we controlled for several covariates, including age, gender, race, household income, education, smoking status, BMI, KL grade, injury/surgery, OA status of the other knee, PASE score, alcohol consumption, total protein intake and total energy intake. In a separate model we additionally adjusted for potential dietary confounding factors related to the OA risk, including daily fibre intake (g/day; fibre from whole grain excluded), total fat intake (g/day; continuous), total vitamin intake (vitamins A, D, E, K, B1, B2 and B6) and minerals (calcium, zinc) [16, 22]. Tests for linear trend were conducted by assigning the median value of whole grain consumption within each quartile group as a continuous variable. To address the potential confounding from knee OA in the contralateral knee, we included a sensitivity analysis including only participants free of OA for both knees at baseline.

To assess each whole grain food intake and OA risk association, we fitted a Cox proportional hazards model that included dark bread, cooked cereals, cold breakfast cereals and tortillas. Each grain food was mutually adjusted for other grain food types to evaluate whether the associations were independent. We also investigated food sources of fibre and their potential association with knee OA risk. We calculated fibre intake by the amount of fibre in the food source (total fibre, cereal fibre, bean fibre, fruit fibre and vegetable fibre).

Data were analysed using SAS 9.4 (SAS Institute, Cary, NC, USA) and statistical significance was defined as P < 0.05.

Results

A total of 2846 OAI participants were included in this study. The mean age was 60.5 years (s.d. 9.2), 42.7% were male and 84.5% were White. Over a 96 month follow-up period, 518 participants (691 knees) developed knee OA, and the average follow-up was 78.48 months (s.d. 27.25). Table 1 describes baseline characteristics according to total whole grain consumption. Compared with those with lower whole grain consumption, participants with higher consumption were older, more likely to be White, had a higher education and lower BMI, consumed less alcohol and were less likely to currently smoke and be depressed.

Table 1.

Baseline characteristics of participants (N = 2846) according to quartiles of whole grain intake

Characteristics	Values	Quartiles of whole grain intake
		Q1 (n = 711)	Q2 (n = 712)	Q3 (n = 712)	Q4 (n = 711)	P-value
Age, mean (s.d.), years	60.48 (9.20)	58.39 (8.65)	60.46 (9.32)	61.19 (9.29)	61.89 (9.16)	<0.01
Male, %	42.66	42.62	42.70	42.70	42.62	1.00
Race, %						0.03
Non-Hispanic White	84.50	80.17	85.39	87.50	84.95
Non-Hispanic Black	12.75	16.88	12.36	9.55	12.24
Other	2.74	2.95	2.25	2.95	2.81
Education, %						<0.01
≤High school	13.43	18.03	11.10	14.04	10.55
College	45.10	46.06	43.12	43.68	47.54
>College	41.47	35.91	45.78	42.28	41.91
Family income ($US), %
≤25 000	11.08	13.42	10.16	11.56	12.09	0.32
25 000–50 000	23.97	22.02	22.53	25.33	25.97
50 000–100 000	37.85	37.10	40.35	37.48	36.42
>100 000	27.10	27.46	26.96	25.63	25.52
Depressed, %	7.84	11.53	7.30	6.74	5.77	<0.01
Smoking status, %						0.12
Never	52.88	47.12	55.06	53.79	55.56
Current	6.39	11.25	5.62	4.78	3.94
Past	40.72	41.63	39.33	41.43	40.51
Alcohol, g/day, mean (s.d.)	7.53 (11.92)	9.89 (15.30)	7.88 (11.37)	6.62 (9.86)	5.72 (10.01)	<0.01
PASE, mean (s.d.)	164.35 (81.57)	166.76 (84.75)	164.65 (82.27)	165.85 (79.13)	160.16 (80.04)	0.18
BMI, kg/m2, mean (s.d.)	27.85 (4.50)	28.55 (4.63)	28.08 (4.50)	27.54 (4.33)	27.25 (4.42)	<0.01
KL grade (index knee) = 1, %	36.05	35.02	36.24	35.11	37.83	0.36

Average daily whole grain intake: Q1, 0.65 g/day; Q2, 1.73 g/day; Q3, 2.78 g/day; Q4, 4.59 g/day.

In primary multivariable-adjusted analysis, we observed that higher total whole grain consumption was significantly associated with a lower risk of knee OA [hazard ratio (HR)_Q4vsQ1 = 0.66 (95% CI 0.52, 0.84), HR_{quartile 3vs1} = 0.67 (95% CI 0.52, 0.85), HR_{quartile 2vs1} = 0.80 (95% CI 0.64, 1.00); P for trend <0.01]. After adjusting for vitamins, minerals and other dietary factors, the observed association was consistent [HR_{quartile 4vs1} = 0.67 (95% CI 0.52, 0.88), HR_{quartile 3vs1} = 0.67 (95% CI 0.52, 0.86), HR_{quartile 2vs1} = 0.81 (95% CI 0.65, 1.01); P for trend <0.01] (Table 2). No effect modifications by age, sex or BMI were observed. By contrast, no significant association was observed for refined grain consumption (Table 2). In the sensitivity analysis including only participants free of OA for both knees at baseline, the findings were consistent.

Table 2.

HRs and 95% CIs of incident radiographic knee OA according to quartiles of grain consumption (N = 2846)

Quartiles	Cases, n	Person-years	Model 1a		Model 2b
			HR (95% CI)	P for trend	HR (95% CI)	P for trend
Whole grains
1	152	4053	1.00	<0.01	1.00	<0.01
2	133	4257	0.80 (0.64, 1.00)		0.81 (0.65, 1.01)
3	117	4247	0.67 (0.52, 0.85)		0.67 (0.52, 0.86)
4	116	4256	0.66 (0.52, 0.84)		0.67 (0.52, 0.88)
Refined grains
1	135	4160	1.00	0.85	1.00	0.55
2	130	4269	0.90 (0.70, 1.17)		0.95 (0.74, 1.24)
3	135	4211	0.92 (0.72, 1.18)		0.97 (0.77, 1.28)
4	118	4173	0.97 (0.76, 1.23)		1.07 (0.83, 1.38)

^aAdjusted for age, sex, race (African American, White, other), baseline KL grades 0–1, injury/surgery (yes, no), OA status of the other knee, baseline PASE score, total energy intake (kcal/day; continuous), total protein intake (g/day; continuous), income, education, smoking, alcohol intake (g/day; continuous) and BMI (continuous).

^bAdditionally adjusted for vitamins A, D, E, K, B1, B2 and B6, calcium, zinc, total fat intake (g/day) and total fibre intake (g/day).

The associations between specific whole grains and the risk of knee OA are presented in Table 3. For dark bread, after controlling for other whole grain food, there was a significant inverse association for knee OA risk (HR _Q4vsQ1 = 0.68, 95% CI: 0.53, 0.87; HR _Q3vsQ1 = 0.79, 95% CI: 0.63, 0.99; HR _Q2vsQ1 = 0.79, 95% CI: 0.62, 1.02; P trend < 0.01). After adjusting for other dietary factors, the observed associations remained. A modest inverse association was also observed between higher consumption of cooked cereals and knee OA risk.

Table 3.

Hrs and 95% CIs of incident radiographic knee OA according to quartiles of specific types of grain food consumption (N = 2846)

Quartiles	Cases, n	Person-years	Model 1a		Model 2b
			HR (95% CI)	P for trend	HR (95% CI)	P for trend
Dark breadc
1	151	4093	1.00	<0.01	1.00	<0.01
2	128	4161	0.79 (0.62, 1.02)		0.81 (0.63, 1.04)
3	127	4346	0.79 (0.63, 0.99)		0.80 (0.64, 1.01)
4	112	4213	0.68 (0.53, 0.87)		0.69 (0.54, 0.89)
Cooked cerealsc
1	142	4171	1.00	0.04	1.00	0.09
2	132	4171	0.96 (0.76, 1.21)		0.95 (0.75, 1.21)
3	121	4281	0.78 (0.61, 1.00)		0.79 (0.61, 1.02)
4	123	4190	0.83 (0.65, 1.07)		0.85 (0.66, 1.09)
Cold cerealsc
1	140	4139	1.00	0.32	1.00	0.46
2	136	4188	1.03 (0.80, 1.32)		1.04 (0.81, 1.34)
3	123	4161	0.88 (0.69, 1.12)		0.90 (0.70, 1.15)
4	119	4325	0.92 (0.72, 1.19)		0.93 (0.70, 1.22)
Tortillac
1	134	4051	1.00	0.98	1.00	0.93
2	132	4140	1.03 (0.79, 1.35)		1.05 (0.80, 1.37)
3	127	4330	1.04 (0.79, 1.36)		1.05 (0.80, 1.38)
4	125	4292	1.00 (0.77, 1.30)		1.02 (0.78, 1.33)

^aAdjusted for age, sex, race (African American, White, other), baseline KL grades 0–1, injury/surgery (yes, no), OA status of the other knee, baseline PASE score, total energy intake (kcal/day; continuous), total protein intake (g/day; continuous), income, education, smoking, alcohol intake (g/day; continuous) and BMI (continuous).

^bAdditionally adjusted for vitamins A, D, E, K, B1, B2 and B6, calcium, zinc, total fat intake (g/day) and total fibre intake (g/day).

^cEach grain food was mutually adjusted for specific grain food type.

As for secondary analyses, we observed a significant association between higher cereal fibre intake and reduced knee OA risk [Table 4; HR_{quartile 4vs1} = 0.61 (95% CI 0.46, 0.81), P for trend <0.01]. After adjusting for other dietary factors, the association remained [HR_{quartile 4vs1} = 0.65 (95% CI 0.49, 0.87), P for trend = 0.01]. We also observed that higher bean fibre intake was significantly associated with an increased risk of knee OA. No significant associations were observed for total fibre intake or fruit or vegetable fibre intake with knee OA risk.

Table 4.

HRs and 95% CIs of incident radiographic knee OA according to quartiles of fibre intake (N = 2846)

Quartiles	Cases, n	Person-years	Model 1a		Model 2b
			HR (95% CI)	P for trend	HR (95% CI)	P for trend
Total fibre
1	129	4103	1.00	0.42	1.00	0.85
2	132	4229	0.91 (0.72, 1.16)		0.94 (0.74, 1.20)
3	123	4272	0.91 (0.71, 1.17)		0.97 (0.74, 1.26)
4	134	4209	0.88 (0.66, 1.18)		0.96 (0.71, 1.31)
Cereal fibrec
1	135	4082	1.00	<0.01	1.00	0.01
2	137	4219	0.77 (0.61, 0.98)		0.79 (0.62, 1.00)
3	129	4252	0.78 (0.61, 1.01)		0.81 (0.63, 1.04)
4	117	4260	0.61 (0.46, 0.81)		0.65 (0.49, 0.87)
Bean fibrec
1	116	4150	1.00	0.12	1.00	0.08
2	136	4292	1.16 (0.91, 1.47)		1.20 (0.95, 1.53)
3	126	4208	1.17 (0.91, 1.51)		1.19 (0.92, 1.55)
4	140	4163	1.35 (1.03, 1.77)		1.38 (1.04, 1.82)
Vegetables/fruits fibrec
1	133	4177	1.00	0.51	1.00	0.73
2	116	4184	0.83 (0.65, 1.06)		0.86 (0.67, 1.10)
3	131	4324	0.82 (0.64, 1.06)		0.85 (0.65, 1.10)
4	133	4128	0.86 (0.66, 1.12)		0.89 (0.67, 1.19)

^aAdjusted for age, sex, race (African American, White, other), baseline KL grades 0–1, injury/surgery (yes, no), OA status of the other knee, baseline PASE score, income and education.

^bAdditionally adjusted for total energy intake (kcal/day; continuous), total protein intake (g/day; continuous), smoking, alcohol intake (g/day; continuous) and BMI (continuous).

^cEach fibre intake was mutually adjusted for other fibre types.

Discussion

In this 96 month follow-up study we observed that total whole grain consumption was inversely associated with knee OA risk. The observed association was independent of demographic and socio-economic factors, clinical factors and other dietary factors related to OA. We also observed a stronger association of dark bread than other whole grain food with knee OA risk. A similar association was observed for grain fibre with knee OA risk. To the best of our knowledge, this is the first prospective study to report a significant association between whole grain consumption and knee OA risk. These findings have crucial implications for potential dietary prevention of knee OA development.

Numerous studies have demonstrated the potential health benefits of long-term consumption of whole grain foods [31–33]. Increased whole grains intake has shown positive effects on major chronic diseases, including some cancers [5, 34], cardiovascular diseases [35], type 2 diabetes [27], overweight and obesity [36]. The health benefits of whole grain food include reducing serum LDL cholesterol, regulating postprandial blood glucose levels and modulating gut microbiota [37, 38]. In a previous study, we found that more compliance with a prudent dietary pattern was inversely associated with the risk of knee OA, with whole grains being an essential component [22]. However, no investigators have examined the prospective association between whole grain consumption and knee OA risk in a large cohort study. Possible explanations for the observed association warrants further discussion. Long-term whole grain rye and wheat consumption have been associated with a lower level of inflammation biomarkers [39]. Whole grain foods, as natural sources of soluble fibre β-glucans, reduce serum cholesterol [40, 41]. Numerous clinical trials have also revealed the lipid- and cholesterol-lowering capacity of β-glucans from whole grain oat and barley [42, 43]. A 12 week randomized controlled trial reported that a daily serving of whole grain pasta affected inflammatory and glycemia-related markers for healthy obese or hyperglycaemic volunteers [44].

Additionally, the observed association might be associated with inflammatory markers and gut microbiota variations modulated by the long-term consumption of healthful ingredients from whole grain food. Evidence is emerging supporting that whole grain intake is closely related to gut microbiota variation and immune and inflammatory markers. A 6 week randomized trial revealed that consumption of whole grain has a modest positive effect on the composition of microbiota and short-chain fatty acid producer Lachnospira, effector memory T cells and the acute innate immune response [9]. Similarly, a randomized crossover trial with 28 healthy volunteers showed that 4 week treatments of whole-grain barley, brown rice or an equal mixture of the two produced a positive impact on faecal microbial ecology and blood markers of inflammation, glucose and lipid metabolism, in which health benefits might be attributed to modulation of the gut microbiome [45]. Several human intervention studies also suggested the benefits of cereal fibres and their active subfractions on gut microbiota composition [46].

Key molecules and regulators affected by variations in gut microbiota composition usually accompany the development and progression of OA. A large, deeply phenotyped population-based cohort showed that higher OA-related knee pain was significantly associated with Streptococcus sp. and intestinal microbiome β diversity [47]. In addition, dysregulation of intestinal microorganisms affects metabolites, for instance, short-chain fatty acids, bile acids, trimethylamine N-oxide, regulating mitochondrial function, metabolism, biogenesis, autophagy and redox reactions in chondrocytes to maintain bone tissue homeostasis [48]. Recent studies have also found that short-chain fatty acid–responsive receptors (GPR41 and GPR43) play a crucial role in maintaining gut microbiota-derived signals into the host cell microenvironment [48]. Investigations of chondrocyte transcriptomics suggest that these two core molecules regulate histone H3K9 acetylation and PGC1-α signalling to influence inflammation and chemotaxis production in chondrocytes from osteoarthritic cartilage [49].

In secondary analyses we found an inverse association between higher cereal fibre intake and knee OA risk than with other fibre sources. Limited numbers of prospective studies have assessed the risk of knee OA associated with dietary fibre consumption and its source. Findings from the OAI and Framingham Offspring OA Study showed that higher total fibre intake was related to a lower risk of symptomatic OA, while the association with radiographic OA was unclear [16]. Another longitudinal study revealed that a high intake of dietary total or grain fibre (25 g/day) was associated with a lower likelihood of moderate–severe knee pain over 8 years [50]. These findings generally supported our results. However, we observed that bean fibre intake might be associated with an increased risk of knee OA, but this association did not reach statistical significance, suggesting that the finding might be due to chance.

The present study’s strengths include using data from a large multicentre prospective study with a long-term follow-up, carefully collecting whole grain contents from various food sources and repeated assessments of knee radiographs up to 96 months. There are also some limitations. First, the dietary data were obtained at baseline. Lacking repeated records of dietary data might reduce the precision of the observed association. However, FFQ measures a long-term diet that is less likely to change significantly within several years and using quartiles in the analysis may reduce the influence of outliers. Although we carefully adjusted for multiple dietary and lifestyle confounding factors possibly associated with OA, residual confounding inevitably exists. Finally, our study participants from the OAI cohort may not represent the general population in the USA and other countries. However, it is unlikely that the underlying biological mechanisms differ substantially in other populations. Nonetheless, future studies are warranted to replicate our results in other cohorts.

Conclusion

In summary, our findings revealed a significant inverse association of whole grain consumption with knee OA risk. In addition, higher consumption of dark bread and grain fibre was significantly associated with a lower risk of knee OA. These findings provide new hints that further support whole grains as a potential nutritional strategy to prevent knee OA.

Data availability statement

All data generated and analysed in this study are available upon reasonable request to the corresponding author.

Funding

This research was supported by the National Institutes of Health (NIH) National Institute of Arthritis and Musculoskeletal and Skin Diseases (R01 AR074447 A1) and partially by generous donations to the Tupper Research Fund at Tufts Medical Center. The OAI is a public–private partnership comprised of 5 contracts (N01-AR-2–2258, N01-AR-2–2259, N01-AR-2–2260, N01-AR-2–2261 and N01-AR-2–2262) funded by the NIH, a branch of the Department of Health and Human Services, and conducted by the OAI Study Investigators. Private funding partners include Pfizer, Novartis, Merck Research Laboratories and GlaxoSmithKline. Private sector funding for the OAI is managed by the Foundation for the NIH.

Disclosure statement: The authors have declared no conflicts of interest.