Functional disability is a major predictor of death among elderly people. From age 75 years and onwards, one in ten will lose their independence in performing basic activities with each year of life(Reference Gill, Williams and Tinetti1). The quality of the diet plays a role in overall mortality(Reference Knoops, Groot de, Fidanza, Alberti-Fidanza, Kromhout and van Staveren2), and carbohydrate foods with a low glycaemic index (GI) and glycaemic load (GL) may be of particular importance for elderly people, for whom impaired carbohydrate metabolism and insulin resistance are common conditions(Reference Teuscher, Reinli and Teuscher3). A high-GI diet is associated with the risk of acute myocardial infarction among subjects 60 years of age or above but in not those under 60 years old(Reference Tavani, Bosetti, Negri, Augustin, Jenkins and La Vecchia4). No studies have yet investigated the impact of dietary GI/GL on the physical performance of elderly people.
We hypothesized that a diet with a lower dietary GI/GL is associated with better physical performance. The physiological mechanism of this hypothesis is not completely clear, although it has been suggested that a low dietary GI provides a lower glycaemic response which, in turn, may decrease insulin secretion and longer-lasting glucose availability(Reference Jenkins, Wolever, Taylor, Barker, Fielden, Baldwin, Bowling, Newman, Jenkins and Goff5). An improved metabolic homeostasis may improve physical performance(Reference Kirwan, Cyr-Campbell, Campbell, Scheiber and Evans6). Dietary GL is an indirect measure of total insulin demand and represents both the quality (GI) and the quantity of dietary carbohydrates(Reference Brand-Miller, Thomas, Swan, Ahmad, Petocz and Colagiuri7).
Based on data from the Survey in Europe on Nutrition and the Elderly, a Concerted Action (SENECA), we investigated the associations between dietary GI/GL and physical performance at baseline and 6 years later.
Methods
The SENECA is a longitudinal, multi-centre study in men and women, randomly selected from birth cohorts of 1913 to 1918. The study was initiated in 1988/89 and followed up in 1993 and again in 1999(Reference De Groot and van Staveren8). Non-eligible subjects were those who lived in a psycho-geriatric nursing home, who were not fluent in the country’s language, and those who were not able to answer questions independently. In the initial study, the mean response rate of eighteen centres was 51 %, varying from 37 % to 62 % between centres(Reference van’t Hof, Hautvast, Schroll and Vlachonikolis9). Of the about 2600 participants from the initial study, 69 % participated in the study in 1993 and 58 % participated in the study in 1999(Reference Haveman-Nies, de Groot, Burema, Cruz, Osler and van Staveren10). About half of the male population and a quarter of the female population died during the 10-year follow-up(Reference Haveman-Nies, de Groot, Burema, Cruz, Osler and van Staveren10). A Physical Performance Test (PPT) was introduced in 1993 and repeated in 1999. The present study comprises eight of nine towns/centres participating at both follow-up studies: Yverdon-les-Bains, Switzerland; Roskilde, Denmark; Betanzos, Spain; Romans, France; Padua, Italy; Culemborg, The Netherlands; Vila France de Xira, Portugal; and Marki, Poland. Sixteen subjects from the Portuguese centre in Coimbra were excluded due to a possible examiner effect, resulting in diverging PPT results between the subjects in Coimbra and the rest of the towns/centres in 1999. Standing height was measured without shoes to the nearest 0·1 cm, and body weight was measured in underwear to the nearest 0·5 kg. A total of 765 subjects completed a PPT, a diet assessment, a physical assessment and a general questionnaire in 1993. Of these, 357 subjects completed the PPT in 1999. Between 1993 and 1999, 158 of 765 died, 180 dropped out due to hospitalization, moving away, being on holiday or were unwilling to continue to participate, two had errors in the PPT, and sixty-eight did not state any reason for not participating.
Physical Performance Test
The PPT is a validated instrument to measure simple overall body functions among older subjects with a relatively high functional status(Reference Reuben and Siu11). The test, conducted in the homes of the elderly people, consisted of the performance of seven tasks: (i) writing a sentence; (ii) simulating eating with beans on a spoon; (iii) lifting a book and putting it on a shelf; (iv) putting on and removing a jacket; (v) picking up a coin from the floor; (vi) turning 360°; and (vii) walking 15 m. A trained observer counted the seconds used to complete the first six tasks and noted the stability and continuity whereby the seventh task was performed(Reference Schroll, Bjornsbo, Ferrt and Livingstone12). Tasks 1–6 were given scores from 0 to 4, and task 7 was given a score from 0 to 3. The total PPT score ranged from 0 to 27, with a higher score indicating a better physical performance. Reasons for not completing the PPT have been described in detail elsewhere(Reference Schroll, Bjornsbo, Ferrt and Livingstone12).
Dietary assessments
In 1993, the individual dietary intake was collected using a validated diet history method, consisting of a check-list of foods(Reference Nes, van Staveren, Zajkas, Inelmen and Moreiras-Varela13). Food items were related to the meal patterns of the local community with the previous month as the reference period. The diet history method has shown a general acceptable agreement compared with a 3 d weighted record method among SENECA towns/centres(Reference van Staveren, Burema, Livingstone, van den Broek and Kaaks14). Food data were classified into forty-eight food groups of the EURO-code system(Reference Kohlmeier15). Twenty-eight food groups comprised enough carbohydrates to be included in the calculation of dietary GI and GL. Food GI and available carbohydrates (CHO) were obtained from international tables(Reference Foster-Powell, Holt and Brand-Miller16) if three principles were met: (i) the studies were implemented on a minimum of six subjects; (ii) the test foods contained 50 g of available carbohydrates; and (iii) the measuring period was 2 h for healthy subjects and 3 h for subjects with diabetes(17). Dietary GI was expressed as a percentage of the overall glycaemic response elicited by white bread: 
, where i is each of the twenty-eight carbohydrate food groups and W is the total food weight. The dietary GL was estimated by multiplying the dietary GI by a weighted intake of available carbohydrates of each food item(17).
Statistical analysis
Linear regression models were used to study cross-sectional and prospective associations between dietary GI/GL quintiles and physical performance at baseline and 6 years later. Analyses were not stratified by sex, as means of PPT scores, dietary GI and GL were similar with equal variations among men and women. Three models were applied:
1. An unadjusted model with the inclusion of baseline PPT scores in the prospective analyses.
2. A semi-adjusted model with additional inclusion of age, BMI, physical activity (grouped as much less, less or averagely active compared with peers v. more or much more active compared with peers), self-perceived general health ranging from 1 to 5 (where 1 = very poor, 2 = poor, 3 = fair, 4 = good and 5 = very good health) and number of chronic diseases (grouped as no disease/one disease v. more than one disease).
3. A fully adjusted model with additional inclusion of town/centre as a random effect.
All statistical analyses were carried out with the SAS statistical software package version 9·1 (SAS Institute Inc., Cary, NC, USA) and graphs were plotted in Sigma Plot 10 (Systat Software Inc., London, UK).
Results
The cross-sectional analyses included 765 subjects, 353 men and 412 women. Characteristics of the study sample are described in Table 1. Despite 38 % of the participants suffering from more than one chronic disease, 90 % generally considered themselves in fair or (very) good health. Dietary GI, dietary GL and PPT scores varied significantly across towns/centres, all P < 0·0001 (Table 2).
Table 1 Subject characteristics in the Survey in Europe on Nutrition and the Elderly, a Concerted Action in 1993 (n 765)
PPT, Physical Performance Test; GI, glycaemic index; GL, glycaemic load.
*Versus more or much more active compared with peers.
†Ordinal variable from 1 to 5: 1 = very poor health, 2 = poor health, 3 = fair health, 4 = good health and 5 = very good health.
‡Versus subjects with no disease or one disease.
§n 357.
∥White bread as the reference food with GI = 100.
Table 2 Dietary glycaemic index (GI), dietary glycaemic load (GL) and Physical Performance Test (PPT) score distributed by eight towns/centres of the Survey in Europe on Nutrition and the Elderly, a Concerted Action in 1993 (n 765)
*White bread as the reference food.
†Significant difference in GI and GL between towns/centres (Kruskal–Wallis or χ 2 test): P < 0·0001.
Cross-sectional and prospective analyses are presented in three adjustment models in Table 3. In the unadjusted analysis, mean baseline PPT scores decreased significantly across ascending dietary GI quintiles. With adjustment for age, BMI, physical activity, self-perceived health, chronic diseases and town/centre, the association became non-significant (Table 3). The magnitude of the association was higher in the unadjusted analysis than in the fully adjusted model (Fig. 1(a)). In the unadjusted analysis, the difference in dietary GI between the first quintile and the fifth quintile corresponded to a difference of 0·68 in PPT score.
Table 3 Regression estimates of changes in Physical Performance Test (PPT) scores in 1993 (n 765) and 1999 (n 357) with changes of one quintile of dietary glycaemic index (GI)/glycaemic load (GL) in 1993. Data from the Survey in Europe on Nutrition and the Elderly, a Concerted Action
β, regression estimate; see, standard error of the β estimate.
*Unadjusted model in 1993: adjusted for PPT score in 1993, where outcome is PPT score in 1999.
†Adjusted for same as Model 1 plus age, BMI, self-reported physical activity, self-perceived health and number of chronic diseases.
‡Adjusted for same as Model 2 plus town/centre.
§Significant association between dietary GI/GL and PPT score (t test): P < 0·05.
Fig. 1 Mean scores of the Physical Performance Test (PPT), with standard error of the mean represented by vertical bars, plotted as a function of dietary glycaemic index (GI) quintiles (a) and dietary glycaemic load (GL) quintiles (b) in unadjusted (· · · ● · · ·) and adjusted (· · · ○ · · ·) associations: Survey in Europe on Nutrition and the Elderly, a Concerted Action in 1993 (n 765). Adjusted associations include age, BMI, self-reported physical activity, self-perceived health, number of chronic diseases and town/centre. The linear trend across GI quintiles was P = 0·05 unadjusted and P = 0·08 adjusted and across GL quintiles was P = 0·05 unadjusted and P = 0·92 adjusted
In the unadjusted analysis, mean baseline PPT scores decreased significantly across ascending dietary GL quintiles (Table 3). This association remained significant after adjustment for age, BMI, physical activity, self-perceived health and chronic diseases, but became non-significant with additional adjustment for town/centre. The magnitude of the association was higher in the unadjusted analysis than in the fully adjusted model (Fig. 1(b)). In the unadjusted analysis, a difference in dietary GL between the first quintile and the fifth quintile corresponded to a difference of 0·64 in PPT score.
Three hundred and fifty-seven subjects of the baseline study sample had also completed the PPT in 1999. During the 6 years of follow-up, mean PPT score decreased significantly by 2·7 points from 22·6 to 19·9. Dietary GI and dietary GL were not significantly associated with PPT score 6 years later in the unadjusted analysis or after adjustment as indicated above (Table 3).
Discussion
Our results suggest that dietary GI/GL might play a significant role in physical performance among 75–80-year-old Europeans. We found a difference of 0·68 in PPT score between subjects in the first quintile and the fifth quintile of dietary GI, corresponding to an 18-month decrease in the PPT scores observed among the SENECA cohort. Equivalently, we found a difference of 0·64 in PPT score between subjects in the first quintile and the fifth quintile of dietary GL, corresponding to a 17-month change in the PPT scores observed among the SENECA cohort. This means that those consuming a diet with a high dietary GI/GL may have a 17–18-month accelerated age-related decrease in physical performance compared with those consuming a diet with low dietary GI/GL. No attempt was made to examine whether a high physical performance was caused by postprandial decreases in blood glucose elicited from a lower dietary GI/GL or high diet quality in itself(Reference Davis, Miller and Mitchell18).
A borderline-significant association between dietary GI and physical performance at baseline remained after adjustment for several confounders. In addition, it seems that town/centre may have a particularly significant effect on the association between dietary GL and physical performance, as the magnitude of the association decreased substantially after we accounted for the possible clustering effect of geographical location. This could be caused by the large variation in habitual food intake across countries(Reference Schroll, Carbajal, Decarli, Martins, Grunenberger, Blauw and de Groot19). From a previous study of SENECA subjects, we know that total energy intake was overestimated in the Polish centre and underestimated among men in the French centre(Reference van Staveren, Burema, Livingstone, van den Broek and Kaaks14). This may partly explain the strong effect of town/centre on the association between dietary GL and physical performance. Another SENECA study has found that the total energy intake of the elderly people in some towns/centres decreased more than 1 MJ/d between 1988/89 and 1993(Reference Moreiras, van Staveren, Amorim Cruz, Carbajal, de Henauw, Grunenberger and Roszkowski20). Such changes in habitual diet prior to baseline in our study may have weakened the causality of the cross-sectional association between dietary GI/GL and physical performance. Smoking, socio-economic status and dietary factors, including macronutrient intake, dietary fibre and alcohol intake, were not significant confounders in the present study.
We did not find a significant association between dietary GI/GL and physical performance 6 years later. It is likely that changes in health and social circumstances could have caused changes in the habitual diets between baseline and follow-up, which, in turn, may have attenuated the studied associations. However, the diet was not assessed at the follow-up study in 1999. Another explanation for the non-significant prospective results could be that subjects were more homogeneous than previously, and that the statistical power was weakened due to a reduced sample size in 1999 compared with 1993.
One major strength of the present study was the application of an objective test of physical performance, validated for use among healthy older subjects(Reference Brach, van Swearingen, Newman and Kriska21). Measuring simple body functions makes it less likely that subjects have compensated for reduced strength by changing movements, which may be the case when measuring daily activities(Reference Brach, van Swearingen, Newman and Kriska21). On the other hand, our study may be limited by the use of the diet assessment method, which was not constructed for studying dietary GI. However, the mean dietary GI was of the same magnitude among the SENECA cohort and elderly Dutch men from the Zutphen Elderly Study. In both studies the diets were assessed using a cross-check dietary history method(Reference van Dam, Visscher, Feskens, Verhoef and Kromhout22). Another caveat is that the international GI tables are based on studies of healthy young subjects and food items primarily from America and Australia. Therefore, international differences in food ingredients of habitual food items, e.g. wheat bread, may have resulted in misclassification of GI/GL. Despite a decreased glucose metabolism among elderly subjects, the GI ranking of food items may still be comparable to those of younger subjects(Reference Teuscher, Reinli and Teuscher3). A low dietary GI would still decrease the postprandial glycaemic response among elderly subjects(Reference Kaplan and Greenwood23). Finally, dietary intake, physical activity, self-perceived health and chronic diseases were all self-reported measures, and hence memory loss or subjective judgements may have influenced results.
In conclusion, a higher dietary GI/GL was associated with a lower concurrent physical performance among Europeans at 75–80 years of age. These associations did not remain significant after adjustment for age, BMI, physical activity, self-perceived health, chronic diseases and town/centre. Neither dietary GI, nor dietary GL, was associated with physical performance 6 years later.
Acknowledgements
The study was supported by the FREJA Programme and the Danish Research Agency. All co-authors have approved the submission of the manuscript and there are no conflicts of interest. Access to the SENECA database was facilitated by B.L.H. In cooperation with B.L.H., K.S.B. and I.T., I.K. planned the study purpose and interpreted the study results. K.S.B. was involved in the collection of the Danish SENECA data. I.K. performed the statistical data analyses with assistance of statistician Peder Frederiksen. We thank the SENECA investigators for skilful examination of participants.
