On 2017 Mar 19, Harri Hemila commented:

Problems in study inclusions, in data extraction, and in the scales may bias the meta-analysis on vitamin C and post-operative atrial fibrillation (POAF)

Hu X, 2017 state in their methods that “studies that met the following criteria were included: (1) randomized controlled trials (RCTs) of adult patients who underwent cardiac surgery; (2) patients randomly assigned to receive vitamin C or placebo …”. However, Hu X, 2017 included the study by Carnes CA, 2001 although that was not an RCT, instead “an age- and gender-matched control group (not receiving ascorbic acid) was retrospectively selected”. In addition, the Hu X, 2017 meta-analysis did not include the data of 2 rather large US trials that found no effect of vitamin C against POAF and thus remained unpublished leading to publication bias, see Hemilä H, 2017. Furthermore, Hu X, 2017 claimed that “funnel plots showed no evidence of publication bias”, but the existence of the 2 unpublished US studies refutes that statement.

Furthermore, Altman DG, 1998 pointed out that “the odds ratio [OR] should not be interpreted as an approximate relative risk [RR] unless the events are rare in both groups (say, less than 20-30%)”. However, in the Fig. 2 of Hu X, 2017, the lowest incidence of POAF in the placebo groups was 19% and 6 out of 8 studies had incidence of POAF over 30% in their placebo groups. In such a case the OR does not properly approximate RR, and the authors should have calculated the effect on the RR scale instead.

In their Fig. 4, Hu X, 2017 state that the mean duration of intensive care unit (ICU) stay in the vitamin C group was 24.9 hours in Colby JA, 2011. However, Colby JA, 2011 reported in their Table 1 that the duration of ICU stay was 249.9 hours, i.e. 10 times greater. Evidently, such an error leads to bias in the pooled estimate of effect, but also leads to exaggeration of the heterogeneity between the included trials.

Hu X, 2017 calculated the effect of vitamin C on the duration of ICU stay and of hospital stay on the absolute scale, i.e. on days, although there were substantial variations in the placebo groups, and thus the relative scale would have been much more informative Hemilä H, 2016. As an illustration of this difference between the scales, Hemilä H, 2017 calculated that the effect of vitamin C on hospital stay in days was significantly heterogeneous with I² = 60% (P = 0.02). In contrast, the effect of vitamin C on hospital stay on the relative scale was not significantly heterogeneous with I² = 39% (P = 0.09). The lower heterogeneity on the relative scale is explained by the adjustment for the baseline variations in the studies.

Hu X, 2017 write “compared with placebo group, vitamin C administration was not associated with any length of stay, including in the ICU”. However, Hemilä H, 2017 calculated that there was strong evidence from 10 RCTs that vitamin C shortened ICU stay in the POAF trials by 7% (P = 0.002).

Hu X, 2017 also concluded that vitamin C did not shorten the duration of hospital stay, whereas Hemilä H, 2017 calculated that vitamin C shortened hospital stay in 11 POAF trials by 10% (P = 10^-7 ).

Although the general conclusion of Hu X, 2017 that vitamin C has effects against POAF seems reasonable, there is very strong evidence of heterogeneity in the effect. Five trials in the USA found no benefit, discouraging further research in the USA. However, positive findings in less wealthy countries suggest that the effect of vitamin C should be further studied in such countries, Hemilä H, 2017.

On 2017 Mar 19, Harri Hemila commented:

Problems in study inclusions, in data extraction, and in the scales may bias the meta-analysis on vitamin C and post-operative atrial fibrillation (POAF)

Hu X, 2017 state in their methods that “studies that met the following criteria were included: (1) randomized controlled trials (RCTs) of adult patients who underwent cardiac surgery; (2) patients randomly assigned to receive vitamin C or placebo …”. However, Hu X, 2017 included the study by Carnes CA, 2001 although that was not an RCT, instead “an age- and gender-matched control group (not receiving ascorbic acid) was retrospectively selected”. In addition, the Hu X, 2017 meta-analysis did not include the data of 2 rather large US trials that found no effect of vitamin C against POAF and thus remained unpublished leading to publication bias, see Hemilä H, 2017. Furthermore, Hu X, 2017 claimed that “funnel plots showed no evidence of publication bias”, but the existence of the 2 unpublished US studies refutes that statement.

Furthermore, Altman DG, 1998 pointed out that “the odds ratio [OR] should not be interpreted as an approximate relative risk [RR] unless the events are rare in both groups (say, less than 20-30%)”. However, in the Fig. 2 of Hu X, 2017, the lowest incidence of POAF in the placebo groups was 19% and 6 out of 8 studies had incidence of POAF over 30% in their placebo groups. In such a case the OR does not properly approximate RR, and the authors should have calculated the effect on the RR scale instead.

In their Fig. 4, Hu X, 2017 state that the mean duration of intensive care unit (ICU) stay in the vitamin C group was 24.9 hours in Colby JA, 2011. However, Colby JA, 2011 reported in their Table 1 that the duration of ICU stay was 249.9 hours, i.e. 10 times greater. Evidently, such an error leads to bias in the pooled estimate of effect, but also leads to exaggeration of the heterogeneity between the included trials.

Hu X, 2017 calculated the effect of vitamin C on the duration of ICU stay and of hospital stay on the absolute scale, i.e. on days, although there were substantial variations in the placebo groups, and thus the relative scale would have been much more informative Hemilä H, 2016. As an illustration of this difference between the scales, Hemilä H, 2017 calculated that the effect of vitamin C on hospital stay in days was significantly heterogeneous with I² = 60% (P = 0.02). In contrast, the effect of vitamin C on hospital stay on the relative scale was not significantly heterogeneous with I² = 39% (P = 0.09). The lower heterogeneity on the relative scale is explained by the adjustment for the baseline variations in the studies.

Hu X, 2017 write “compared with placebo group, vitamin C administration was not associated with any length of stay, including in the ICU”. However, Hemilä H, 2017 calculated that there was strong evidence from 10 RCTs that vitamin C shortened ICU stay in the POAF trials by 7% (P = 0.002).

Hu X, 2017 also concluded that vitamin C did not shorten the duration of hospital stay, whereas Hemilä H, 2017 calculated that vitamin C shortened hospital stay in 11 POAF trials by 10% (P = 10^-7 ).

Although the general conclusion of Hu X, 2017 that vitamin C has effects against POAF seems reasonable, there is very strong evidence of heterogeneity in the effect. Five trials in the USA found no benefit, discouraging further research in the USA. However, positive findings in less wealthy countries suggest that the effect of vitamin C should be further studied in such countries, Hemilä H, 2017.