Thirteen placebo-controlled comparisons have examined the therapeutic effect of zinc lozenges on the duration of common cold episodes of natural origin (Table

and Fig

). The total number of common cold episodes in these trials was 1407. All the 13 comparisons were double-blind, although this feature was not a selection criterion. Because of double-blinding, all trials used allocation concealment. Weissman

*et al*. [

28] used consecutive allocation, but all the other trials were randomized. All studies examined young and middle-aged adults, except the Macknin

*et al*. [

27] trial, which examined schoolchildren.

In Table , the trials are ordered by the total daily quantity of elemental zinc obtained from the lozenges. There is a seven-fold variation in the total daily dose of zinc. There is also a considerable variation in the results. Seven comparisons found a statistically significant benefit from zinc lozenges, but six did not.

Smith

*et al*. [

19] did not observe a difference between the study groups in the duration of colds; however, they did find a significant reduction in the severity scores on days 4 to 7 of treatment with P(1-tail)=0.01. Eby

*et al*. [

7] did not report the mean or median duration, but they did report the number of participants who had no symptoms at the end of the 7-day trial, an outcome used for calculating the P-value in Table

. The duration of colds given in Table

for the Eby trial is based on the exponential model which they used to estimate the time at which half of their patients were cured.

Table shows that a substantial proportion of the variation in the results can be explained by the daily zinc dosage. None of the five comparisons that used less than 75 mg/day of zinc found an effect of zinc lozenges, whereas seven of the eight comparisons which used over 75 mg/day of zinc found a statistically significant benefit, although the benefit in the Smith *et al*. trial was restricted to the symptom severity at the late phase of the colds.

The P-values of the individual trials are combined by using the Fisher method in Table . Combining the P-values of all the 13 comparisons provides very strong evidence that the zinc lozenge and placebo groups differ over all the trials. However, the benefit of zinc is restricted to trials where the dose was greater than 75 mg/day. A significant effect by zinc lozenges is seen separately in three high-dose trials where zinc acetate was used and in five high-dose trials which used zinc salts other than acetate.

Fig. () shows the forest plot of all 13 zinc lozenge comparisons. There is a highly significant heterogeneity between the 13 trials on the basis of both the χ^{2} and I^{2}-tests, with χ^{2}(12 df) = 109 and I^{2} = 89%. The trials are divided into the low and high dose subgroups as in Table and the two subgroups are considerably different in their estimate of the zinc lozenge effect.

In the low-dose trials, there is no evidence of heterogeneity and all the low-dose comparisons are consistent with no effect of zinc lozenges. In the eight high-dose trials, the zinc lozenges reduce the duration of colds by 32% (95% CI: 27% to 37%) but there is strong evidence of heterogeneity within this high-dose subgroup (χ^{2}(7 df) = 46 and I^{2} = 85%) (Fig. ).

Pooling the three high dose (>75 mg/day) zinc acetate trials gives a mean effect of 42% reduction in the duration of colds and no heterogeneity is seen between these trials (Table

). Five high dose (>75 mg/day) trials used zinc salts other than acetate [

7,

19,

20,

23,

24]. None of them reported the SD value, but SD was estimated from the published survival curves (Supplementary Material

3) [

7,

19,

23,

24] or the reported t-value [

20]. Pooling these five non-acetate trials gives a mean effect of 20% (95% CI: 12% to 28%) reduction in the duration of colds. However, there is significant heterogeneity between these five non-acetate trials (χ

^{2}(4 df) = 25.3, P = 0.0001; I

^{2} = 84%).

| **Table 3**Pooling the Results of the High Dose Zinc Acetate Trials |

Sensitivity analysis by the methodological quality of the trials was not carried out because all trials were double-blind, which also means that all used concealed allocation. One trial used consecutive allocation [

28], but in a double-blind trial it is not reasonable to assume that consecutive allocation would lead to systematic bias between the study groups. Furthermore, the trial [

28] had low dose of zinc and therefore its exclusion would strengthen, and not weaken, the evidence that zinc lozenges differ from the placebo.

All three trials which used zinc acetate in doses higher than 75 mg/day (Table

) were methodologically rigorous randomized trials [

21,

22,

25]. In the Petrus trial, only one participant was lost from follow-up [

22]. In the first Prasad trial, two participants in the placebo group dropped out on day 2 [

25], whereas there were no drop-outs in the second Prasad trial [

21]. No sensitivity analysis was done in this subgroup.

There is substantial heterogeneity within the five high-dose non-acetate trials. Sensitivity analysis was done leaving out the Eby

*et al*. [

7] and Smith

*et al*. [

19] trials, which excluded a large number of randomized participants, 45% and 36%, respectively. Furthermore, over half of the common cold durations needed to be imputed to include these trials in Fig. (

). When these two trials were excluded from the high-dose non-acetate subgroup, the remaining three trials lead to 22% (95% CI: 11 to 32%) decrease in the duration of colds. This confidence interval is essentially the same as that for all the five non-acetate trials, see above. Thus, exclusion of these two trials has no influence on conclusions.