For this cross-sectional study 206 volunteers living independently at home were recruited from August 2009 to September 2010 in the region of Nürnberg (Germany). Potential participants were sought through a newspaper advertisement and via personal contact in a day clinic and a rehabilitation center. In order to be included, participants needed to be 75 years or older, not suffering from any illness that profoundly impacted their diet and should not show signs of significant cognitive impairment (Mini Mental State Examination ≥ 24 out of 30 points [10]). The assessments took place either at the study site or participants were visited at home, if they were not able or willing to attend the study clinic. This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures were approved by the ethics committee of the Friedrich-Alexander-Universität Erlangen-Nürnberg. Written informed consent was obtained from all subjects.
Sample characteristics
The living situation was assessed as self-reported “living alone” or “not living alone”. The educational level of participants with only elementary school or no degree was defined as “low”, “medium” for those who attended a secondary school and “high” for participants with a university entrance diploma or higher degrees.
Participants’ height and weight were measured standing upright without shoes in light clothing. BMI was calculated for each subject as weight [kg]/height2 [m2].
The questionnaire on instrumental activities of daily living (IADL, 8 questions, max. 8 points) of Lawton and Browdy [11] was used to assess the degree of dependency in everyday life. A lower score designates a higher level of dependency. The participants’ answers on the IADL items dealing with dependency in going shopping and cooking meals were separately evaluated and documented as “goes shopping independently” and “cooks independently” vs. “needs help with shopping” and “needs help with cooking”. The use of medication was recorded as “more than three medications” or “less than three medications”. The Charlson Comorbidity Index (CCI) was used to assess comorbid conditions. From 19 diseases, weighted with 1, 2, 3 or 6 points, that have been found to increase mortality, a sum-score is calculated, with a higher score pointing to a higher mortality risk [12]. Reported chewing and swallowing difficulties were also documented.
Assessment of frailty
We used the frailty definition of Fried et al. [13] and therefore assessed the following five criteria: weight loss (self-reported, more than 4.5 kg in the last year), exhaustion (self-reported feeling that everything was an effort or that one could not “get going” more than 2 times a week), low grip strength (Jamar dynamometer, men ≤ 29–32 kg, women ≤ 17–21 kg stratified by BMI quartiles of the original study sample of Fried et al. [13]), low walking speed (depending on gender and height > 6–7 sec/ 4.57 m,) and low physical activity (men < 1.6 kJ (383 kcal)/ week, women < 1.1 kJ (270 kcal)/ week) estimated with the short form of the Minnesota Leisure Time Activities Questionnaire [14]. The cut off values for grip strength, walking speed and physical activity were derived from the lowest sex specific quintiles of the original study population of Fried et al. [13]. Subjects without any of these five attributes were categorized as non-frail, those with one or two positive criteria as pre-frail and those with three or more as frail.
Nutritional assessment
In a personal interview usual food intake was estimated using a slightly modified form of the food frequency questionnaire (FFQ) of the German part of the European Prospective Investigation into Cancer and Nutrition[15], which consists of 103 food items. Within this tool questions on the usual consumption of foods and food groups during the last 12 months are asked based on standard portion sizes (e. g. 1 cup, 1 piece, 1 teaspoon per month/ week/ day). Additionally there are questions on the kinds of fats used and the use of dietary supplements. The modifications mentioned above affected the definition of 12 food items to comply within our research objectives (e.g. subdividing the category “fish” into three categories of fish with different contents of fat and protein). Furthermore, the categories “bacon” and “salty snacks” were added, as they may contribute considerably to energy intake. Also a question on the consumption of unrefined cereals was added.
46 items of the FFQ were identified as main protein sources i. e. all foods derived from animal products (meat, egg, milk, fish), cereals (e. g. bread, rice, pasta) and protein rich vegetables (potatoes, legumes, soy). For these main protein sources the usual time(s) of consumption (morning, noon, evening) was asked in addition to the frequency of consumption.
From standard portions and frequencies of consumption, all items were converted to g/d. Daily energy and protein intake were calculated using the German nutrient database “Bundeslebensmittelschlüssel” (BLS II.3 [16]). Average daily protein intake is expressed as grams per day (g/d), grams per kg body weight (g/kg BW) and as percentage of daily energy intake (E%). Energy intake is expressed as kJ/d and kJ/kg BW. The amount of protein ingested per meal was ascertained by summing up the amounts of protein of the main protein sources for each meal. If more than one mealtime was indicated, an equal distribution of the portions over the indicated mealtimes was postulated.
Data analysis and statistics
For all statistical analyses SPSS 20.0 (IBM) software was used.
Sample characteristics are presented as median (min-max.) for continuous variables and as percent for categorial variables. The distribution of the prevalence of participants’ characteristics in the three frailty groups was tested for significant differences by χ2 testing.
Differences in continuous sample characteristics, daily protein intake (g, g/kg BW, E%) as well as in distribution of protein intake (%) over the three mealtimes in non- frail, pre-frail and frail participants were tested for significance by Kruskall-Wallis test. A coefficient of variation (CV = SD/ mean value) of protein intake (g/meal) in the morning, at noon and in the evening was calculated for every participant to estimate the unevenness of the distribution of protein intake over the day. The CV is a dimensionless, relative measure of statistical dispersion. A CV of zero connotes a total evenness of the protein intake over the day, i. e. the same amount of protein is ingested in the morning, at noon and in the evening. The more uneven the distribution is, the higher is the individual CV of protein intake. CV is presented as median (min.-max.) for each of the three frailty groups and compared by Kruskal-Wallis testing. Distribution of the CV of protein intake in the single, dichotomous frailty criteria was compared by Mann-Whitney-U testing.
The risk of being frail or pre-frail vs. non frail and the risk of each single frailty criterion, respectively, in the 2nd, 3rd and 4th quartile of protein intake (g/kg BW) vs. the 1st quartile (lowest intake) was calculated as odds ratios (OR) accompanied by 95% confidence intervals by multinomial logistic regression analyses. Confounding covariates were identified by ‘manual backward elimination’ with exclusion if an initially included factor was both insignificant and did not cause a change-in-estimate of > 10% of the exposure of interest.