Authors' reply on the systematic review and does-response meta-analysis of fruit and vegetable consumption in relation to mortality
CCBYNCOpen access
Rapid response to:
Research
Fruit and vegetable consumption and mortality from all causes, cardiovascular disease, and cancer: systematic review and dose-response meta-analysis of prospective cohort studies
To the editor:
We appreciate very much the enormous interest from the globe in our recent study published in BMJ.1
We concur with Ellen M Storm2 that there were fewer studies evaluating the health benefits of fruit and vegetable consumption at the level of greater than 5 servings per day. Therefore, more studies are warranted in the future to confirm the association of fruit and vegetable consumption at levels of higher than 5 servings per day with risk of mortality. However, from public health point of view, the debate on the exact amount is not relevant at present, because the average consumption of fruit and vegetable is far below 5 servings per day in the general population.3 Consistent with our findings, the World Health Organization recommends 5 servings (400 g) of fruit and vegetables per day as the population goal for chronic disease prevention.4 Compared with the recommended level of fruit and vegetable consumption, the current average population level has much room to improve.
Both Liang Zhou5 and Zhihao Liu6 raised a concern of omitting a number of “eligible” studies in our meta-analysis. However, because Mann et al7 and Ness et al8 did not report data on person-years or total number of participants for each category of fruit and/or vegetable consumption, we were unable to include them in our dose-response meta-analysis using the generalized least squares trend estimation (GLST) method.9-11 The study by Lo et al12 was excluded because it reported daily expenditures on rather than consumption of fruit and vegetables.
Nechuta et al13 reported on the association between combined consumption of fruit and vegetable and mortality from the Shanghai Women’s Health Study. However, we did include a more recent article14 from the same cohort which examined the risk of mortality of fruit and vegetable separately. Including Nechuta et al13 in our meta-analysis did not materially alter our results or conclusions: the pooled hazard ratio (95% confidence interval) associated with each portion increase in fruit and vegetable consumption was 0.96 (0.94 to 0.97; P=0.001) for all-cause mortality. In addition, we still observed a threshold at around five servings daily consumption of fruit and vegetables, after which there was no further reduction in risk.
In our meta-analysis, two of the included studies reported fruit juice as part of fruit consumption.15 16 Liang Zhou5 argued that we should remove them from the analysis because there was evidence showing divergent associations of fruit and fruit juice consumption on type 2 diabetes.17 18 However, because there is no evidence showing any difference between fruit and fruit juice with regard to the risk of mortality, we do not believe there is sufficient rationale to exclude these two studies from our main analysis. Nonetheless, we conducted a sensitivity analysis by excluding these two studies and found almost identical results; the pooled hazard ratio of all-cause mortality was 0.94 (95% confidence interval 0.90 to 0.99; P=0.002) for each additional serving a day of fruit.
Suhail A. Doi19 questioned the random-effects models used in our meta-analysis, and suggested the use of the inverse variance heterogeneity (IVhet) and the quality effects (QE) models developed by the author. However, to our knowledge, the DerSimonian-Laird random-effects model20 21 is the standard method used in meta-analyses, and therefore it was used in our study.
We agree with Chibo Liu22 that the pooled HR associated with each unit increment of fruit and vegetable consumption may not be optimal in the context of curvilinear associations. Therefore, we presented both the dose-response curve plot and forest plot in our original paper.
We used the previously described definition23 for each serving of fruit (80 g) and vegetables (77 g), because these serving sizes were calculated on the basis of real weighting and averaging commonly consumed fruit and vegetables,23 rather than based on arbitrary numbers.
We conducted stratified analyses but did not perform a formal meta-regression to identify potential sources of heterogeneity because a robust conclusion from meta-regression requires a large number of included studies.24 Also, for the evaluation of publication bias, we were actually short of statistical power because of the limited number of included studies.25 In addition, the Egger’s test, which is based on regression method, may have a high false positive rate.25 26
Finally, we have been informed about a slight error in our paper. From one of the studies (Leenders et al, Am J Epidemiol 2013;178:590-602) included in the meta-analysis we inadvertently used data on fruit consumption and all cause mortality for women only rather than for the whole population. We have re-run the analyses using the correct hazard ratio for overall mortality of 1.00 (95% confidence interval 0.99 to 1.01) for 100 g/day fruit consumption. The revised pooled hazard ratio for all cause mortality should be 0.94 (0.89 to 0.98; P=0.006) for an increment of one serving of fruit a day, which is almost identical to that reported in the paper (0.94, 0.90 to 0.98; P=0.002) and hence does not affect the conclusions.
Wang Xia, Department of Maternal and Child Health Care, School of Public Health, Shandong University, Jinan, China
Wei Bao, Epidemiology Branch, Division of Intramural Population Health Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Rockville, MD 20852, USA
Frank B. Hu, Departments of Nutrition and Epidemiology, Harvard School of Public Health, Boston, MA 02115, USA
References
1. Wang X, Ouyang Y, Liu J, Zhu M, Zhao G, Bao W, et al. Fruit and vegetable consumption and mortality from all causes, cardiovascular disease, and cancer: systematic review and dose-response meta-analysis of prospective cohort studies. BMJ 2014;349:g4490.
2. Storm EM. Re: Fruit and vegetable consumption and mortality from all causes, cardiovascular disease, and cancer: systematic review and dose-response meta-analysis of prospective cohort studies. 2014. http://www.bmj.com/content/349/bmj.g4490/rr/761891.
3. Hall JN, Moore S, Harper SB, Lynch JW. Global variability in fruit and vegetable consumption. Am J Prev Med 2009;36:402-09 e5.
4. World Health Organization. Promoting fruit and vegetable consumption around the world. http://www.who.int/dietphysicalactivity/fruit/en/index2.html.
5. Zhou L. The omission of eligible publications may bias the conclusions. 2014. http://www.bmj.com/content/349/bmj.g4490/rr/762437.
6. Liu Z. Insufficient literature was included in the meta-analysis by Wang et al. 2014. http://www.bmj.com/content/349/bmj.g4490/rr/762200.
7. Mann JI, Appleby PN, Key TJ, Thorogood M. Dietary determinants of ischaemic heart disease in health conscious individuals. Heart 1997;78:450-5.
8. Ness AR, Maynard M, Frankel S, Smith GD, Frobisher C, Leary SD, et al. Diet in childhood and adult cardiovascular and all cause mortality: the Boyd Orr cohort. Heart 2005;91:894-8.
9. Berlin JA, Longnecker MP, Greenland S. Meta-analysis of epidemiologic dose-response data. Epidemiology 1993:218-28.
10. Greenland S, Longnecker MP. Methods for trend estimation from summarized dose-response data, with applications to meta-analysis. American journal of epidemiology 1992;135:1301-09.
11. Orsini N, Bellocco R, Greenland S. Generalized least squares for trend estimation of summarized dose-response data. Stata J 2006;6:40-57.
12. Lo YT, Chang YH, Wahlqvist ML, Huang HB, Lee MS. Spending on vegetable and fruit consumption could reduce all-cause mortality among older adults. Nutr J 2012;11:113.
13. Nechuta SJ, Shu XO, Li HL, Yang G, Xiang YB, Cai H, et al. Combined impact of lifestyle-related factors on total and cause-specific mortality among Chinese women: prospective cohort study. PLoS Med 2010;7:e1000339.
14. Zhang X, Shu XO, Xiang YB, Yang G, Li H, Gao J, et al. Cruciferous vegetable consumption is associated with a reduced risk of total and cardiovascular disease mortality. Am J Clin Nutr 2011;94:240-6.
15. Sahyoun NR, Jacques PF, Russell RM. Carotenoids, vitamins C and E, and mortality in an elderly population. Am J Epidemiol 1996;144:501-11.
16. Whiteman D, Muir J, Jones L, Murphy M, Key T. Dietary questions as determinants of mortality: the OXCHECK experience. Public Health Nutr 1999;2:477-87.
17. Xi B, Li S, Liu Z, Tian H, Yin X, Huai P, et al. Intake of fruit juice and incidence of type 2 diabetes: a systematic review and meta-analysis. PLoS One 2014;9:e93471.
18. Li S, Miao S, Huang Y, Liu Z, Tian H, Yin X, et al. Fruit intake decreases risk of incident type 2 diabetes: an updated meta-analysis. Endocrine 2014.
19. Doi SA. Re: Fruit and vegetable consumption and mortality from all causes. 2014. http://www.bmj.com/content/349/bmj.g4490/rr/761969.
20. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177-88.
21. Jackson D, White IR, Thompson SG. Extending DerSimonian and Laird's methodology to perform multivariate random effects meta-analyses. Stat Med 2010;29:1282-97.
22. Liu C. Fruit and vegetable consumption and mortality. 2014. http://www.bmj.com/content/349/bmj.g4490/rr/762303.
23. He FJ, Nowson CA, MacGregor GA. Fruit and vegetable consumption and stroke: meta-analysis of cohort studies. Lancet 2006;367:320-6.
24. Thompson SG, Higgins JP. How should meta-regression analyses be undertaken and interpreted? Stat Med 2002;21:1559-73.
25. Sterne JA, Gavaghan D, Egger M. Publication and related bias in meta-analysis: power of statistical tests and prevalence in the literature. J Clin Epidemiol 2000;53:1119-29.
26. Harbord RM, Egger M, Sterne JA. A modified test for small-study effects in meta-analyses of controlled trials with binary endpoints. Stat Med 2006;25:3443-57.
Rapid Response:
To the editor:
We appreciate very much the enormous interest from the globe in our recent study published in BMJ.1
We concur with Ellen M Storm2 that there were fewer studies evaluating the health benefits of fruit and vegetable consumption at the level of greater than 5 servings per day. Therefore, more studies are warranted in the future to confirm the association of fruit and vegetable consumption at levels of higher than 5 servings per day with risk of mortality. However, from public health point of view, the debate on the exact amount is not relevant at present, because the average consumption of fruit and vegetable is far below 5 servings per day in the general population.3 Consistent with our findings, the World Health Organization recommends 5 servings (400 g) of fruit and vegetables per day as the population goal for chronic disease prevention.4 Compared with the recommended level of fruit and vegetable consumption, the current average population level has much room to improve.
Both Liang Zhou5 and Zhihao Liu6 raised a concern of omitting a number of “eligible” studies in our meta-analysis. However, because Mann et al7 and Ness et al8 did not report data on person-years or total number of participants for each category of fruit and/or vegetable consumption, we were unable to include them in our dose-response meta-analysis using the generalized least squares trend estimation (GLST) method.9-11 The study by Lo et al12 was excluded because it reported daily expenditures on rather than consumption of fruit and vegetables.
Nechuta et al13 reported on the association between combined consumption of fruit and vegetable and mortality from the Shanghai Women’s Health Study. However, we did include a more recent article14 from the same cohort which examined the risk of mortality of fruit and vegetable separately. Including Nechuta et al13 in our meta-analysis did not materially alter our results or conclusions: the pooled hazard ratio (95% confidence interval) associated with each portion increase in fruit and vegetable consumption was 0.96 (0.94 to 0.97; P=0.001) for all-cause mortality. In addition, we still observed a threshold at around five servings daily consumption of fruit and vegetables, after which there was no further reduction in risk.
In our meta-analysis, two of the included studies reported fruit juice as part of fruit consumption.15 16 Liang Zhou5 argued that we should remove them from the analysis because there was evidence showing divergent associations of fruit and fruit juice consumption on type 2 diabetes.17 18 However, because there is no evidence showing any difference between fruit and fruit juice with regard to the risk of mortality, we do not believe there is sufficient rationale to exclude these two studies from our main analysis. Nonetheless, we conducted a sensitivity analysis by excluding these two studies and found almost identical results; the pooled hazard ratio of all-cause mortality was 0.94 (95% confidence interval 0.90 to 0.99; P=0.002) for each additional serving a day of fruit.
Suhail A. Doi19 questioned the random-effects models used in our meta-analysis, and suggested the use of the inverse variance heterogeneity (IVhet) and the quality effects (QE) models developed by the author. However, to our knowledge, the DerSimonian-Laird random-effects model20 21 is the standard method used in meta-analyses, and therefore it was used in our study.
We agree with Chibo Liu22 that the pooled HR associated with each unit increment of fruit and vegetable consumption may not be optimal in the context of curvilinear associations. Therefore, we presented both the dose-response curve plot and forest plot in our original paper.
We used the previously described definition23 for each serving of fruit (80 g) and vegetables (77 g), because these serving sizes were calculated on the basis of real weighting and averaging commonly consumed fruit and vegetables,23 rather than based on arbitrary numbers.
We conducted stratified analyses but did not perform a formal meta-regression to identify potential sources of heterogeneity because a robust conclusion from meta-regression requires a large number of included studies.24 Also, for the evaluation of publication bias, we were actually short of statistical power because of the limited number of included studies.25 In addition, the Egger’s test, which is based on regression method, may have a high false positive rate.25 26
Finally, we have been informed about a slight error in our paper. From one of the studies (Leenders et al, Am J Epidemiol 2013;178:590-602) included in the meta-analysis we inadvertently used data on fruit consumption and all cause mortality for women only rather than for the whole population. We have re-run the analyses using the correct hazard ratio for overall mortality of 1.00 (95% confidence interval 0.99 to 1.01) for 100 g/day fruit consumption. The revised pooled hazard ratio for all cause mortality should be 0.94 (0.89 to 0.98; P=0.006) for an increment of one serving of fruit a day, which is almost identical to that reported in the paper (0.94, 0.90 to 0.98; P=0.002) and hence does not affect the conclusions.
Wang Xia, Department of Maternal and Child Health Care, School of Public Health, Shandong University, Jinan, China
Wei Bao, Epidemiology Branch, Division of Intramural Population Health Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Rockville, MD 20852, USA
Frank B. Hu, Departments of Nutrition and Epidemiology, Harvard School of Public Health, Boston, MA 02115, USA
References
1. Wang X, Ouyang Y, Liu J, Zhu M, Zhao G, Bao W, et al. Fruit and vegetable consumption and mortality from all causes, cardiovascular disease, and cancer: systematic review and dose-response meta-analysis of prospective cohort studies. BMJ 2014;349:g4490.
2. Storm EM. Re: Fruit and vegetable consumption and mortality from all causes, cardiovascular disease, and cancer: systematic review and dose-response meta-analysis of prospective cohort studies. 2014. http://www.bmj.com/content/349/bmj.g4490/rr/761891.
3. Hall JN, Moore S, Harper SB, Lynch JW. Global variability in fruit and vegetable consumption. Am J Prev Med 2009;36:402-09 e5.
4. World Health Organization. Promoting fruit and vegetable consumption around the world. http://www.who.int/dietphysicalactivity/fruit/en/index2.html.
5. Zhou L. The omission of eligible publications may bias the conclusions. 2014. http://www.bmj.com/content/349/bmj.g4490/rr/762437.
6. Liu Z. Insufficient literature was included in the meta-analysis by Wang et al. 2014. http://www.bmj.com/content/349/bmj.g4490/rr/762200.
7. Mann JI, Appleby PN, Key TJ, Thorogood M. Dietary determinants of ischaemic heart disease in health conscious individuals. Heart 1997;78:450-5.
8. Ness AR, Maynard M, Frankel S, Smith GD, Frobisher C, Leary SD, et al. Diet in childhood and adult cardiovascular and all cause mortality: the Boyd Orr cohort. Heart 2005;91:894-8.
9. Berlin JA, Longnecker MP, Greenland S. Meta-analysis of epidemiologic dose-response data. Epidemiology 1993:218-28.
10. Greenland S, Longnecker MP. Methods for trend estimation from summarized dose-response data, with applications to meta-analysis. American journal of epidemiology 1992;135:1301-09.
11. Orsini N, Bellocco R, Greenland S. Generalized least squares for trend estimation of summarized dose-response data. Stata J 2006;6:40-57.
12. Lo YT, Chang YH, Wahlqvist ML, Huang HB, Lee MS. Spending on vegetable and fruit consumption could reduce all-cause mortality among older adults. Nutr J 2012;11:113.
13. Nechuta SJ, Shu XO, Li HL, Yang G, Xiang YB, Cai H, et al. Combined impact of lifestyle-related factors on total and cause-specific mortality among Chinese women: prospective cohort study. PLoS Med 2010;7:e1000339.
14. Zhang X, Shu XO, Xiang YB, Yang G, Li H, Gao J, et al. Cruciferous vegetable consumption is associated with a reduced risk of total and cardiovascular disease mortality. Am J Clin Nutr 2011;94:240-6.
15. Sahyoun NR, Jacques PF, Russell RM. Carotenoids, vitamins C and E, and mortality in an elderly population. Am J Epidemiol 1996;144:501-11.
16. Whiteman D, Muir J, Jones L, Murphy M, Key T. Dietary questions as determinants of mortality: the OXCHECK experience. Public Health Nutr 1999;2:477-87.
17. Xi B, Li S, Liu Z, Tian H, Yin X, Huai P, et al. Intake of fruit juice and incidence of type 2 diabetes: a systematic review and meta-analysis. PLoS One 2014;9:e93471.
18. Li S, Miao S, Huang Y, Liu Z, Tian H, Yin X, et al. Fruit intake decreases risk of incident type 2 diabetes: an updated meta-analysis. Endocrine 2014.
19. Doi SA. Re: Fruit and vegetable consumption and mortality from all causes. 2014. http://www.bmj.com/content/349/bmj.g4490/rr/761969.
20. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177-88.
21. Jackson D, White IR, Thompson SG. Extending DerSimonian and Laird's methodology to perform multivariate random effects meta-analyses. Stat Med 2010;29:1282-97.
22. Liu C. Fruit and vegetable consumption and mortality. 2014. http://www.bmj.com/content/349/bmj.g4490/rr/762303.
23. He FJ, Nowson CA, MacGregor GA. Fruit and vegetable consumption and stroke: meta-analysis of cohort studies. Lancet 2006;367:320-6.
24. Thompson SG, Higgins JP. How should meta-regression analyses be undertaken and interpreted? Stat Med 2002;21:1559-73.
25. Sterne JA, Gavaghan D, Egger M. Publication and related bias in meta-analysis: power of statistical tests and prevalence in the literature. J Clin Epidemiol 2000;53:1119-29.
26. Harbord RM, Egger M, Sterne JA. A modified test for small-study effects in meta-analyses of controlled trials with binary endpoints. Stat Med 2006;25:3443-57.
Competing interests: No competing interests