Article Text

Original research
Fried-food consumption and risk of cardiovascular disease and all-cause mortality: a meta-analysis of observational studies
Free
  1. Pei Qin1,2,
  2. Ming Zhang1,2,
  3. Minghui Han3,
  4. Dechen Liu3,
  5. Xinping Luo1,
  6. Lidan Xu4,
  7. Yunhong Zeng5,
  8. Qing Chen6,
  9. Tieqiang Wang7,
  10. Xiaoliang Chen7,
  11. Qionggui Zhou1,
  12. Quanman Li3,
  13. Ranran Qie3,
  14. Xiaoyan Wu1,
  15. Yang Li1,
  16. Yanyan Zhang1,
  17. Yuying Wu1,
  18. Dongsheng Hu1,2,
  19. Fulan Hu1,2
  1. 1Department of Biostatistics and Epidemiology, Shenzhen University Health Science Center, Shenzhen, Guangdong, China
  2. 2Guangdong Provincial Key Laboratory of Regional Immunity and Diseases, Department of Pathology, Shenzhen University School of Medicine, Shenzhen, China
  3. 3Department of Epidemiology and Health Statistics, Zhengzhou University, Zhengzhou, Henan, China
  4. 4Department of Nutrition, The Second Affilicated Hospital of Shenzhen University, Shenzhen, Guangdong, China
  5. 5Department of Health Management, University of Chinese Academy of Sciences Shenzhen Hospital, Shenzhen, China
  6. 6Department of Mental Health, Bao'an Chronic Diseases Prevent and Cure Hospital, Shenzhen, China
  7. 7Department of Infectious Disease Control and Prevention, Shenzhen Guangming District Center for Disease Control and Prevention, Shenzhen, China
  1. Correspondence to Fulan Hu, Shenzhen University Health Science Center, Shenzhen 518060, China; hufu1525{at}163.com

Abstract

Objective We performed a meta-analysis, including dose–response analysis, to quantitatively determine the association of fried-food consumption and risk of cardiovascular disease and all-cause mortality in the general adult population.

Methods We searched PubMed, EMBASE and Web of Science for all articles before 11 April 2020. Random-effects models were used to estimate the summary relative risks (RRs) and 95% CIs.

Results In comparing the highest with lowest fried-food intake, summary RRs (95% CIs) were 1.28 (1.15 to 1.43; n=17, I2=82.0%) for major cardiovascular events (prospective: 1.24 (1.12 to 1.38), n=13, I2=75.7%; case–control: 1.91 (1.15 to 3.17), n=4, I2=92.1%); 1.22 (1.07 to 1.40; n=11, I2=77.9%) for coronary heart disease (prospective: 1.16 (1.05 to 1.29), n=8, I2=44.6%; case–control: 1.91 (1.05 to 3.47), n=3, I2=93.9%); 1.37 (0.97 to 1.94; n=4, I2=80.7%) for stroke (cohort: 1.21 (0.87 to 1.69), n=3, I2=77.3%; case–control: 2.01 (1.27 to 3.19), n=1); 1.37 (1.07 to 1.75; n=4, I2=80.0%) for heart failure; 1.02 (0.93 to 1.14; n=3, I2=27.3%) for cardiovascular mortality; and 1.03 (95% CI 0.96 to 1.12; n=6, I2=38.0%) for all-cause mortality. The association was linear for major cardiovascular events, coronary heart disease and heart failure.

Conclusions Fried-food consumption may increase the risk of cardiovascular disease and presents a linear dose–response relation. However, the high heterogeneity and potential recall and misclassification biases for fried-food consumption from the original studies should be considered.

  • epidemiology
  • meta-analysis

Data availability statement

Data are available upon reasonable request.

Statistics from Altmetric.com

Introduction

Cardiovascular disease (CVD) remains the leading cause of mortality worldwide.1 Dietary patterns are key factors for CVD and mortality. Fried foods are popular worldwide and constitute one part of the Western dietary pattern. Although increasing studies have reported the adverse effects of the Western dietary pattern,2 it remains unclear whether fried-food consumption increases the risk of CVD and mortality.

Frying can modify the nutrient composition such as dehydration of foods and generation of trans-fatty acids.3 4 Also, frying can increase the energy of foods and make foods more appetising, leading to excess intake of energy-dense foods. Concerns about the effect of fried foods on cardiovascular health are justified. Several epidemiological studies showed fried-food consumption associated with increased risk of cardiovascular risk factors such as obesity,5 type 2 diabetes mellitus6 and hypertension.7 However, inconsistent findings have been reported on the association between fried-food consumption and CVD, with some studies suggesting a positive association8 9 and others showing no clear association.6 10 Few studies have explored the relation between fried-food consumption and mortality and shown contradictory results, with some studies showing increased risk11 12 and others showing no association.10 13–15 Moreover, the dose–response association of fried-food consumption and CVD and all-cause mortality remains unknown, although the risk estimates have been reported by different frequencies of fried-food consumption in most studies.

Clarifying the controversy around the association between fried-food consumption and CVD and mortality is important to provide evidence for dietary advice to reduce the risk. Therefore, we conducted a meta-analysis of observational studies to explore the association between fried-food intake and risk of CVD and all-cause mortality in the adult general population and quantify the dose–response relation.

Methods

This meta-analysis was registered in PROSPERO (CRD42020184509) and reported according to the Meta-analysis Of Observational Studies in Epidemiology group.16

Literature search strategy, study selection, data extraction and quality assessment

Details of search strategy, inclusion criteria, data extraction and quality assessment are provided in online supplemental file 1. Systematic literature review search terms and strategy are mentioned in online supplemental table 1.

Data synthesis and analysis

Summary relative risks (RRs) and 95% CIs were estimated by the random-effects model.17 We used generalised least squares regression to estimate study-specific dose–response estimates for each additional serving/week in fried-food intake18 and random-effects model to pool the estimates.17 Restricted cubic splines, with three knots at the 10th, 50th and 90th percentiles of the distribution, were used to examine the linear/non-linear dose–response association. Pnon-linearity was calculated by testing the null hypothesis that the coefficient of the second spline is equal to 0.19 Details of data transformation and calculation, heterogeneity, subgroup analysis, sensitivity analysis, meta-regression analysis and publication bias are mentioned in online supplemental file 2. All analyses were performed with Stata V.12.1 (StataCorp, College Station, Texas, USA).

Results

Literature search

Figure 1 summarises the study selection. Nineteen articles were included in the meta-analysis.

Figure 1

Flow chart of study selection for the meta-analysis. RR, relative risk.

Study characteristics

The characteristics of included studies are mentioned in table 1 and online supplemental table 2. A total of 17 studies (4 case–control,20–23 12 cohort6 8–11 14 24–28 and 1 nested case–control29) were included in the meta-analysis of CVD (562 445 participants and 36 727 major cardiovascular events (MCEs)). Six cohort studies10–15 investigated the association with all-cause mortality (754 873 participants and 85 906 deaths), with a median follow-up of 9.5 years. Online supplemental table 3 shows the results of quality assessment. The mean Newcastle-Ottawa Quality Assessment Scale (NOS) score was 6.33 for cohort and 6.60 for case–control studies on MCEs and 7.00 on all-cause mortality (all cohort). About half of the included studies had good quality and the others had fair quality. Most studies were comparable in different cohorts and had a sufficient follow-up, except that one cohort study had a 3-year follow-up.9

Table 1

Characteristics of studies included in meta-analysis of associations of fried-food intake with risk of cardiovascular disease and mortality

Table 2

Subgroup analyses of major cardiovascular disease events and all-cause mortality for the highest versus lowest category of fried food intake

Fried-food consumption and MCEs

For MCEs,6 8–11 14 20–29 summary RR (95% CI) was 1.28 (1.15 to 1.43) for the highest versus lowest category, with high heterogeneity (I2=82.0%, Pheterogeneity <0.001; figure 2). Publication bias was detected by Egger’s test (p<0.001) and visual inspection of funnel plots (online supplemental figure 1). After applying the trim-and-fill method, the association remained significant (RR 1.16; 95% CI 1.03 to 1.31).

Figure 2

Forest plot for the highest versus lowest fried food intake and risk of major cardiovascular disease events, coronary heart disease, stroke, heart failure, cardiovascular mortality and all-cause mortality. RR, relative risk. The study by Cahill et al reported two separate cohort studies (the Nurses' Health Study and the Health Professionals Follow-Up Study).

In the subgroup analysis, the positive association with MCEs remained consistent in most subgroups (table 2). We found evidence of heterogeneity among subgroups stratified by number of cases (p=0.032) and adjustment for BMI (p=0.048) and physical activity (p=0.005), with a stronger association for studies with <1000 cases and without adjustment for BMI or physical activity (table 2). On sensitivity analysis, the RRs remained consistent (online supplemental table 10).

In the dose–response analysis, the association was linear (Pnon-linearity=0.158) (figure 3) and the risk increased by 3% (RR 1.03; 95% CI 1.01 to 1.04; I2=75.6%, Pheterogeneity <0.001) for each additional fried-food serving/week, based on 13 studies (figure 4). Publication bias was detected by Egger’s test (p<0.001). Results on subgroup analysis and sensitivity analysis were similar to the highest versus lowest analysis (online supplemental tables 4 and 11).

Figure 3

Dose–response association of fried food intake and risk of major cardiovascular disease events, coronary heart disease, stroke, heart failure, cardiovascular mortality and all-cause mortality modelled by restricted cubic splines.

Figure 4

Forest plot for fried food intake (per one additional serving/week) and risk of major cardiovascular disease events, coronary heart disease, stroke, heart failure, cardiovascular mortality and all-cause mortality. RR, relative risk. The study by Cahill et al reported two separate cohort studies (the Nurses' Health Study and the Health Professionals Follow-Up Study).

Fried-food consumption and coronary heart disease

Eleven studies (three case–control,20–22 seven cohort6 9 10 14 24 25 and one nested case–control29) investigated the association between fried-food intake and CHD risk (26 252 cases and 361 173 participants). Summary RR was 1.22 (95% CI 1.07 to 1.40; I2=77.9%, Pheterogeneity <0.001) for the highest versus lowest category (figure 2). Publication bias was detected by Egger’s test (p<0.001) and visual inspection of funnel plots (online supplemental figure 1). After applying the trim-and-fill method, the association became non-significant (RR 1.15; 95% CI 0.97 to 1.36).

On subgroup analyses, the findings remained consistent across most subgroups (online supplemental table 5). We found significant heterogeneity between subgroups on adjustment for physical activity (p=0.017), with a stronger association among studies without adjustment for physical activity. Sensitivity analysis showed consistent association (online supplemental table 10).

In the dose–response analysis, the association was linear (Pnon-linearity=0.212, n=7 studies6 9 10 14 22 25) (figure 3) and the risk increased by 2% (RR 1.02; 95% CI 1.01 to 1.02; I2=0, Pheterogeneity=0.726) for each additional fried-food serving/week (figure 4). No publication bias was detected by Egger’s test (p=0.314). Results on subgroup analysis and sensitivity analysis were similar to the highest versus lowest analysis (online supplemental tables 5 and 11).

Fried-food consumption and stroke

Four studies (three cohort14 25 26 and one case–control23) (10 103 cases and 74 388 participants) investigated the association with risk of stroke. Summary RR was 1.37 (95% CI 0.97 to 1.94; I2=80.7%, Pheterogeneity=0.001) for the highest versus lowest category (figure 2). No evidence of publication bias was shown by Egger’s test (p=0.092).

The results were unstable for the subgroup analyses (online supplemental table 6). On sensitivity analysis, excluding one study at a time except for the study by Larsson and Wolk,25 the results remained consistent, with a non-significant association (online supplemental table 10). When removing the study by Larsson et al,25 the association became significant (RR 1.57; 95% CI 1.18 to 2.07) (online supplemental table 10).

Despite evidence of a potential linear relation by restricted cubic spine (Pnon-linearity=0.565, n=3 studies14 25 26) (figure 3), the risk with each additional fried-food serving/week was non-significant (RR 1.13; 95% CI 0.95 to 1.34; I2=74.6%, Pheterogeneity=0.020) (figure 4). Publication bias was not evident by Egger’s test (p=0.079). Results on subgroup analysis and sensitivity analysis were similar to the highest versus lowest analysis (online supplemental tables 6 and 11).

Fried-food consumption and heart failure

Four cohort studies8 25 27 28 (6085 cases and 158 544 participants) investigated the association with risk of heart failure (figure 2). The summary RR was 1.37 (95% CI 1.07 to 1.75; I2=80.0%, Pheterogeneity=0.002) for the highest versus lowest category. No evidence of publication bias was found by Egger’s test (p=0.176).

The results were unstable on subgroup analyses of heart failure (online supplemental table 7). Sensitivity analysis, excluding one study at a time except for the study by Larsson et al,25 resulted in a non-significant association (online supplemental table 10).

In the dose–response analysis, the association was linear (Pnon-linearity=0.266, n=4 studies8 25 27 28) (figure 3) and the summary RR was 1.12 (95% CI 1.01 to 1.23) for each additional fried-food serving/week, with considerable heterogeneity (I2=91.6%, Pheterogeneity <0.001) (figure 4). The publication bias was not significant by Egger’s test (p=0.067). Results on subgroup analysis and sensitivity analysis were similar to the highest versus lowest analysis (online supplemental tables 7 and 11).

Fried-food consumption and cardiovascular mortality

Three cohort studies11 14 25 (176 279 participants and 4294 deaths) investigated the association with risk of cardiovascular mortality (figure 2). The summary RR was 1.03 (95% CI 0.93 to 1.14; I2=27.3%, Pheterogeneity=0.253) for the highest versus lowest category. No evidence of publication bias was found by Egger’s test (p=0.923).

The results remained consistent across most subgroups (online supplemental table 8). Sensitivity analysis showed similar results (online supplemental table 10).

Although the restricted cubic spine showed a potential linear trend (Pnon-linearity=0.161, n=3 studies11 14 25) (figure 3), the risk for each additional fried-food serving/week was non-significant (RR 1.00; 95% CI 0.99 to 1.01; I2=0, Pheterogeneity=0.517) (figure 4). No publication bias was detected by Egger’s test (p=0.242). Results on subgroup analysis and sensitivity analysis were similar to the highest versus lowest analysis (online supplemental tables 8 and 11).

Fried-food consumption and all-cause mortality

The summary RR for all-cause mortality was 1.03 (95% CI 0.96 to 1.12) for the highest versus lowest intake category, with no evidence of heterogeneity (I2=38.0%; Pheterogeneity=0.153; figure 2). No publication bias was detected by Egger’s test (p=0.656).

We found a non-significant association across most subgroups, but a positive association in cohort studies with follow-up ≥10 years (RR 1.05; 95% CI 1.00 to 1.11) (table 2). On sensitivity analysis, the RRs remained consistent (online supplemental table 10).

Although the restricted cubic spine presented a potential linear trend (Pnon-linearity=0.951; n=5 studies10–12 14 15) (figure 3), the risk for each additional fried-food serving/week was non-significant (RR 1.00; 95% CI 0.97 to 1.02; I2=76.3%, Pheterogeneity=0.002) (figure 4). No publication bias was detected by Egger’s test (p=0.988). Results on subgroup analysis and sensitivity analysis were similar to the highest versus lowest analysis (online supplemental tables 9 and 11).

Discussion

In our meta-analysis, we quantitatively assessed the associations between fried-food consumption and risk of CVD and all-cause mortality by comparing the highest with lowest consumption categories and analysing a linear/non-linear dose–response association. With high consumption of fried food, risk of MCEs increased 28%, CHD 22% and heart failure 37%, with non-significant risk of stroke, cardiovascular mortality or all-cause mortality. The associations persisted across most subgroups stratified by various study and participant characteristics. Furthermore, we found a linear association between fried-food consumption and MCEs, CHD, heart failure and the risk per each additional serving/week (one serving equals 114 g) significantly increased 3%, 2% and 12%, respectively.

Our findings provide further evidence that fried-food consumption is associated with MCEs and CHD and the risk increases by 2%~3% per each additional serving/week, with subgroup and sensitivity analyses showing consistent findings. Our meta-analysis included several studies that only investigated one type of fried-food consumption (ie, fried fish, fried potato or fried snack) but not the total fried food, which may underestimate the association. The subgroup analysis showed increased risk of CVD with consumption of total fried food and fried fish, which suggests that the adverse effect of fried-food consumption on CVD may not differ by fried-food type. However, we did not find a significant association for other fried-food types, which was estimated by pooling the risk estimates of one study reporting a non-significant association between fried potato intake and cardiovascular events25 and another study reporting more than onefold increased risk of acute stroke with fried snack intake.23 The reason for the non-significant association may be the limited number of studies, so more studies are needed to investigate the association of different types of fried foods with CVD. Also, no study has investigated the relation of total fried-food intake and stroke, and only one study investigated cardiovascular mortality,11 which may be another reason for the non-significant association. Therefore, current evidence supports the positive association between fried-food intake and MCEs and CHD, but further investigation is needed to examine the association with heart failure, stroke and cardiovascular mortality.

When comparing the highest with lowest fried-food intake, we found moderate heterogeneity among studies for MCEs. Meta-regression revealed that number of cases and adjustment for BMI and physical activity may contribute to the observed heterogeneity, but there was still substantial heterogeneity within subgroups, which indicates the existence of other sources of heterogeneity. Nevertheless, sensitivity analyses excluding one study at a time did not alter the pooled results. Egger’s tests and funnel plots revealed publication bias, but the application of the trim-and-fill method did not change the effect size, suggesting that the positive association of fried-food consumption and MCEs observed was not affected by the publication bias. Similarly, we found moderate heterogeneity for studies of CHD but no heterogeneity in subgroups stratified by physical activity, so adjustment for physical activity was the main source of heterogeneity. Furthermore, although we observed no heterogeneity for studies of stroke, heart failure and cardiovascular mortality, the small number of studies investigating stroke and cardiovascular mortality may lead to unstable risk estimates.

Findings from studies of fried-food consumption and all-cause mortality have been inconsistent, and our meta-analysis observed a non-significant association. However, the available evidence for all-cause mortality is limited, leading to reduced statistical power, non-significant summary risk estimates and unstable results in subgroup analyses. Another explanation may be that only two studies10 11 focused on total fried-food consumption and all-cause mortality and other studies11 13–15 explored the individual fried-food types. Of note, restricting studies with follow-up >10 years yielded a positive association between fried-food intake and all-cause mortality with no heterogeneity. The finding may suggest that it may take long observation period to observe a positive association between fried-food consumption and all-cause mortality, and future cohort studies exploring the association should have a long follow-up.

The mechanisms for the association between fried-food consumption and CVD are not well understood. First, fried food typically contains high amounts of dietary fat and leads to excess energy intake, which may increase the risk of CVD.30 The Southern Community Cohort Study (SUN)31 32 showed a 37% and 10% increased risk of general obesity and central adiposity, respectively, when comparing fried-food consumption ≥4 and ≤2 times/week. Second, a higher intake of trans-fatty acids that generated from hydrogenated vegetable oils due to frying has been found associated with increased risk of CVD.33 34 Third, frying can increase levels of cholesterol oxidation products35 and dietary advanced glycation endproducts,36 which are involved in the inflammatory response and oxidative stress in CVD.37 38 Fourth, fried food such as fried chicken and French fries are processed as restaurant foods that are usually high in added sodium, and high sodium intake increased CVD risk.39 Moreover, people who consume fried food frequently may drink sugar-sweetened beverages and consume other high-energy fast foods, which in turn causes obesity, diabetes and CVD.40–42

This meta-analysis revealed linear dose–response findings except for the highest versus lowest intake-category meta-analysis of the association of fried-food consumption with CVD and all-cause mortality, which can help to quantify the associations and test the shape of these possible associations. To our knowledge, no meta-analysis has explored the association, although contradictory findings have been reported by observational studies. Furthermore, especially for MCEs, CHD and all-cause mortality, the meta-analysis involved a large sample size of participants and cases/deaths and had a sufficiently long follow-up period.

There are some limitations noted. First, the small number of studies for stroke, heart failure, cardiovascular mortality and all-cause mortality reduced the statistical power to detect an association, destabilised the associations on sensitivity analyses and limited the explanations of subgroup analyses. Second, some unmeasured or imperfectly measured confounders may imply residual confounding. Third, fried-food consumption was self-reported in most studies, which may imply recall bias and misclassification of exposure. Measurement error is difficult to avoid in observational studies when measuring fried-food intake, which have different types and large variations. Fifth, no study defined the serving size of total fried food but a standard serving was assumed based on only three popular fried-food types, which may overestimate or underestimate the empirical intake levels. Given that the types of fried food consumed may vary by geographical areas, ethnic population and other factors, future studies could report the corresponding serving size so that we can better understand the dose–response relation.

Our study provided evidence for the adverse effects of consuming fried food on CVD and can be useful for dietary guidelines. WHO suggested limiting fried-food consumption to reduce the amount of total fat intake and industrially produced trans-fatty acid intake for a healthy diet.43 However, no dietary guideline is approved for the specific effect of fried-food consumption on CVD. The association between fried-food consumption and risk of stroke, heart failure, cardiovascular mortality and all-cause mortality has not been firmly established, so more studies are warranted to investigate the association.

Conclusions

Our meta-analysis indicates that fried-food consumption is associated with increased risk of CVD. The findings may support public health recommendations to control fried-food intake for preventing CVD. However, the high heterogeneity, and potential recall and misclassification biases on fried-food consumption from the original studies should be considered when interpreting the findings of this meta-analysis. More research is needed to determine the association of fried-food intake and risk of stroke, heart failure, cardiovascular mortality and all-cause mortality.

Key messages

What is already known on this subject?

  • Increasing number of studies has explored the association of fried-food consumption and risk of cardiovascular events and all-cause mortality; however, the association remains contradictory and dose–response association is unknown.

What might this study add?

  • In this meta-analysis, fried-food consumption was significantly associated with increased risk of major cardiovascular events, coronary heart disease and heart failure.

  • A linear association was observed for the relation of fried-food consumption and major cardiovascular events, coronary heart disease and heart failure.

How might this impact on clinical practice?

  • These findings may have important implications for guideline recommendations regarding the adverse effects of fried-food intake on cardiovascular disease.

Data availability statement

Data are available upon reasonable request.

Ethics statements

Acknowledgments

The authors would like to thank Laura Smales (BioMedEditing) for proof-reading the manuscript.

References

Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

Footnotes

  • PQ and MZ contributed equally.

  • Contributors PQ, FH, MZ and DH designed research; PQ and MH conducted the meta-analysis and drafted the manuscript; PQ and MH analysed the data; DL, XL, LX, YZ, QC, TW, XC, HL, QZ, QL, RQ, XW, YL, YZ, YW, FH, MZ and DH revised the manuscript. FH had primary responsibility for final content. All gave final approval and agreed to be accountable for all aspects of work ensuring integrity and accuracy.

  • Funding This work was supported by the National Natural Science Foundation of China (grant numbers 81373074, 81402752 and 81673260); the Natural Science Foundation of Guangdong Province (grant number 2017A030313452); the Medical Research Foundation of Guangdong Province (grant number A2017181) and the Science and Technology Development Foundation of Shenzhen (grant numbers CYJ20140418091413562, JCYJ20160307155707264, JCYJ20170412110537191 and JCYJ20170302143855721).

  • Competing interests None declared.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.