Effects of Intake of Apples, Pears, or Their Products on Cardiometabolic Risk Factors and Clinical Outcomes: A Systematic Review and Meta-Analysis

ABSTRACT Apples and pears contain nutrients that have been linked to cardiovascular health. We conducted a systematic review and meta-analysis to summarize related research. Medline, Cochrane Central, and Commonwealth Agricultural Bureau databases were searched for publications on apple or pear intake and cardiovascular disease (CVD)/ cardiometabolic disease (CMD). Studies in adults (healthy or at risk for CVD) that quantified apple or pear intake were included. Random-effects models meta-analysis was used when ≥3 studies reported the same outcome. In total, 22 studies were eligible including 7 randomized controlled trial, 1 nonrandomized trial, and 14 prospective observational studies. In RCTs, apple intake significantly decreased BMI, but made no difference in body weight, serum lipids, blood glucose, or blood pressure. In observational studies, apple or pear intake significantly decreased risk of cerebrovascular disease, cardiovascular death, type 2 diabetes mellitus, and all-cause mortality. No association was reported for cerebral infarction or intracerebral hemorrhage. In conclusion, apple or pear intake significantly decreased BMI and risk for CVD outcomes.


Introduction
Fruit intake is associated with a decreased risk of cardiovascular disease (CVD) events and risk factors in several epidemiological studies (1)(2)(3)(4)(5). The Dietary Guidelines for Americans 2015-2020 report recommends a high intake of fruit, as part of a healthy eating pattern for the prevention of chronic disease (6). Few Americans eat adequate servings of fruit to adhere to this recommendation; however, apples are the most consumed and the fourth least expensive type of fresh fruit in the United States (7)(8)(9). A better understanding of the health effects of apple intake as well as intake of fruits with similar nutrient content, such as pears, aside from total fruit, on CVD could be useful in informing federal nutrition guidance regarding fruit, considering apples' popularity and accessibility.
CVD risk factors and events are highly prevalent in the United States, and the health and financial burdens of this disease, warrant the investigation of prevention through diet (10). Each year, nearly 1 in every 6 US healthcare dollars is spent on treatment for CVD (11). Direct healthcare costs attributable to CVD were $193 billion and the associated costs due to productivity loss were $123.5 billion in 2012 (11). The health benefits associated with fruit consumption could result in considerable cost savings, through possible reductions of medications, invasive interventions, and lost productivity.
Apples and pears contain several bioactive compounds, including flavonoids, dietary fiber, and antioxidants, that have been individually associated with decreased risk for CVD risk factors and events (12)(13)(14)(15). Intakes of dietary fiber and antioxidants from fruits have been found to be significantly associated with a decreased risk for CVD (14)(15)(16)(17). In an analysis of the Cancer Prevention Study II, men and women with the highest intake of flavonoids had an 18% risk reduction for CVD mortality (12).
While the cardiovascular health benefits of the bioactive compounds found in apples and pears are widely recognized, the effects of intakes of whole apples and pears and their products has remained somewhat inconclusive. Several studies have found significant CVD risk reductions due to apple or pear intake (16,(18)(19)(20)(21), but not all studies (22). A recent systematic review and meta-analysis of observational studies indicated a protective relationship between apple and pear intake and risk of type 2 diabetes mellitus (T2DM), a highly prevalent and costly risk factor for CVD (23). To the best of our knowledge, a systematic review and meta-analysis on apple and pear intake and CVD has not been previously conducted, and this study aims to fill this gap in the literature.

Methods
We conducted a systematic review of published literature evaluating the effects of apple intake on CVD and cardiometabolic disease (CMD) risk factors and events. We developed a single causal pathway, or analytical framework, depicting the potential association between intake of apples, pears, and their products and CVD/CMD risk factors and outcomes to guide our review (Supplemental Figure 1). The systematic review results were reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis statement (24).

Data sources and study eligibility
We conducted a comprehensive literature search in MEDLINE, Cochrane Central, and Commonwealth Agricultural Bureau abstracts from 1946 through August 2019 for publications that measured apple and/or pear intake and CVD/CMD clinical outcomes and risk factors in adults (Supplemental Table 1). No language restriction was applied during searches. Citations were screened in duplicate using the predefined study eligibility criteria and discrepancies were resolved by consensus in group conferences.

Study inclusion criteria.
We included prospective cohorts and intervention trials conducted in adults ≥18 y of age that quantified the amount of apple or pear intake. We included studies that examined pear or combined apple and pear intake because of similarities in nutrient content between apples and pears and because apple and pear intake is often combined in analysis of nutrition data (51). Studies were eligible if the population was either healthy or had CVD/CMD risk factors (i.e., hypertension, hyperlipidemia, metabolic syndrome, or diabetes) at baseline. Studies with no apple or pear intake or low intake of apples or pears as comparators were accepted. The clinical outcomes of interest included any CVD, acute coronary syndrome, myocardial infarction, ischemic heart disease, stroke, coronary artery disease, atrial fibrillation, all-cause mortality, cardiovascular mortality, cerebrovascular disease, cerebral infarction, and heart failure. CMD risk factors of interest included metabolic syndrome, T2DM, blood glucose, hypertension, systolic blood pressure (SBP), diastolic blood pressure (DBP), hypercholesterolemia, blood lipids [total cholesterol, high density lipoprotein (HDL) cholesterol; low density lipoprotein (LDL) cholesterol , verylow density lipoprotein (VLDL) cholesterol , triglycerides, LDL:HDL cholesterol , and total:HDL cholesterol ), body weight, BMI, and waistto-hip ratio.

Study exclusion criteria.
We excluded studies that evaluated apple pectin or apple pomace. We also excluded studies in children and pregnant women. The following study designs were excluded: retrospective, cross-sectional, case reports, and single-arm (interventions with no control group), mixed intervention, pharmacokinetic, in vitro, and cell-culture studies. In addition to the above common eligibility criteria, we established additional criteria specific to the study design.

Eligibility criteria for intervention trials.
In our analysis of intervention trials, we included studies with known doses of apples, pears, and their products. The minimum intervention duration was at least 1 wk for recorded blood glucose levels and at least 3 wk for other risk factor outcomes. Studies with <5 subjects per arm were excluded.

Eligibility criteria for cohort studies.
In our analysis of cohort studies, we included studies with reported intake of apple, apple products, pear, pear products, or combined apple and pear. Studies that reported multivariable results adjusted for any potential confounders were eligible. Studies with at least 6 mo of followup time were included.

Data extraction and quality assessment
Data from each study were extracted independently by 1 of 6 investigators and reviewed and confirmed by ≥1 other team member. Any conflicts regarding extraction were resolved in team discussions. The extracted data table included study design, intervention description, population characteristics, methods for controlling for potential confounders or effect modifiers, outcomes, and results depicting associations between apple or pear intake and the specific outcome(s) of interest.
We assessed the methodologic quality of each study based on predefined criteria, in accordance with the Agency for Healthcare Research and Quality's suggested methods for systematic reviews (25). Study quality was determined in duplicate and discrepancies were resolved by consensus in group discussion. We applied risk of bias items in the modified Newcastle-Ottawa Scale (26) for observational studies, the Cochrane risk of bias for clinical trials (27), and nutrition-specific items for a critical appraisal of micronutrient systematic reviews for both clinical trials and observational studies (28).

Data synthesis
Analyses were conducted separately for intervention trials and observational studies. We performed random-effects model meta-analyses when similar data from ≥3 studies were available (29). For intervention trials, we combined net differences [net change = (apple intake final − apple intake initial ) − (control final − control initial )] for continuous outcomes. We tested between-study heterogeneity with the Q statistic (significant when P < 0.10) and quantified its extent with I 2 . I 2 values of 25%, 50%, and 75% were considered low, moderate, or high heterogeneity, respectively (30).
For observational studies, we synthesized RRs comparing the extreme categories of apple or pear intake (highest compared with lowest, as defined within each study) provided that the categories corresponded to similar doses of intake across studies. We performed sensitivity analyses when there were studies reporting various similar doses of apple or pear intake (including a comparison of medium intake to low intake) or if they reported data for subgroups, such as sex. All statistical analyses were conducted in Stata version 14 (StataCorp). Forest plots were created using R version 3.3 (R Core Team).
For studies that reported quantities by serving size, we considered a serving equivalent to 1 medium-sized apple or pear. For a study that reported apple intake as catechin content, we used the USDA Food Composition Database (51) to obtain the total catechin composition of a medium-sized apple and back calculated the amount of apple consumed for each quartile.
All included studies that could not be combined quantitatively in meta-analyses were summarized in narrative form and in tables that tabulated the important features of the study populations, design, intervention, outcomes, and results. Summary tables were organized by outcomes of interest.

Results
Database searches identified 1834 citations. Full-text articles of 66 citations that were accepted in abstract screening were retrieved and reviewed against eligibility criteria. Full-text screening yielded 28 eligible articles (Figure 1). Of 28 eligible articles, 11 reported intervention trial data, and 17 reported prospective cohort data. No foreign language publications met the inclusion criteria.

Intervention trials
Eight intervention trials (7 RCTs and 1 nonrandomized trial) in 11 publications that enrolled 596 participants were eligible. Trials were conducted in Brazil (n = 1), Denmark (n = 1), Germany (n = 1), Iran (n = 1), Ireland (n = 1), Norway (n = 1), and the United States (n = 1), and for 1 trial the location was not reported. The duration of the trials ranged from 4 to 20 wk. Two trials included only men, 3 included only women, and 3 trials included both men and women. Of 7 RCTs, 5 were parallel-arm randomized trials, and 2 were crossover trials. Included participants in RCTs varied considerably; they were healthy (31,32), nondiabetic (33), or had increased risk of CVD/CMD (19,32,34,35). The apple or pear interventions and their comparators varied across RCTs. Apple or pear intake ranged from 75 to 900 g/d, and comparators included no apple or pear, control beverages with equivalent amounts of calories and fructose, dried plums, oat cookies, and kiwis. In addition to 12-wk time points, long-term data at 6 and 12 mo were reported for 1 RCT. Meta-analyses of RCTs were conducted for the following outcomes: body weight 36)], BMI ], HDL cholesterol [4 RCTs (19,31,33,36)], total cholesterol (TC) [5 RCTs (19,[32][33][34][35]], LDL cholesterol [4 RCTs (19,31,33,36)], and triglycerides (TG) [5 RCTs (19,31,33,36)]. No metaanalysis was conducted for SBP, DBP, waist circumference, glucose, or waist:hip, LDL:HDL cholesterol, total cholesterol:HDL cholesterol, or glucose:insulin ratios, as these outcomes were reported in <3 RCTs. Baseline details for intervention trials can be found in Table 1, metaanalysis results are tabulated in Table 2, the risk of bias of each study is listed in Supplemental Table 2, and meta-analysis results of trials reporting the effect of apple in percentage net change are tabulated in Supplemental Table 3.

Body weight-related outcomes Body mass index.
Three RCTs with a total of 229 participants (31)(32)(33) reported the effect of apple intake on BMI. Meta-analysis of the 3 trials comparing apple with a variety of controls found a significant decrease in BMI (summary net change:−0.39; 95% CI: −0.59, −0.20) without heterogeneity (I 2 0.0%; P = 0.934) (Figure 2).

Waist-to-hip ratio.
One crossover RCT conducted in Denmark in 23 healthy men and women (36) measured the effect of apple intake on waist-to-hip ratio and found no difference between groups. A separate crossover RCT conducted among 43 adults at risk for CVD reported the effect of pear intake on waist-to-hip ratio (35). Waist-to-hip ratio in the group   consuming pears was significantly reduced compared to that of the control group at both 6-and 12-wk time points.
In the single nonrandomized trial of 70 men that used a time point of 16 wk, there was a significantly lower total cholesterol in the apple intake group versus the control group (37).
In the single nonrandomized trial of 70 men that compared consumption of 2 apples per d with consumption of 3 apples per wk for 16 wk found a significant increase in HDL cholesterol with apple intake compared with control (38). Triglycerides. Five RCTs (19,(31)(32)(33)36) reported the association between apple intake and TG. A meta-analysis of RCTs with a total of 324 subjects compared apple intake with a heterogeneous control group and found no difference in TG (net change: 8.91 mg/dL; 95% CI: −10.32, 28.14; I 2 = 59.1%, P = 0.044). Sensitivity analysis using 6-mo (net change: 9.66 mg/dL; 95% CI: −9.05, 28.36; I 2 = 54.0%) and 12-mo  Other serum lipid outcomes. One parallel RCT (19) reported the effect of apple intake for 8 wk on VLDL cholesterol in 46 hyperlipidemic, overweight men. This RCT found that there was no difference in VLDL cholesterol between apple and control groups. Two RCTs (19,31) that reported on the effect of apple intake on LDL cholesterol: HDL cholesterol found no significant difference between groups.
Total cholesterol: HDL cholesterol. Two RCTs (31, 36) that included a total of 206 healthy participants found no difference in total cholesterol: HDL cholesterol between apple intake compared with control groups.
Apolipoprotein B. One RCT (19) that included 46 hyperlipidemic men measured the effect of apple intake on apolipoprotein B. There was no significant difference in apolipoprotein B concentrations between the apple and control groups.
Another RCT (36), which included 23 healthy men and women, found a statistically significant reduction for both SBP (net change: One crossover RCT (35) reported the effect of pear intake on SBP. Among 43 adults at risk for CVD, SBP tended to be reduced at 12 wk in the pear intake group while no changes were observed in the control group. A second publication reporting results for the same RCT (38) reported the effect of pear intake on SBP and DBP. Among 36 adults at risk for CVD, SBP was significantly lower than baseline levels among the pear group, while no changes were observed for the control group. No changes were observed in DBP for either group.
Glucose metabolism outcomes. Three RCTs (32,36,39) with a total of 236 participants reported the effect of apple intake on insulin levels and found no difference in insulin levels between groups. A meta-analysis was not possible because 1 trial only reported qualitative results (39). Two RCTs (32,39) that reported the effect of apple intake on blood glucose levels and 1 RCT (32) that reported on glucose:insulin found no difference between groups.
Cohort studies. A total of 14 cohorts were evaluated in 17 publications and the following outcomes were examined: acute coronary syndrome [1 cohort (20) (50)]. Studies were conducted in the United States (n = 5), Denmark (n = 1), Finland (n = 1), Sweden (n = 3), The Netherlands (n = 2), Australia (n = 1), and the United States/China (n = 1). Three cohorts reported results for at ≥1 outcome stratified by sex. Six cohorts included only women, 2 included only men, and 7 included both men and women. Three publications reported combined results for Nurses' Health Study (NHS), Nurses' Health Study II (NHS II), and Health Professionals Follow-Up Study (HPFS) and stratified their results by cohort. Nine of 14 cohorts reported results for apple and pear intake and the remaining 5 cohorts reported only apple intake. Studies varied in the length of follow-up and in the amount of apple or pear intake (high, medium, or low intake). Follow-up ranged from 4 to 28 y. High apple or pear intake ranged from to 35.2 to more than 100 g/d, medium apple or pear intake ranged from 0 to 100 g/d, and low apple or pear intake ranged from 0 to 35.2 g/d. Baseline details for studies can be found in Table 3, the meta-analysis results are tabulated in Table 4, and risk of bias is listed in Supplemental Table 4.
Acute coronary syndrome. The Danish Diet, Cancer and Health Study cohort (20) followed 55,338 men and women between the ages of 50 and 64 for a median of 7.7 y and reported a significantly decreased risk of acute coronary syndrome (ACS) [incidence rate ratio (IRR): 0.78, 95% CI: 0.64, 0.95] in men with high apple intake. However, no association was observed in women (IRR: 0.93, 95% CI: 0.67, 1.29). Additional analyses found that for every 25 g/d increase in apple intake, men had a decreased risk of ACS (IRR: 0.97, 95% CI: 0.94, 0.99), but women did not (IRR: 0.97, 95% CI: 0.93, 1.01).
Meta-analyses were conducted using the random-effects model. I 2 is an indicator of between-comparison heterogeneity.

FIGURE 4
Meta-analysis of observation studies reporting the effect of apple consumption dose on cerebrovascular disease.
Cerebrovascular disease. Four studies (17,21,40,41) reported an association between apple intake and cerebrovascular disease, including total and acute stroke. All studies included both men and women, and only 1 study (17) stratified results by men and women (reporting 2 data points).
Meta-analysis of 3 studies [with a total of 139,507 participants (17,21,41)] found no difference in risk of intracerebral hemorrhage (summary adjusted RR: 0.92; 95% CI: 0.76, 1.12; I 2 = 0.0%) in subjects at the highest compared with subjects at the lowest quantile of apple intake.
All-cause mortality. Meta-analysis of 3 studies with a total of 55,625 participants (18,46,48) found a significant decrease in all-cause mortality (summary adjusted RR: 0.85; 95% CI: 0.77, 0.92) with low heterogeneity (I 2 = 0.0%; P = 0.49) when comparing subjects with the highest intake with subjects with the lowest intake of apple ( Figure 5). Death was ascertained using data from the Central Statistical Office of Finland (46), the Swedish Bureau of Statistics and Cause of Death Registry (48), or the Western Australian Mortality Database (18).

Discussion
This systematic review has demonstrated significant reduction in the risk of adverse cardiovascular outcomes associated with apple intake in cohort studies and for the outcome of BMI in RCTs. To the best of our knowledge, this is the first meta-analysis of the impact of apple intake on a comprehensive list of CVD/CMD risk factors and outcomes in cohort studies and in RCTs. Our review found that higher intake of apple, pear, or combined apple and pear was associated with a significant decrease in BMI in RCTs, and a significant decrease in the risk of cerebrovascular disease, CVD/CMD mortality, diabetes, and all-cause mortality in observational studies. There was no difference between apple/pear and control groups for the outcomes of body weight, HDL cholesterol, LDL cholesterol, total cholesterol, and TG in meta-analysis of RCTs, or the risk of cerebral infarction and intracerebral hemorrhage in meta-analysis of cohort studies. While the meta-analysis included between 3 and 5 studies, the number of included participants varied considerably across RCTs and cohort studies. In RCTs, the number of included participants ranged between 175 and 393, and in cohort studies the number of participants ranged between 55,625 and 284,323. Further, individual studies did not find a significant change in waistto-hip ratio (n = 1), apolipoprotein b (n = 1), total cholesterol: HDL cholesterol (n = 2), glucose: insulin (n = 1), insulin (n = 3), and VLDL cholesterol (n = 1). Studies had conflicting results for LDL cholesterol:HDL cholesterol (n = 2), glucose (n = 2), SBP (n = 2), DBP (n = 2), and ACS (n = 1).
The protective association observed in this study between apple intake and diabetes is consistent with observations made in a previous systematic review and meta-analysis (23). Though similar, our estimation of the effect of apple intake on diabetes was stronger than that of the previous review (RR of 0.77 compared to RR of 0.82). Unique to this study, we also conducted a sensitivity analysis on cohort studies that only included women and separate meta-analyses of RCTs (23). Our findings are consistent with those of another review that observed protective associations between increased fruit intake and heart disease outcomes (51). However, our review focused solely on apple, pear, or combined apple and pear exposure, while other investigators reviewed studies with a variety of fruit exposures. To the best of our knowledge, our findings of the effect of apple intake on other CVD/CMD risk factors were not addressed in previous systematic reviews and meta-analyses.
The significant association that we demonstrate in this review between apple and pear intake and cardiovascular outcomes in cohort studies is consistent with studies demonstrating the health benefits of apples, possibly due to the presence of phytochemicals like flavonoids, anthocyanins and other antioxidants (13). Apples and pears are a rich source of flavonoids, which have documented cardiovascular benefits (12). The significant association observed between BMI reduction and apple intake in the RCTs is consistent with the results of the metaanalyses of cohort studies in that reduction in BMI correlates strongly with improved cardiovascular outcomes (52). However, we did not observe significant associations among other cardiovascular risk factors with apple intake, including serum lipid concentrations.

FIGURE 6
Meta-analysis of observational studies reporting the effect of apple and pear consumption dose on on type 2 diabetes.
In the results reported above, we combined apple and pear intake, because of similarities in nutrient content between apples and pears (53). We examined the validity of this approach in both the RCTs and the cohort studies. In the RCTs, only 1 study that was included in meta-analysis reported results for apple and pear as one of the arms (34). Repeating the meta-analyses for the apple-only subgroup showed a similar trend for the remaining 6 studies that were included in this meta-analysis (Supplemental Table 6). Nine out of 14 cohort studies grouped apple and pear intake together into 1 exposure category. When we conducted a subgroup analysis of the 5 studies that studied only apple intake, meta-analyses could be performed only for T2DM incidence and all-cause mortality. For the other outcomes, there were not enough studies available for apple-only analyses. For the 2 outcomes mentioned, the results had similar trends (Supplemental Table 5) compared to the results obtained when including studies where apple and pear intake was combined.
In the meta-analyses reported for cohort studies above, we combined studies regardless of whether the studies controlled for diet or not. However, when we performed a subgroup analysis using only those studies that controlled for diet, the results showed trend similar to that seen when all studies were included (Supplemental Table 7).
The limitations of this review are largely reflective of the quality of primary studies as well as the consistency of data collection and analysis methods in individual studies. First, many of the included observational studies used data collection and outcome assessment methods that are prone to biases, including self-report for dietary and outcome assessment as well as record linkage for outcome assessment (54). Second, the populations examined in some of the included cohort studies were restricted to nurses or other health professionals who may not be representative of the average individual, potentially limiting the generalizability of the results. There is also a potential for a substitution effect in that subjects who consumed more apples may also have a diet that may substantially vary from other subjects who do not consume apples or other plant-based foods. We observed that in the cohort studies, we have done an analysis controlling for diet where we found the trends to be similar even after controlling for diet. In the randomized clinical trials, the assumption was made that the diets in the various arms were equivalent. Nevertheless, this substitution effect may be a limitation of this meta-analysis. While the majority of cohort studies controlled for diet and lifestyle variables in their analyses, the effect of dietary patterns or overall fruit intake may have confounded the associations from the few studies which did not control for these variables. Lastly, some RCTs failed to describe randomization techniques, varied in the types of controls used to compare the effects of apple intake, and failed to provide sufficient information regarding participants' compliance with the intervention.

Conclusion
Except for a significant decrease in BMI, meta-analyses of apple or pear intake in RCTs did not show a statistically significant change in risk factors for CVD/CMD. Observational studies demonstrated an association between apple, pear, or combined apple and pear intake and a decreased risk of acute coronary disease, cerebrovascular disease, CVD mortality, diabetes, and all-cause mortality suggesting that apple intake may be of benefit in CVD/CMD prevention.