Intervention

Participants in the treatment group received a daily oral tablet that contained amounts of vitamins and minerals very similar to those found in most commercially available multivitamin and mineral supplements. Table 1 lists the exact ingredients of each tablet. Participants in the placebo group received a daily oral tablet containing calcium, 120 mg; magnesium, 100 mg; and vitamin B₂ (riboflavin), 3.4 mg. These ingredients were chosen so that the tablet would look and smell as similar as possible to the complete multivitamin and mineral supplement and would cause similar changes in the color of participants' urine. Tishcon Corp. (Westbury, New York) prepared the supplement and placebo pills specifically for this study. All participants were required to take the tablets daily for 1 year.

Data Collection

Baseline measurements included sociodemographic characteristics, clinical data, nutritional status, level of stress, health behaviors, and quality of life. All sociodemographic data were dichotomized and included age (45 to 64 years or 65 years), sex, ethnicity (black or white), and years of education (less than high school or high school graduate and beyond). Clinical data included body mass index (normal, <25 kg/m²; overweight, 25 to 29.9 kg/m²; obese, 30 kg/m²). Hemoglobin A_1c level was measured among participants with diabetes. We also assessed history of vitamin use in the year before the study, previous influenza immunization, and whether the participant had had at least one infection in the previous year.

Nutritional status was assessed with a 3-day food diary that participants kept at baseline and at 6 months. At randomization, participants were given detailed written and verbal instructions on how to fill out the food diary, and the study nutritionist reviewed each completed diary with the participant. Analysis was performed by using Nutritionist V, version 2.0 (First DataBank, Inc., San Bruno, California). Nutrient deficiency, for the purpose of this study, was defined as intake below the 33rd percentile of the recommended daily allowance (RDA) for zinc; selenium; iron; folic acid; or vitamins A, C, E, or B₆. Selection was based on research indicating that even moderate deficiencies of these individual nutrients can impair markers of immune function (3-6, 15-17). Our choice of the 33rd percentile was somewhat arbitrary. There is no standard cut-point for micronutrient deficiency as it relates to impairment of immune function because the role of each micronutrient depends on many other factors, such as age, chronic disease, other nutrient deficiencies, and bioavailability. Because many variables are unique to each individual and sample, previous studies demonstrating an immune function benefit from supplementation have defined deficiency post hoc (11, 16).

Stress, which has been shown to influence susceptibility to minor infection (18, 19), was measured by using a 10-item scale that assessed perceived stress during the previous month (20). Each question was measured on a scale of 0 to 4 points, yielding a possible range of 0 to 40 points. For health behaviors, we included smoking status (current, past, or never) and measures of routine physical and recreational exercise (very or moderately active vs. inactive), using items from the National Health and Nutrition Examination Survey II.

The primary end point of our study was incidence of participant-reported infection. Infection data were obtained from symptom checklist diaries recorded daily by the study participants. Detailed instructions on keeping the diaries were given to each participant at the start of the study, and participants were encouraged to call with any questions if they became uncertain. Each participant brought his or her diary to every quarterly study visit, and all aspects of the recorded data were reviewed with participants for clarity, accuracy, and completeness. Using this record of illness, the study physician then assigned a specific diagnosis and duration. Standard criteria were used to diagnose common adult infectious illnesses (upper respiratory tract infection, lower respiratory tract infection, influenza-like syndrome, gastrointestinal infection, and urinary tract infection). Symptom checklist records that fit into the above categories prompted further detailed interview by the physician to confirm or exclude other infectious and noninfectious entities (for example, allergic rhinitis and medication-induced diarrhea).

Infection incidence was determined by using both continuous and categorical variables. First, we calculated the mean number of total "symptomatic days" due to infection per year. Because many study participants (n = 54) reported no infection and the data were therefore considerably skewed, the dichotomous variable of infection versus no infection proved more meaningful. We also assessed the significance of infections by ascertaining "absentee" days, that is, days of infection-related illness that resulted in missed work or inability to perform planned activities.

The secondary end point, quality of life, was assessed at baseline and at 12 months. Proxy measures of physical and mental health were obtained by using the Medical Outcomes Study 12-Item Short Form (SF-12) (21). We computed total scores using weighted-item response categories similar to those used to score the 36-Item Short-Form Health Survey (22). Adverse effects, adherence, and any substantial change in medical status were assessed at the quarterly visits.

Statistical Analysis

Descriptive statistics for baseline covariates were assessed by using frequencies for dichotomous variables and medians and interquartile ranges for continuous variables. Chi-square tests and t-tests were used to evaluate the differences between the treatment and placebo groups across all categorical and continuous baseline covariates, respectively. These analyses were conducted for all 158 participants recruited to the study and for the 130 participants who followed the protocol for the entire study year.

Twenty-eight participants withdrew from the study between collection of the baseline data and initial evaluation of the study end points. Because no data were available on the study end points for these participants, we evaluated the effect of attrition using "worst-case" and "best-case" scenarios. First, we assumed all participants who withdrew had an infection and analyzed the data using an intention-to-treat analysis. We then repeated the analysis assuming that all participants who withdrew did not have an infection. Neither of these assumptions affected the directionality or statistical significance of our results. Thus, we proceeded with our analysis using the total number of participants who completed the trial (n = 130). All analyses were conducted by using Stata 7.0 (Stata Corp., College Station, Texas).

For our primary outcomes, we calculated the absolute and relative risk for infection and infection-associated absenteeism for all persons who completed the 12-month trial (n = 130), according to study group, age, and diabetes status. A Mantel–Haenszel test of homogeneity was performed to test for differences between the age and diabetes subgroups. We also conducted univariate and multivariate analyses using ordinary least-squares regression to evaluate the effect of the supplement on physical and mental quality of life while controlling for age, diabetes status, nutritional deficiency, and physical activity.

Role of the Funding Source

The funding source had no role in the design, conduct, or reporting of the study.

Results

Top Editors' Notes Methods Results Discussion Author & Article Info References

Participant Characteristics

Table 2 summarizes the characteristics of the study participants. The treatment group was more physically active and better nourished than the placebo group but was similar with respect to baseline covariates. The overall sample was predominately female, white, and between the ages of 45 and 64 years. Approximately 10% of participants had not finished high school. Two thirds of the sample were overweight or obese, and approximately 30% had type 2 diabetes mellitus. Approximately 18% of the total sample were classified as nutritionally deficient. Slightly more than half of the participants in both the placebo and treatment groups remembered having had at least one infection in the previous year.

Twenty-eight persons did not complete the trial (15 in the treatment group vs. 13 in the placebo group). Participants who withdrew were more likely to be 45 to 64 years of age and nondiabetic at baseline, but no other differences were noted (Figure). The most commonly reported reason for withdrawal was the inconvenience of keeping a daily diary or attending quarterly physician visits. One serious adverse event, a cardiac arrest, occurred during the trial; the affected participant was successfully resuscitated and continued in the study.

Infection Outcomes

No infections requiring hospitalization occurred during the study period. Forty-two percent of participants had upper respiratory tract infections, 19% had influenza-like syndromes, 12% had gastrointestinal infections, 7% had lower respiratory tract infections, and fewer than 2% had urinary tract infections or miscellaneous infections. Twenty percent of persons experienced more than one type of infection over the study year.

Seventy-three percent of the placebo group experienced one or more infection-related illnesses versus 43% of the treatment group (P < 0.001). Similarly, 57% of the placebo group reported illness-related absenteeism versus 21% of the treatment group (Table 3). Analysis within the age and diabetes subgroups showed that the differential effect of vitamin supplementation on infection was found primarily among participants with diabetes. Ninety-three percent of diabetic participants in the placebo group experienced an infectious illness versus only 17% of diabetic participants in the treatment group (relative risk, 0.18 [95% CI, 0.07 to 0.44]). Similarly, 89% of diabetic participants in the placebo group reported one or more absentee days compared with 0% of diabetic participants in the treatment group. Participants without diabetes did not differ by study group.

We did not find an effect of age on infection. However, the lack of a statistically significant effect among participants older than 65 years of age does not allow us to draw any meaningful conclusions because only 33 participants were in this age group. Participants younger than 65 years of age experienced an apparent reduction in infection because of vitamin supplementation (78% in the placebo group vs. 43% in the treatment group); however, the difference in outcome between the two age groups was not statistically significant (P > 0.2) (Table 3). At baseline, diabetic participants were more likely than nondiabetic participants to be deficient in one or more micronutrients (33% vs. 19%; P = 0.06) (Table 4).

Treatment Effects on Quality of Life

The quality-of-life scores in our study were similar to those in persons who reported minor medical problems in a validation study of the SF-12 (21). In multivariate analyses, vitamin supplementation had no effect on physical or mental health scores (difference, 1.02 [CI, –2.22 to 4.2] and 0.98 [CI, –1.97 to 3.94], respectively). No additional analyses of quality of life were conducted within the age and diabetes subgroups because treatment had no main effect on outcome.

Discussion

Top Editors' Notes Methods Results Discussion Author & Article Info References

The purpose of this study was to assess the effect of a typical "one-a-day" multivitamin and mineral supplement on infection rate and perceived quality of life in a fairly broad sample of relatively healthy adults. We found no difference between the treatment and placebo groups in physical and mental health measures of quality of life. Furthermore, within the SF-12 instrument, the specific items of "general health" and "energy" were analyzed separately and showed no difference in treatment effects between the study groups. Thus, it seems that multivitamin and mineral supplements do not enhance health or energy level any more than does placebo.

Our trial, which was performed in a sample of middle-aged persons, demonstrated a benefit in incidence of infection. However, this benefit was almost entirely observed in participants with diabetes, for whom the magnitude was dramatic. Correction of micronutrient deficiencies would be the most biologically plausible explanation for our results. When we defined nutritional deficiency as intake below the 33rd percentile of the RDA, participants who were diabetic at baseline were more likely than nondiabetic participants to be deficient in one or more micronutrients. It remains uncertain, however, whether these small differences in nutritional status were enough to account for the dramatic difference in outcome. This uncertainty is due not only to the limitations of our study but also to inherent difficulties in defining micronutrient deficiency.

Most of the previous research on vitamin and mineral supplementation has evaluated its effect on various markers of immune function (3-10, 15, 16). A few studies have evaluated clinical outcomes, although mostly in severely compromised or malnourished persons. Research in samples of relatively healthy elderly persons, however, has shown that even mild nutrient deficiencies can impair immune response (3-6, 12, 13, 16). Such subclinical deficiencies are common in community-dwelling elderly persons and probably account for some of the impaired immunity in this population (11-13, 15-17). Whether these data can be extrapolated to other minimally compromised groups, such as middle-aged participants with stable chronic disease processes, is unknown.

Only two published trials have evaluated the effect of a multivitamin and mineral supplement on clinical infection in apparently healthy samples of community-dwelling elderly persons. Chavance and colleagues (14) found no difference in incidence of infection, but their trial was methodologically flawed. Infection data were gathered by using personal recall at 2-month intervals, and the study was too short (4 months) to exclude a treatment effect or to examine seasonal variation. In addition, persons with chronic illness or possible micronutrient deficiency were excluded.

Chandra's 12-month study (11) of 96 healthy elderly persons showed a statistically significant difference in the mean number of days with an infection-related illness (23 days vs. 48 days) in favor of the group that received the vitamin and mineral supplement. No reported results showed an effect on severity of illness (other than number of days antibiotics were taken). In addition, participants in the treatment group had statistically significant improvement in several reliable indices of immune function. Nutritional status was determined by using biochemical measurements, and deficiency in a given nutrient was defined as a blood concentration value below the fifth percentile in a previously established reference range developed by the same researcher from a similar sample. Although the study groups had equal prevalence of deficiency at baseline, the treatment group had less severe deficiencies of vitamin A, beta-carotene, vitamin B₆, vitamin C, iron, and zinc. However, the author noted some statistically significant improvements in markers of immune function even among persons who were not deficient by the same criterion (12). This illustrates the important unresolved issue of how to define micronutrient deficiency. In our trial, we used the criterion of intake below the 33rd percentile of the RDA for zinc; selenium; iron; or vitamins A, C, E, or B₆. These micronutrients were specifically chosen because studies have demonstrated that even minor deficiencies lead to impairment in indices of immune function (3-6, 15-17).

One of the limitations of our study was that we determined micronutrient status with a 3-day food diary, which has only a moderate correlation with more accurate and expensive methods. Nevertheless, even if the quantitation of intake were highly accurate, there remains the challenging problem of defining deficiency. The relationship between intake of a nutrient and levels of the same nutrient in serum or tissue depends on numerous bioavailability variables. Defining deficiency on the basis of response to supplementation depends on the specified response end point, numerous physiologic characteristics of the study sample, and biological interconnections and interactions among various micronutrients. All of these factors disallow a simple association between a specific nutrient deficit (however defined) and a given physiologic or clinical outcome. We believe, however, that the diabetic persons we studied had substantial preexisting micronutrient deficiencies. They were also less likely than nondiabetic participants to be white, to have a high school degree, and to have previously used vitamins and were more likely to have experienced illnesses before the trial, to be obese, and to have perceived stress. Lower socioeconomic status and poor health are strongly associated with poorer nutritional status.

Although substantial research shows that certain infections are associated with diabetes (for example, foot ulcer infections, urinary tract infections, and unusual fungal infections), little published work addresses common upper respiratory tract infections, which accounted for the largest proportion of sick days in our trial. The type and severity of impaired immunity in diabetic persons have been studied to some extent, but scant literature exists on ways in which to improve such impairment, other than possibly by improving glycemic control (23-25). Some research has suggested that antioxidant systems are impaired in diabetes, thus providing a theoretical basis for benefit from a supplement with antioxidant potential (26).

Age was the other prespecified subgroup in this trial. No effect of supplementation was noted among those 65 years of age and older. Because few elderly participants were included, however, even a large effect may have been missed, as shown by the large CI (Table 3).

Unlike most researchers performing studies of nutritional supplements, we tested the blinding of our drug intervention. Although we went to great effort to have the pills look and smell as identical as possible, most participants correctly guessed what they were taking. Nevertheless, we doubt that this partial unblinding substantially influenced the results. First, at least some differential effect of supplementation would be expected in all of the subgroups. However, in the largest subgroup (nondiabetic persons), no difference was seen between those who received vitamins and those who received placebo. Second, in the group in which treatment had the largest effect on infection rate, unblinding had no effect on the secondary outcome of interest, quality of life.

Infection type and severity could have varied by season of the year, but randomization did not differ by season, and the 12-month study duration provided equal exposure to seasonal variability for all participants. It should be reiterated that the infections prevented in this study were minor, primarily respiratory tract infections and influenza-like illnesses. Nonetheless, because such infections are very common, they account for a large share of the productive days lost in industrialized societies (27, 28).

Our findings suggest that in certain diabetic samples, perhaps those with a high prevalence of micronutrient deficiency, daily use of a multivitamin and mineral supplement can decrease infection frequency. These dramatic results warrant testing in a larger clinical trial, both for confirmation and to disentangle the effects of chronic disease status from those of other important factors, such as ethnicity and obesity, that serve as proxies for socioeconomic status and poorer nutrition. Multivitamin and mineral supplements are convenient and relatively inexpensive. If our results are confirmed in a larger trial, the widespread implementation of this preventive measure could have a substantial economic impact and could ease the burden of suffering in our society.