Knee osteoarthritis (OA) is the most prevalent joint disease worldwide and is associated with pain, stiffness and loss of function. However, no disease-modifying treatments are available(Reference Kolasinski, Neogi and Hochberg1). Although knee OA is commonly diagnosed among middle-aged or older population, the disease process can start during earlier life. Studies have found that the prevalence of knee pain can even exceed 30 % among adults aged 30–40 years(Reference Antony, Jones and Venn2). The knee symptoms may be one of the early risk factors of knee OA in later life(Reference Antony, Jones and Jin3). A systematic review indicated that the evidence from well-conducted case–control studies supported knee pain was a predisposing factor of knee OA, though there was a paucity of well-designed cohort studies(Reference Thomas, Wood and Selfe4). Thus, identifying factors affecting knee symptoms among adults aged about 30–40 years may be important for developing prevention strategies and management of knee OA.
Diet is important in disease prevention (e.g. CVD), as the foods consumed in daily life could have beneficial or detrimental effects on our health(Reference Ros, Martinez-Gonzalez and Estruch5–Reference Askari, Daneshzad and Bellissimo8). Two approaches are generally employed in data analysis of diet: hypothesis-oriented approaches (reflecting the adherence to guidelines or recommendations, e.g. dietary score) and exploratory approaches, which focus on study-specific data, for example, principal component analysis(Reference Schulze and Hoffmann9,Reference Gerber10) . Hypothesis-oriented approaches have advantages over exploratory approaches as they are based on existing knowledge of optimal diet and provide clear information about healthy levels(Reference Wilson, Blizzard and Gall11). Several studies have reported associations between daily diet assessed by dietary score and knee OA in older adults with knee OA or at risk of OA. Cross-sectionally, adherence to a Mediterranean diet was associated with lower knee pain(Reference Veronese, Stubbs and Noale12) and lower prevalence of knee OA(Reference Veronese, Stubbs and Noale13); longitudinally, adherence to a Mediterranean diet was associated with lower risk of knee pain worsening and lower incidence of symptomatic knee OA(Reference Veronese, Koyanagi and Stubbs14). However, no studies have assessed the effect of diet on knee symptoms in younger adults, who are an important target group for disease prevention. Moreover, the Mediterranean diet index was defined according to dietary habits(Reference Willett, Sacks and Trichopoulou15), but not a national dietary guideline which represents the most recommended diet based on the current evidence.
Therefore, we aimed to describe associations between diet quality (as assessed by adherence to national dietary guidelines) in adolescence and adulthood and knee symptoms in adulthood.
Methods
Participants
In 1985, The Australian Schools Health and Fitness Survey (ASHFS) was conducted to provide benchmark data on the health and fitness of Australian schoolchildren/adolescents (7–15 years, n 8498) with a nationally representative sample. Two-stage probability sampling was used for the ASHFS to randomly select schools and then children/adolescents within age groups in schools. Students aged 10–15 years completed the 24-h diet record in adolescence and represent the adolescent cohort in the current analysis. The Childhood Determinants of Adult Health (CDAH) Study was a 20-year follow-up of children/adolescents who participated in the ASHFS. During 2004–2006, participants (aged 26–36 years) completed the food frequency questionnaire (FFQ) and food habits questionnaire (FHQ) and attended clinics (n 2410) which were located at sites in major cities and regional centres around Australia and represent the adulthood cohort in the current analysis. The CDAH Knee Cartilage Study was a 4–5 year follow-up of the CDAH study. During 2008–2010, participants residing in metropolitan Melbourne and Sydney were invited to attend a computer-assisted telephone interview as part of the CDAH Knee Cartilage Study. The exclusion criteria included: being pregnant, having other forms of arthritis, having contraindications of MRI, and finally 449 participants (aged 31–41 years) participated and completed the knee symptom questionnaire. The detailed reasons for non-participation in the CDAH Knee Cartilage Study included: 1646 did not reside in Melbourne or Sydney, 235 did not respond or refused, three were pregnant during the CDAH Knee Cartilage Study, two had rheumatoid arthritis, thirteen had contraindications of MRI and sixty-two withdrew. The details of enrolment of ASHFS, CDAH and the CDAH Knee Cartilage Study have been published elsewhere(Reference Antony, Jones and Venn2,Reference Gall, Jose and Smith16) . In the current analysis, 275 of the 449 with a knee symptom questionnaire had adolescent diet measurements when they were aged 10–15 years in ASHFS, 399 had eligible adult diet measurements (fourteen were pregnant when diet data were collected and thirty-six missed > 10 % FFQ item responses or key FHQ responses), resulting in 240 with eligible diet measurements in both adolescence and adulthood (Fig. 1).
This study was approved by the Southern Tasmania Health and Medical Human Research Ethics Committee, the Monash University Human Research Ethics Committee and the Northern Sydney and Central Coast Area Health Human Research Ethics Committee. All participants provided written informed consent.
Anthropometric measurements
Weight was measured to the nearest 0·5 kg in 1985 and 0·1 kg at follow-up (with shoes and bulky clothing removed) using a scale (Heine). Height was measured to the nearest 0·1 cm (with shoes and socks removed) using a stadiometer (Invicta). BMI was calculated as weight in kg divided by height in metres squared.
Dietary measurements
In ASHFS, participants recorded the time and estimated amount of each food or drink item consumed during a 24-h period. Trained data collectors showed the adolescents how to fill out the food record, and each adolescent was interviewed on collection of the records to check and clarify the entries. The paper questionnaires were manually processed in 1985 to provide the gram weight and the kJ energy content of each food or beverage item. The energy content of each item was determined using a specially compiled database of the nutrient composition of Australian foods(Reference English, Cashel and Lewis17). The data for each item were used for this current study to calculate the proportion of a standard serving as defined in the 2013 Australian Dietary Guidelines(18). For example, if a participant reported consuming 60 g of toast at breakfast in the ASHFS food record, this equates to 1·5 standard 40-g servings of bread. In CDAH, participants completed a 127-item FFQ and an FHQ and the paper questionnaires were scanned using Teleform (version 9.0). Each frequency reported in the FFQ was assumed to be a standard serving defined in the 2013 Australian Dietary Guidelines(18). For example, if a participant reported eating breakfast cereal once per day, this was assumed to be one standard 30-g serving of grains. Dietary Guidelines Index (DGI) and total energy intake were calculated based on the consumed servings. The DGI comprises nine indicators, and the maximum possible score was 100. A higher score indicated higher diet quality (higher adherence to Australian Dietary Guidelines). Seven indicators, worth 10 points each, related to recommended minimum intakes (dietary variety, vegetables, fruit, grains, lean meats and alternatives, low-fat dairy products and alternatives, water). For example, a participant scored 10 for fruit intake if they consumed at least two servings of fruit as guideline recommended. Participants could receive proportional scores for partially meeting the recommendations. Two indicators were for lower intake of discretionary foods (20 points), including foods high in saturated fat (e.g. cream), alcohol, added salt (e.g. bacon), and added sugars (e.g. cake) and limiting saturated fats (10 points). The details of DGI measures have been published elsewhere(Reference Wilson, Blizzard and Gall11). The changes in diet quality from adolescence to adulthood were represented by the score changes of DGI and its components (scores in adulthood minus the corresponding scores in adolescence).
Physical activity measurements
In ASHFS, participants self-reported past week duration and frequency of discretionary sport or exercise (leisure activity), walking and cycling to and from school (transport activity), school physical education and school sport. For each activity, frequency was multiplied by duration to estimate min/week and activities were summed to estimate total physical activity. In CDAH, physical activity was assessed using the long version of the International Physical Activity Questionnaire. Participants were asked to report the total time and frequency of occupational, commuting, domestic and leisure activity during the past week. Physical activities were calculated by multiplying frequency by duration to estimate min/week of vigorous, moderate and walking activity. Time spent in each domain was summed to estimate total physical activity.
Knee symptom measurements
The Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) is widely used for evaluating clinically important symptoms in OA patients(Reference McConnell, Kolopack and Davis19). WOMAC has also shown adequate responsiveness, content and construct validity for evaluating knee complaints in adolescents and adults(Reference Heintjes, Bierma-Zeinstra and Berger20). In the CDAH Knee Cartilage Study, participants completed a questionnaire, including WOMAC. Knee pain, stiffness and dysfunction during the past 30 d were assessed on a scale of 0–9 for each subscale, where 0 indicated no complaints and 9 indicated the maximum intensity of the complaint. WOMAC pain, stiffness and dysfunction were assessed in 5, 2 and 17 subscales, respectively. WOMAC scores were calculated by adding the score of subscales in each domain. The maximum possible scores of pain, stiffness and dysfunction were 45, 18 and 153, respectively. A non-zero score indicates the presence of knee symptom. Knee injury history was also recorded in the questionnaire.
Statistical analyses
Mean (standard deviation), number (percentage) and median (interquartile range) were used to describe the characteristics of the participants. Zero-inflated negative binomial regression analyses were used to assess the associations between diet quality and knee symptoms and estimate the mean ratios, as there are exceeded zeros in the knee symptom data and the sample variances significantly exceeded the sample means (over-dispersion)(Reference Hu, Pavlicova and Nunes21,Reference Rose, Martin and Wannemuehler22) . Age, sex, BMI, physical activity, total energy intake and knee injury history were included as potential confounders based on biological plausibility. A P-value <0·05 (2-tailed) was considered as statistical significance. All statistical analyses were performed in Stata (Texas, USA), version 16.0.
Results
Among the participants included in data analysis, the average adolescent and adult age with dietary data was 12·6 (sd 1·8) and 30·9 (sd 2·8) years, and the average adolescent and adult DGI was 46·5 (sd 12·6) and 55·4 (sd 11·3), respectively. The prevalence of knee pain, stiffness and dysfunction (at mean age 35·4 years) was 35·1 % (30·4 %, 40·0 %), 31·6 % (27·0 %, 36·4 %) and 39·9 % (35·1 %, 44·8 %), respectively (Table 1).
WOMAC, Western Ontario and McMaster Universities Osteoarthritis Index.
Overall adolescent DGI was not associated with knee pain, stiffness or dysfunction in adulthood (Table 2). Similarly, most DGI components (vegetable, food variety, fruit, grain, lean protein, dairy products, water, limiting saturated fat) were not associated with adult knee symptoms, although higher score for lower intake of discretionary foods in adolescence was significantly associated with lower knee pain and dysfunction in adulthood (Table 2).
DGI, Dietary Guidelines Index.
Bold denotes statistical significance, P < 0·05.
* Adjusted for age, sex, BMI, physical activity, total energy intake in adolescence and knee injury history in adulthood.
Overall adult DGI and some DGI components (vegetable, food variety, grain, lean protein, lower intake of discretionary foods) were not associated with knee pain, stiffness or dysfunction in adulthood (Table 3). However, higher score for limiting saturated fat in adulthood was significantly associated with lower WOMAC pain, stiffness and dysfunction (Table 3). In addition, higher score for drinking water was significantly associated with lower stiffness and higher score for fruit intake was significantly associated with lower dysfunction (Table 3). In contrast, higher DGI score for dairy intake was significantly associated with higher WOMAC pain, stiffness and dysfunction (Table 3).
DGI, Dietary Guidelines Index.
Bold denotes statistical significance, P < 0·05.
* Adjusted for age, sex, BMI, physical activity, total energy intake and knee injury history in adulthood.
The changes of overall DGI and most of the component scores (vegetable, food variety, grain, lean protein, dairy products, water, lower intake of discretionary foods) from adolescence to adulthood were not associated with knee pain, stiffness or dysfunction in adulthood (Table 4). However, the change of component score for limiting saturated fat from adolescence to adulthood was associated with lower stiffness, and the change score for fruit intake was associated with lower dysfunction in adulthood (Table 4).
DGI, Dietary Guidelines Index.
Bold denotes statistical significance, P < 0·05.
* Adjusted for sex, age, BMI, physical activity, total energy intake in adolescence and adulthood and knee injury history in adulthood.
Discussion
This is the first study describing the associations between diet quality in adolescence and adulthood and knee symptoms in adulthood. We found overall DGI and a large number of DGI components in adolescence and adulthood and their changes from adolescence to adulthood were not associated with knee symptoms in adulthood. However, several DGI components (lower intake of discretionary foods in adolescence, limiting saturated fat, fruit intake and water intake in adulthood) were associated with lower knee symptoms, whereas higher DGI score for dairy products in adulthood was associated with higher knee symptoms. Moreover, the change of DGI score from adolescence to adulthood for limiting saturated fat was associated with lower stiffness, and the change of DGI score for fruit intake was associated with lower dysfunction.
We did not find significant associations between the overall DGI in adolescence and adulthood and knee symptoms in adulthood. The negative results may be because the effects from the different DGI components were different. Although some components (e.g. limiting saturated fat in adulthood) were associated with lower knee symptoms, the effects may be neutralised by another component (e.g. dairy products) which has the opposite effects. Our results were consistent with a previous review, which suggested that different nutrients could have antagonistic effects on a chronic disease(Reference Gerber10).
We found that eating only lower amounts of discretionary foods in adolescence, independent of BMI, was associated with lower knee pain and dysfunction in adulthood, which has not been reported in previous studies. Consuming greater amounts of discretionary foods has been associated with metabolic diseases (e.g. type 2 diabetes) in young adults(Reference Jaworowska, Blackham and Davies23), and the associations may even persist after adjustment for BMI(Reference Krishnan, Coogan and Boggs24,Reference Duffey, Gordon-Larsen and Steffen25) . Knee OA shares many causal pathways with these metabolic diseases(Reference Fernandes and Valdes26). Thus, eating lower amounts of discretionary foods may have beneficial effects on knee joint by reducing activation of causal pathways for metabolic diseases. These beneficial effects were only associated with adolescent diets not adult diets, even though the sample size was larger in adulthood than that in adolescence. The reason is unclear but may suggest that eating lower discretionary foods is particularly important in adolescence and could have long-term effects.
We found limiting saturated fat in adulthood was associated with lower knee pain, stiffness and dysfunction in adulthood. The underlying mechanism may be the decreased detrimental effects from saturated fat and/or increased beneficial effects from unsaturated fat on knee joint. Although there are no similar studies based on younger adult populations, some studies have reported that intake of saturated fat was associated with knee structural signs of OA in middle-aged healthy adults(Reference Wang, Davies-Tuck and Wluka27) and knee OA structural progression in patients with radiographic knee OA(Reference Lu, Driban and Xu28). In addition, intake of unsaturated fat could have protective effects on knee structures(Reference Loef, Schoones and Kloppenburg29) and symptoms (pain and function)(Reference Hill, March and Aitken30,Reference Gruenwald, Petzold and Busch31) among patients with knee OA.
We found that intake of fruit in adulthood was associated with lower knee dysfunction in adulthood, which is partly consistent with previous trials. The trials reported that freeze-dried strawberry powder(Reference Schell, Scofield and Barrett32) and freeze-dried blueberry powder(Reference Du, Smith and Avalos33) could reduce pain and dysfunction in adults with knee OA. However, we did not find evident effects on pain or stiffness; the reason may be that not all fruits have the same analgesic effects as berries(Reference Schell, Scofield and Barrett32). We also found that water intake in adulthood was associated with low stiffness in adulthood. Drinking plenty of water has been associated with lower risk of chronic diseases, such as type 2 diabetes(Reference Carroll, Davis and Papadaki34), suggesting its potential benefits on metabolic fitness. The underlying mechanism of our results warrants further studies but may have some shared metabolic pathways between diabetes and knee OA(Reference Courties, Sellam and Berenbaum35).
Our finding that dairy score was associated with higher knee pain, stiffness and dysfunction should be interpreted cautiously. Previous studies reported that milk consumption (any kind) was associated lower prevalence of symptomatic knee OA cross-sectionally(Reference Kacar, Gilgil and Tuncer36) and less radiographic OA progression (decrease of joint space width) longitudinally(Reference Lu, Driban and Duryea37). However, the effects from specific kinds of milk (full-fat and skimmed milk) were not identified. A third study reported that the consumption of full-fat dairy, but not skimmed dairy, was associated with lower prevalence of knee OA(Reference Denissen, Boonen and Nielen38). In our DGI calculation, the dairy score was made up of 2 parts: 5 points were allocated for amounts of dairy products (any kind), and 5 for whether it was reduced fat(Reference Wilson, Blizzard and Gall11). Therefore, participants consuming full-fat dairy would get lower dairy scores than those consuming skimmed dairy; however, they may get beneficial effects on knee symptoms from consuming full-fat dairy.
We also found the change of score for limiting saturated fat was associated with lower stiffness and the change of score for fruit intake was associated with lower dysfunction. As diet was measured with a 24-h food record in adolescence and an FFQ in adulthood, it is not clear if the change scores represent the improvements of diet quality or are the results of different dietary measurement methods. Moreover, the sample size decreased in these analyses due to fewer participants having eligible diet measurements in both adolescence and adulthood. This means that we have less power to find statistically significant results, although some effect sizes in our results were relatively large (e.g. association between the score change for limiting saturated fat and knee pain). Therefore, further studies are needed to verify and interpret our results.
We observed that the significant associations between diet components and knee symptoms were largely evident in adulthood. This is in line with our previous study, where the effect of adult adiposity measures on knee structures in adults were much more evident than that of adolescent adiposity measures(Reference Meng, Venn and Eckstein39). Adolescent diet could be changeable during growth, so it may only have a few residual effects. Moreover, the single 24-h food record may not fully reflect the daily adolescent diet.
Our study has some strengths. First, this study used the 25-year prospective data from adolescence to adulthood and this is the first study focusing on knee symptoms in adults who are important in knee OA prevention. Second, the use of DGI represented the adherence to the Australian dietary guidelines, and the core recommendations of these national guidelines are consistent with most dietary guidelines worldwide(Reference Herforth, Arimond and Alvarez-Sanchez40). Third, our results have been adjusted for important potential confounders, including total energy intake and BMI. Some limitations of our study should be considered. First, the current sample size was determined by the available data from original cohort (ASHFS). A formal power calculation for sample size was not performed because this was a secondary analysis of the data collected in the main study. We only had a modest sample size due to the low proportion of participants who could be included in analyses. In particular, the adolescent dietary measures were only collected among adolescents aged 10–15 years, but not the whole cohort (aged 7–15 years). Participants in the current study were 1·8-year older and 0·8 kg/m2 higher in BMI than those in the remainder of the ASHFS, whereas female proportions were comparable. Thus, the generalisability of our results may be limited, and further confirmatory studies are needed. Second, we only collected knee symptom data once, so we were unable to describe longitudinal changes in knee symptoms. Third, the dietary data were collected using different measurements (the 24-h food record for adolescence and the 127-item FFQ for adulthood), which necessitated different methods of scoring the DGI components. The different scoring methods may introduce bias in calculating the longitudinal changes of DGI. In adolescence, only a single 24-h food record was taken, whereas multiday food records may be more accurate, as they may be more reflective of daily diet. Reassuringly, this method has been suggested to be extremely valuable in collecting children’s diet data despite its flaws(Reference Foster and Bradley41). In adulthood, the FFQ only collected data on frequency of food consumption and each frequency was assumed to be a standard serving, this may lead the inaccuracy in quantifying the food intake. However, the assignment of a constant portion size has been validated in epidemiological studies, though it may result in a reduction of interindividual variance(Reference Noethlings, Hoffmann and Bergmann42). Fourth, we did not collect the data regarding family history of knee OA and bleeding disorders history, which may have impacts in the development of knee OA, so we were unable to assess their potential effects on our findings.
In conclusion, the overall diet quality in adolescence and adulthood was not associated with knee symptoms in adulthood. However, some diet components such as limiting saturated fat in adulthood may affect later knee symptoms.
Acknowledgements
Special thanks to the participants who made this study possible. The roles of Liz O’Loughlin, Judy Hankin in collecting the data and Marita Dalton in managing the database are gratefully acknowledged.
This work was supported by the National Health and Medical Research Council of Australia (Grant number: 490006). The funders had no role in the design, analysis or writing of this article.
All authors participated in the study conception and design, acquisition of data, analysis and interpretation of data, preparation of manuscript, approved the manuscript for submission and publication and agreed to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
There are no conflicts of interest.