The authors of this Research paper have informed us of a coding error in their analysis programmes that generated the results for this study (BMJ 2013;347:f5001, doi:10.1136/bmj.f5001). Specifically, the total fruit intake variable included both whole fruits and fruit juices in the Nurses’ Health Study and Health Professionals Follow-up Study, but the authors incorrectly treated it as whole fruit in their initial analyses. Consequently, some of the values recorded for the hazard ratios and 95% confidence intervals of type 2 diabetes for whole fruit intake and fruit juice intake are incorrect. The authors state that these erroneous values do not affect the conclusions of their study or their estimates for individual fruits, and would also like to clarify that the total fruit intake variable was correctly coded in the Nurses’ Health Study II. The corrections that should be made to this paper as a consequence of the authors’ coding mistake are listed below.
In both the results section of the abstract and the third paragraph of the long results section, the pooled hazard ratio (95% confidence interval) of type 2 diabetes for every three servings/week of total whole fruit consumption should have read 0.98 (0.97 [not “0.96”] to 0.99). This inaccurate lower confidence interval is also listed in table 2, where it has been corrected in the republished version below. In addition, this error appeared in the print version, which can be seen online in the “Section PDF,” under the paragraph “Main results and the role of chance” and in the figure.
In the results section of the abstract, the penultimate sentence should have read that the “pooled hazard ratio for the same increment in fruit juice consumption was 1.07 (1.04 to 1.09), not “1.08 (1.05 to 1.11).” Readers should also note that this incorrect hazard ratio is listed in table 4, where it has been corrected in the republished version below. This inaccurate value for the hazard ratio also appeared in the print version, which can be seen online in the “Section PDF,” under “Main results and the role of chance” and in the figure.
Under the methods section, the seventh sentence in the paragraph “Assessment of fruit consumption” should have been truncated to read: “We calculated total whole fruit consumption by summing the consumption levels of the 10 individual fruits.”
In the first paragraph of the results section, the last two sentences contain some inaccuracies. The penultimate sentence should have noted that the average rate of loss to follow-up for each two year follow-up cycle was 0.7% [not “0.8%”] for <4 servings/week and 0.7% for ≥3 servings/day of total fruit consumption in the Nurses’ Health Study. While the last sentence should have read that these respective values were 0.9% and 1.3% [not “1.0% and 1.1%”] in the Health Professionals Follow-up Study.
In the last sentence of the second paragraph in the results section, the Spearman correlation coefficients for total whole fruits in relation to the modified alternate healthy eating index score should have been 0.29 [not “0.22”] for the Nurses’ Health Study, 0.29 for the Nurses’ Health Study II, and 0.32 [not “0.28”] for the Health Professionals Follow-up Study.
In the last paragraph of the results section, the first sentence should have noted that “replacing each three servings/week of fruit juice consumption with the same amount of total or individual whole fruits, the risk of type 2 diabetes in the pooled analysis was 5% (95% confidence interval 3% to 7%) [not “7% (95% confidence interval 4% to 9%)”] lower for total whole fruits.”
In the figure in the online version, the bar for “Total fruits” should have depicted a hazard ratio and 95% confidence interval of 0.95 (0.93 to 0.97) for the “Relative risk of type 2 diabetes.”
Lastly, tables 1, 2, and 4 all contain some incorrect values. In table 1, the values for the Nurses’ Health Study and the Health Professionals Follow-up Study were incorrect. A corrected version of table 1 is below.
In table 2, the values for the Nurses’ Health Study, the Health Professionals Follow-up Study, and the pooled results contained some inaccuracies. A corrected version of table 2 is below.
In table 4, the hazard ratios and 95% confidence intervals for the variable of fruit juice should have been different. A corrected version of table 4 is below.
Cite this as: BMJ 2013;347:f6935