What does this journal article contribute to the discussion? Background: The health benefits of organic foods are unclear the estimate for phosphorus; phosphorus levels were significantly higher than in conventional produce, although this difference is not Pu the health effects of or rpose: To ganic and conventional foods. review evidence comparing dinically significant. The risk for contamination with detectable pes- ticide residues was lower among organic than conventional produce (risk difference, 30% [CI,-37% to-23%), but differences in risk for exceeding maximum allowed limits were small. Escherichia coli contamination risk did not differ between organic and conventional produce. Bacterial contamination of retail chicken and pork was Data Sources: MEDLINE January 1966 to May 2011), EMBASE, CAB Direct, Agricola, TOXNET, Cochrane Library Canuary 1966 to May 2009), and bibliographies of retrieved articles. Study Selection: English-language reports of comparisons of orcmmon but unrelated to farming method. However, the risk for ganically and conventionally grown food or of populations consum-isolating bacteria resistant to 3 or more antibiotics was higher in ing these foods. conventional than in organic chicken and pork (risk difference, 33% [CI, 21% to 45% Data Extraction: 2 independent investigators extracted data on methods, health outcomes, and nutrient and contaminant levels. Limitation: Studies were heterogeneous and limited in number and publication bias may be present Data Synthesis: 17 studies in humans and 223 studies of nutrient and contaminant levels in foods met inclusion criteria. Only 3 of the Conclusion: The published literature lacks strong evidence that human studies examined clinical outcomes, finding no significant organic foods are significantly more nutritious than conventional differences between populations by food type for allergic outcomesfoods. Consumption of organic foods may reduce exposure to (eczema, wheeze, atopic sensitization) or symptomatic Campylo- pesticide residues and antibiotic-resistant bacteria. bacter infection. Two studies reported significantly lower urinary pesticide levels among children consuming organic versus conven- tional diets, but studies of biomarker and nutrient levels in serum, urine, breast milk, and semen in adults did not identify dinically meaningful differences. All estimates of differences in nutrient and Ann Intem Med.2012:157 348-366 contaminant levels in foods were highly heterogeneous except for For author affiliations, see end of text. Primary Funding Source: None. www.annals.org etween 1997 and 2010, U.S. sales of organic foods purchase organic foods for different rea- sons, including concerns about the effects of conventional practices on the environment, human health, and are tast- Consumers increased from $3.6 to $26.7 billion (1,2). Although prices vary, consumers can pay up to twice as much for organic than conventional foods (3-5) farming animal welfare and perceptions that organic f Organic certification requirements and farming pra er than their conventional alternatives (2, 10-13). tices vary worldwide, but organic foods are generally grown without synthetic pesticides or fertilizers or routine use of antibiotics or growth hormones (6, 7). Organic livestock and safety characteristics of organic and conventional are fed organically produced feed that is free of pesticides and animal byproducts and are provided access to the out- of organic and conventional foods have summarized stud doors, direct sunlight, fresh air, and freedom movement (7). In addition, organic regulations typically require thatels without assessing the statistical significance of those dif- organic foods are processed without irradiation or chemical food additives and are not grown from genetically modified organisms (6, 8). The International Federation of Organic Agriculture Movements endorses the principles of “health, ecology, fairness, and care (9) he purpose of this study is to comprehensively syn thesize the published literature on the health, nutritional, foods. Previous reviews comparing the nutritional content om of movetes narratively (13-18), reported differences in nutrient lev- ferences or weighting outcomes by sample size (19-22), or considered only harms (23) Data Sources and Searches With a professional librarian, we developed search strategies for 7 databases: MEDLINE anuary 1966 to May 2011), EMBASE, CAB Direct, Agricola, TOXNET, and Cochrane Library Canuary 1966 to May 2009) with such terms as organic, vgetable, fruit, and berf (Supple- 348 62012 American College of Phyicians Organic Venus Conventional Foods REVIEW ment 1, available at www.annals.org) and reviewed bibli- ples contaminated and summary standardized mean differ- reporting mean nutrient or harm levels. Differences were calculated as organic minus con- ventional (for example, a positive number indicates more ences (SMDs) for studies Peer-reviewed, English-language studies, regardless of design, were eligible for inclusion if they reported a com- We performed tests of homogeneity (Q statistic and P statistic) on all summary effect sizes. Homogeneity was indicated if p was less than 25% and P value for the Q statistic was greater than 0.010. If the 2 tests agreed, we report only the F statistic otherwise, we report results for foods grown organically and conventionally or a compara- tive evaluation of nutrient levels or bacteria, funal, or mlk Gncluding raw milk), or egs grown organi- both. We used funnel plots to assess publication bias ( foods, those that evaluated samples from livestock es We qualitatively summarized studies not reporting infor- infor- mation on variance and excluded studies not reporting any gastrointestinal tracts, and those that did not report mation about variance or results of statistical tests (24-34). Organic practices included biodynamic farming and were defined by investigators’ stated adherence. Studies merely comparing the effects of organic and nonorganic fon on variance or statistical testing. All analyses version 2 (Biostat, Englewood, New Jersey). Because of the fertiliza large number of comparisons (22 for produce and 31 for tion practices were ineligible unless they specified that the meat, poultry, milk, and eggs), we report adjusted P values for summary estimates using the Sidak formula for mult ple comparisons. For each reported summary effect size, we we excluded studies of such foods as recombinant bovineted 1 study at a time to assess the influence of each somatotropin-free milk and grass-fed beef unsfoodindividual study on summary effects and omitted outdliers organic farming practices (28, 32, 33, 35-47). Similarly, that were more than 1 order of magnitude larger or smaller than others. We explored heterogeneity by conducting sub- group analyses by food type, organic standard used, testing method, and study design when at least 3 studies examined One author abstracted data on study methods (for ample, design; food tested; sample size; organic testing methods; harvest season; and cultivar, breed, or standard;these subgroups. We limited our analyses of bacterial contamination to foodborne pathogens monitored by the Centers for Disease report only the P statistic; otherwise, we report results for poultry, milk (including raw milk), or eggs grown organ cally and conventionally. We excluded studies of processed foods, those that evaluated samples from livestock feces or atroestnal tracts, and those that did not report infor- mation on variance both. We used funnel plots to assess publication bias (48 mation about variance or results of statistical tests (24-34). Organic practices included biodynamic farming and were defined by investigators’ stated adherence. Studies merely comparing the effects tion practices were ineligible unless they specified that the We qualitatively summarized studies not reporting infor- and excluded studies not reporting any information on variance or statistical testing. All analyses completed by using Comprehensive Meta-analysis, version 2 (Biostat, Englewood, New Jersey). Because of the fertiliza large number of comparisons (22 for produce and 31 for meat, poultry, milk, and eggs), we report adjusted Pvalues for summary estimates using the Sidak formula for multi- ple comparisons. For each reported summary effect size, we were of organic and nonorganic receiving organic fertilizer was grown by using organic farming practices (28, 32, 33, 35-47). Similarly we excluded studies of such foods as recombinant bovine omitted 1 study at a time to assess the influence of each somatotropin-free milk and grass-fed beef unless the foodindividua study on summary effects and omitted outliers production was reported to be organic. Data Extraction and Quality Assessment that were more than 1 order of magnitude larger or smaller than others. We explored heterogeneity by conducting sub- group analyses by food type, organic standard used, testing method, and study design when at least 3 studies examined One author abstracted data on study methods (for ex ample, design; food tested; sample sizes organic standard: testing methods; harvest season; and cultivar, breed, or population studied) and end points (Supplement 2, avail- able at www.annals.org). At least 1 additional author veri- fied all abstracted data; discrepancies were resolved with discussion. If 2 or more studies presented the same data from a single population or the s farm included these data only once in our analyses these subgroups. We limited our analyses of bacterial contamination to foodborne pathogens monitored by the Centers for Disease Control and Prevention’s FoodNet (49) (for example, Campylobacter, Listeria, Salmonella, and Escherichia coli). However, given the potential for transfer of antibiotic re- sistance between species, we included all human pathogens (for example, Staprylococcus aureus) in the analyses of – We defined quality criteria a priori and evaluatedthe tic resistance extent to which included human population studies speci- fied the organic standard used, evaluated the amount of organic foods consumed in diets, linked reported outcomes include such studies with health outcomes, obtained institutional review board approval and participant consent, and were not funded bymodels and used this an organization with a come. For the studies that directly evaluated the study foods, we evaluated the extent to which each study speci fied the organic standard used, used the or processing method for both groups, reported sample size, used equal sample size in both groups, and were funded by organizations with a financial interest in the briaceae fa study outcome. We also evaluated the extent to which the organic-conventional cultivar or breed, grown on neighboring farms, and har- vested during the same season. Data Synthesis and Analysis Studies frequently reported several results per outcome (for example, mean vitamin C level in years 1 and 2). To only once in our analyses, we com- bined the results within each study by using random-effects study-level summary effect in our i financial interest in the study outummary Similarly, several studies (50-53) reported multiple re- sults for resistance to the same antibiotic by examining same harvesting or different bacteria (for example, Salmonella and Campylo- zebcter). To include these studies only once in each effect size calculation, we used results for pathogens in the Entero family (for example, Salmonella for the main analyses and the alternate species (for example, Campylo- comparison pairs were of the same in scnsitu Among the produce studies, several studies that other- summary effect size cal- culations did not report sample sizes. To avoid discarding e, we assumed that they had a sample size of 3 (a wise could have been included in We calculated summary effect sizes by using random effects models for outcomes with at least 3 studies reportcommon sample size among the smaller studies). In sensi- ing data: summary risk differences (RDs) and summarytivity analyses, we varied this to 10, the median sample size prevalence rates for studies reporting the number of samamong studies. This alternate assumption did not change Table I. Summary of Benefits: SMD of Nutrient Levels Found in Organic Versus Conventional Fruits, Vegetables, and Grains Summary of All Idenated Studies Results of Meta-analys Compansons. C-sons Organic, nt Conventional, es, C arsons Studies. Studies Ogank Conventional SMD(95%얘 Describing Sample Sample Size, Sample Size, strawbenes, and tomatoes 12 28 1141 1306 050(005to095) 048 Yes00%) 1.14 (0.13 to 2.42) 083 Yes(91%) -009(-0,0 to0SD 100 Yes OGS) 045(0.3000 1為 100 Yes(87%) 061001 to 122) 068 Yes(84%) 082 (0 44 to 120 -d001 No0%) 060 01 to 130, 066 Yes(81%) 030(-047 to 108) 100 Yes(90%) (-320to062) 100 Yes(83%) 0/9( 187 to029) 0gy Yes(83%) 245020to46n 052 Yes 04%) 264041 to486) 036 Yes 193%) -019(-168to 1.31 100 Yes9) 12 a-Tocopherol Foods studied peaches, peans, plums, com, cabbage, canots, and olive ofl 37 15 484 500 Phosphonus Foods studied carrots, celery, com plums, onions, and potatoes 12 13 Foods studed wheat, banana, plum, tomato, soybeans, grape juice, and eggplant 73 QuercetinFoods sudied: plum, tomatoes, bell peppers, grapes, grape leaves, lettuce, strawbernies, and black currant 1% 1% Total tavanck Foods虫dnt appes, Foods studed apples, peaches, 34 ape kMes,盘antenes, dicory, and black amants loul phenob %, pkim, bd peppers, bemes, tomatoes, dicory. lve al, gtape leaves, orrges, strawbemes, bok dot ketuce, leafy 401 401 1030.47 to 159) 000 Yes (67%) SMDstandardaed mean difference *Allsumeuary dini moasures pon d anea softandone as models. Amoug saudíes exann ning nutritonal unt nt. я ules with null incompletidly (hence, they wese eduded from synthees). The eseption to this rule was among studies tended to RP rt rado ารเปซ (hence, they wese added f om suinmary dla1.akulation) tended lo find sign.kandy higher keb of summary eflect ses sitivity analyses wee performed, in which studies not reporning sample sine were and subesoup analyses wese done fresh v. dry woight. Findings did not substantially change with the sensitivity The number of comparisons in which a statistically signiicant diierence was with higher levels in the organic g eveds in the coeventionalproup las de ฬามไม่ธ included fr cach statistical analysis. I The diflerence between mcan nutrient levl in oeganic minus tha in conventional divided by the pooled SDs hus, a positive (ainc) number indias highr owe) All summary P valaes are adjnted P vall the winter (but not organic meat in general) to be a risk crossover study reported a statistically significant reduction factor for illness due to Campylobacter infection (odds rain serum total homocysteine levels, phosphorus levels, and tio, 6.86 [CI, 1.49 to 31.691) fat mass after 2 weeks on an organic Mediterranean diet The remaining studics examined differences in thecompared with a conventional Mediterrancan diet but did um, urine, breast milk, and semen of persons consuming not describe the magnitude or clinical significance of these organic and conventional diets. We found no studies comreductions (67). Another crossover study found that or paring pesticide levels among adult consumers of organic ganic diets were associated with higher urinary excretion of versus conventional foods Seven studies evaluated serum and urine antioxidant levels or immune system markers; of these found no consistent differences in plasma or urine tion (56). carotenoids, polyphenols, vitamins E and C content, low density lipoprotein cholesterol, antioxidant activity, ability of women from the Dutch KOALA (Child, Parent, and to protect against DNA damage, immune system markers,Health: Lifestyle and Genetic Constitution) Birth Cohort or semen quality between participants consuming oranic consuming predominantly organic versus conventional and conventional diets (4, 57, 62, 65, 66, 69). All were meat and dairy products (58, 68). They found no differ- randomized, controlled trials except the study of semenence in the amount of total fatty acids in the breast milk of quality (a prospective cohort study)(54). One prospectivemothers who consumed meat and dairy products of which in and kacmpferol but not other polyphenols and no difference in 7 of 8 serum markers of antioxida- Two cross-sectional studies examined the breast milk Table 2. Summary of Harms: RD or SMD in Harms in Organic Versus Fruits, Vegetables, and Grains Summary of All Identified Studies Studies, n Any detectable pesticide residue E coli contamination Cadmium leve Lead level Mercury level Anenic level DON level OTA level OTA ochratoxin A RD risk diffenencs SMD sandandard mean diffesence and conventional was identified with lower levels in the difference with lower (availiable at www.annals.o) lsts the stadies indaded for each staristical analysis minus thar in the conventional groups thes, a positive (ive) number indiates mote (e) between mean contaminant level in organic minus that in coeventionall divided by the pooled All RDs are absolute RDs SMD s the .. one of stulis inch.ld the ode syndnas “ndudes a data ser (US, Departrnent of Agriculture’s P stode Data Progam) chat over ample sources wih a history of violations Hence, pevalence estimates may ovenstate peevalence of contamination in both organ % 195% Cl-2.92% to 7.18%). P < 0.001; p-0%). ##Fow cadmmm. kuL mmary, arsenic. DON, and OTA kvek these arethe sample sines instead ofthe number ofcontaminated unpks divided by the total number of ly produced versus mothers of which less than (58, 68). In subanalyses, controlled and 29% (44 studies) sampled food grown on commercial farms. Among animal product studies, 11% (8 studies) were conducted on experimental farms and 56% more than 90% were 50% were organically they found higher levels of 2 beneficial fatty acids (conju- (40 studies) surveyed farms. Of the 37 milk studies in- gated linoleic acid and trans-vaccenic acid) in the breast cluded, 7 examined pasteurized milk and 30 examined raw milk of mothers consuming and meat products versus alternatives (58) Studles of Nutrient and Contaminant Levels in Organic Versus Conventional Foods ly organic dairy milk (Supplement 4). consuming Forty-six percent (102 studies) of included studies Figure 2 (bottom panel). The most common standards Two hundred twenty-three studies of foods met nlu standards (43 studies), International Federation of Organic sion criteria: 153 studies of fruits, vegetables, and grains Agriculture Movements or other association standards (34 and 71 studies of meats, poultry, milk, and eggs ( study tudies), and U.S. Department of Agriculture standards reported on both types of foods [1891) (Supplement 4 (22 studies). The most common standards among produce available at www.annals.org). Seventy percent (157 studies) tudies were from organic associations country-specific were from Europe, and 21% (47 studies) were from the European regulatory standards were most common among United States or Canada. Study methods varied: Among animal product studies (Figure 1). Sixty-eight percent (151 studies) reported that harvest- in which potential confounders (for example, ng or processing methods were the same for both groups weather, geography, or plant cultivar) of the relationship teremaining studies largely did not describe harvesting or between method of cultivation and nutrient levels were processing methods (such as in studies that examined retail studies, 52% (80 studes) were on