Health effects of protein intake in healthy elderly populations: a systematic literature review

The purpose of this systematic review is to assess the evidence behind the dietary requirement of protein and to assess the health effects of varying protein intake in healthy elderly persons in order to evaluate the evidence for an optimal protein intake. The literature search covered year 2000–2011. Prospective cohort, case–control, and intervention studies of a general healthy population in settings similar to the Nordic countries with protein intake from food-based sources were included. Out of a total of 301 abstracts, 152 full papers were identified as potentially relevant. After careful scrutiny, 23 papers were quality graded as A (highest, n=1), B (n=18), or C (n=4). The grade of evidence was classified as convincing, probable, suggestive, or inconclusive. The evidence is assessed as: probable for an estimated average requirement (EAR) of 0.66 g good-quality protein/kg body weight (BW)/day based on nitrogen balance (N-balance) studies and the subsequent recommended dietary allowance (RDA) of 0.83 g good-quality protein/kg BW/day representing the minimum dietary protein needs of virtually all healthy elderly persons. Regarding the optimal level of protein related to functional outcomes like maintenance of bone mass, muscle mass, and strength, as well as for morbidity and mortality, the evidence is ranging from suggestive to inconclusive. Results from particularly prospective cohort studies suggest a safe intake of up to at least 1.2–1.5 g protein/kg BW/day or approximately 15–20 E%. Overall, many of the included prospective cohort studies were difficult to fully evaluate since results mainly were obtained by food frequency questionnaires that were flawed by underreported intakes, although some studies were ‘calibrated’ to correct for under- or over-reporting. In conclusion, the evidence is assessed as probable regarding the EAR based on N-balance studies and suggestive to inconclusive regarding an optimal protein intake higher than the estimated RDA assessed from N-balance studies, but an exact level cannot be determined. Potentially adverse effects of a protein intake exceeding 20–23 E% remain to be investigated.

T he present literature review is part of the fifth version of the Nordic Nutrition Recommendations (NNR5) project with the aim of reviewing and updating the scientific basis of the fourth edition of the NNR issued in 2004 (1). The NNR5 project is mainly focused on a revision of those areas in which new scientific knowledge has emerged since the fourth edition, with special relevance for the Nordic setting. A number of systematic literature reviews form the basis for establishment of dietary reference values in NNR5. The present expert group was established to systematically review studies regarding nitrogen balance (N-balance) and protein quantity and sources (animal versus vegetable) associated with health outcomes among healthy elderly populations.
In 2002, the Institute of Medicine (IoM) published the US dietary reference values for protein (2) that was mainly based on a meta-analysis of N-balance studies by Rand et al. (3) to estimate protein requirement. This meta-analysis included only one study that reported individual data on protein requirements of older subjects (4). They reported no statistically significant variations by adult age or sex.
The meta-analysis by Rand et al. was also taken into account in the NNR4 protein requirement assessment. While the US recommendation was expressed as 0.83 g (page number not for citation purpose) ae Review Article Food & Nutrition Research 2014. # 2014 Agnes N. Pedersen and Tommy Cederholm. This is an Open Access article distributed under the terms of the Creative Commons Attribution-Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. good-quality protein/kg body weight (BW)/day, the NNR4 recommendation was given as 10Á20 energy percent (E%) from protein, which allowed for an overall macronutrient intake distribution, as well as adaptation to the Nordic dietary habits. The Nordic recommended protein intake of 10Á20 E% was considered adequate to meet the requirement for protein, including essential amino acids for healthy elderly persons too.
In 2007, the WHO/FAO/UNU published their most recent recommendation on protein requirement (5), which was also based on the Rand et al. meta-analysis (3), but with increased requirements for most essential amino acids, which made a certain level of protein quality necessary. In 2012, the European Food Safety Authority (EFSA) published their Dietary Reference Values for protein based on N-balance studies (6), again mainly the Rand meta-analysis (3). These recommendations were the same for the healthy elderly as for the general populations. Both the WHO and the EFSA Panel considered several health outcomes associated with protein intake, but data were found to be insufficient to establish Dietary Reference Values.
Thus, until now, recommendations on protein requirements have been based on N-balance studies only. The recommendations of an optimal protein intake in relation to health outcomes are not clear. The present evidence on the relationship between protein intake and health outcomes has, however, not been based on systematic literature reviews.
The purpose of this systematic review was to re-assess the evidence for the dietary requirement of protein based on N-balance studies, and to assess the health effects of varying protein intake in healthy elderly populations based on prospective observational cohort or caseÁ control studies and randomized controlled studies.

Methods
The process for conducting the systematic review is described in detail in the guidelines devised by the NNR5 working group (7). Briefly, the key characteristics of the systematic review are 1) Definition of the research questions to be answered. 2) Definition of the eligibility criteria.
3) A systematic search that attempts to identify all studies that would meet the eligibility criteria. 4) A systematic selection and evaluation of the included papers. 5) Construction of summary tables of the studies. 6) Rating the evidence and formulate conclusions.

Research questions
The main protein expert group for the adult population made the research questions in cooperation with other relevant expert groups including the Elderly Group, see Appendix 1.

Eligibility criteria
We included studies reporting protein intake from foods, but excluded studies using isolated protein as supplements, as well as studies based on the intake of certain amino acids.

Population
Studies of general healthy elderly populations with a mean age of ]65 years in settings similar to the Nordic countries were included, while studies with disabled/frail elderly were excluded. But since the review deals with elderly people, the described studies are likely to include people with various health problems such as sarcopenia, cardiovascular risk factors like hypertension and dyslipidemia, or some diseases that did not hinder free-living. Studies without Caucasians or with Caucasians as a minority group were excluded. Secondary prevention studies, studies that addressed adiposity or obesity, and studies on athletes were also excluded.

Study type and design
Observational studies: prospective cohort studies and caseÁcontrol studies were included, while cross-sectional studies were excluded. Studies were also excluded if length of follow-up was obviously too short to reliably assess the stated outcome. No studies with less than 1-year follow-up were included.
For controlled intervention studies, the required length of study depended on outcome; for N-balance studies the length was set to at least 14 days in accordance with a recent meta-analysis (3). Single meal postprandial studies were excluded. Required number of participants depended on outcome and power calculations.
Publication language had to be English or any of the Nordic languages.

Publication type
Original articles, meta-analyses, and systematic reviews were included. Narrative reviews were examined to ensure that all relevant studies were included.
Time period for publication 2000 onward.

Search method and terms
The protein expert group that addressed the adult population defined the search terms relevant for both adults and the elderly in collaboration with a librarian, see Appendix 2. The databases used were PubMed and SweMed (the latter was used to identify Nordic papers not published in PubMed). The main search included the period January 2000 to January 2011. An additional Agnes N. Pedersen and Tommy Cederholm search was run in Medline through the PubMed platform in January 2012 in order to update the search with the most recent papers published from January to December 2011.

Selection and evaluation of papers
Screening of all the papers was carried out by the two authors. The first screening included 301 abstracts sent from the main protein group. All articles suggested by at least one of the two experts were ordered as full text papers. The second screening included 152 full text papers, and papers suggested by at least one expert were included in the quality assessment. The quality assessment was done according to the principles in the guidelines (7). Briefly, a quality assessment tool specific for the study type was used to grade the papers as A0high quality study with very low level of potential bias; B0some bias, but not enough to invalidate the results; C 0significant bias and weaknesses that may invalidate the results. Then, after the quality grading, evidence tables were constructed with a description and the quality assessment of each study. Finally, for each evaluated outcome with more than one article, the grading of evidence was based on summary tables and a four-class grading: convincing (high), probable (moderate), suggestive (low), and inconclusive (insufficient). The minimum requirement for suggestive was two studies showing an association, and no conflicting results. If some studies showed a nonsignificant (neither positive nor negative) association, it was decided that for suggestive evidence the number of studies showing an association was required to be at least two higher than those showing no association.

Results
The total search included 301 abstracts. The initial main search included 267 abstracts screened for eligibility ( Fig. 1). Of these, 137 were selected and ordered as full text papers, including narrative reviews. The additional search (JanuaryÁDecember 2011) resulted in 34 abstracts of which 15 were ordered as full text papers, including narrative reviews. Thus, 152 full text papers were included in the second screening.
A total of 23 papers were quality graded, including two additional papers identified through reference lists from the included papers and the narrative reviews ( Fig. 1).
The reasons for exclusion of the 131 full text papers are shown in Appendix 3.

(34) abstracts
First screening 137 (15) (3) included 19 N-balance studies of eucaloric diets with at least three test protein intakes given. Only one study (4) in the meta-analysis reported individual data on requirements of older subjects. The meta-analysis found no significant differences in requirements between adult age, sex, or source of dietary protein, but they also stated that the data did not provide sufficient power to detect possible differences. The median estimated nitrogen requirement was 105 mg N/kg BW/day corresponding to 0.66 g goodquality protein/kg BW/day, and based on the 97.5th percentile, the estimated recommended dietary allowance (RDA) was set to 0.83 g good-quality protein/kg BW/day. In 2001, Campbell et al. (8) conducted a small 14-week strictly controlled metabolic study with 10 healthy elderly men and women who were provided 0.8 g protein/kg BW in a eucaloric diet. There was no young control group. Steady state was reached at week 2 indicating that the protein intake was adequate for the participants. But hereafter the urinary nitrogen excretion decreased and was associated with a loss in mid-thigh muscle area. The authors suggested that the protein intake might have been marginally inadequate and result in longer-term accommodation in skeleton muscle.
In a controlled metabolic study by Morse et al. (10), 11 healthy elderly women were provided diets with three different protein intake levels, that is, low protein diet: 0.5 g/kg BW; medium: 0.75 g/kg BW; high: 1.0 g/kg BW during three 18-day periods with a minimum 1-week habitual diet between the periods. N-balance was determined at week 2 and 3 of each diet. Mean adequate protein allowance was estimated to be 0.90 g/kg BW at week 2 and 0.76 g/kg BW at week 3, but the urinary nitrogen excretion decreased between week 2 and 3 indicating that a steady state was not reached yet.
In 2008, Campbell et al. (9) tested young versus old, and men versus women in a high quality controlled metabolic study with a low protein (0.5 g/kg BW), medium protein (0.75 g/kg BW) and high protein (HP) (1.0 g/kg BW) diet. The included elderly women were apparently the same as in the study by Morse et al. (10). After 18 days, the N-balance was not different between the four groups, and the estimated requirement expressed per kg BW was not significantly different for the young versus old, or men versus women. Mean protein requirement was lower for older women versus older men, but expressed per kg fat-free mass (FFM) there was no significant difference. For all subjects combined, the adequate protein allowance was estimated to be 0.859 0.21 g/kg BW/day, and not statistically different from the estimate of 0.83 g/kg BW/day, as suggested by Rand et al. (3).
A short-term study was also included because of an HP intake in the test meal versus usual protein (UP) intake (11). Young men and women (UP 1.04 g/kg BW and HP 2.08 g/kg BW) versus old men and women (UP 0.89 g/kg BW and HP 1.79 g/kg BW) were tested for 10 days on each diet in a cross-over design. There was no age related difference in N-balance. Still, there was concern about an HP diet corresponding to ca. 24 E% in the elderly because of a potentially negative effect on the kidney function expressed as lack of increase in glomerular filtration rate (GFR) from a habitual low GFR value among the elderly participants.
In summary, the evidence is assessed as probable regarding the estimated average requirement (EAR) of 0.66 g good-quality protein/kg BW per day and the subsequent RDA of 0.83 g good-quality protein/kg BW per day for all adult age groups, including the elderly, based on N-balance studies ( Table 1).
The evidence of potential adverse effects of a HP diet (ca. 24 E%) based on only one study is regarded as inconclusive (Table 1).

Protein intake and muscle mass
The evaluation of the association between protein intake and muscle mass among healthy elderly persons is based on one randomized controlled trial (12), quality graded as B, and two prospective cohort studies (13, 14), quality graded as B and C, respectively, see Appendix 4, Table 2.
In a small study with 12 healthy elderly women, Castaneda et al. (12) provided two weight-maintaining diets with marginal protein (0.45 g/kg BW) and adequate protein (0.92 g/kg BW), respectively. This corresponded to an estimated protein E% of ca. 6 E% and 13 E%, respectively. After 10 weeks, the women with the marginal protein intake had statistically significant decreased muscle mass from 1790.9 to 14.790.8 kg, and decreased muscle fiber cross-sectional area (CSA) (type I fibers) as well as a decrease in the protein-sensitive anabolic mediator insulin-like growth factor-1, IGF-1, while the women with the adequate protein intake achieved increased muscle fiber CSA and IGF-1.
In the Health ABC study, Houston et al. (13) measured total lean mass (LM) and non-bone appendicular LM (aLM) with DXA in community-dwelling older men and women. Total, animal and vegetable protein intake were measured with a food frequency questionnaire (FFQ) at baseline and expressed in energy adjusted g/day and E%. The mean daily protein intake was 0.9 g/kg BW, and after 3-year follow-up the mean decrease in LM and aLM was 0.6891.94 and 0.4891.08 kg, respectively. Those in the highest quintile of protein intake ( :19 E%) lost 40% less LM and aLM compared to the lowest quintile (:11 E%). The change in LM and aLm was statistically significant related to total and animal protein intake, but not to vegetable protein intake. However, there was no statistically significant difference in loss of LM between quintiles among the subgroup of weight stable participants (49.5%).
As part of a prospective randomized controlled cohort trial of supplemental calcium to prevent fractures, Meng et al. (14) analyzed the association between total protein intake and body composition after 5 years, expressed as LM and bone mineral content (BMC) in elderly women. They found that those in the highest tertile of total protein intake (T3:!1.6 g/kg BW or 20 E%) compared to the lowest tertile (T1: 0.85 g/kg BW or 18 E%) had higher LM (T3: 37.494.8 kg versus T1: 35.594.5 kg) and BMC, independent of age, body size, energy intake and physical activity. The energy intake in the lowest tertile of protein intake ( :5.3 MJ) indicates underreporting, and thus the actual protein intake is difficult to assess.
The evidence is assessed as suggestive regarding a positive relation between muscle mass and total protein intake in the range of 13Á20 E% ( Table 2).

Protein intake and bone health
The evaluation of the association between protein and bone health is based on two randomized controlled trials, eight prospective cohort studies, and one caseÁcontrol study, see Appendix 4, Table 3.
Bone mineral density/bone mineral content One intervention study (15), quality graded as C, and three cohort studies (14,16,17), quality graded as C, B and B, respectively, were identified on the association between protein and bone mineral density (BMD) or BMC.
In a randomized controlled study,  exchanged carbohydrates isocalorically with meat resulting in a HP diet (HP: 2498 E%) and compared this with a low protein diet (LP: 1693 E%) among 32 elderly men and women. After 9 weeks, the HP group had increased BMC and IGF-1 compared to the LP group. The reported dietary intakes indicated underreporting (an energy intake of ca. 6 MJ), thus the actual protein intake is unknown.
The association between total protein intake and the relation to BMD after 1-year follow-up was assessed among Australian elderly women (16). Compared to the lowest tertile of protein intake, the two highest tertiles (!0.84 g/kg BW) together was associated with higher BMD at the hip. There was no baseline measurement of BMD.
As part of a prospective randomized controlled cohort trial of supplemental calcium to prevent fractures,  Meng et al. (14) analyzed the association between total protein intake and body composition after 5 years, expressed as LM and BMC in elderly women. They found that those in the highest tertile of total protein intake (!1.6 g/ kg BW or 20 E%) compared to the lowest tertile (0.85 g/ kg BW or 18 E%) had a 6% higher BMC. The energy intake in the lowest tertile of protein intake (:5.3 MJ) indicates underreporting, and thus the actual protein intake is difficult to assess.
The Rancho Bernardo study among men and women (17) reported an inverse association between vegetable protein intake (energy adjusted g/day) and BMD after 4 years. Among women only, there was a positive association between animal protein intake and 4-year-BMD for every 15 g/day increase.
The evidence is assessed as suggestive regarding a positive association between protein intake and BMD, but the protein intake level is unknown due to underreporting (Table 3).

Bone loss
One intervention study (18), quality graded as B, and four prospective cohort studies, two quality graded as B (17,19), and two quality graded as C (20, 21), were identified on the association between protein intake and bone loss, see Appendix 4, Table 3.
In a randomized controlled trial with calcium and vitamin D supplementation versus placebo, Dawson-Hughes and Harris (18) compared habitual dietary protein intake with a 3-year change in BMD. The highest tertile (20 E% or 1.2 g/kg BW) was associated with less bone loss compared to lowest tertile (14 E% or 1.1 g/kg BW), but only in the intervention group. The habitual mean intake in the placebo group was 871 mg calcium and ca. 7 mg vitamin D.
In the Framingham Osteoporosis Study, Hannan et al. (19) examined 4-year bone loss (change in BMD) among men and women, and found an inverse relation for bone loss to total and animal protein intake expressed as E%, but no statistically significant association to vegetable protein. Mean total protein intake was 16 E% and mean animal protein intake accounted for 10 E%. Expressed in quartiles, the lowest quartile of total protein intake (7.3Á13.5 E% or 0.21Á0.71 g/kg BW) had the greatest bone loss, and the highest quartile (19.9Á27.4 E% or 1.23Á2.78 g/kg BW) showed the lowest bone loss. A similar association was seen for animal protein intake.
In the Rancho Bernardo study among men and women (17), the annual change in BMD during 4 years did not show a statistically significant association to baseline protein intake.
In an osteoporosis intervention study (the STOP IT trial), 96 women of the placebo group were followed prospectively for 3 years (20). Baseline protein intake was expressed as quartiles of protein intake in E%, whereas BMD was measured at baseline and after 3 years. The lowest quartile of total protein intake ( :13 E%) was not associated with bone loss when compared to the highest quartile ( :20 E%).
The Study of Osteoporotic Fractures in postmenopausal women (21) expressed the protein intake in an animal/ vegetable (A/V) ratio. 72% of the total protein intake was from animal sources. They found a positive association between the A/V ratio and the rate of bone loss after 3.6 years.
The evidence is assessed as inconclusive regarding the relation of protein intake to bone loss (Table 3).

Bone fracture
Two prospective cohort studies (21, 22), quality graded as C, and B, respectively, and a caseÁcontrol study (23), quality graded as B, were identified on the association between protein and risk of fractures, see Appendix 4, Table 3.
The Study of Osteoporotic Fractures in postmenopausal women (21) expressed the protein intake in an animal/ vegetable (A/V) ratio. 72% of the total protein intake was from animal sources. They found an increased risk of hip fractures after 7 years related to the A/V ratio. When the model was adjusted for BMD, the relation of A/V ratio to fracture risk became non-significant. The protein intake assessment was flawed by a substantial underreporting, that is, :5 MJ.
In the Framingham Osteoporosis Study (22), men and women had their risk of hip fractures assessed after 11.6 person years. The authors found that the upper three quartiles of total protein intake expressed as energy adjusted g/day, compared to the lowest quartile, were associated with a 37% lower risk of hip fracture.
In a caseÁcontrol study from 1997 to 2001 including men and women stratified by age groups 50Á69 or 70Á89 years (23), the intake data (obtained after an average of 4.2 months) of total, animal and vegetable protein E% were expressed in quartiles. The lowest quartile of total protein intake was 5.6Á13.9 E% and the highest quartile was 17.4Á30.8 E%. Animal protein intake was about 2/3 of the total intake. Highest versus lowest quartile of total, animal and vegetable protein were all inversely related to risk of osteoporotic hip fractures, but only among the age group 50Á69 years.
The evidence is assessed as inconclusive regarding the relation of protein intake to risk of fractures (Table 3).
Only one study was identified on the association between protein intake and risk of falls (24) and thus, the evidence is assessed as inconclusive. This study found no statistically significant associations between falls and total animal and vegetable protein intake after 12 months follow-up; see Appendix 4, Table 3.  Protein intake and physical training The evaluation of the association between protein and physical training is based on only two randomized controlled trials (25, 26), both quality assessed as B, see Appendix 4, Table 4. In a strictly controlled metabolic study by Campbell et al. (25), all subjects were assured to have a regular protein intake of 0.8 g protein/kg BW per day in euenergetic menus in order to maintain BW stability, whereas they were randomized to resistance training or a sedentary behavior for 12 weeks. The 17 old adults (mean age about 65 years) who trained were able to increase strength and their muscle mid-thigh area (measured with CT), whereas fat-free mass and body water (measured with deuterium dilution) decreased. The sedentary group had a decrease in mid-thigh muscle area by an intake of 0.8 g protein/kg BW ( :10 E%).
The study by Haub et al. (26) showed that the effect on strength and muscle mass by resistance training for 12 weeks in 21 men aged 6595 years during weight stability did not differ with protein source, that is, beef-containing diet or lacto-ovo-vegetarian diet. The diets altogether provided more than 1 g protein/kg BW per day (range 1.03Á1.17 g/kg BW). Both groups had increased muscle mass and strength.
Due to few studies, the evidence is assessed as inconclusive regarding the relation of total protein intake and sources of protein (animal versus vegetable protein) to muscle mass and body composition in combination with resistance training.

Protein intake and various outcomes
For some outcomes, only one study was identified on the association to protein intake and thus, the evidence is assessed as inconclusive; see Appendix 3, Table 5.

Blood pressure
The Rotterdam prospective cohort study (27), quality graded as B, studied the association between risk of hypertension and intake of energy adjusted tertiles of total animal and vegetable protein among persons ]55 years of age without hypertension at baseline. The lowest tertile of total protein intake was 70915 g/day (14 E%) and the highest was 97919 g/day (19 E%). They found that among persons ]70 years of age animal protein intake was related to increased risk of hypertension after 6 years follow-up, otherwise no statistically significant associations were observed.

Frailty
In the Women's Health Initiative Observational Study (WHI-OS), quality graded as B, Beasley at al studied the risk of incident frailty among 24,471 non-frail elderly women (28). The assessed protein intake with the use of FFQ was calibrated according to a study with objective biomarkers of energy intake (doubly labeled water) and protein intake (urinary nitrogen). These results demonstrated that the FFQ reported energy intake was considerably underestimated, that protein was modestly underestimated, and thus, that protein E% was overestimated. After three years, 13.5% were classified as frail. Protein intake, expressed as both uncalibrated and calibrated intake, was inversely associated to frailty, but the association was stronger for the calibrated intake.

Mortality
The British National Diet and Nutrition Survey of people aged 65 years and over (29), quality graded as C, assessed dietary intake at baseline in 1994Á95 and associated the protein intake in g/day to all-cause mortality 14 years after. They used a 4-day weighed dietary record, but there was no further information in the paper about the method or validation. Obviously there was under-reporting among the women (energy intake B6 MJ). The study reported a decreased risk of all-cause mortality associated with higher protein intake.

Discussion
The main findings of this systematic review on protein intake and the relation to health outcomes in healthy elderly populations comparable to the Nordic populations, are that the evidence is assessed probable regarding the EAR of 0.66 g good-quality protein/kg BW/day based on N-balance studies and the subsequent estimated RDA of 0.83 g good-quality protein/kg BW/day representing the basis for an estimation of the minimum dietary protein needs of virtually all healthy elderly persons. This corresponds to an average intake of approximately 10 E% from protein. The estimation of an optimal level of protein intake based on the evidence from randomized controlled trials, caseÁcontrol studies and prospective cohort studies of the relations of protein intake to functional outcomes (e.g. maintenance of bone mass, muscle mass and strength), morbidity, and mortality are ranging from suggestive to inconclusive, but altogether pointing in the direction of beneficial effects from a somewhat higher intake than the minimum need.
It should be noted that the grading of the evidence was only based on studies from 2000 up to and including 2011, and for some outcomes inclusion of earlier studies might have resulted in different grading. The most recent recommendations of protein intake from USA (2) and EFSA (6) are based on N-balance studies, while the relation between health outcomes and protein intake was considered insufficient to establish reference values (6) or recommendations (2). Studies with total, animal or vegetable protein were included in this review while studies at amino acid level were not included. The usual diet in the Nordic countries is considered unlikely to be limited in their content of essential amino acids, and thus, we did not regard it relevant to make an update of the comprehensive work by WHO/FAO/UNU expert group from 2007 about amino acid requirements (5). We included mainly long-term studies with only healthy old adults under free-living conditions. Postprandial (acute/single meal studies) and short-term studies, and also studies with protein isolates may not reflect the effect from ad libitum long-term dietary habits and/or mechanisms like adaptation. Studies including both healthy persons and persons with disease risk factors show different outcomes on nutrition exposures, e.g. greater reductions in LDL-cholesterol values among hypercholesterolemic individuals compared to individuals with normocholesterolemia after intervention with casein or soy protein (30, 31), or greater reductions in blood pressure among those with hypertension compared to normotensives after partial substitution of carbohydrate with protein (32).
N-balance studies N-balance remains the method of choice for determining protein requirement in adults. Nevertheless, there are limitations in this method that need to be addressed, such as being related to the accuracy of the measurements, the short duration of the studies and the difficulty in interpretation of the results. Rand et al.'s meta-analysis (3) included 27 N-balance studies with six studies performed in persons aged 63 and older. From those, Rand et al. selected only studies that provided individual data for persons studied at three or more levels of protein intake. Thus, Rand's estimation was based on 19 Nbalance studies with 221 'young' individuals and only 14 'old' individuals from a study by Uauy et al. (4). Rand et al. found no statistically significant differences between 'young' and 'old', sex or source of dietary protein, but the data did not provide sufficient power to detect possible differences. The median nitrogen requirement was 130.5 mg per kg BW per day in the old versus 103.9 mg per kg BW per day in the younger group corresponding to a difference of 0.17 g protein per kg BW per day or a 26% higher requirement. The conclusion that Rand et al. draw from their analysis was that the healthy elderly may have a somewhat higher requirement, but that there was not enough evidence to give age-specific recommendations (3).
The objective of the high quality graded N-balance study by Campbell et al. (9) was to study the effect of age on the EAR. They found no difference in the EAR between the young and old participants, and the calculated adequate protein allowance of 0.85 g good-quality protein per kg BW per day for all participants combined was not statistically different from the RDA estimated in Rand's meta-analysis (3).
An earlier study by Campbell et al. (8) also found 0.8 g protein/kg BW to be sufficient to obtain N-balance among 10 elderly people. A short-term study (11) with only 10 days on each diet was also included in the current systematic review because a HP (22Á24 E%) intake was tested against a usual protein (11Á12 E%) intake in young as well as elderly healthy participants. The net daily N-balance increased equally in the young and older participants on the HP diet.
From these studies, we assess the evidence as probable regarding the estimated EAR of 0.65 g good-quality protein/kg BW/day and the resulting RDA of 0.83 g good-quality protein/kg BW/day as adequate to meet the minimum dietary protein needs of virtually all healthy elderly persons.
WHO/FAO/UNU (5) based their protein requirement for elderly people on Rand et al.'s meta-analysis (3) and also on a reassessment of the N-balance studies among the elderly who were not included in the meta-analysis and, supported by additional studies with measurements of metabolic demand, WHO/FAO/UNU concluded, that 'these data lend confidence to the overall conclusion drawn from the nitrogen balance data that the physiological protein requirement does not increase with age'. Based on Rand et al. (3), Campbell et al. (8), and studies before 2000 including some short-term studies, EFSA (6) concluded that 'the available data are insufficient to specifically determine the protein requirement in older adults and that at least the same level of protein intake as for young adults is required for older adults. As sedentary older adults have a lower energy requirement the protein to energy ratio of this subgroup is higher than for younger adults'.
Protein intake Á muscle mass Long-term health and functional outcomes related to various protein intakes are likely of greater potential importance than N-balance studies that only give shortterm indications on what intake is needed to prevent loss of protein mass, that is, mainly muscle mass, and bone mass. It has to be emphasized that the processes of dietary-based losses of muscle mass and strength, that is, sarcopenia, and bone, that is, osteoporosis, are extremely slow. This means that it may take many years before these losses are: (1) possible to identify with current measurement techniques (e.g. DXA); and (2) lead to compromised function, morbidity or mortality. Moreover, there is no strong consensus on what are validated or accepted outcome variables, as well as which are the most reliable biological marker of protein status.
Prevention of sarcopenia, that is, the age-related loss of muscle mass, strength and function is highly relevant. Advanced sarcopenia is a part of physical frailty and thus, associated with increased likelihood of falls and impairment in the ability to perform activities of daily living (33). The protein intake in the prevention of sarcopenia is one of many possible intervention strategies in question.
The studies included in this systematic review also point at a caveat whether the allowance represents a sufficient intake level in relation to loss of muscle mass. Campbell et al. (8) distinguish between adaption and accommodation. Adaptation refers to metabolic changes that occur in response to changes in protein intake and result in the establishment of a new steady state without a compromise or loss in physiological function, whereas accommodation refers to metabolic changes due to the decreased protein intake to establish a new steady state but at the cost of compromised physiological functions. During the small 14-week study, Campbell et al. found that urinary nitrogen excretion decreased after week 2, and the change was associated with a loss of skeletal muscle (mid-thigh muscle area) which points at a longerterm accommodation in skeletal muscle. Additionally, Morse et al. (10) found a decrease in nitrogen excretion between week 2 and week 3 in an 18-day study, but they found no changes in fat-free mass.
Based on a small randomized controlled trial (12) and two prospective cohort studies (13, 14) with 3Á5 years follow-up, we assessed the evidence of a relation between prevention of loss of muscle mass and a total protein intake in the range of 13Á20 E% to be suggestive. We excluded one study (34) with 'borderline' elderly (mean age 6297 years) according to our eligibility criteria, but the inclusion of this study would not have changed our assessment of evidence. In a strictly controlled metabolic study with focus on N-balance (8) and on resistance training (25), Campbell et al. found that a protein intake of 0.8 g /kg BW ( :10 E%) during 14 weeks resulted in a loss of mid-thigh muscle area in the sedentary control group during BW stability. In the Health ABC Study (13), that is the first longitudinal study to examine the role of dietary protein on changes in body composition by using state-of-art-body-composition measurements, the mean protein intake was 0.9 g/kg BW and the mean 3-year loss of lean body mass was 0.6891.9 kg. Subjects in the highest quintile of protein intake ( :19 E%) lost less LM compared to those in the lowest quintile (:11 E%). But of notice is that there was no statistically significant association between total protein intake and a 3-year loss of muscle mass adjusted for physical activity in the 49.5% of participants that were weight stable, which points to the necessity of a sufficient energy intake among the elderly to keep up the protein stores.
The EFSA Panel (6) considered that in healthy adults the available data on the effects of dietary intake on muscle mass and function did not provide evidence to be considered as a criterion for setting a population reference intake (PRI) and further they stated that 'There are no data showing that an additional intake of protein would increase muscle mass in different age groups who are in nitrogen balance, including subjects undertaking endurance or resistance exercise'.

Protein intake Á bone health
During aging there is an average loss of both muscle and bone mass in the population, however, with very wide variations between individuals.
The role of dietary protein on bone health has been controversial. On the one hand, urinary calcium loss is increased by HP intakes, while on the other hand protein increases calcium absorption or bioavailability, which questions the net effect of HP diets on calcium economy and the effect on bone health (35). Any negative effect of protein may also be opposed by an increase in the protein sensitive anabolic mediator insulin-like growth factor 1, IGF-1.
We assess the evidence as suggestive for an association between protein intake and BMD. The evidence regarding the relation of protein to an overall effect on bone health based on studies with bone loss, fracture risk and falls as outcomes are according to our view point inconclusive. Thus, the conservative interpretation is in line with EFSA (6) that found the available evidence regarding protein and bone health to be insufficient.
The positive association between protein intake and BMD/BMC was based on one intervention study (15), and three prospective cohort studies (14, 16, 17), but even though the effect was observed with protein intakes above approximately 0.85 g/kg BW or 16Á18 E%, the actual protein intake level was unknown due to underreporting. In the 5-year cohort study by Meng et al. (14), where lean body mass as well as BMC was measured they found that the association between protein intake and BMC disappeared after further adjustment for LM. Thus, the authors concluded that the protein effect on bone mass may be partly mediated by its effect on muscle. In the randomized controlled study with calcium and vitamin D supplementation by Dawson-Hughes and Harris (18), the highest tertile of protein intake (20 E% or 1.2 g/kg BW) was associated with less bone loss compared to the lowest tertile (14 E% or 1.1 g/kg BW), but only in the intervention group. Interestingly, the habitual mean intake in the placebo group was 871 mg calcium and approximately 7-mg vitamin D, which closely corresponds to the NNR from 2004 (1). Thus, the possible effect of protein intake on bone health may depend on an intake level of calcium and vitamin D above the current recommendations.
A prospective cohort study (22) and a caseÁcontrol study (23) found total protein intake associated with a decreased risk of fractures, but in the caseÁcontrol study it was only statistically significant among 50Á69-year-old participants, while the association was non-significant among the 70Á79-year-old participants (23). In the Study of Osteoporotic Fractures in postmenopausal women (21), a high A/V ratio was associated with an increased risk of fractures.
Overall, the animal protein intake showed both positive (17,19,23) and negative (21) effects on bone health, while no detrimental effect of total protein intake was observed. No upper limit could be detected for total protein intake in relation to bone health. The study by Hannan et al. (19) even found that elderly with total protein intake up to three times RDA (1.2Á2.8 g/kg BW or 18Á27 E%) had the least bone loss after controlling for known confounders, including weight changes.

Protein intake Á physical exercise
Only few studies address the issue of regular protein intake for optimal effect of physical exercise in healthy old adults. We included two American medium term RCTs (25, 26) that evaluate if the amount or the quality of the protein intake have different effects on muscle mass gain or strength by Â3 months of resistance training.
Due to the few studies, we assess the evidence to be inconclusive.
In addition, there are studies that use supplementation of protein hydrolysates in order to increase the protein intake during periods of resistance training. Such studies fall outside the scope of this review of effects by regular food, but they still provide understanding of potential effects of combined training and increased protein intake, and are thus mentioned briefly. A recent meta-analysis by Cermak et al. from 2012 (36) combined the results from six studies performed in 'older' subjects, and it was concluded that protein supplementation Á provided either by supplement or via the habitual diet corresponding to 42930 g Á increased the effects on muscle mass by an additional 38% as well as a 33% greater increase in muscle strength (1-RM leg press strength). It should be observed that the authors set the distinction between young and old at 50 years of age. In all but one study participant, average age was below 65 years. However, Verdijk et al. (37) evaluated the effect of 20 g extra protein given three times/week in conjunction with exercise during 12 weeks in 26 healthy men (average age 7292 years). No further effect on muscle mass or strength of the protein supplementation was observed. Neither was an effect on total protein intake by the supplementation observed, that is, intake being constant at 1.1 g/kg/day in both groups. Interestingly, the separate studies in the meta-analysis by Cermak et al. could not themselves establish evidence for positive effects by supplementation, mainly due to underpowered study designs.
There are two short-term studies that also should be mentioned regarding timing. In the study by Andrews et al. (38), fifty 60Á69 year-old subjects were provided a post-exercise drink with 0.4 g protein/kg/day during 12 weeks leading to a range of daily protein intake from 0.72 to 1.35 g/kg. Variability in mean daily protein intake was not associated with a change in LM. In contrast to these results, Esmarck et al. (39) reported that a protein supplementation of 10 g given directly in connection with the training session had positive effects on muscle hypertrophy, mean fiber area and strength, whereas a corresponding intake 2 h after the training session could not provide these effects.
Protein intake Á various outcomes Some outcomes, that is, hypertension, frailty, and mortality, were only addressed in one article, and thus, the evidence was assessed as inconclusive.
Potential risks Potential risks, e.g. kidney damage or hypertension, of HP intakes need to be addressed. The N-balance study by Walrand et al. (11) found that baseline GFR was lower in the old participants compared to the young participants. The study reported that a HP intake corresponding to 24 E% did not increase GFR in the old participants. This is probably due to the reduced kidney function in elderly, since patients with mild-to-moderate chronic kidney disease also do not show the usual protein-induced increase in GFR (40). The increase in GFR is a normal physiological adaption to increased protein intake (41) but it is also an important component of the hyperfiltration theory of Brenner (42) due to its presumed effect of increasing glomerular pressure. It is also an important component of the hyperfiltration theory that protein overloading of remnant nephron mass should be avoided.
In a prospective cohort study (27), it was reported that animal protein intake increased the risk of hypertension only among persons ]70 years of age after 6 years follow-up.

Methodological problems
Except from the strictly controlled metabolic studies it has generally been difficult to assess the actual protein intake in the observational studies due to misreporting (underreporting). The FFQ is a widespread dietary assessment method in observational studies. Based on a thorough calibration study, Beasley et al. (28) concluded that FFQ better assess nutrient consumption as a fraction of total energy intake than absolute nutrient consumption. Thus, it may be more appropriate to conclude from protein E% than protein intake expressed in g per day or per kg BW.
Still, taken together studies on protein intake and longterm outcomes related to muscle mass (8, 12Á14, 25), bone mass (14Á16, 18), and frailty (28) altogether indicate that there are suggestive evidence for positive effects of a higher intake than the 0.83 g/kg BW/day recommendations based on N-balance studies only.

Conclusion
The evidence is assessed as probable for an EAR of 0.66 g protein/kg BW/day based on N-balance studies and the subsequent RDA of 0.83 g good-quality protein/kg BW/ day (i.e. approximately 10 E%) representing the minimum dietary protein needs of virtually all healthy elderly persons.
There are still only few studies in this particular population and no new data justify a modification from the current estimated requirement.
The progressive loss of muscle mass and function (sarcopenia) as well as osteoporosis are true problems of the old population. It is likely to assume that these processes are too slow for short-term N-balance studies as well as short-term intervention studies to discover potentially beneficial effects from a slightly increased protein intake. The estimation of an optimal level of protein intake based on the evidence from randomized controlled trials, caseÁcontrol studies and prospective cohort studies of the relations of protein intake to functional outcomes (e.g. maintenance of bone mass, muscle mass and strength), morbidity, and mortality are ranging from suggestive to inconclusive. In particular, some long-term prospective studies indicate that an intake of up to at least 1.2Á1.5 g protein/kg BW/day (i.e. approximately 15Á20 E%) is safe and may have beneficial effects. Still, adequate enough data do not exist to estimate this optimal intake of protein based on the main physiological end points in the elderly.
Overall, many of the included prospective cohort studies were difficult to fully evaluate since results mainly were obtained by FFQs that were flawed by underreported intakes, despite some studies were 'calibrated' to correct for under-or over-reporting.
The overall impression from the included studies in the systematic review is that the optimal protein intake may be higher than the estimated RDA assessed from N-balance studies, whereas an exact level cannot be determined.
Regarding harmful effects of a HP intake, the evidence is considered as inconclusive. It cannot be ruled out that a HP intake corresponding to approximately 24 E% or 2 g protein/kg BW and day may affect kidney function negatively among old adults.