Works Cited
Milne, Anne C., Alison Avenell, and Jan Potter. “Meta-Analysis: Protein and Energy Supplementation in Older People.” Annals of Internal Medicine 144.1 (2006): 37-48. ProQuest. Web. 1 Oct. 2014.
Meta-Analysis: Protein and Energy Supplementation in Older People Anne C. Milne, MSc; Alison Avenell, MD; and Jan Potter, MBChB Background: Protein and energy undernutrition is common in older people, and further deterioration may occur during illness. Purpose: To assess whether oral protein and energy supplementa tion improves clinical and
nutritional outcomes for older people in the hospital, in an institution, or in the community. Data Sources: Cochrane Central Register of Controlled Trials (CEN TRAL), MEDLINE, EMBASE,
HealthStar, CINAHL, BIOSIS, and CAB abstracts. The authors included English- and non-English-language studies and hand-searched journals, contacted manufacturers, and sought information from trialists. The date of the most recent search of CENTRAL and MEDLINE is June 2005. Study Selection: Randomized and quasi-randomized controlled tri als of oral protein and energy
supplementation compared with placebo or control treatment in older people. Data Extraction: Two reviewers independently assessed trials for inclusion, extracted data, and assessed trial quality. Differences were resolved by consensus. Data Synthesis: Fifty-five trials were included (n = 9187 randomly tions (Peto odds ratio, 0.72 [95% Cl, 0.53 to 0.97]) and reduced mortality (Peto odds ratio, 0.66 [CI, 0.49 to 0.90]) for those un dernourished at baseline. Few studies reported evidence that suggested any change in mortality, morbidity, or function for those given supplements at home. Ten trials reported gastrointestinal disturbances, such as nausea, vomiting, and diarrhea, with oral supplements. Limitations: The quality of most studies, as reported, was poor, particularly for concealment of allocation and blinding of outcome assessors. Many studies were too small or the follow-up time was too short to detect a statistically significant change in clinical out come. The clinical results are dominated by 1 very large recent trial in patients with stroke. Although this was a high-quality trial, few participants were undernourished at baseline. Conclusions: Oral nutritional supplements can improve nutritional status and seem to reduce mortality and complications for under nourished elderly patients in the hospital. Current evidence does not support routine supplementation for older people at home or for well-nourished older patients in any setting. assigned participants). For patients in short-term care hospitals who were given oral supplements, evidence suggested fewer complica-Ann Intern Med. 2006:144:37-48. For author affiliations, see end of text.www.annals.OIJ ndernutrition among older people is a continuing source of concern (1, 2). Older people have longer periods of illness and longer hospital stays (3), and data show that up to 55% of elderly
hospitalized patients are undernourished at admission (4, 5). Malnutrition is asso ciated with
poorer recovery in a broad range of patients and conditions (6- 8). However, poor nutritional
status may be a marker for severity of existing medical conditions, and whether improving
nutritional status with oral protein and energy supplementation can improve acute or chronic
medical conditions is not clear. Recent systematic reviews examining the potential benefits of nutritional supplementation in older
people in clude Stratton and colleagues’ review (9) of randomized and nonrandomized trials (166
trials; 7630 patients) across all disease groups and settings, which concluded that nu tritional
supplementation had positive effects on nutri tional outcomes and mortality in elderly people and,
in some cases, clinical and functional benefits. Potter’s meta analysis (10) of 18 trials that
included older patients both in the hospital and in the community suggested a statisti cally
significantly lower mortality for the supplemented group (odds ratio, 0.61 [95% CI, 0.45 to 0.82]).
A recent update of a Cochrane review by Avenell and Handoll (11) of nutritional supplementation for
hip fracture care in older people found some evidence that oral protein and energy feeds (evaluated
by 8 trials) reduced unfavorable outcome (death or complications) but did not observe a
demonstrable effect on deaths alone. Overall, the evidence was weak because of methodologic defects in the reviewed studies. Oral nutritional supplements are
widely prescribed for older people both in the hospital and in the commu nity. We undertook a
systematic review of randomized tri als of oral protein and energy supplementation to assess
clinical and nutritional outcomes for older people who are offered supplements in different
settings.
METHODS
We identified studies and performed the analyses ac cording to the Cochrane method (12). The
search included the following databases: Cochrane Central Register of Controlled Trials
(CENTRAL) (Issue 2, 2005),
MEDLINE (1966 to June 2005), EMBASE (1980 to
March 2004), HealthStar (1975 to March 2001), CI-
See also:
Print
Editors’ Notes ………………………… 38
Editorial comment. …………………….. 59
Web-Only Appendix Tables Appendix Figures CME quiz
Conversion of figures and table into slides
© 2006 American College of Physicians 137
roduced with permission of the copyright owner. Further reproduction prohibited without
permission.
REVIEW Protein and Energy Supplementation in Older People
Context
Physicians sometimes recommend nutritional supplementa tion for sick, older persons.
Contribution
This review summarizes 55 trials of protein and energy supplementation in people older than 65
years of age. Compared with placebo or no supplementation, nutritional supplements did not affect
morbidity or mortality in peo ple living in the community. Among older and undernour ished
hospitalized patients, supplements sometimes re duced mortality and complications, such as
infections,
poor wound healing, and pressure sores. Oral supplements also sometimes caused nausea, vomiting,
and diarrhea.
Cautions
Many trials were small or had short follow-up times and used outcome assessors who knew
which patients took supplements.
-The Editors
NAHL (1982 to March 2004), BIOSIS (1985 to March 2004), and CAB abstracts (1973 to March 2004). We
in cluded English- and non-English-language studies. We also hand-searched nutrition journals and
reference lists and contacted oral nutritional supplement manufacturers.
We included randomized or quasi-randomized trials
with an minimum intervention of 1 week. Groups of study participants had to have a minimum average
age of 65 years. We included all patient groups, with the exception of people in critical care or
those who were recovering from cancer treatment. We included commercial supplements, other
milk-based supplements, and fortification of normal food sources. We excluded studies of specially
designed immunomodulatory supplements or supplements of spe cific amino acids. The full
description of the search strategy is available elsewhere (13). We contacted trialists for fur
ther information on ambiguous numerical data and to al low trial quality to be more accurately
assessed.
We examined the following outcomes as prespecified in our protocol: all-cause mortality, number of
people with morbidity or complications, length of hospital stay, func tional status, participants’
perceived quality of life, percent age change in weight, percentage change in mid-arm mus cle
circumference, acceptance of the supplement, and adverse effects. We included trials that reported
at least 1 relevant clinical outcome measure. Two reviewers indepen dently extracted outcome data
from the included trials and performed quality assessment of trials. We used a 10-item quality
assessment checklist, which is based on the quality assessment tool of the Cochrane Bone, Joint and
Muscle Trauma Group (14), to rate studies between 0 and 2 points for each item, including
assessment of allocation conceal ment, intention-to-treat analysis, and blinding of outcome
assessors. We resolved all differences by discussion.
3813 January 20061 Annals of internal Medi6nel Volume 144 • Number 1
Statistical Analysis
We combined data for the meta-analysis for the di chotomous variables of mortality and
complications and adverse effects by using RevMan 4.2 software (Cochrane Collaboration, Oxford,
United Kingdom). Low event rates pose particular problems for summarizing data in a system atic
review. Default use of a correction for continuity or simply adding 0.05 to each cell when counts
are less than
5.00 tends to produce biased estimates. Many methods are
recommended in the literature (15). The widely available Peto method (16) produces estimates
without the need for 0-cell corrections, and it produces unbiased estimates when equal numbers of
patients are in each group (17). For each study, we calculated Peto odds ratios and combined the
results by using fixed-effects models with 95% confidence limits. We calculated weighted mean
difference and 95% Cis for length of hospital stay, percentage weight change, and percentage
mid-arm muscle circumference change by using a fixed-effects model. We explored heterogeneity be
tween comparable trials with the test (18) by using greater than 50% as the cutoff value for
statistically signif icant heterogeneity. When evidence suggested heterogene ity, we applied a
random-effects model.
The trials reported body weight and anthropometric
measures in several ways. For meta-analyses of weight change and mid-arm muscle circumference
change, we se lected the mean and SD of the percentage weight change during the trial period
because of their clinical relevance (19). When the percentage weight change was not avail able, we
calculated the difference between the initial and final body weight, expressed as a percentage of
baseline weight and an SD of 1Oo/o inferred. This SD was conser vative and was at the upper limit
of any observed result. If baseline weight was not reported, we assumed a standard value of 60 kg.
As in Potter and colleagues’ study (19), we chose mid-arm muscle circumference as the
anthropometry measure because it is a measure of muscle. When this was not described in the trial,
we derived it from the mid-arm circumference or mid- upper arm circumference and tri ceps
skinfold by using a standard formula (20).
We performed prespecified subgroup analyses of the mortality data by comparing 1) baseline
nutritional status as defined by the investigators (nourished or undernour ished), 2) mean age
(<75 years or :=:::75 years), 3) amount of kilojoules provided in the supplement (< 1674
kJ [<400 kcal] or :=:::1674 kJ [:=:::400 kcal]), 4) duration of
intervention (<35 days or :=:::35 days), and 5) patient health (well or unwell).
We performed an exploratory subgroup analysis for mortality on the basis of diagnostic group (hip
fracture, chest conditions, stroke, and congestive heart failure), ge riatric conditions (trials
that included frail patients with a variety of conditions), and perioperative surgical patients.
We also stratified the trials by setting (short-term care hospital, long-term care institutions
[including nursing homes], and home in the community) because we sus-
roduced with permission of the copyright owner. Further reproduction prohibited without
permission.
Protein and Energy Supplementation in Older People REVIEW
peered differences in the nature and duration of the inter vention in different settings.
Prespecified sensitivity analy sis included only trials that reported clearly concealed
randomization. We evaluated the potential for publication bias by using a funnel plot. In
addition, we performed a sensitivity analysis to address possible heterogeneity be tween findings
of small and large trials, the latter having more than 100 participants in each group.
Role of the Funding Sources
The Medical Research Council, United Kingdom; Chief Scientist Office of the Scottish Executive
Health De partment, United Kingdom; and the Student Awards Agency for Scotland, United Kingdom,
funded the study. The funding sources had no role in the design, conduct, or reponing of the study
or in the decision to submit the paper for publication.
Figure 1. Flow chart for study selection.
Potentially relevant publications Identified through electronic searching. bibliographies, and
authors
(n = 34 870)
Papers retrieved for more detailed evaluation
(n = 242)
Papers exduded on the basis of title and abstract (generally because of lack of suitability of
study design, age group, or intervention)
(n = 34 628)
RESULTS
Description of Studies
From more than 34 000 titles or abstracts screened, we included 55 studies in our review (Figure
1). The 55 stud ies (21-75) recruited 9187 participants (Appendix Tables 1 and 2, available
at www.annals.org). Nearly half of the participants were from the recent Feed Or Ordinary Diet
(FOOD) trial (37) of oral nutritional supplements for pa tients with stroke. Most trial
participants (74%; 25 trials) were hospitalized inpatients. Fewer patients were at home
I Papers included I
(n= 83)
I Awaiting further infonmatton I
1 (n = 14)
I Excluded after evaluation
I (n = 145)”
Studies reporting particular outcomes
in the community (16%; 21 trials) or in long-stay, elderly care, or continuing care wards or
nursing homes ( 1Oo/o; 9 trials). Overall, most participants were patients with stroke (45%; 2
trials) or were mixed groups with various geriatric conditions (42%; 33 trials). We also included
trials of pa tients with hip fracture (7%; 10 trials), patients with chronic obstructive
pulmonary disease (5%; 7 trials), sur gical patients (1o/o; 2 trials), and patients with
congestive
heart failure (<1o/o; 1 trial).
The source of funding was unclear for most studies. Eleven trials were coauthored by an employee of
the man ufacturer of the oral supplement or were fully funded by the manufacturer (Appendix Table
1, available at www
.annals.org). The interventions in the trials aimed to pro vide between 175 kcal (732 kJ) and 1000
kcal (4.2 MJ) and between 10 g and 36 g of protein daily. Fifteen trials provided less than 1674 kJ
(400 kcal) per day, 30 trials provided 1674 kJ or more ( 400 kcal) per day, and 10 trials did not
specify the supplemented energy value. Most supplements included vitamins and minerals. The inter
vention period ranged from 10 days to 18 months and was 35 days or more in 33 trials, was less than
35 days in 12 trials, and was unspecified in 10 trials. Eight studies, in cluding the FOOD trial
(37), provided supplements until hospital discharge (estimated mean ranging from 12 days to 38
days). The duration of follow-up was usually the same as that of the intervention.
Seventeen trials, including the FOOD trial (37), re-
I Primary studies Mortality (n = 25)
(n= 55) Morbidity and complication (n = 20)
Length of stay (n = 11) Adverse effects (n = 18) Weight change (n = 38)
Anm muscle circumference (n = 15)
*Main reasons were study not randomized, intervention did not meet inclusion criteria (nasogastric
feeding, high-protein vs. low-protein , early vs. late introduction of feeding, and
immunomodulatory supplements) , or participants did not fit the inclusion criteria (too young and
patients with cancer) .
ported that supplements were well-accepted by most pa tients, although this was often not defined
or was variously defined. Other studies, particularly those offering supple ments over longer
periods of time, reported major prob lems with adherence for 24% to 45% of participants (24, 41,
62, 71, 59). Total energy and protein intake were, however, substantially greater than
nonsupplemented in take in nearly all studies, although Fiatarone and colleagues
(36) highlighted that the increase in intake from the sup plements may be partially offset by a
reduction in normal food intake. When reported, completeness of follow-up varied between 100% and
27% of those randomly as signed. Participant withdrawal or dropout was 25% or higher in 12 of the
55 trials.
3 January 20061 Annals of Internal Medicine IVolume 144 • Number 1139
roduced with permission of the copyright owner. Further reproduction prohibited without permission.
REVIEW Protein and Energy Supplementation in Older People
Methodologic Quality of Included Studies
Full details of the quality assessment are available in the Cochrane review (13). Appendix Table 2
(available at www
.annals.org) presents total scores. The trials had low scores,
with only 27 of 55 trials achieving 50% or more of the max imum quality score. Sixteen studies
confirmed adequate con cealment of allocation, and 22 studies reported intention-to treat
analysis or we could perform intention-to-treat analysis. The quality was poorest with regard to
blinding. Only 9 stud ies dearly reported the blinding of outcome assessors.
Outcomes
We suspected heterogeneity because of the differences in the nature and duration of the
intervention in different set tings. We therefore grouped trials post hoc for analysis by setting
(that is, short-term care hospital, long-term care insti tutions [including nursing homes], and
community-dwelling elders). The duration of the intervention was 8 weeks or more in 20%, 55%, and
81% of trials set in the hospital, in long term care, and in the community, respectively. Three
hospital based interventions continued at home after discharge.
Mortality
At the time of last follow-up, which was usually when supplementation was discontinued, nutritional
supplemen tation was associated with reduced mortality from a global analysis of25 trials (6852
randomly assigned participants), which was borderline statistically significant (Peto odds ra tio,
0.86 [CI, 0.74 to 1.00]) (Figure 2). For patients in short-term care hospitals, mortality was not
statistically sig nificantly reduced (Peto odds ratio, 0.88 [CI, 0.74 to 1.04]), unless only
undernourished patients were included (Peto odds ratio, 0.66 [CI, 0.49 to 0.90]) (Figure 3). A
reduction in mortality from the analysis of patients in long-term care was also not statistically
significant (Peto odds ratio, 0.65 [CI, 0.41 to 1.02]). For participants in long-term care, trials
were too small and were too few in number to examine the effect of supplementation in nour ished
and undernourished older people. Evidence did not suggest a reduction in mortality for people
living at home regardless of nutritional status (Peto odds ratio, 1.05 [CI,
0.57 to 1.95]). We found no statistically significant heter ogeneity within any setting (I2 = 0%
to 12.6%).
Results of the subgroup analysis suggested improved survival with supplementation in undernourished
people (17 trials; 2093 participants; 3 trials providing separate results for nourished and
undernourished patients) (Peto odds ratio, 0.73 [CI, 0.56 to 0.94]), when people were 75 years of
age or older (18 trials; 1611 participants) (Peto odds ratio, 0.64 [CI, 0.49 to 0.85]), when people
were offered 1674 kJ or more per day in the supplement (15 trials; 6157 participants) (Peto odds
ratio, 0.85 [CI, 0.73 to 0.99]), and when participants were not well (22 trials; 6630 participants)
(Peto odds ratio, 0.86 [CI, 0.74 to 1.00]). The test for heterogeneity was less than 2% for all
subgroups.
4013 January 20061 Ann.ll, of I nrc·rna I :’vkdicincl Volume 144 • Number 1
The results for mortality were statistically significant when we included only trials with dearly
concealed random ization (12 trials; 5991 participants) (Peto odds ratio, 0.84 [CI, 0.72 to
0.98]). Appendix Figure 1 (available at www
.annals.org) and our sensitivity analysis do not suggest that
small, positive trials were over-represented (21 trials; 1464 participants) (Peto odds ratio, 0.87
[CI, 0.59 to 1.29]) com pared with larger trials (4 trials; 5388 participants) (Peto odds ratio,
0.86 [CI, 0.73 to 1.01]). Our post hoc subgroup anal yses for mortality based on diagnostic group
found statistically significant results for trials that included patients with various geriatric
conditions (15 trials; 2313 participants) (Peto odds ratio, 0.69 [CI, 0.52 to 0.92]). No evidence
suggested a change in survival with oral supplements from trials of pa tients with stroke (2
trials; 4063 participants) (Peto odds ratio,
0.92 [CI, 0.76 to 1.11]) or hip fracture (5 trials; 269 partici pants) (Peto odds ratio, 0.88 [CI,
0.41 to 1.89]). The data were too limited to undertake meta-analyses for other diag nostic
groups.
Morbidity and Complications
Twenty trials provided data on morbidity and complica tions (Appendix Table 2, available at
www.annals.org). Global meta-analysis of 19 trials (5508 participants) reporting participants with
infective complications (38, 61, 62); incom
plete wound healing (30, 35); total pressure sores (37, 50);
total complications, excluding deaths (26, 33, 44, 53, 68);
illness that led to discontinuation (42, 54, 63, 74); exacerba tion of chronic obstructive
pulmonary disease (67, 69); and hospitalization (52) suggested fewer complications, although this
was not statistically significant (Peto odds ratio, 0.82 [CI,
0.65 to 1.03]). Hospitalized patients who were given supple ments had a statistically significant
decrease in complications (Peto odds ratio, 0.72 [CI, 0.53 to 0.97]). Supplementation did not have
a statistically significant effect on morbidity or complications in people in long-term care (Peto
odds ratio,
0.92 [CI, 0.56 to 1.52]) or at home (Peto odds ratio, 1.01 [CI, 0.63 to 1.64]) (Figure 4). Subgroup
analyses based on diagnostic group suggested a reduced risk for complications with supplementation
only for patients with hip fracture (4 trials; 147 participants) (Peto odds ratio, 0.48 [CI, 0.24
to 0.96]). In most cases, including the FOOD trial (37), out come assessors for complications were
not blinded to treat ment allocation.
Adverse Effects
Most trials did not report adequate methods for assessing potential adverse effects. In most trials
that discussed adverse effects with supplements (18 trials), no comparison with the control group
was performed. Of these trials, 10 reported some problems with tolerance and side effects and 8
reported no adverse effects (Appendix Table 2, available at www.annals
.org). Meta-analysis of 6 trials (477 participants) that reported participants with adverse effects
in both groups suggested a statistically significant effect on gastrointestinal disturbances,
roduced with permission of the copyright owner. Further reproduction prohibited without
permission.
Protein and Energy Supplementation in Older People REVIEW
Figure 2. Meta-analysis of mortality.
Study, Year (Reference)
Treatment Group,
Control Group,
Peto OR
PetoOR
nIn nIn
(95’Yc Cl)
(95’Yc0 Cl
Short-term care hospital
Delmi et al., 1990 (33) 6/27
10/32
0.64 (0.2Q-2.00)
Madigan, 1994 (54) 4/18
0/12
6.42 (0.78-53.07)
Hankins, 1996 (44) 2/17
4/14
0.35 (0.06-2.05)
Volkert et al., 1996 (71) 4/35
8/37 &
