Potato intake and incidence of hypertension: A dose-response meta-analysis

Milad Nazarzadeh *; The collaboration center of meta-analysis research (ccMETA), Iranian Research Center on Healthy Aging, Sabzevar University of Medical Sciences, Sabzevar, Iran

Zeinab Bidel; The collaboration center of meta-analysis research (ccMETA), Iranian Research Center on Healthy Aging, Sabzevar University of Medical Sciences, Sabzevar, Iran

Alireza Mosavi Jarrahi; Faculty of Health Sciences, Simon Fraser University, BC, Canada - Department of Social Medicine, School of Medicine, Shahid Beheshti University of Medical Sciences, Tehran, Iran

* Corresponding author: Milad Nazarzadeh, The collaboration center of meta-analysis research (ccMETA), Iranian Research Center on Healthy Aging, Sabzevar University of Medical Sciences, Sabzevar, Iran; Mobile: +989113486561; Email:

Dear Editor,

We read with great interest the study of “Potato intake and incidence of hypertension: results from three prospective US cohort studies” by Borgi et al. [1]published in the BMJ. Despite high quality methodology and large sample size, taking into account certain statistical limitation in the paper, multivariate modeling of nonlinear dose-response through categorization of the exposure variableor entering quadratic dummy variable in the model may lead to bias inthe interpretation of the results [2]. One of the best alternative statistical methods is restricted cubic spline regression which is available in many statistical package [3]. The Borgi et al. used classic meta-analysis method to pool the adjusted hazard ratios (HRs) without considering potential bias due to non-linear dose response relationship between different types of potato serving and hypertension. Considering the potential for a non-liner dose response, we reanalyzed and used Greenland and Longnecke method of dose-response meta-analysis with unite of dose serving per week. The non-linearity of dose response was modeled using restricted cubic spline with three knots at the 25th, 50th, and 75th percentiles of potato consumption.The model demonstrated the non-linearity of the dose-response relationship for the all models (P-value of non-linearity < 0.001) (Figure 1). Our non-significant finding while generally support the same finding as the authors did for potato chips and “baked, boiled, or mashed” potato, however, we showed that the “baked, boiled, or mashed potatoes and French fries” (BBMFF)have a highly significant curvilinear dose-response relationship with hypertension (Figure 1). The non-linearity of dose response may weaken the conclusion made by authors and deserve serious attention.

References

1. Borgi L, Rimm EB, Willett WC, et al. Potato intake and incidence of hypertension: results from three prospective US cohort studies. BMJ (Clinical research ed) 2016;353:i2351 doi: 10.1136/bmj.i2351[published Online First: Epub Date]|.

2. Royston P, Altman DG, Sauerbrei W. Dichotomizing continuous predictors in multiple regression: a bad idea. Statistics in medicine 2006;25(1):127-41 doi: 10.1002/sim.2331[published Online First: Epub Date]|.

3. Orsini N, Greenland S. A procedure to tabulate and plot results after flexible modeling of a quantitative covariate. Stata Journal 2011;11(1):1-29

Figure 1. Non-linear dose-response analyses of different type of potato consumption and risk of hypertension in meta-analysis (P-value of non-linearity for all the four panel were < 0.001). HR: hazard ration; CI: 95% confidence interval