Annually, more than half of the deaths are due to CVD and cancer worldwide(1,Reference Dehghan, Mente and Rangarajan2) . Healthy diet plays a prominent and profound role in these conditions. Yogurt, a dairy product, made from the fermentation of lactic acid in milk by Lactobacillus bulgaricus and Streptococcus thermophilusis, is among the most widely consumed fermented foods(Reference Kok and Hutkins3). Due to the close contribution of gut microbiota to human health and the effect of yogurt consumption on gut microbiota, its consumption might affect health outcomes. Moreover, yogurt provides energy and beneficial compounds such as proteins, minerals, multivitamins and conjugated linoleic acid that may jointly favour long-term health(Reference Fernandez and Marette4). However, it is also a source of saturated fats and added sugar, which is presumed to increase the risk of CVD and mortality(Reference Dehghan, Mente and Rangarajan2,Reference Ding, Li and Qi5) .
Several epidemiological studies have previously examined the association between yogurt consumption and risk of chronic diseases(Reference Astrup6). Regular consumption of yogurt has been associated with a lower risk of hypertension(Reference Rice7), type 2 diabetes(Reference Chen, Sun and Giovannucci8) and certain cancers(Reference Wu, Zeng and Huang9,Reference Sapkota, Hsu and Zaridze10) . However, findings on the contribution of yogurt consumption to longevity are limited and conflicting(Reference Godos, Tieri and Ghelfi11). While some prospective studies have indicated an inverse association between yogurt intake and overall mortality(Reference Farvid, Malekshah and Pourshams12,Reference Schmid, Song and Zhang13) , some others have failed to reach such findings(Reference Praagman, Franco and Ikram14,Reference Michels, Willett and Vaidya15) . Moreover, different associations were reported between yogurt consumption and mortality outcomes in men and women(Reference Schmid, Song and Zhang13,Reference Goldbohm, Chorus and Galindo Garre16) . In an earlier meta-analysis of prospective cohort studies, high yogurt intake was not related to a lower risk of all-cause, CVD and cancer mortality(Reference Gao, Jia and Chen17). However, this review had several limitations. For example, it missed several large and some smaller studies(Reference Praagman, Franco and Ikram14,Reference Goldbohm, Chorus and Galindo Garre16,Reference Park, Mitrou and Kipnis18–Reference Sakauchi, Khan and Mori22) . In addition, two ineligible studies were included in the analysis(Reference Bongard, Arveiler and Dallongeville23,Reference Dik, Murphy and Siersema24) . The exposure in one of these studies was a combination of dairy products not just yogurt(Reference Bongard, Arveiler and Dallongeville23), and in the other study, patients with cancer were included(Reference Dik, Murphy and Siersema24). Furthermore, several studies have since been published(Reference Schmid, Song and Zhang13,Reference Pala, Sieri and Chiodini25–Reference Nakanishi, Homma and Osaki28) . Given the above-mentioned points, a comprehensive review and meta-analysis is required to quantitatively summarise earlier studies in this regard. The present study was therefore performed to comprehensively examine the association between the contribution of yogurt consumption to the risk of mortality from all causes, CVD and cancer based on earlier publications and to do a dose–response meta-analysis of cohort studies in this regard.
Methods
The present study was performed based on Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines(Reference Liberati, Altman and Tetzlaff29).
Search strategy
All relevant studies were identified by searching PubMed/Medline, ISI Web of Science and Scopus databases, without any language restriction, from inception to August 2022. Two independent authors (SN and FN) developed and performed the literature search. The following keywords were used in our search strategy: ((‘dairy products‘(Mesh) OR ‘dairy products’(tiab) OR ‘dairy’(tiab) OR ‘yogurt’(Mesh) OR ‘yogurt’(tiab) OR ‘yoghurt” (tiab)) AND (‘mortality’(tiab) OR ‘death’(tiab) OR ‘fatal’(tiab) OR ‘survival’(tiab) OR ‘mortality’(Subheading) OR ‘neoplasm’(tiab) OR ‘cancer survivor’(tiab) OR ‘cardiovascular disease’(tiab) OR ‘coronary disease’(tiab) OR ‘myocardial ischemia’(tiab) OR ‘coronary artery disease’(tiab) OR ‘myocardial infarction’(tiab) OR ‘stroke’(tiab) OR ‘mortality’(Mesh) OR ‘death’(Mesh) OR ‘neoplasms’(Mesh) OR ‘cancer survivors’(Mesh) OR ‘cardiovascular diseases’(Mesh) OR ‘coronary disease’(Mesh) OR ‘myocardial ischemia’(Mesh) OR ‘coronary artery disease’(Mesh) OR ‘myocardial infarction’(Mesh) OR ‘stroke’(Mesh))). In addition, the bibliographies of the retrieved articles and previous systematic and narrative reviews were also scanned manually to identify potential publications.
Inclusion criteria
Two independent authors (SN and FN) screened the title and abstract of publications found in the systematic search to identify eligible studies. The publications were considered for inclusion in this meta-analysis if they (1) were cohort studies (2) reported hazard ratios (HR) or relative risks (RR) and their corresponding 95 % CI for the association between yogurt consumption and mortality from all causes, CVD and cancer. If multiple papers were published using the dataset of a single cohort, we included the one with the most comprehensive information.
Exclusion criteria
We excluded ecologic studies, non-original research papers (reviews, letters, editorials or commentaries) and meta-analyses. We also did not include unpublished studies as well as those studies where individuals were included based on the existence of a specific disease at baseline. In addition, if a study reported the risk estimates for yogurt consumption in combination with other dairy foods, it was not included in our analysis.
Data extraction
Using a standardised form, pairs of authors (HT and SN) independently reviewed the title and abstract of all articles and extracted the required data. In case of any disagreements, consensus was reached. For each study, we recorded the following information: the first author’s name, year of publication, study location, mean age or age range, sex, sample size, length of follow-up, type of outcome, number of deaths from all causes, CVD and cancer, dietary assessment method, the amount of yogurt consumption in each category, adjustment for confounding variables in multivariable analysis as well as RR and 95 % CI of mortality for each category of yogurt consumption. We extracted RR with the most adjusted model. For one study that reported multivariable models with and without additional adjustment for dietary Ca intake, we selected the multivariable model without adjustment for dietary Ca intake.
Risk of bias assessment
Quality assessments were performed in duplicate by two independent reviewers (SN and FN). The Newcastle–Ottawa scale was used to assess the quality of each study(Reference Margulis, Pladevall and Riera-Guardia30). The total score of Newcastle–Ottawa scale ranges between 0 and 9. Studies achieving nine points were considered to provide the highest quality.
Data synthesis and analysis
Reported RR and HR and their 95 % CI for comparison of the highest v. the lowest categories of yogurt consumption were applied to estimate log RR and HR ± se. If a study reported risk estimates per standard deviation (sd) or per unit increment in yogurt consumption, the following method was used to translate per sd or per unit increment risk estimate to the top compared with the bottom categories of population baseline distribution of yogurt values. Briefly, we calculated the difference between the means/medians of the highest and lowest categories of yogurt consumption in other publications included in the analysis. Then, the mean difference between the mean/median of the highest and lowest categories was calculated. Finally, per-sd or per unit increase risk estimate was transformed to per ‘calculated mean/median difference’ and was included in the analysis. When the exact amount of sd was not presented in the published paper, we assumed a scaling factor of 2·18 times the log risk estimate for a 1-sd increase in yogurt consumption(Reference Danesh, Collins and Appleby31). The analyses were conducted using a random-effects meta-analysis. Cochran’s Q test and I2 were used to estimate between-study heterogeneity. A value of I 2 more than 50 % was considered as substantial heterogeneity(Reference Naghshi, Sadeghi and Willett32). When a study provided results by sex, we first pooled these estimates using a fixed-effects model and included the pooled value in the main analysis, but sex-specific results were presented separately in subgroup analyses. For studies that reported results separately for CHD and stroke mortality or cancer subtypes, we used the method developed by Hamling et al(Reference Hamling, Lee and Weitkunat33) to combine the risk estimates, and the obtained pooled risk estimate was included in the main meta-analysis. For studies that did not provide the information required to apply the Hamling method, we used a fixed-effects model to pool the results. Yogurt intake was converted into serving/d based on standard portion sizes developed by the US Department of Agriculture (i.e. one serving = 244 g)(34).
In case of between-study heterogeneity, subgroup analysis was done based on study location, sex, follow-up duration, methods used for assessing dietary intakes and adjustment for confounding factors (BMI and energy intake) to find potential sources of heterogeneity. The possibility of publication bias was examined using the Egger’s and Begg’s tests as well as visual inspection of funnel plots(Reference Duval and Tweedie35). The sensitivity analysis was conducted to find the effect of any specific study on the pooled effect size.
A method suggested by Greenland and Longnecker(Reference Greenland and Longnecker36) and Orsini et al(Reference Orsini, Bellocco and Greenland37) was used to examine dose–response analysis. We computed study-specific slopes (linear trends) and 95 % CI from the natural logarithm of the RR, or HR, and their corresponding CI across categories of yogurt consumption. In this method, information on the number of deaths, number of participants and the effect sizes with the variance estimates for ≥ 3 quantitative categories of exposure was required. We used the median or mean amount of each category. For studies that reported the range of yogurt consumption rather than mean, the midpoint of the upper and lower limits was used to determine the amount of yogurt consumption. If the highest and lowest categories of yogurt consumption were open-ended, the width of the adjacent interval was used to calculate an upper or lower cut-off value.
A two-stage generalised least-squares trend estimation method was applied to assess a linear dose–response association of each additional serving of yogurt consumption with mortality risk. In this method, the study-specific slope lines are estimated and combined with a random-effects model to obtain an overall average slope(Reference Orsini, Bellocco and Greenland37). To assess a non-linear dose–response association, restricted cubic splines with three knots at fixed percentiles of 10, 50 and 90 % of the distribution were applied(Reference Greenland and Longnecker36). First, a restricted cubic spline model with a generalised least-squares trend estimation method was calculated after taking into account the correlation within each set of reported RR/HR. Second, all the study-specific estimates were combined with the use of the restricted maximum likelihood method in a multivariate random-effects meta-analysis(Reference Orsini, Bellocco and Greenland37,Reference Jackson, White and Thompson38) . A probability value for non-linearity was evaluated by null hypothesis testing in which the coefficient of the second spline was considered equal to zero. All statistical tests were performed using the Stata software version 14 (Stata Corp. College Station, TX). For all tests, P-values less than 0·05 were considered statistically significant.
Results
Literature search
As presented in Fig. 1, our search initially identified 16 392 publications, of which 14 638 were considered after removal of duplicate articles. Of these, 14 608 articles were again excluded, because they were irrelevant to the study objective. Finally, the full text of thirty papers were assessed, of them twelve publications were excluded: one reported risk estimates for theoretical substitutions between dairy products and mortality outcomes(Reference Laursen, Thomsen and Beck39), two was performed on colorectal cancer(Reference Dik, Murphy and Siersema24) and myocardial infarction patients(Reference Cruijsen, Jacobo Cejudo and Küpers40), three were conducted on the same population with the same outcome(Reference Ding, Li and Qi5,Reference Michels, Willett and Vaidya15,Reference Mazidi, Mikhailidis and Sattar41) , five studies had reported the risk estimates for combination of dairy products, not yogurt separately(Reference Bongard, Arveiler and Dallongeville23,Reference Dalmeijer, Struijk and van der Schouw42–Reference Silva, Giatti and Diniz45) , and one study had insufficient data(Reference Louie, Flood and Burlutsky46). Finally, we considered data from seventeen cohort studies (eighteen publications)(Reference Dehghan, Mente and Rangarajan2,Reference Farvid, Malekshah and Pourshams12–Reference Praagman, Franco and Ikram14,Reference Goldbohm, Chorus and Galindo Garre16,Reference Park, Mitrou and Kipnis18–Reference Sakauchi, Khan and Mori22,Reference Pala, Sieri and Chiodini25–Reference Nakanishi, Homma and Osaki28,Reference Bonthuis, Hughes and Ibiebele47–Reference Soedamah-Muthu, Masset and Verberne50) . One cohort study in Japan had three publications on colorectal(Reference Kojima, Wakai and Tamakoshi19), stomach(Reference Tokui, Yoshimura and Fujino21) and ovarian(Reference Sakauchi, Khan and Mori22) cancer mortality. The article of Schmid et al(Reference Schmid, Song and Zhang13) was published on two cohort studies including Nurses’ Health Study (NHS) and Health Professionals Follow-up Study (HPFS). Out of these, eleven publications had considered all-cause mortality, ten were about CVD mortality and thirteen about cancer mortality.
Fig. 1 Flow diagram of study selection
Characteristics of included studies
Table 1 provided detailed information for each eligible study. Among seventeen cohort studies published between 2004 and 2022 with a follow-up period of 6–32 years, four (three publications) were conducted in the USA(Reference Schmid, Song and Zhang13,Reference Park, Mitrou and Kipnis18,Reference Lin, Gui and Liang27) , one in the United Kingdom(Reference Soedamah-Muthu, Masset and Verberne50), three in the Netherlands(Reference Praagman, Franco and Ikram14,Reference Goldbohm, Chorus and Galindo Garre16,Reference Praagman, Dalmeijer and van der Schouw49) , one in Australia(Reference Bonthuis, Hughes and Ibiebele47), five (seven publications and three of them were reports from one study) in Japan(Reference Kojima, Wakai and Tamakoshi19–Reference Sakauchi, Khan and Mori22,Reference Lu, Sugawara and Matsuyama26,Reference Nakanishi, Homma and Osaki28,Reference Matsumoto, Ishikawa and Nakamura48) , one in Italy(Reference Pala, Sieri and Chiodini25), one in Iran(Reference Farvid, Malekshah and Pourshams12) and the study of Dehghan et al(Reference Dehghan, Mente and Rangarajan2) on population from twenty-one different countries. The sample size of the included studies varied from 1529 to 293 888 participants with an age range of 18–90 years. Overall, 896 871 individuals with 75 791 cases of all-cause mortality, 20 554 cases of cancer mortality and 14 623 cases of CVD mortality were included. One publication included only men(Reference Park, Mitrou and Kipnis18), one included only women(Reference Sakauchi, Khan and Mori22), and the others included both men and women. However, statistical analyses were separately done for males and females in six papers. In most studies, yogurt consumption was evaluated using FFQ, although one study had applied food recall(Reference Lin, Gui and Liang27). All studies had adjusted for age. Moreover, most publications had controlled for BMI (n 14), physical activity (n 11), smoking status (n 14), alcohol intake (n 13), energy intake (n 11) and other dietary variables (n 10). Based on the Newcastle–Ottawa scale score(Reference Margulis, Pladevall and Riera-Guardia30), thirteen papers which had a total score above the median (≥ 7) were defined to be high quality (online Supplementary Table 1).
Table 1 Characteristics of included studies on the association between yogurt consumption and mortality in adults aged > 18 years
RR, relative risk; F, female; g/d, gram(s) per d; HR, hazard ratio; serv, serving; NSCS, Nambour Skin Cancer Study; HPFS, Health Professionals Follow-Up Study; NHS, Nurses’ Health Study; GCS, Golestan Cohort Study; PURE, Prospective Urban Rural Epidemiology; EPIC-NL, European Prospective Investigation into Cancer and Nutrition – the Netherlands; RS, Rotterdam Study; NIH-AARP, National Institutes of Health-American Association of Retired Persons; JACC, Japan Collaborative Cohort Study; JMS, Jichi Medical School; NLCS, Netherlands Cohort Study; NHANES, National Health and Nutrition Examination Survey.
Findings from the systematic review
Out of twelve cohorts (eleven publications) investigating the association between yogurt consumption and overall mortality, four studies reported an inverse association(Reference Dehghan, Mente and Rangarajan2,Reference Farvid, Malekshah and Pourshams12,Reference Lin, Gui and Liang27,Reference Nakanishi, Homma and Osaki28) and six showed no significant association(Reference Goldbohm, Chorus and Galindo Garre16,Reference Pala, Sieri and Chiodini25,Reference Lu, Sugawara and Matsuyama26,Reference Bonthuis, Hughes and Ibiebele47,Reference Praagman, Dalmeijer and van der Schouw49,Reference Soedamah-Muthu, Masset and Verberne50) . However, in the publication of Schmid et al(Reference Schmid, Song and Zhang13), an inverse association was seen in women (NHS) but not men (HPFS). Moreover, in the study of Goldbohm et al.(Reference Goldbohm, Chorus and Galindo Garre16), yogurt consumption was inversely associated with all-cause mortality among men but not women. In terms of CVD mortality, yogurt consumption was not associated with a lower risk of CVD mortality in nine cohorts(Reference Schmid, Song and Zhang13,Reference Praagman, Franco and Ikram14,Reference Pala, Sieri and Chiodini25–Reference Nakanishi, Homma and Osaki28,Reference Bonthuis, Hughes and Ibiebele47,Reference Praagman, Dalmeijer and van der Schouw49) . However, two studies(Reference Farvid, Malekshah and Pourshams12,Reference Goldbohm, Chorus and Galindo Garre16) revealed that yogurt consumption was inversely related to deaths from CVD. For the association of yogurt consumption and cancer mortality, one publication(Reference Nakanishi, Homma and Osaki28) reported an inverse association and nine did not find any significant association(Reference Farvid, Malekshah and Pourshams12,Reference Park, Mitrou and Kipnis18–Reference Sakauchi, Khan and Mori22,Reference Pala, Sieri and Chiodini25,Reference Mazidi, Mikhailidis and Sattar41,Reference Matsumoto, Ishikawa and Nakamura48,Reference Praagman, Dalmeijer and van der Schouw49) . However, Schmid et al(Reference Schmid, Song and Zhang13) found an inverse association between yogurt intake and risk of cancer mortality among women (NHS) but not men (HPFS).
Findings from the meta-analysis
The association between yogurt intake and risk of overall mortality was investigated in twelve cohort studies (eleven publications)(Reference Dehghan, Mente and Rangarajan2,Reference Farvid, Malekshah and Pourshams12,Reference Schmid, Song and Zhang13,Reference Goldbohm, Chorus and Galindo Garre16,Reference Pala, Sieri and Chiodini25–Reference Nakanishi, Homma and Osaki28,Reference Bonthuis, Hughes and Ibiebele47,Reference Praagman, Dalmeijer and van der Schouw49,Reference Soedamah-Muthu, Masset and Verberne50) , which included a total of 476 160 participants with 75 791 cases. Comparing the highest v. lowest intakes of yogurt, the pooled RR for all-cause mortality was 0·93 (95 % CI: 0·89, 0·98), indicating a significant inverse association between yogurt consumption and overall mortality. There was evidence of moderate between-study heterogeneity (I2 = 47·3 %; P = 0·03) (Fig. 2).
Fig. 2 Forest plot for the association between yogurt consumption and risk of all-cause mortality in adults aged ≥ 18 years by comparing the highest and lowest categories of yogurt intake. RR, relative risk; NSCS, Nambour Skin Cancer Study; HPFS, Health Professionals Follow-Up Study; NHS, Nurses’ Health Study; GCS, Golestan Cohort Study; PURE, Prospective Urban Rural Epidemiology; EPIC-NL, European Prospective Investigation into Cancer and Nutrition-Netherland; NLCS, Netherlands Cohort Study; NHANES, National Health and Nutrition Examination Survey
Eleven cohort studies (ten publications)(Reference Farvid, Malekshah and Pourshams12–Reference Praagman, Franco and Ikram14,Reference Goldbohm, Chorus and Galindo Garre16,Reference Pala, Sieri and Chiodini25–Reference Nakanishi, Homma and Osaki28,Reference Mazidi, Mikhailidis and Sattar41,Reference Bonthuis, Hughes and Ibiebele47,Reference Praagman, Dalmeijer and van der Schouw49) were included in the analysis of yogurt consumption and CVD mortality. These studies involved a total of 331 261 participants, among them 14 623 mortality cases were found. The pooled RR for CVD mortality was 0·89 (95 % CI: 0·81, 0·98), indicating a significant inverse association, with a low between-study heterogeneity (I2 = 33·2 %; P = 0·13) (Fig. 3).
Fig. 3 Forest plot for the association between yogurt consumption and risk of CVD mortality in adults aged ≥ 18 years by comparing the highest and lowest categories of yogurt intake. RR, relative risk; NSCS, Nambour Skin Cancer Study; HPFS, Health Professionals Follow-Up Study; NHS, Nurses’ Health Study; GCS, Golestan Cohort Study; EPIC-NL, European Prospective Investigation into Cancer and Nutrition-Netherland; RS, Rotterdam Study; NLCS, Netherlands Cohort Study; NHANES, National Health and Nutrition Examination Survey
The association between yogurt consumption and risk of cancer mortality was assessed in twelve studies (thirteen publications)(Reference Farvid, Malekshah and Pourshams12,Reference Schmid, Song and Zhang13,Reference Park, Mitrou and Kipnis18–Reference Sakauchi, Khan and Mori22,Reference Pala, Sieri and Chiodini25–Reference Nakanishi, Homma and Osaki28,Reference Matsumoto, Ishikawa and Nakamura48,Reference Praagman, Dalmeijer and van der Schouw49) . These studies included a total of 741 973 participants with 20 926 deaths. Combining the RR from these publications, we failed to find any significant association between yogurt consumption and risk of cancer mortality (pooled RR: 0·96; 95 % CI: 0·89, 1·03). No significant between-study heterogeneity was also observed (I2 = 26·5 %; P = 0·18) (Fig. 4).
Fig. 4 Forest plot for the association between yogurt consumption and risk of cancer mortality in adults aged ≥ 18 years by comparing the highest and lowest categories of yogurt intake. RR, relative risk; HPFS, Health Professionals Follow-Up Study; NHS, Nurses’ Health Study; GCS, Golestan Cohort Study; EPIC-NL, European Prospective Investigation into Cancer and Nutrition-Netherland; NIH-AARP, National Institutes of Health-American Association of Retired Persons; JACC, Japan Collaborative Cohort Study; JMS, Jichi Medical School; NHANES, National Health and Nutrition Examination Survey
Findings from the dose–response analysis
Out of twelve studies investigating the association between yogurt consumption and overall mortality, eleven studies that provided sufficient information(Reference Dehghan, Mente and Rangarajan2,Reference Farvid, Malekshah and Pourshams12,Reference Schmid, Song and Zhang13,Reference Goldbohm, Chorus and Galindo Garre16,Reference Pala, Sieri and Chiodini25,Reference Lu, Sugawara and Matsuyama26,Reference Nakanishi, Homma and Osaki28,Reference Bonthuis, Hughes and Ibiebele47,Reference Praagman, Dalmeijer and van der Schouw49,Reference Soedamah-Muthu, Masset and Verberne50) were included in the dose–response analysis. Each additional serving/d yogurt consumption was inversely associated with risk of all-cause mortality (pooled RR: 0·93; 95 % CI: 0·86, 0·99, I2 = 63·3 %) (online Supplementary Fig. 1). There was evidence of non-linear association between yogurt consumption and all-cause mortality (P-non-linearity < 0·001), and there was no further reduction in risk above 0·5 serving/d (Fig. 5).
Fig. 5 Non-linear dose–response association of yogurt consumption (based on serving/d) with risk of all-cause, CVD and cancer mortality in adults aged ≥ 18 years. The solid lines indicate the spline model. The dashed lines present the 95 % CI
Ten studies(Reference Farvid, Malekshah and Pourshams12–Reference Praagman, Franco and Ikram14,Reference Goldbohm, Chorus and Galindo Garre16,Reference Pala, Sieri and Chiodini25,Reference Lu, Sugawara and Matsuyama26,Reference Nakanishi, Homma and Osaki28,Reference Bonthuis, Hughes and Ibiebele47,Reference Praagman, Dalmeijer and van der Schouw49) were included in the dose–response association of yogurt consumption and CVD mortality. Each additional serving/d yogurt consumption was associated with a 14 % lower risk of CVD mortality (pooled RR: 0·86; 95 % CI: 0·77, 0·97, I2 = 36·6 %) (online Supplementary Fig. 2). There was evidence of departure from linearity (P-non-linearity = 0·009), and there was no further reduction in risk above 0·5 serving/d (Fig. 5).
Out of twelve studies on the association between yogurt consumption and risk of cancer mortality, seven studies(Reference Farvid, Malekshah and Pourshams12,Reference Schmid, Song and Zhang13,Reference Pala, Sieri and Chiodini25,Reference Lu, Sugawara and Matsuyama26,Reference Nakanishi, Homma and Osaki28,Reference Praagman, Dalmeijer and van der Schouw49) were included in the dose–response analysis. Each additional serving/d of yogurt consumption was not associated with cancer mortality (pooled RR: 0·95; 95 % CI: 0·85, 1·07, I2 = 53·3 %) (online Supplementary Fig. 3). There was no evidence of non-linear association (P-non-linearity = 0·08) (Fig. 5).
Subgroup and sensitivity analyses, and publication bias
To examine the sources of between-study heterogeneity, we performed subgroup analysis. Table 2 shows findings for the different subgroups. None of the variables could explain the between-study heterogeneity, except for study location, which appeared to be the main factor responsible for this heterogeneity. The observed associations remained almost unchanged after control for important potential confounders, including BMI and energy intake. Based on sensitivity analysis, we found that the overall association did not affect by any individual study. Visual inspection of the funnel plot as well as findings from Begg’s and Egger’s tests revealed no evidence of publication bias in the analyses.
Table 2 Stratified analyses on associations of yogurt consumption with risk of mortality from all causes, CVD and cancer in adults aged ≥ 18 years
RR, relative risk.
* Number of relative risks.
† Obtained from the random-effects model.
‡ Inconsistency – the percentage of variation across studies due to heterogeneity.
§ Obtained from Q test.
Discussion
The findings from the current systematic review and meta-analysis of seventeen cohort studies demonstrated a significant inverse association between yogurt consumption and mortality from all causes and CVD. However, we failed to find any statistically significant association between yogurt consumption and cancer mortality. Each additional serving of yogurt consumption per d was associated with a lower risk of mortality from all causes and CVD.
Although the association between total dairy consumption and risk of mortality has been widely examined, less attention has been paid to the association between yogurt consumption and mortality. To the best of our knowledge, this is the most comprehensive meta-analysis of cohort studies that investigated the association of yogurt consumption with the risk of mortality. In the present study, we found a significant inverse association between yogurt consumption and risk of all-cause mortality. Our findings are comparable with a meta-analysis on the association between dairy intake and mortality. A meta-analysis of prospective studies, including 636 726 participants, reported a significant inverse association between consumption of fermented dairy products and risk of all-cause mortality(Reference Mazidi, Mikhailidis and Sattar41). In contrast, no association was observed between yogurt intake and overall mortality in a meta-analysis of prospective cohorts(Reference Guo, Astrup and Lovegrove51). Similar results were also found in another meta-analysis on yogurt consumption and risk of all-cause mortality(Reference Gao, Jia and Chen17). The different findings might be explained by some methodological limitations of that meta-analysis(Reference Gao, Jia and Chen17). The authors included an ineligible study in which the RR was reported for the combination of yogurt and cottage cheese, not yogurt only(Reference Bongard, Arveiler and Dallongeville23). They also included a prospective study conducted on cancer patients(Reference Dik, Murphy and Siersema24). In addition, we included further studies(Reference Schmid, Song and Zhang13,Reference Pala, Sieri and Chiodini25–Reference Nakanishi, Homma and Osaki28) that appeared after the release of those meta-analyses(Reference Gao, Jia and Chen17,Reference Guo, Astrup and Lovegrove51) .
In the current study, we found that yogurt consumption was significantly and inversely associated with the risk of CVD mortality. Unlike our findings, yogurt intake was not related to CVD mortality in the latest meta-analysis in this regard(Reference Gao, Jia and Chen17). Lack of such an association was also observed in another meta-analysis of prospective cohort studies(Reference Guo, Astrup and Lovegrove51). It should be noted that several large cohort studies(Reference Schmid, Song and Zhang13,Reference Pala, Sieri and Chiodini25–Reference Nakanishi, Homma and Osaki28) have been published since the release of previous meta-analyses(Reference Gao, Jia and Chen17,Reference Guo, Astrup and Lovegrove51) . This might explain the diverse findings. With regard to cancer mortality, our findings were in agreement with a prior meta-analysis examining the association with yogurt consumption(Reference Gao, Jia and Chen17). Lu et al(Reference Lu, Chen and Niu52) in a meta-analysis of eleven population-based cohort studies reached no significant association between consumption of dairy products and risk of cancer mortality. It is possible that the association between yogurt consumption and cancer mortality is dependent on the cancer type and characteristics of the study population. Other potential factors including different lifestyle-related factors and dietary intakes might need to be further explored in future studies.
Although the underlying mechanisms behind the inverse association between yogurt consumption and mortality risk are not fully understood, one possible explanation is alteration in the gut microbiome. Yogurt carries the beneficial bacteria into the gut-promoting immune functions through which it can improve human health. Lisko et al.(Reference Lisko, Johnston and Johnston53) demonstrated a fluctuation in the diversity of gut microbiota after a short-term period of yogurt intake in healthy subjects, indicating that regular consumption of yogurt may favourably affect the gut microbiota. However, the positive effects of yogurt intake are not limited to bacterial quantity and diversity. In fact, the beneficial metabolites produced by bacteria such as SCFA may also play an important role in its beneficiary effect(Reference Schmid, Song and Zhang13,Reference Lisko, Johnston and Johnston53) .
Regular consumption of fermented yogurt products in experimental studies has been associated with an increased level of HDL-cholesterol(Reference Kiessling, Schneider and Jahreis54) as well as lower levels of total cholesterol and LDL-cholesterol. Therefore, a reduction in the ratio of LDL-cholesterol:HDL-cholesterol, an index of atherogenicity, might explain its inverse association with mortality(Reference Agerholm-Larsen, Bell and Grunwald55). Moreover, fermented dairy products serve as an important dietary source of vitamin K2 (menaquinone), which can in turn stimulate β-cell proliferation and improve insulin sensitivity(Reference Walther, Karl and Booth56). Yogurt intake was longitudinally associated with less weight gain and lower waist circumference, supporting the growing evidence that changes in gut bacteria may affect weight gain(Reference Sayon-Orea, Martínez-González and Ruiz-Canela57).
Health-promoting effects of fermented yogurt products may be attributed to the biosynthesis or release of bioactive peptides with antihypertensive, antimicrobial, antioxidative and immune-modulatory properties(Reference Schmid, Song and Zhang13,Reference Michels, Willett and Vaidya15) . Findings from clinical trials have also indicated that consumption of yogurt may be effective in reducing chronic inflammation(Reference Pei, DiMarco and Putt58,Reference Labonté, Couture and Richard59) . Ca content of yogurt might also play a role in its inverse association with mortality. Ca appears to interact with SFA, forming fatty acid-insoluble soaps, consequently reducing SFA absorption, lowering TAG concentrations and improving the HDL-cholesterol:LDL-cholesterol ratio. Diets rich in Ca were associated with beneficial changes in blood pressure and lowering risk of stroke(Reference Fernandez, Panahi and Daniel60,Reference Bucher, Cook and Guyatt61) . Taken together, these findings support the notion that yogurt consumption may be effective in reducing risk of all-cause and CVD mortality.
The present meta-analysis has several strengths. A large sample size (896 871 participants and 75 791 cases) provides an adequate level of statistical power to detect the associations of interest. Moreover, findings were adjusted for numerous potential confounding variables in the included studies. Publication bias that can affect the results of studies is possible in any meta-analysis, but we found no evidence of such bias. Furthermore, a dose–response analysis in the current study adds to the present literature. However, a number of limitations should be considered. First, there was not sufficient information available to investigate the association of yogurt consumption with risk of mortality based on its fat content. Second, the present meta-analysis was based on observational studies; therefore, causality cannot be inferred. Third, given the use of FFQ as a method of dietary assessment in most included studies, measurement errors and misclassification of participants in terms of yogurt intake cannot be ignored. Fourth, the units of yogurt consumption were different across studies. Fifth, most included studies had measured yogurt consumption only at study baseline. Sixth, some studies did not provide sufficient information for the dose-dependent meta-analysis. Moreover, nutrient content of yogurt is dependent on various factors such as animals’ diet, food fortification, biosynthesis and physicochemical conditions, which might be different across studies.
In conclusion, we found an inverse association between yogurt consumption and risk of all-cause and CVD mortality; however, we failed to find any significant association with cancer mortality. To shed light on this issue, it seems that further studies, particularly pooled analyses, are required.
Acknowledgements
Acknowledgements: Not required. Financial support: This meta-analysis was funded by the Endocrine Research Center of Tabriz University of Medical Sciences, Tabriz, Iran (grant No 70988). The funder had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or the decision to submit the manuscript for publication. Authorship: S.N. and F.N. contributed to literature search and data analysis. H.T. and S.N. contributed to data extraction and manuscript drafting. M.N. and H.T. contributed to study conception and data analysis. AS contributed to study conception, manuscript drafting, data analysis and approving the final manuscript. All authors acknowledge the full responsibility for the analyses and interpretation of the report. Ethics of human subject participation: Not required.
Conflict of interest:
The authors declared no conflicts of interest.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S1368980022002385
Open access
