A Randomized, Double-Blind, Placebo-Controlled Trial of Nutritional Supplementation During Acute Illness

Study Population 

We randomized hospitalized patients aged 65 years or more to a normal hospital diet plus placebo (n = 222) versus a normal hospital diet plus oral supplements daily for 6 weeks (n = 223). Inclusion criteria were age 65 years or more, able to swallow, and able to sign an informed written consent form. Patients excluded from the study were those who had undergone gastric surgery, with diagnosed malabsorption or morbid obesity (body mass index > 40), in a coma, with diagnosed severe dementia (abbreviated mental test < 6) and malignancy, living in an institution, and already taking supplements. The study was approved by the Barnsley Research Ethics Committee, and informed written consent was obtained from each patient.

Treatment and Placebo 

The supplement group received 2 bottles (200 mL each) of oral nutritional supplement daily at 8:00 a.m. and 12 noon in addition to the standard hospital diet for a period of 6 weeks. The composition of the supplement was such as to provide 995 kcal for energy and 100% of the Reference Nutrient Intakes for vitamins for a healthy older person. The placebo was identical to the supplement but contained no protein or micronutrients and with a minimum calorie content (60 kcal). Samples of the trial placebo, supplements, and other commercially available food supplements were piloted in 15 volunteers. No patients, research assistant, investigator, or any other medical or nursing staff could distinguish the placebo from the supplements during the study.

Dietary intakes of the study population were measured using a validated food diary.11 All patients had prescribed supplement bottles collected during the study period, and leftover supplements were measured.

Clinical Assessment and Follow-Up 

The main outcome measures were 6 months of disability (Barthel score), non-elective readmission and length of stay in hospital, discharge destination (own home or institution), morbidity (infective complications), and mortality. Disability was assessed using the Barthel score on a 20-point scale. The Barthel scores 10 functions on a scale from 0 (fully dependent) to 20 (independent).12 Nutritional status was assessed from anthropometric, hematologic, and biochemical data. All anthropometric measurements were performed by S. F. using standard methods with intraobserver’s differences assessed before the commencement of the study. The interassay coefficient of variation was 2% for serum albumin, 7% for transferrin, 12.6% for red cell folate, and 8.4% for plasma vitamin B12. The interassay coefficient of variation was 3.9% for C-reactive protein concentration.

Randomization 

The randomization sequence was generated by the trial statistician; concealed in sequentially numbered, sealed opaque envelopes; and kept in a clerical office at a different city. Contact was made by telephone for a trial medication number, and patient baseline details were recorded in the database.

Sample Size Calculation 

In patients who have had a stroke, we previously showed that a 1 g/L lower serum albumin concentration in the hospital is associated with a 1.13-fold increase in the risk of death (95% confidence interval [CI] 1.01-1.27).3, 4 Therefore, a sample size of 438 (219 in the supplements group and 219 in the control group) would allow the detection of a true mean difference in serum albumin of 1 g/L (assuming a within-group standard deviation of 3.5 g/L and 10% loss to follow-up), with 80% power and 5% (2-sided) significance. This sample size also would allow for the detection of a 1.5-point difference in disability, as measured by the Barthel score at 6 months (assuming a within-group standard deviation of 5.0) with 80% power and 5% (2-sided) significance.

Quality Assurance 

The trial statistician carried out an interim and end-of-study blind validation of collected data. Two researchers independently checked hospital records for all routinely collected outcome measures. Agreement was 99%.

Analyses 

All statistical analyses were done with blinding maintained and on an intention-to-treat basis with a P value of less than .05 regarded as statistically significant. A Cox proportional hazards model was used to examine the 6-month risk of non-elective readmission and mortality between the 2 groups after controlling for a number of non-nutritional factors such as age, disability, chronic illness, drugs, smoking, and tissue inflammation.

Baseline Characteristics of Study Population 

The 2 groups were comparable with respect to baseline disability, smoking, drug and alcohol intake, dietary intake, and nutritional status (P > .05). Although the patients in the supplement group were older and had higher C-reactive protein and serum ferritin concentrations, lower transferrin concentrations, and a high prevalence of chronic diseases compared with the placebo group, these baseline differences did not reach statistical significance except for serum ferritin and transferrin concentrations (Table 1). Figure 2 shows adherence to supplement and placebo treatments expressed as the percentage of patients consuming drinks, divided into quartiles. Nausea was the most common side effect, reported by 22% of the patients with no difference between the supplement and placebo groups.

Table 1. Baseline Characteristics of Patients

			Placebo		Supplements		P Value
			n	%	n	%
	Gender	Male	119	(53.6%)	115	(51.6%)	.7
		Female	103	(46.4%)	108	(48.4%)
	Smoking	Never smoked	58	(26.7%)	79	(36.4%)
		Ex-smoker	116	(53.5%)	100	(46.1%)
		Current smoker	43	(19.8%)	38	(17.5%)	.1
	Residence	Own home	141	(65.0%)	123	(56.7%)	.1
	before	Own home with carer	19	(8.8%)	28	(12.9%)
	hospitalization	Sheltered housing	35	(16.1%)	40	(18.4%)
		Assistance at home	22	(10.1%)	26	(12.0%)
	Chronic illness/patient		1.7		1.9		.6
	Drugs/patient		3.5		3.5		.9
	Admission diagnosis
	IHD (MI and angina)		29	(13.3%)	30	(13.8%)	.9
	Fracture neck of femur		9	(4.1%)	9	(4.1%)	.9
	Syncope		2	(.9%)	7	(3.2%)	.1
	Atrial fibrillation		8	(3.7%)	11	(5.1%)	.5
	Stroke		12	(5.5%)	5	(2.3%)	.1
	Fall		13	(6.0%)	12	(5.5%)	.8
	COPD		25	(11.5%)	39	(18.0%)	.1
	Chest infection		21	(9.6%)	18	(8.3%)	.6
	Urinary tract infection		5	(2.3%)	9	(4.1%)	.3
	Elective surgery knee		14	(6.4%)	13	(6.0%)	.9
	Elective surgery hip		11	(5.0%)	8	(3.7%)	.5
	Heart failure		17	(7.8%)	19	(8.8%)	.7
	Miscellaneous		52	(24%)	37	(17%)	.1
			Mean	SD	Mean	SD
	Age (y)		76.3	(6.1)	77.1	(6.3)	.2
	Barthel score		16.1	(4.5)	15.8	(4.7)	.9
	CRP (mg/L)		48.8	(66.8)	55.6	(79.6)	.5
	Dietary intake (kcal)		1653	(399)	1672	(345)	.1
	BMI (kg/m2)		25.2	(4.1)	25.0	(4.3)	.6
	Weight (kg)		66.7	(13.1)	65.7	(13.7)	.4
	MUAC (cm)		28.3	(3.5)	28.0	(3.8)	.5
	TSF (mm)		15.8	(6.6)	15.5	(6.5)	.9
	Albumin (g/L)		37.8	(4.6)	37.9	(4.7)	.6
	Transferrin (g/L) (baseline)		2.2	(.5)	2.1	(.5)	.05
	Ferritin (μg/L)		169.6	(234.6)	181.8	(199.3)	.03

SD = standard deviation; IHD = ischemic heart disease; MI = myocardial infarction; COPD = chronic obstructive pulmonary disease; CRP = C-reactive protein; BMI = body mass index; MUAC = mid upper arm circumference; TSF = triceps skinfold.

View Large Image Download to PowerPoint

Figure 2. Adherence to supplements and placebo (maximum 84 drinks).

Nutritional Outcomes 

At the 6-month follow-up there were no statistically significant differences between the supplement and placebo groups in body weight, body mass index, mid upper arm circumference, triceps skinfold, or transferrin. Serum albumin concentration increased significantly in the supplement group (Table 2). Both red-cell folate and plasma vitamin B12 concentrations significantly improved in the supplement group compared with the decrease seen in the placebo group (Figure 3).

Table 2. Effect of Supplements on Main Outcome Measures Compared with Placebo at 6 Months

	Outcome Measure		Placebo		Supplements		P Value
			No.	Mean (SD)	No.	Mean (SD)
	Nutritional	Body mass index	106	26(4)	119	26(4)	.6
		Body weight(kg)	106	69(13)	119	69(14)	.9
		MUAC(cm)	106	28.6(3)	119	28.3(4)	.4
		TSF(mm)	106	15.6(7)	119	15.3(6)	.9
		Serum albumin (35-55 g/L)	106	40.5(4)	119	42(4)	.04
		Serum transferrin (2-4 g/L)	106	2.59(0.6)	119	2.4(0.5)	.3
	Clinical	Barthel score	223	18.6(3)	222	18.3(3)	.9
		Length of stay(d)	223	10.1(8)	222	9.4(7)	.2
		No. of infections	223	26(12%)	222	21(10%)	.4
		No. of readmission	223	89(40%)	222	65(29%)	.02
		No. of deaths	223	19(9%)	222	32(14%)	.1

SD = standard deviation; MUAC = mid upper arm circumference; TSF = triceps skinfold.

View Large Image Download to PowerPoint

Figure 3. Effect of supplements on red-cell folate and plasma vitamin B12 concentrations compared to placebo.

Clinical Outcomes 

The proportion of patients readmitted to the hospital at 6 months was significantly lower in patients randomly assigned to nutritional supplements (29%) compared with those in the placebo group (40%) (P < .05). Figure 4 shows the risk of readmission for supplement and placebo groups over the 6-month period. Cox regression analysis showed that the risk of non-elective readmission in the 6-month follow-up period was significantly lower in the supplement group than in the placebo group after adjustment for other clinical risk indicators, with a hazard ratio of 0.68 (95% CI, 0.49-0.94). No other clinically important difference between the supplement and placebo groups was found in length of stay in hospital, infections, disability, discharge destination, or mortality (Table 2). The death rate was higher at 6 months in the supplement group (32/223; 14%) compared with the placebo group (19/222; 9%) (P value = .06). After adjustment for other clinical risk indicators, Cox regression suggests that the hazard or risk of dying, up to 6 months postrandomization, was 1.65 times higher (95% CI, 0.93-2.92, P = .09) in the supplement group compared with the placebo group.

View Large Image Download to PowerPoint

Figure 4. Kaplan-Meier curve for 6-month readmission.

Strengths and Weaknesses 

We believe this study to be the first randomized, double-blind, placebo-controlled trial to compare the effect of a 6-week protein, energy, and micronutrient supplementation on the medium-term clinical outcome of hospitalized elderly patients. Despite the modest adherence to the full amount of prescribed supplements, we compared the group randomly allocated to supplements with the group allocated to placebo. This approach, in addition to being unbiased, provides a pragmatic answer to the question of primary clinical interest in this trial, for example, does nutritional supplementation of older patients during acute illness and convalescence/rehabilitation periods have a clinical benefit? Adherence to nutritional supplements by older people is known to be poor even when the supplements are prescribed in the drug charts and issued by the nursing staff to patients with their prescribed medications and with continuous prompting from the research dietician.16, 17

Future research should focus on understanding the complex factors that determine the intake of food in the aging population, such as the effect of improved palatability on stimulating food intake and screening tools for acceptability of such products by the elderly.

We cannot exclude the possibility that the decreased readmission rates as a result of nutritional support is a chance finding; however, it is also possible that it is a real one given the improvement in micronutrient status and the fact that mild subclinical nutritional deficiencies are known to be common even in relatively healthy older persons.15 Our inclusion criteria and baseline characteristics suggest that our study population represent a better-nourished group of patients. By contrast, those who were excluded because of severe illness, dementia, supplement intake, or living in institutional care were more likely to be undernourished and may therefore benefit most from nutritional support.