Cancer is widely recognized as one of the leading public health issues worldwide. According to the GLOBOCAN estimates, in 2018 there were 18.1 million new cases of cancer and it contributed to the death of 9.6 million people [1]. Despite a decreasing trend in cancer mortality observed in recent years, it is still the second most common cause of death world-wide, second to cardiovascular diseases (CVD) [2]. Taking into account trends of recent years, e.g., in Europe and the US, there seems to be a transition concerning the distribution of these two main causes of death. It is reasonable to speculate that cancer will replace CVD as the major cause of death in years to come. According to recent data provided by the PURE study group, this has already happened in a number of high- as well as middle-income countries in adults aged 35–70 years [2]. Until 2040, the global burden of neoplasms is going to rise by more than half [3]. Due to simultaneous improvements in diagnosis and treatment approaches, there will be a substantial increase in the number of cancer survivors as well [4].

Irrespective of site-specific details in the pathogenesis of tumors, up to more than 90% of cancers are considered to be attributable to modifiable risk factors such as tobacco smoking, excessive body weight, physical inactivity, alcohol consumption, infectious agents, environmental pollution, and suboptimal diet [5, 6]. The latter is made responsible for about 5–10% of total cancer cases [5, 7, 8].

According to the World Cancer Research Fund (WCRF), high consumption of fruits, vegetables and whole grains, as well as low intake of red and processed meat can lower cancer risk. As food items and nutrients are consumed in combination, dietary patterns have been successfully implemented as a tool to assess the additive or synergistic effect of food in nutritional epidemiology [9, 10].

With regard to prevention of non-communicable diseases, one of the most well-represented dietary patterns in literature is the Mediterranean diet (MedDiet) [11]. The MedDiet is a plant-based pattern characterized by high amounts of fruits, vegetables, nuts, legumes, fish, cereals including whole grains, and extra-virgin olive oil, at the same time reducing intake of red, processed meat, eggs and dairy [12]. An additional component is a moderate intake of red wine [12]. A large body of clinical and epidemiological studies have observed the protective effect of the MedDiet on cardiovascular disease, diabetes, obesity as well as cancer [13].

We previously conducted a systematic review and meta-analysis on the association between adherence to the MedDiet and risk of cancer, which was followed by two updates [14,15,16]. In the last update, we were able to pool data from 83 studies (including randomized controlled trials, cohort and case–control studies) showing an inverse association between the highest MedDiet adherence category and the risk of cancer mortality as well as incidence of breast, colorectal, gastric, liver, head and neck, and prostate cancer [16]. Although little time has passed, since then, we decided to synthesize the available data in another update due to the following reasons. Since the publication of the latest version of the review, the number of new reports from cohort and case–control studies has increased substantially [17,18,19]. Additionally, some of the new studies focus on cancer subtypes not previously included in our reports [20, 21]. Moreover, we wanted to expand our findings by assessment of the certainty of evidence, which is rarely evaluated in nutrition research evidence syntheses.

Therefore, the aim of this review was to enhance our previous findings on adherence to the MedDiet pattern and risk of cancer mortality, site-specific cancer and all-cause as well as cancer mortality among cancer survivors. Additionally, we aimed to assess the certainty of evidence for identified comparisons.


The protocol for previous versions of this review was published in PROSPERO International Prospective Register of Systematic Reviews (CRD42013004382). This update of the systematic review was planned and conducted according to the standards of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement [22].

Search strategy

Two electronic databases, PubMed (from August 2017 to April 2020) and Scopus (from January 2017 to April 2020), were searched with no limitations to publication language. Following search terms were adopted for PubMed and Scopus: (“Mediterranean diet” OR “Mediterranean” OR “diet” OR “dietary pattern” OR “dietary score” OR “dietary adherence”) AND (“cancer” OR “neoplasm” OR “neoplastic disease” OR “survivors” OR “recurrence”) AND (“prospective” OR “follow-up” OR “cohort” OR “longitudinal” OR “case–control”). References from identified articles, systematic reviews and meta-analyses were screened for potential eligibility.

Study selection

Two reviewers (J.M. and A.D.) independently evaluated the eligibility of studies with any disagreements resolved by discussion with the third reviewer (L.S.). In contrast to previous versions of this review, we expanded our analyses with all-cause mortality among cancer survivors. Studies were included if they fulfilled the following criteria: (1) randomized controlled trials (RCTs), prospective cohort, case–cohort, nested case–control, or case–control studies, (2) conducted in adult population (aged ≥ 18 years) which (3) assessed association between adherence to MedDiet and (4) risk of cancer mortality, site-specific cancer, all-cause, cancer mortality or cancer reoccurrence among cancer survivors. If several reports from a single study were available, the one with longer follow-up or a larger number of participants/cases was selected.

Data extraction

After completing selection of eligible studies, two reviewers (J.M. and L.S.) extracted the following data: (1) name of first author, (2) country, (3) study name, (4) study design, (5) outcome, (6) population size, (7) number of cases, (8) length of the study follow-up, (9) age at entry, (10) sex, (11) composition of the MedDiet score and its range, (12) adjustment set and (13) multivariable risk estimates (odds ratio (OR), risk ratio (RR) or hazard ratio (HR) comparing groups of highest and lowest adherence to MedDiet) with corresponding 95% confidence intervals (CI). If a study presented several risk estimates, the one with maximal adjustment was chosen. If separate results for men and women or different cancer subtypes were presented in a study, the estimates were pooled using a fixed-effects model.

Certainty of evidence assessment

To evaluate the certainty of evidence for associations between adherence to MedDiet and cancer outcomes in cohort studies and RCTs, the NutriGrade tool was adopted [23]. This tool is based on the following 9 items: (1) risk of bias, study quality, and study limitations (maximum 2 points for cohorts or 3 points for RCTs); (2) precision (maximum 1 point); (3) heterogeneity (maximum 1 point); (4) directness (maximum 1 point); (5) publication bias (maximum 1 point); (6) funding bias (maximum 1 point); (7) study design (+ 2 points—only for RCTs); (8) effect size (maximum 2 points—only for cohort studies); and (9) dose–response (maximum 1 point—only for cohort studies). Risk of bias domain was assessed using a checklist created by authors of the tool. Four categories based on the total score were used to interpret the certainty of evidence: very low (0 to < 4 points), low (4 to < 6 points), moderate (6 to < 8 points) and high (≥ 8 points).

Statistical analysis

The meta-analysis was conducted by pooling the multivariable-adjusted RRs, HRs or ORs of the highest compared with the lowest MedDiet adherence category using a random-effects model with the DerSimonian–Laird method [24]. Outcomes in the meta-analysis were assumed to be ORs in case-control studies and  RRs in prospective studies and RCTs. Using an inverse variance method, the standard error (SE) for the log-transformed OR/RR was calculated and interpreted as an estimated variance of log-transformed OR/RR to weight each study [24]. Included studies were categorized according to the following clinical outcomes: (1) cancer mortality, (2) biliary tract cancer, (3) bladder cancer, (4) blood cancer, (5) breast cancer, (6) colorectal cancer, (7) endometrial cancer, (8) esophageal cancer, (9) gallbladder cancer, (10) gastric cancer, (11) glioma, (12) head and neck cancer, (13) liver cancer, (14) ovarian cancer, (15) pancreatic cancer, (16) prostate cancer, (17) respiratory cancer, (18) skin cancer, (19) all-cause mortality, (20) cancer mortality, and (21) cancer reoccurrence among cancer survivors. Estimates from case–control, cohort studies and RCTs were compared separately. Joint estimates for observational studies were obtained by pooling together data from case–control and cohort studies in the same model. Additional analyses were conducted for associations between individual components of the MedDiet and overall cancer risk:

  • Alcohol (within the range vs. higher consumption)

  • Cereals (higher vs. lower consumption)

  • Dairy (lower vs. higher consumption)

  • Fish (higher vs. lower consumption)

  • Fruit (higher vs. lower consumption)

  • Legumes (higher vs. lower consumption)

  • Meat (lower vs. higher consumption)

  • Nuts (higher vs. lower consumption)

  • Olive oil (higher vs. lower consumption)

  • Vegetables (higher vs. lower consumption)

  • Whole grains (higher vs. lower consumption)

I2 statistic and Cochran’s Q test were used to evaluate the heterogeneity between studies. For the I2 value greater than 50% indicated a substantial statistical heterogeneity [25]. Subgroup analyses were conducted only for prospective cohort studies, for comparisons which included ≥ 10 studies and were stratified for sex (male/female), geographical location (Mediterranean/non-Mediterranean countries) and type of MedDiet score (Trichopoulou MedDiet score [12]/Fung MedDiet score [26]). For breast cancer, pooled risk estimates were additionally compared by menopausal status (premenopausal/postmenopausal) and receptor expression (ER/PR/HER/mixed). Furthermore, analysis for colorectal cancer risk was run separately for anatomical location (proximal colon/distal colon/rectum).

For comparisons with ≥ 10 studies, small-study effects, such as publication bias, were explored by funnel plots and Egger’s regression test, as recommended by Cochrane Collaboration [27]. All analyses were conducted in Review Manager version 5.3 (Nordic Cochrane Center, Copenhagen, Denmark) and R version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria) with the “metafor” package [28].


Database search and study characteristics

The updated literature search revealed 3720 publications after removal of duplicates from different databases. Additionally, 83 studies identified in previous versions of this systematic review were re-considered [29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111]. After title-abstract screening, 137 articles were assessed for eligibility and 20 articles were excluded at this step (ESM Table 1). Details of the study search and selection process were presented as a PRISMA-compliant flowchart in ESM Fig. 1.

Main characteristics of studies identified in the updated search are summarized in Table 1. Overall, 117 studies (with 12 case–control [112,113,114,115,116,117,118,119,120,121,122,123], 26 cohort [17,18,19,20,21, 124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144], five case–cohort [145,146,147,148,149], and one RCT (corrected report) [111] not identified in previous versions of this review) pooling 3,202,496 participants were included in the update [17,18,19,20,21, 29,30,31,32,33,34, 36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63, 65,66,67,68,69,70,71,72, 80,81,82,83,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98,99,100,101,102,103,104,105,106,107,108,109,110,111,112,113,114,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146,147,148,149].

Table 1 General characteristics of newly added case–control, case-cohort, cohort studies and randomized controlled trials identified in the updated literature search

Definitions of Mediterranean diet

The majority of 44 newly included studies assessed adherence to the MedDiet using predefined dietary scores. Two main definitions of MedDiet used in the included studies referred to scores by Trichopoulou [12] and Fung [26]. Fewer reports adopted scores by Sofi [150] and Buckland [55]. Differences between scores concerned mostly cut-off points for moderate alcohol consumption and a way of handling healthy fat intake. Five added studies derived MedDiet scores from principal component analysis [113, 114, 116, 122, 125].

Corresponding risk estimates were based on comparison of extreme quantiles (top quintile/quartile/tertile versus bottom) [17, 18, 113,114,115,116, 122, 124,125,126, 128,129,130,131,132,133,134,135,136,137,138,139,140], fixed cut-off points [20, 21, 112, 117,118,119,120, 123, 141,142,143,144,145,146,147,148,149], per standard deviation [121], per-tertile [127] or per-20 percentile increase in the MedDiet score [19]. Majority of studies used MedDiet scores evaluated in the baseline, whereas one study reported risk in the context of a 12-year change of adherence to the dietary pattern [19].

Main outcomes

According to the different clinical outcomes, risk of cancer mortality was evaluated in 18 cohort studies and one RCT (n = 71,145 cases); breast cancer risk in 12 cohort, one RCT (n = 35,373 incident cases) and 11 case–control studies (n = 10,615 prevalent cases); colorectal cancer risk in nine cohort, one case–cohort (n = 26,185 incident cases) and seven case–control studies (n = 9683 prevalent cases); prostate cancer risk in five cohort, one case–cohort (n = 36,006 incident cases) and five case–control studies (n = 2466 prevalent cases); respiratory cancer risk in four cohort and one case–cohort studies (n = 12,730 incident cases); gastric cancer risk in three cohort, one case–cohort (n = 2343 incident cases) and three case–control studies (n = 1517 prevalent cases); liver cancer risk in three cohort (n = 1274 incident cases) and one case–control study (n = 518 prevalent cases); bladder in three cohort (n = 5844 incident cases) and one case–control study (n = 690 prevalent cases); pancreatic cancer risk in two cohort, one case–cohort (n = 1436 incident cases) and one case–control study (n = 688 prevalent cases); blood cancer risk in two cohort (n = 3614 incident cases) and two case–control studies (n = 691 prevalent cases); esophageal cancer in one cohort, one case–cohort (n = 1181 incident cases) and one case–control study (n = 304 prevalent cases); head and neck in one cohort (n = 1868 incident cases) and eight case–control studies (n = 4601 prevalent cases); endometrial cancer in one cohort (n = 1392 incident cases) and three case–control studies (n = 2355 prevalent cases); biliary tract (n = 163 incident cases), gallbladder (n = 77 incident cases), ovarian (n = 696 incident cases), skin cancer (n = 1436 incident cases) and glioma risk (n = 2313 incident cases) in one cohort study, respectively. Among cancer survivors, eight cohort studies summarized all-cause mortality (n = 4883 cases), cancer-specific mortality in four cohort studies (n = 1790 cases), and cancer reoccurrence in one cohort study (n = 92 cases).

Pooled estimates from random-effects models are summarized in Table 2 and corresponding forest plots are presented in ESM Figs. 2–22. Highest versus lowest adherence to the MedDiet was associated with a lower risk of cancer mortality in cohort studies (RRcohort: 0.87, 95% CI 0.82–0.92; I2 = 83%), but not in one RCT (RRRCT: 0.75, 95% CI 0.17–3.33, I2 = NA). Among cancer survivors, there was no association between the adherence to the MedDiet and cancer mortality risk (RRcohort: 0.96, 95% CI 0.82–1.11; I2 = 0%); however, an inverse association was observed in relation to all-cause mortality (RRcohort: 0.75, 95% CI 0.66–0.86, I2 = 41%). An inverse association of breast cancer with highest adherence to the MedDiet was found in one RCT (RRRCT: 0.41, 95% CI 0.19–0.87, I2 = NA) and observational studies (RRobservational: 0.94, 95% CI 0.90–0.97, I2 = 31%). However, considering separate designs there was a risk reduction in case–control studies (ORcase–control: 0.87, 95% CI 0.82–0.93, I2 = 6%), but not in cohort studies (RRcohort: 0.97, 95% CI 0.94–1.00, I2 = 0%). Regarding colorectal cancer, the highest adherence to the MedDiet was linked to a reduced risk (RRobservational: 0.83, 95% CI 0.76–0.90, I2 = 82%; ORcase–control: 0.64, 95% CI 0.52–0.79, I2 = 89%, RRcohort: 0.92, 95% CI 0.87–0.99, I2 = 50%). Furthermore, inverse associations between adherence to the MedDiet and risk of head and neck (RRobservational: 0.56, 95% CI 0.44–0.72; I2 = 91%, ORcase–control: 0.54, 95% CI 0.40–0.72, I2 = 92%; RRcohort: 0.73, 95% CI 0.60–0.89, I2 = NA), bladder (RRobservational: 0.87, 95% CI 0.76–0.98; I2 = 38%, ORcase–control: 0.66, 95% CI 0.47–0.93, I2 = NA; RRcohort: 0.89, 95% CI 0.81–0.97, I2 = 11%), gastric (RRobservational: 0.70, 95% CI 0.61–0.80; I2 = 52%, ORcase–control: 0.63 95% CI 0.53–0.75, I2 = 34%; RRcohort: 0.77, 95% CI 0.64–0.92, I2 = 44%), liver (RRobservational: 0.64, 95% CI 0.54–0.75, I2 = 0%; ORcase–control: 0.51, 95% CI 0.34–0.77, I2 = NA; RRcohort: 0.67, 95% CI 0.56–0.80, I2 = 0%), and respiratory (RRcohort: 0.84, 95% CI 0.76–0.94; I2 = 42%) cancers were found, respectively. Consistently no effect of adhering to the MedDiet was observed in relation to blood (RRobservational: 0.94, 95% CI 0.88–1.02, I2 = 0%; ORcase–control: 0.89, 95% CI 0.68–1.18, I2 = 0%; RRcohort 0.95, 95% CI 0.88–1.02, I2 = 0%) and prostate cancer (RRobservational: 0.98, 95% CI 0.93–1.04, I2 = 39%; ORcase–control: 0.76, 95% CI 0.52–1.13, I2 = 6%; RRcohort 0.98, 95% CI 0.94–1.02, I2 = 28%). Additionally no associations were observed for endometrial, esophageal and pancreatic cancer in observational studies (RRobservational: 0.67, 95% CI 0.41–1.11, I2 = 91%; RRobservational: 0.64, 95% CI 0.35–1.16, I2 = 81%; RRobservational: 0.80, 95% CI 0.60–1.06, I2 = 79%, respectively) with contrary findings from cohort (RRcohort: 0.98, 95% CI 0.82–1.17, I2 = NA; RRcohort: 0.85, 95% CI 0.67–1.09, I2 = 0%; RRcohort: 0.92, 95% CI 0.81–1.05, I2 = 0%, respectively) and case–control studies (ORcase–control: 0.58, 95% CI 0.35–0.95, I2 = 77%; ORcase–control: 0.26, 95% CI 0.13–0.52, I2 = NA; ORcase–control: 0.48, 95% CI 0.35–0.66, I2 = NA, respectively).

Table 2 Pooled relative risk of cancer mortality, site-specific cancers and outcomes among cancer survivors for highest versus lowest adherence to Mediterranean dietary pattern

Subgroup analysis

None of the effect estimates was modified by the type of MedDiet score or geographical localization of study. Both menopausal status neither receptor expression pattern did not change the effect estimate for breast cancer. By specifying anatomical location of colorectal cancer, the general inverse association was re-established for distal colon and rectum (RR: 0.88, 95% CI 0.79 to 0.96, I2 = 0% and RR: 0.86, 95% CI 0.75–0.98, I2 = 42%, respectively), but not for proximal colon (RR: 1.01, 95% CI 0.93–1.09, I2 = 0%). The corresponding effect estimates are summarized in ESM Tables 2–4.

Components of the MedDiet and risk of cancer

Summary risk ratios for the components of the MedDiet score are presented in Fig. 1. We found an inverse association between alcohol consumption within the recommended range compared to higher consumption (RR: 0.92, 95% CI 0.87–0.97), whole grain intake (RR: 0.93, 95% CI 0.88–0.98), fruit intake (RR: 0.94, 95% CI 0.91–0.97) as well as vegetable intake (RR: 0.96, 95% CI 0.94–0.98) and overall cancer risk. No associations were identified for cereals, dairy, fish, legumes, meat, nuts, and olive oil.

Fig. 1
figure 1

Pooled risk ratios of individual Mediterranean diet components and overall cancer risk

Publication bias

The results of Egger’s linear regression test did not support the presence of publication bias for cancer mortality (P = 0.55), breast cancer (P = 0.94), and colorectal cancer (P = 0.74) following comparison between highest and lowest adherence to the MedDiet. Funnel plots were created for analyses including at least 10 studies. Visual inspection of the plots suggested low asymmetry for colorectal cancer, as well as moderate asymmetry for cancer mortality and breast cancer, implying that publication bias might be affecting these associations (ESM Figs. 23–25).

Certainty of evidence

Application of the NutriGrade tool to the results from cohort studies resulted in moderate certainty of evidence for cancer mortality and colorectal cancer risk. Low certainty of evidence was found for incidence of bladder, blood, breast, gastric, liver, prostate and respiratory cancer as well as all-cause mortality among survivors. In RCTs certainty of evidence for breast cancer and cancer mortality was low. The credibility of findings for remaining site-specific cancers, cancer mortality and reoccurrence among cancer survivors was rated as very low, suggesting very low confidence in effect estimates (Table 2, ESM Table 5).


In this updated systematic review, we meta-analysed current evidence on the association between adherence to MedDiet pattern and the risk of cancer. We identified 44 new studies, which provided data for an additional one million participants. The present analysis confirmed our previous findings on the inverse association of adherence to MedDiet on cancer mortality and colorectal cancer risk [16]. Contrary to our earlier reports, we observed conflicting findings between case–control and cohort studies for breast cancer [16]. Lack of association in cohort studies might suggest that significant findings found in case–control studies could be explained by bias linked to study design. Therefore, we cannot conclude on presence of inverse association between MedDiet and breast cancer risk. Our finding corresponds with statement from Continuous Updated Project by the WCRF suggesting too limited evidence to draw a conclusion on the relationship of healthy dietary patterns and breast cancer [10]. For the first time, we were able to observe an inverse association between adherence to the MedDiet and bladder, gastric and lung cancer incidence, as well as all-cause mortality among cancer survivors. Moreover, the present report included new cancer subtypes such as skin cancer and glioma, as well as identified new studies for those comparisons represented previously by a single study. The certainty of the evidence, evaluated for the first time in these series of reviews, was judged as “moderate” for cancer mortality and colorectal cancer and “very low” to “low” for other cancer subtypes. The NutriGrade scoring system did consider only meta-analyses of RCTs and cohort studies [23], but not case–control studies. Similarly, the evidence which was the basis for the 3rd WCRF report, considered only RCTs, cohort studies, and nested case–control studies. Individual case–control studies were not anymore considered due to limitation such as recall bias [10]. Considering the fact that in the current update we were able to identify only two RCTs with a very limited sample size, a major focus when interpreting our results should be put on findings from cohorts with a supportive role of case–control studies.

In 2014, Fardet and Rock referred to analyses of single nutrients or food groups as a reductionist approach not adequate in studies on the preventive effects of nutrition in chronic diseases such as cancer [151]. In contrast, dietary patterns may take into account synergistic and antagonistic interactions between the components of a food matrix, thus yielding a holistic net effect of diet [9]. Adherence to high-quality diets, such as the Healthy Eating Index (HEI) or the alternate HEI was inversely associated with cancer risk by approximately 15% [152]. In our analyses, the beneficial associations of the complete MedDiet on cancer was only reflected by inverse associations between overall cancer risk and fruit, vegetables, whole grains, and moderate alcohol intake, but not for fish, dairy and nuts (Fig. 1). Nevertheless, these observations may provide some insights to explain the mechanisms of action of MedDiet components/bioactive substances [153].

Fruits, vegetables, legumes and whole-grain products are a rich source of dietary fibre. Strong evidence from observational studies suggests a protective role of fibre mostly against colorectal cancer [10]. However, higher intake of dietary fibre was linked to a reduced risk of several other types of neoplasms including breast, gastric and lung cancer as well [154,155,156]. Gut microbiota reduces digested fibre to short-chain fatty acids such as butyrate, which helps to maintain proper function of the intestinal epithelium, as well as to reduce oncogenic potential by inducing cell apoptosis [157]. Some experimental studies demonstrated a direct interaction between fibre and pattern recognition receptors modulating immune anti-tumour response [158]. By increasing stool bulk, fibre can also dilute and slow absorption of potential carcinogens [158]. In addition, fruits and vegetables provide a variety of phytochemicals with potential anti-cancer effects. Bioactive substances such as carotenoids, flavonoids, stilbenes, coumarins and tannins can act synergically to increase antioxidative capacity and reduce cell oxidative damage [159, 160]. Furthermore, these compounds were shown to inhibit signal-transducing pathways, cell proliferation and oncogene expression, as well as to induce cell-cycle arrest [161]. A meta-analysis of prospective observational studies yielded an inverse association between antioxidative phytochemicals intake or their plasma/serum levels and risk of cancer [162]. Whole grains contain alk(en)ylresorcinols, benzoxanizoids and phytosteroids, which exerted an inhibitory effect on model human cancer cells [163]. Frequent consumption of whole grains was observed to lower risk of cancer mortality and incidence [164,165,166].

Apart from providing protective compounds, adherence to the MedDiet pattern decreases exposure to potential carcinogens by omitting intake of detrimental food items. Thus, extensive consumption of red and processed meat are associated with an increased risk of cancer, especially colorectal [165, 167]. Both food groups are a potential source of N-nitroso compounds, polycyclic aromatic hydrocarbons, and heterocyclic amines known to be cancerogenic [168,169,170]. A recent meta-analysis suggested that the above-mentioned chemicals are associated both with increased risks of colorectal and gastric cancers [171, 172].

Alcohol predominately in the form of red wine represents the most controversial ´food group´ within the context of the associations of the MedDiet on cancer. Increased ROS synthesis, suppressed anti-tumour immune response as well as metabolization of ethanol into DNA-damaging acetaldehyde may all explain the positive association between alcohol intake and cancer [125, 126]. The 3rd WCRF report indicated that there is a “convincing” grade of evidence for a positive association between alcohol intake and risk of upper aerodigestive, breast, colorectal or liver cancers, irrespective of the type of drink [10]. Definitions of moderate alcohol intake differ between the various MedDiet scores. According to Trichopoulou et al., consumption of up to 50 g/days for men and 25 g/days for women in form of red wine is considered as moderate [12], whereas Fung et al. [92] set cut-off points at 25 g/days and 15 g/days, respectively. Potential anti-tumorigenic effects of red wine are attributed to its polyphenolic content, especially resveratrol [173]. Although our results suggested a small reduction in overall cancer risk for alcohol intake within the range, compared to higher alcohol consumption, the benefit from light-to-moderate consumption of wine on cancer risk in observational studies is inconclusive [174,175,176]. Risk estimates for several cancers based on MedDiet scores including alcohol did not differ from those simply adjusted for total alcohol intake [122, 145,146,147,148,149]. Consumption of wine together with meals is a part of the cultural heritage in Mediterranean countries, but it is less common in other countries [177]. Therefore, the promotion of wine drinking in countries, where it is not a habit seems pointless, as small benefits do not exclude potential harm.

As already stated in the previous versions of this systematic review, a major limitation of our findings is the inconsistency of the definition of the MedDiet pattern [16]. Initially, the phrase was coined on the basis of observation made in several communities in the Mediterranean basin in the 1960s. Dietary intake has changed significantly since that time, which was stated in follow-up reports from the Seven Countries Study [178]. Therefore, MedDiet should rather be considered as a set of local variants based on cultural setting, food price and availability [177]. Consequently, dietary indices adopted in nutritional epidemiology as a means to quantify adherence to the MedDiet show substantial differences both in the composition of score as well as cut-off points for specific components. A recent umbrella review identified 74 different MedDiet scores used among studies eligible for systematic reviews and meta-analyses [13]. Popular definitions such as traditional MedDiet or alternate MedDiet indices use cut-off points based on median intake in studied populations [12, 26], which may substantially differ between studied populations. For example, median intake of vegetables for men in the Italian subcohort of the EPIC-InterAct study was 291 g/days, whereas the respective value in the Swedish subcohort was only 123 g/days [179]. A potential tool to address dissimilarities between MedDiet scores is the adoption of country-specific food environments [177]. For instance, olive oil, especially EVOO is rarely consumed in the US and northern Europe; therefore, MUFA-to-SFA ratio represents a more suitable measure of healthy fat intake [177]. However, little is known whether the use of these correction factors may result in equivalent preventive effects of MedDiet against cancer. Future studies need to focus on the use of literature-cased cut-off points for food groups as well as on the question whether different adaptations of MedDiet will yield comparable health-related outcomes.

Another limitation is the fact that pooled estimates presented in this review are based predominately on cohort studies set in Europe and the US, whereas single reports covered data from Asia. Uneven distribution of geographical locations might contribute to increased heterogeneity of data due to differences in cancer prevalence, genetic factors, or the burden of environmental risks, which can modify the effect of diet. However, stratifying analyses for study location did not affect the identified risk estimates in the present study.

Our results are based on observational rather than experimental data, which limits the interpretation of our findings with respect to causality. The use of randomized controlled trials in nutrition is limited by the inability to maintain high compliance during the long term of follow-up. Therefore, the use of data from prospective cohort studies is reasonable. To increase trust in our estimates, we did not consider case–control studies, which are prone to recall bias.

Most of the included studies constructed risk models using scores derived from food frequency data assessed during recruitment to study. Diet quality can substantially change during a long follow-up period. Thus, baseline adherence to the dietary pattern does not have to represent the true exposure. A particular strength of our systematic review is the large number of included studies as well as corresponding cancer cases. Another advantage of our analysis was the use of the NutriGrade tool. While assessing the certainty of evidence is key to construct evidence-based recommendations, it is rarely adopted in meta-analyses in nutrition research. To our best knowledge, this systematic review represents the first summary of trustworthiness of associations between adherence to the MedDiet and risk of cancer.


In conclusion, this systematic review and meta-analysis provides an updated body of evidence on the association between adherence to the MedDiet and risk of cancer. Our results suggest that highest adherence to the MedDiet was inversely associated with risk of cancer mortality in the general population, and all-cause mortality among cancer survivors as well as colorectal, head and neck, respiratory, gastric, liver and bladder cancer risks. However, the very low to moderate certainty of evidence found in this update requires a conservative interpretation of our findings.