Routine protein energy supplementation in adults: systematic reviewBMJ 1998; 317 doi: http://dx.doi.org/10.1136/bmj.317.7157.495 (Published 22 August 1998) Cite this as: BMJ 1998;317:495
- Jan Potter, consultant physiciana,
- Peter Langhorne, senior lecturerb,
- Margaret Roberts, consultant physiciana
- aVictoria Geriatric Unit, Victoria Infirmary NHS Trust, Glasgow G41 3DX
- bAcademic Section of Geriatric Medicine, Department of Medicine, University of Glasgow, Glasgow Royal Infirmary Trust, Glasgow G4 0SF
- Correspondence to: Dr J Potter, Department of Medicine for the Elderly, Mansionhouse Unit, Victoria Infirmary NHS Trust, Glasgow G41 3DX
- Accepted 22 August 1998
Objectives : To determine whether routine oral and enteral nutritional supplementation can improve the weight, anthropometry, and survival of adult patients.
Design : Systematic review of randomised controlled trials of oral or enteral protein supplementation in adults. Trials were identified from Medline (Silver Platter 3.11, 1966-96), reference lists of identified studies and review articles, and communication with feed manufacturers.
Subjects : Randomised controlled trials comparing oral or enteral protein supplementation with no routine supplementation. All trials of adult subjects were included except those addressing nutrition in pregnancy.
Main outcome measures : Change in body weight and anthropometry (mid-arm muscle circumference), and all cause case fatality recorded at the end of scheduled follow up. body weight and anthropometry were analysed as the weighted mean difference and 95%confidence intervals of the percentage change in these variables. Case fatality was analysed with odds ratio and 95% confidence intervals.
Results : 32 eligible reports (2286 randomised patients) published between February 1979 and July 1996 were identified, of which 30 (93.8%) (2062 randomised patients) reported outcomes of interest. Case fatality data were available for 1670 (81%) patients, and continuous variable data for up to 1607 (78%) patients. The treatment group receiving routine nutritional supplementation showed consistently improved changes in body weight and anthropometry compared with controls; weighted mean difference 2.06% (95%confidence interval 1.63% to 2.49%) and 3.16% (2.43% to 3.89%)respectively. The pooled odds ratio for death in the treatment group was 0.66 (0.48 to 0.91, 2P<0.01). Apparent benefits were observed in several prespecified subgroups of patients, treatment settings, and interventions, but were not evident if trials with less robust methodology were excluded.
Conclusions : Routine oral or enteral supplementation seems to improve the nutritional indices of adult patients, but there are insufficient data in trials which meet strict methodological criteria to be certain if mortality is reduced. Benefits were not restricted to particular patient groups. Further large pragmatic randomised controlled trials of routine nutritional supplementation are justified.
Undernutrition is common in patients admitted to hospital, and hospita sation frequently results in further nutritional depletion
Undernutrition is associated with increased morbidity and mortaty, but c nicians remain to be convinced that routine nutritional supplementation improves outcomes
This systematic review indicates that significant improvements in nutritional status and reductions in case fata ty occurred when protein calorie supplements were routinely given to adults in several cnical situations
Conclusions were influenced by the methodological quaty of the primary trials
Further large pragmatic randomised trials are justified
Malnutrition is a common and underrecognised problem in hospital patients.1–4 Furthermore, illness and hospitalisation are frequently associated with negative energy balance and further deterioration in nutritional status.5 A recent survey of admissions to a general hospital reported a prevalence of malnutrition of 27% to 46% across various hospital specialties.2 Many studies have reported distinct associations between undernutrition and impaired immune function, increased sepsis, impaired wound healing, impaired muscle function and strength, and increased mortality. 1 6–11
When the high prevalence and potentially deleterious effects of undernutrition are considered it is not surprising that many trials have examined the effects of nutritional supplementation in various patient groups. Several trials have shown that poor immune function and poor muscle function can be reversed by nutritional supplementation. 6 7 9 12–16 However, there is no practical consensus among clinicians on the value of routine nutritional supplementation 2 4 17 or on how this could be achieved.
We evaluated the existing evidence on the effectiveness of routinely prescribed oral or enteral protein energy supplements (table A on website) in improving body weight, anthropometry, and survival of adult patients.
Subjects and methods
To determine whether the routine provision of oral or enteral protein energy supplementation improved outcome in adult patients we established the following inclusion criteria for trials: (a) randomised controlled trial, (b) oral or enteral protein energy supplementation, (c) control group receiving placebo or no intervention, and (d) human adult subjects (including all age groups and baseline nutritional states but excluding trials in pregnancy).
Identification of trials
We conducted a Medline search (Silver Platter 3.11) from January 1966 to November 1996. To identify the maximum number of randomised trials we used a broad search stategy with the mesh word nutrition, which was not restricted to English language citations. Where publication type “trial” and study group “human” were available these were also selected. In addition we carried out manual reference searching of all identified articles and reviews on nutritional supplementation. We also asked colleagues and manufacturers of supplement feeds to identify any unpublished material.
Most studies could be excluded from reading the abstract. Studies that could not clearly be excluded in this way were reviewed. The assessment of trial eligibility was done by two independent assessors (JP and MR) who reviewed the introduction and methods blinded to the results and discussion. Disagreements between assessors were decided by an independent reviewer (PL).
We gathered baseline information for each trial on the number and age of patients studied, diagnoses and severity of illness, the type and duration of the intervention, study design, method of randomisation, completeness of follow up, and outcome measures recorded.
The primary trials reported body weight and anthropometric measures in several ways. To allow us to collate standardised information on the change in body weight during the trial periods we selected the mean and SD of the percentage change in weight. This strategy was used because we believed it had clinical relevancethat is, reflected the degree of weight changeand was likely to be available from many trials. Where percentage weight change was not available we calculated the difference between the initial and final body weight, expressed as a percentage of baseline weight, and inferred a SD of 10%. The SD value was a conservative one that was at the upper limit of any of the observed results. If a baseline weight was not reported we assumed a standard value of 60 kg, which applied to all patients regardless of their baseline nutritional status. We chose mid-arm muscle circumference as the anthropometry measure. Where this was not described in a trial we derived it from the mid-upper arm circumference and triceps skin fold thickness using standard formulas.18 Anthropometry data were then pooled as per weight data.
A fixed effects approach (Peto method) was used to calculate the odds ratio and 95% confidence intervals for case fatality,19 and the findings wereconfirmed using an alternative (random effects) approach.20 Weighted mean difference and 95% confidence intervals were calculated using a fixed effect approach for changes in weight and anthropometry measures. These results were confirmed using alternative measures (standard effect sizes) and statistical approaches (random effects model).21
We carried out analyses of prespecified subgroups on the basis of certain patient and intervention characteristics. Patient characteristics included: patient group (healthy volunteers or ill patients), baseline body mass index (<25th centile or >25th centile), mean study population age (<70 or >70 years), specialty group (medical or surgical specialties), and underlying disease (malignant or non-malignant). Intervention characteristics included: method of delivery and type of nutritional supplement provided (oral sip feed, oral natural feed, nasogastric feed, percutaneous endoscopic gastrostomy tube), quantity of supplemented calorie intake (<400 or >400 calories per day), and duration of intervention (<35 and >35 days).
We identified 94 potentially eligible trials from the abstract, and of these, 62 were excluded: 24 (25.5%) were not randomised controlled trials, 13 (13.8%) considered total parenteral nutrition, 19 (20.2%) did not use a control group as defined by our inclusion criteria, and six (6.4%) were perinatal trials. Therefore 32 (34%) trials fulfilled all entry criteria. Two (6.2%) of these trials 22 23 did not report any outcomes of interest (table B on website), leaving 30 (32%) trials for analysis (table 1). No unpublished trials were identified that fulfilled our inclusion criteria.
A total of 2062 patients were available for analysis from the 30 trials. These trials covered a wide range of clinical variables including inpatients, outpatients, surgical and medical disorders, malignant and non-malignant diseases, and young and elderly groups. In addition, although all the trials used either oral or enteral nutritional supplementation they varied in the route of delivery, the amount of additional kilocalories given, and the duration of intervention (table 1). Twenty (66.7%) trials evaluated oral supplementation, seven (23.3%) nasogastric tube feeding, and three (10.0%) percutaneous endoscopic gastrostomy feeding. Six (20%) trials used a stratified randomisation design according to aspects of the patients' clinical characteristics. The individual strata of these trials have been analysed separately.
The methodological characteristics of the trials varied (table 1); nine (30.0%) had clearly concealed randomisation and complete follow up (category A), 21 (70.0%) did not report the randomisation procedure of which 11 (52.4%) had complete follow up (category B), and 10 (33.3%) had incomplete follow up (category C). Only four (13.3%) trials reported a clearly blinded assessment of outcomes.
Change in weight
Twenty six (86.6%) trials provided data on weight change for 1607/1648 (97.5%) patients. The absolute weight change tended to be negative particularly in studies incorporating surgical interventions or treatment of malignancies. In almost all trials, however, there was a greater percentage weight gain or smaller percentage weight loss in the supplemented group than in the controls (fig 1). The pooled weighted mean difference for weight change showed benefit from supplementation (2.06%, 95% confidence interval 1.63% to 2.49%), but was complicated by heterogeneity. The results still showed benefit from supplementation when an alternative random effects20 model was used (3.11%, 2.03% to 4.20%) or if the standardised mean difference was calculated (0.50, 0.40 to 0.60).
The conclusions were similar (weighted mean difference 2.85%, 1.03% to 4.68%) if the analysis was restricted to the most methodologically rigorous trials (category A) or to those trials where no inferences were required regarding baseline weights or SDs (3.39%, 2.12% to 4.66%).
Change in anthropometry
Seventeen (56.7%) trials reported changes in anthropometric measures for 1209/1230 (98.3%) patients. In most trials the treatment group showed improved anthropometric measures compared with the control group. The pooled result showed considerable heterogeneity and gave a pooled weighted mean difference of 3.16% (95% confidence interval 2.43% to 3.89%), which was unchanged when a random effects statistical approach was used (3.27%, 1.74% to 4.80%). Reanalysis using the standardised mean difference confirmed these results (0.36, 0.24 to 0.48).
Conclusions were similar if the analysis was restricted to methodological category A trials (weighted mean difference 3.00%, 1.93% to 4.06%) or trials where no inferences were required regarding baseline weights or SDs (2.73%, 1.81% to 3.66%).
Case fatality —Case fatality data were available from 25 (83.3%)trials (1670 patients). The pooled odds ratio for death by the end of scheduled follow up (fig 2) showed a reduced case fatality in treat- ment compared with control groups of 0.66 (0.48 to 0.91, 2P<0.01), with no significant statistical heterogeneity χ2=11.67; df=13; P>0.2). However, the exclusion of trials which did not meet the highest methodological criteria (category A) reduced this result to a non-significant trend (P>0.1) in favour of supplementation (odds ratio 0.81, 0.44 to 1.50). Comparable results for category B and C trials were 1.48 (0.43 to 5.09) and 0.55 (0.47 to 0.90) respectively. Recalculation of results to include a best and worst case scenario for the missing data from the category B and C trials did not substantially change the conclusions.
Subgroup and sensitivity analysis
Analyses were carried out for several subgroups that met prespecified criteria. Trials that provided inadequate information for inclusion in a subgroup were omitted from the analysis. Subgroups included:(a) patients that were well (community dwelling, healthy volunteers, n=111) or unwell(major operation, acute hospital admission, or chronic long term care resident, n=697), (b) patients that were originally well nourished (body mass index >25th centile, n=510) or undernourished (<25th centile, n=298), (c) patients with malignant (n=210)or non-malignant disease (n=917), (d) mean age of study population >70 years (n=813) or <70 years (n= 731),(e)patients being treated by surgical (n=205) or medical (n=1595) specialists, (f) energy value of treatment given and consumed >400 kcal per day (n=624) or <400 kcal per day (n=184), and (g) duration of intervention >35 days (n=627) or <35 days(n=181).
Within the limitations of the available data, analysis of the subgroups (fig 3) showed that the benefits of routine nutritional supplementation were not restricted to particular subgroups or trials.
In addition to the characteristics considered above, the trials varied widely in the length of scheduled follow up. Although the duration of follow up sometimes differed from the duration of intervention (table 1), follow up was more consistently reported in the primary trials. A sensitivity analysis based on the duration of follow up did not show any clear association with the odds of death (table 2).
Further sensitivity analyses were carried out to assess the potential influence of publication bias and of the assumptions included in our calculations. These analyses indicated that about 1500 patients in trials with neutral results (odds ratio 1) would be sufficient to render the observed reduction in case fatality non-significant (2P>0.05).
We wanted to address the hypothesis that adult patients whose diet was routinely supplemented with additional enteral protein calories would, on average, be more likely to benefit from improved nutritional indices (body weight and anthropometry) and improved survival. If such a benefit could be observed in a diverse group of trials, it would suggest that protein calorie supplementation is generally beneficial. As we were interested in potentially simple and routinely applicable interventions we considered trials of both enteral and oral supplementation. The results suggest that routine supplementation in a variety of patient groups and clinical settings will improve body weight and anthropometry. Weight and anthropometric measures are validated measures of nutritional status,24–28 and improvements in these measures have been shown to be associated with improvements in a number of clinical outcomes. 9 29 30 Positive energy balance and weight gain have been associated with improvements in immune function and sepsis and improvements in muscle bulk resulting in better muscle function, strength, and functional independence. 7 31–35 These improvements could have important benefits on clinical outcomes. This analysis has not, however, shown an unequivocal effect of nutritional supplementation in reducing case fatality; the apparent benefits could be explained by publication bias36 or less reliable trial methodologies.37
Much of the recent controversy over the usefulness of systematic review and meta-analysis38 centres on publication bias—that is, the selective non-publishing of neutral or negative trials. We have attempted to identify unpublished work by several methods including discussions with colleagues and manufacturers of nutritional supplements. It is impossible, however, to guarantee identification of all trials, and a few neutral or negative trials could overturn our conclusions.
Another problem is the ability of the reviewers to assess the methodological quality of the trials, in particular the security of randomisation and the completeness of follow up, both of which may influence results.37 Although all the trials reported that they were randomised, only nine (30.0%) clearly described a secure, concealed procedure (category A). A further 11 (category B) may have used a secure procedure and seemed to have complete follow up; however, it was impossible to ascertain this from the published reports. In the remainder (category C) there were occasionally marked or unexplained discrepancies between the number of patients in the treatment and control arms, or a number of patients were unaccounted for at the end of follow up.
Further limitations are that only a minority (n=4, 13.3%) of trials commented on whether outcome assessments were carried out in a clearly unbiased manner—that is, by observers blinded to treatment allocation. This is important for weight and anthropometry outcomes which may also be biased by the number of subjects unavailable for follow up. It might be speculated that more deaths in the control arm may have led to an underestimate of the degree of weight loss, but it is impossible to know with certainty. Many other secondary outcomes of interest—for example, muscle strength or length of stay, were not reported in most of the papers.
Although the benefits of nutritional supplementation were observed across all subgroups it is interesting to note that rather more elderly people than young adults have been studied, and that for each outcome they seemed to benefit as much as their younger counterparts. This may be important as protein calorie undernutrition is more common in the elderly2 who make up the majority of hospital admissions and frequently have a longer period of illness and longer hospital stay putting them at the greatest risk of continued nutritional depletion.
If our conclusions are not secure beyond reasonable doubt does this really matter? Although no one would argue against providing high quality food in hospitals the question really concerns the routine provision of manufactured nutritional supplements. Only two of the trials studied the use of natural food supplements to achieve improvement in protein and calorie intake. 39 40 Most of the trials used sip feeds, the composition of which varies but generally contains protein and calories in variable proportions but with the same quantities of vitamins and minerals found in the energy equivalent of a well balanced diet. As such it is likely that nutritional supplementation given as sip feeds is not associated with significant problems or side effects. Insertion of feeding tubes does carry a small risk but is often indicated for reasons (for example dysphagia) other than supplement provision alone. Even if the potential risks of nutritional supplementation are low, however, there are still implications in terms of cost or organisation, or both, if nutritional supplementation were to become a routine part of hospital prescribing. These costs would have to be considered against the potential benefits of preventing deteriorations in weight, muscle bulk, function, strength, and immunity, by improved energy balance. 6 7 9 12–16
Oral and enteral protein energy nutritional supplementation may be associated with improvements in weight gain and anthropometry and significant reductions in case fatality. However, there remain considerable uncertainties about these conclusions. We conclude that large pragmatic randomised controlled trials of routine oral or enteral nutritional supplementation are justified.
We thank Dr John Reilly and Professor M Lean for their comments on the manuscript, Catherine Hankey for providing unpublished data, and Fiona Clark for providing information on nutritional products.
Contributors: JP was responsible for the study design, literature search, assessing trial eligibility, data analysis, and drafting the paper; she will act as guarantor for the paper. PL was responsible for the study design, assessing trial eligibility, data analysis, and drafting the paper. MR was responsible for assessing trial eligibility and drafting the paper.
Conflict of interest None.