Data Errors in: Mortality in Randomized Trials of Antioxidant Supplements for Primary and Secondary Prevention: Systematic Review and Meta-analysis

Data Errors: The Review article entitled “Mortality in Randomized Trials of Antioxidant Supplements for Primary and Secondary Prevention: Systematic Review and Meta-analysis” published in the February 28, 2007, issue of JAMA (2007;297[8]:842-857) contained data errors. On page 842 in the “Data Synthesis” section of the abstract, the lower confidence limit for the “multivariate meta-regression analyses showed that low-bias risk trials” that read “1.05” should have read “1.04.”

On page 844, in the first paragraph of the “Results” section, the sentence describing the types of designs used in the study trials that read “Forty trials used parallel-group design, 26 factorial design (23 trials 2 × 2; 2 trials 2 × 2 × 2; 1 trial half replicate of 2 × 2 × 2 × 2), and 2 crossover design,” the parenthetical breakout of the factorial design that attributed “23” to the 2 × 2 design, should have read “22” and number of 2 × 2 × 2 trials, should have been “3.” Accordingly, in Table 2 on page 847, the study design for “Lee et al,⁹⁴ 2005” should read “2 × 2 × 2.”

On page 845 in the first paragraph of the “All Randomized Trials” subsection, the sentence that read “Heterogeneity was not significant (I²=18.6%, P=.10)” should have read “Heterogeneity was significant (I²=18.9%, P=.10).” In the following sentence that begins “Adjusted-rank correlation test (P=.08), but not the regression asymmetry test (P=.26), suggested the bias among trials,” the respective P values should have read “(P=.09)” and “(P=.24).” In the second paragraph of the same subsection, the portion of the sentence that begins on page 845: “Univariate meta-regression analyses revealed significant influences of dose of beta carotene (RR, 1.004; 95% CI, 1.001-1.007; P=.012),” the P value should have been equal to “.014.” In the latter part of the same sentence that falls on page 847, the P value for the dose of selenium that read “P=.002” should have read “P=.001.” In the following part of the sentence, the upper confidence limit that read “1.29” should have read “1.30.” In the third paragraph of the same subsection, on page 847, the P value for the “multivariate meta-regression” for dose of selenium that read “P=.005” should have read “P=.004,” the lower confidence limit for low-bias risk trials that read “1.05” should have read “1.04,” and the P value for the low-bias risk trials in the same sentence that read “P=.005” should have read “P=.006.”

In the first paragraph of the “Bias Risk of Trials” subsection on page 847, the I² value that read “(I²=7.0%)” should have read “(I²=7.5%).” On the same page, in the first and second sentence of the subsection “Antioxidant Supplements Given Singly or in Combination” that read “Beta carotene used singly significantly increased mortality (Table 5). This effect was not significant when combined with other supplements” should have read “Beta carotene used singly or in combination with other antioxidant supplements did not significantly affect mortality.”

In the first sentence of the last paragraph of the same subsection that falls in the third column on page 848 that read “Selenium given singly or in combination with other antioxidant supplements had no significant influence on mortality when analyzed separately (Table 5)” should have been divided into 2 sentences that should have read “Selenium given singly had no significant influence on mortality. Selenium given in combination with other antioxidant supplements significantly decreased mortality (Table 5).”

In Figure 2 on page 851, the denominators for the participant mortality ratios in the “Antioxidants” and “Control” columns were reversed for the “Green et al,⁶⁰ 1999” study. They should have read “15/820” and “22/801,” respectively. The corresponding relative risk (RR) and 95% confidence interval (CI) that read “0.70 (0.36-1.34)” should have read “0.67 (0.35-1.27).” Also in Figure 2, for the “Brown et al,⁶⁵ 2001” study, the mortality ratio that read “17/84” for the “Antioxidant” and “13/76” for the “Control” columns, should have read “1/84” and “1/76,” respectively. The corresponding RR and 95% CI that read “1.18 (0.62-2.27)” should have read “0.90 (0.06-14.2).” Accordingly, the numerators and denominators for total events that read “15 366/99 095” and “9131/81843” should have read “15 350/99 114” and “9119/81 824,” respectively. Tests for heterogeneity that read “χ²₄₆=49.47; P=.34; I²=7.0%” should have read “χ²₄₆=49.73; P=.33; I²=7.5%.” The test for overall effect that read “Z=3.06; P=.002” should have read “Z=2.98; P=.003.”

In Table 5 on page 853, the RR (95% CI) in the “Beta carotene given singly” row that read “1.06 (1.01-1.11)” should have read “1.05 (1.00-1.11)” and the I² value that read “5.4” should have read “11.8.” In the “Beta carotene given in combination with other antioxidant supplements” row, the I² value that read “55.6” should have read “55.5.” In the “Beta carotene given singly or in combination with other antioxidant supplements” row, the CI range that read “(0.96-1.08)” should have read “(0.95-1.07)” and the I² value that read “52.2” should have read “52.5.” In the “Beta carotene given singly or in combination with other antioxidant supplements after exclusion of high-bias risk and selenium trials” row, the I² value that read 36.8” should have read “34.4” In the “Vitamin E given singly” row, the number of study participants that read “47 007” should have read “41 341.” In the “Vitamin E given in combination with other antioxidant supplements” row, the RR that read “1.01” should have read “1.00” and the I² value that read “17.2” should have read “16.9.” In the “Vitamin E given singly or in combination with other antioxidant supplements” row, the I²value that read “2.8” should have read “2.4.” In the “Vitamin E given singly or in combination with other antioxidant supplements after exclusion of high-bias risk and selenium trials” row, the list of references should have included reference 87 and excluded 95.

In the “Vitamin C given in combination with other antioxident supplements” row, the lower confidence limit that read “0.88” should have read “0.87” and the I²  value that read “22.1” should have read “21.7.” In the “Vitamin C given singly or in combination with other antioxidant supplements” row, the RR that read “0.97” should have read “0.96” and the I² value that read “19.4” should have read “18.9.” In the “Selenium given in combination with other antioxidant supplements” row, the upper confidence limit that read “1.01” should have read 1.00” and the I² value that read “9.5” should have read “6.3.” In the “Selenium given singly or in combination with other antioxidant supplements after exclusion of high-bias risk trials” row, the upper confidence limit that read “1.02” should have read “1.01.”

None of the data errors altered the overall results of the study.