On 2017 Mar 26, Harri Hemila commented:

Antioxidants are not equal: an example of the apples and oranges problem

In their network meta-analysis on treatments for contrast medium–induced acute kidney injury (CIAKI), Su X, 2017 pooled different vitamins to a single group of “vitamins and analogues” but in doing so ignored the fact that vitamin C is water soluble, whereas vitamin E is fat soluble, and therefore their effects might be quite different. This is an example of the apples and oranges problem. With an analogous reasoning, studies on ciprofloxacin and penicillin might be pooled together on the basis that they are ”antibiotics”, though they have quite different mechanisms and indications.

Su X, 2017 calculated an odds ratio (OR) = 0.64 for the effect of “vitamins and analogues” but they did not calculate the specific effects of vitamins E and C. From 9 vitamin C trials, Sadat U, 2013 concluded that vitamin C may prevent CIAKI and calculated a risk ratio (RR) = 0.67. From 3 vitamin E trials, Rezaei Y, 2017 calculated that vitamin E decreased the incidence of CIAKI with RR = 0.38 (95%CI 0.24-0.62). On the OR scale, the effect of vitamin E corresponds to OR = 0.34 (95%CI 0.20-0.58).

Thus, vitamin E seems to have a greater effect against CIAKI compared with vitamin C. These two different vitamins should therefore not be pooled into a single group of “vitamins and analogues”, but they should be analyzed separately. The point estimates also suggest that there is greater justification for further research on vitamin E than on vitamin C. Furthermore, vitamins E and C may also interact under some conditions; for example, Hemilä H, 2009 found that vitamin E decreased mortality of older males only when they had a high vitamin C intake, but vitamin E had no effect when vitamin C intake was low.

Finally, Su X, 2017 estimated the effect of “vitamins and analogues” on the OR scale. However, Altman DG, 1998 pointed out that OR should be avoided when events are common. In many CIAKI studies, the proportion of CIAKI cases has been so high that OR gives an exaggerated impression of treatment effect. Su et al. calculated that high-dose statin plus NAC decreased the risk of CIAKI by OR = 0.31 (95%CI 0.14-0.60) and they concluded that “high-dose statin plus NAC or high-dose statin alone were likely to be ranked the best or the second best for preventing CIAKI”. However, the upper 95%CI limit for the effect of vitamin E on the OR scale (0.58) is lower than the upper 95%CI limit for the effect of high-dose statin plus NAC (0.60). Thus, on the basis of trials published so far. there is no basis to consider that these treatments actually differ in efficacy. Furthermore, half of the patients of the three vitamin E trials were concomitantly administered statins, and thus the effect of vitamin E may be at least partly independent of the effects of statins.

Thus, had Su X, 2017 analyzed the vitamin E trials separately, they might have concluded that there is as strong evidence to further study vitamin E as to further study high-dose statin plus NAC.

On 2017 Mar 26, Harri Hemila commented:

Antioxidants are not equal: an example of the apples and oranges problem

In their network meta-analysis on treatments for contrast medium–induced acute kidney injury (CIAKI), Su X, 2017 pooled different vitamins to a single group of “vitamins and analogues” but in doing so ignored the fact that vitamin C is water soluble, whereas vitamin E is fat soluble, and therefore their effects might be quite different. This is an example of the apples and oranges problem. With an analogous reasoning, studies on ciprofloxacin and penicillin might be pooled together on the basis that they are ”antibiotics”, though they have quite different mechanisms and indications.

Su X, 2017 calculated an odds ratio (OR) = 0.64 for the effect of “vitamins and analogues” but they did not calculate the specific effects of vitamins E and C. From 9 vitamin C trials, Sadat U, 2013 concluded that vitamin C may prevent CIAKI and calculated a risk ratio (RR) = 0.67. From 3 vitamin E trials, Rezaei Y, 2017 calculated that vitamin E decreased the incidence of CIAKI with RR = 0.38 (95%CI 0.24-0.62). On the OR scale, the effect of vitamin E corresponds to OR = 0.34 (95%CI 0.20-0.58).

Thus, vitamin E seems to have a greater effect against CIAKI compared with vitamin C. These two different vitamins should therefore not be pooled into a single group of “vitamins and analogues”, but they should be analyzed separately. The point estimates also suggest that there is greater justification for further research on vitamin E than on vitamin C. Furthermore, vitamins E and C may also interact under some conditions; for example, Hemilä H, 2009 found that vitamin E decreased mortality of older males only when they had a high vitamin C intake, but vitamin E had no effect when vitamin C intake was low.

Finally, Su X, 2017 estimated the effect of “vitamins and analogues” on the OR scale. However, Altman DG, 1998 pointed out that OR should be avoided when events are common. In many CIAKI studies, the proportion of CIAKI cases has been so high that OR gives an exaggerated impression of treatment effect. Su et al. calculated that high-dose statin plus NAC decreased the risk of CIAKI by OR = 0.31 (95%CI 0.14-0.60) and they concluded that “high-dose statin plus NAC or high-dose statin alone were likely to be ranked the best or the second best for preventing CIAKI”. However, the upper 95%CI limit for the effect of vitamin E on the OR scale (0.58) is lower than the upper 95%CI limit for the effect of high-dose statin plus NAC (0.60). Thus, on the basis of trials published so far. there is no basis to consider that these treatments actually differ in efficacy. Furthermore, half of the patients of the three vitamin E trials were concomitantly administered statins, and thus the effect of vitamin E may be at least partly independent of the effects of statins.

Thus, had Su X, 2017 analyzed the vitamin E trials separately, they might have concluded that there is as strong evidence to further study vitamin E as to further study high-dose statin plus NAC.