Milk, milk products and lactose intake and ovarian cancer risk: A meta-analysis of epidemiological studies
Article first published online: 28 JUL 2005
Copyright © 2005 Wiley-Liss, Inc.
International Journal of Cancer
Volume 118, Issue 2, pages 431–441, 15 January 2006
How to Cite
Larsson, S. C., Orsini, N. and Wolk, A. (2006), Milk, milk products and lactose intake and ovarian cancer risk: A meta-analysis of epidemiological studies. Int. J. Cancer, 118: 431–441. doi: 10.1002/ijc.21305
- Issue published online: 9 NOV 2005
- Article first published online: 28 JUL 2005
- Manuscript Accepted: 20 APR 2005
- Manuscript Received: 12 JAN 2005
- Swedish Cancer Foundation
- dairy products;
- ovarian cancer;
- systematic review
It has been proposed, on the basis of animal models and ecological studies, that consumption or metabolism of dairy sugar may increase the risk of ovarian cancer. Case-control and cohort studies of the association between lactose and dairy food consumption and ovarian cancer risk, however, have yielded varied findings. We summarized the available literature on this topic using a meta-analytic approach. Random-effects models were used to estimate the summary relative risks (RRsummary). A linear regression analysis of the natural logarithm of the RR was carried out to assess a possible dose-response relationship between lactose intake and ovarian cancer risk. Eighteen case-control and 3 prospective cohort studies were eligible for inclusion in the meta-analysis. The findings of case-control studies were heterogeneous, and, except for whole milk (RRsummary for highest vs. lowest category = 1.27, 95% confidence interval [CI] = 0.97–1.68), do not provide evidence of positive associations between dairy food and lactose intakes with risk of ovarian cancer. In contrast, the 3 cohort studies are consistent and show significant positive associations between intakes of total dairy foods, low-fat milk, and lactose and risk of ovarian cancer. The RRsummary for a daily increase of 10 g in lactose intake (the approximate amount in 1 glass of milk) was 1.13 (95% CI = 1.05–1.22) for cohort studies. In conclusion, prospective cohort studies, but not case-control studies, support the hypothesis that high intakes of dairy foods and lactose may increase the risk of ovarian cancer. © 2005 Wiley-Liss, Inc.
In 1989, Cramer1 reported that per capita milk consumption and lactase persistence (the ability to digest lactose after childhood) was significantly positively correlated with national incidence rates of ovarian cancer. Lactose is a disaccharide found solely in milk and milk products and is cleaved by intestinal lactase to produce galactose and glucose. Galactose has been postulated to increase the risk of ovarian cancer by direct toxicity to the ovarian germ cells and by causing gonadotropin levels to increase, thereby stimulating the proliferation of the ovarian epithelium and eventually to induce ovarian neoplasia.2, 3
The observation by Cramer in 1989 stimulated substantial research into the associations between intakes of milk, milk products, and lactose and ovarian cancer risk. Although many epidemiological studies have examined these relationships, the role of milk, milk products and lactose on ovarian cancer risk remains obscure.
The aim of our study is to assess the association between consumption of milk and milk products and lactose intake and risk of ovarian cancer by conducting a meta-analysis of case-control and cohort studies. We have examined previously the association between dairy food consumption and ovarian cancer risk in the Swedish Mammography Cohort.4 In that analysis, we did not examine consumption of low-fat and whole milk separately (only total milk). That analysis was based on a single dietary assessment. We extend our previous analysis with separate analyses for low-fat and for whole milk and with repeated dietary assessments as well as with more cases to examine the associations of long-term dairy food consumption and lactose intake with ovarian cancer incidence.
Material and methods
Selection of studies
Eligible studies were identified by searching the MEDLINE database for relevant articles published in any language to January 2005 and by hand-searching the reference list of the computer-retrieved articles. For the computer searches we used the following MeSH terms or text words: “dairy products,” “milk” or “lactose” combined with “ovarian cancer” or “ovarian neoplasm”. Studies were included in the meta-analyses if they presented data from a case-control or cohort study on the association between intakes of milk, yogurt, cheese or lactose and incidence of or mortality from epithelial ovarian cancer. To be included, studies had to have expressed their results as an odds ratio (OR) or a relative risk (RR) together with 95% confidence interval (CI), or sufficient data to permit their calculations.
Data extraction and classification
Study characteristics recorded included the name of the first author, study design, publication year, location of the study, years of data collection (case-control studies), duration of follow-up (cohort studies), sample size, type of controls, response rates, measures of exposure and outcome, adjusted covariates and risk estimates with corresponding 95% CI. When several risk estimates were presented, we used the ones adjusted for the greatest number of potential confounders.
The measure of effect of interest is the RR and the corresponding 95% CI. Because the risk of ovarian cancer is low, OR from case-control studies accurately estimate RR. We report all results as RR for simplicity. We quantified associations of dairy foods and lactose intakes with ovarian cancer risk using random-effects models5 of RR comparing the highest with the lowest (the referent) category.
For dose-response meta-analyses of the association between lactose intake and ovarian cancer risk, we used the method proposed by Greenland and Longnecker6 to compute study-specific slopes from the correlated natural logarithm of the RR across exposure categories. Only studies that reported the number of cases and controls and the RR and its variance estimate for 3 or more quantitative exposure categories were included. The median level of lactose intake (g/day) for each category was assigned to each corresponding RR. If data were not available, we estimated the median using the midpoint of each category. An upper bound was not reported in most studies for the category of highest intake, so we assumed it to be the same amplitude as the preceding category. The summary RR estimates were obtained from random-effects models5 applied to the study-specific dose-response slopes.
Statistical heterogeneity among studies was assessed using both the Q statistic and the I2.7 Results were defined as heterogeneous for p-values <0.10. We assessed publication bias by constructing funnel plots, and by 2 formal tests: the Begg-adjusted rank correlation test and the Egger's regression asymmetry test.8 The meta-analyses were carried out using the statistical package Stata (version, 8.0, Stata Corporation, College Station, TX).
Analysis of the Swedish Mammography Cohort
Details about the design and recruitment of women to the Swedish Mammography Cohort, and statistical methods for analysis of dairy food consumption and ovarian cancer risk in this cohort have been described previously.4 Our present analysis is based on follow-up through June 30, 2004. To best represent long-term diet and to reduce random within-person variation, we used repeated measures of diet9 in the current analysis. Specifically, the incidence of ovarian cancer from 1987 (baseline) through 1997 was related to dietary variables assessed on the baseline food-frequency questionnaire and incidence from 1998 through June 30, 2004 was related to the average intake reported on the baseline and on a follow-up questionnaire completed by participants in 1997. The proportional hazards assumptions were satisfied, and the RR with 95% CI were estimated using Cox proportional hazards models, stratified on age in months and the year of entry into the cohort. Multivariate models were also controlled for body mass index, education, parity, oral contraceptive use and intakes of total energy, fruits and vegetables. SAS software was used for these analyses (version 9.1, SAS Institute Inc., Cary, NC).
Results from the Swedish Mammography Cohort
Over an average follow-up of 14.7 years, we ascertained 288 incident cases of invasive epithelial ovarian cancer, including 135 serous tumors. In multivariate analyses, cumulative average intakes of total dairy foods, milk and lactose were positively associated with risk of ovarian cancer (Appendix I). The positive associations seemed to be confined to the serous subtype. For comparison, the multivariate RR for the top compared to the bottom quartile of lactose intake were 1.96 (95% CI = 1.15–3.35; p for trend = 0.009) for serous ovarian tumors and 1.17 (95% CI = 0.74–1.86, p for trend = 0.75) for all other subtypes combined.
Characteristics of studies for meta-analysis
Our search resulted in 22 potential publications. We excluded one study10 because the endpoint was restricted to benign ovarian tumors. Characteristics of the 3 prospective cohort studies (including the Swedish Mammography Cohort)11, 12 and the 18 case-control studies13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30 that met the inclusion criteria are presented in Appendix I (cohort) and in Appendix II (case-control). Two cohort studies were carried out in the United States and one in Sweden. Among the 18 case-control studies, 10 were conducted in the United States, 2 in Canada, and one each in Australia, Italy, Japan, China, Taiwan and Mexico. Population-based controls were used in 11 case-control studies; 7 case-control studies used hospital-based controls. The response rates in the studies that provided this information ranged from 55–100%; the number of ascertained case subject who died before being contacted ranged from 5–27%. Most studies controlled for age, body mass index, parity and use of oral contraceptives. All studies used epithelial ovarian cancer incidence rather than mortality as the endpoint. The Nurses' Health Study11 and the Swedish Mammography Cohort4 were the only studies that reported results separately for histological subtypes of ovarian cancer (serous tumors).
Five case-control and 2 cohort studies reported results on total dairy food consumption in relation to ovarian cancer risk. In all but one of the 7 studies, the RR for the highest compared to the lowest category of dairy food consumption was >1, suggesting a positive association between dairy food consumption and risk of ovarian cancer (Fig. 1). There was significant heterogeneity between the findings of case-control studies, but not of the 2 cohort studies. The summary RR for the highest vs. the lowest category of dairy food consumption was close to null in case-control (RRsummary = 0.99; 95% CI = 0.69–1.43) and 1.66 (95% CI = 1.19–2.31) in cohort studies. There was significant heterogeneity when all studies were considered together (Q = 16.98, p = 0.009; I2 = 64.7%). This, however, was explained by the case-control study by Goodman et al.17 After excluding this study, the overall RR for all studies combined was 1.34 (95% CI = 1.10–1.63), without heterogeneity between studies (Q = 2.54, p = 0.77, I2 = 0%).
Results from 7 case-control studies and one cohort study on consumption of total milk and ovarian cancer risk were heterogeneous (Fig. 2). Significant heterogeneity was also present between the 7 case-control studies. The overall RR was 0.87 (95% CI = 0.68–1.10) for case-control and cohort studies combined.
Eight case-control and 3 cohort studies on skim/low-fat milk consumption and ovarian cancer risk were identified. There was significant heterogeneity present between all studies (Q = 29.80, p = 0.001, I2 = 66.4%), but not between the cohort studies (Fig. 3). The summary RR was 0.80 (95% CI = 0.64–1.01) for case-control and 1.35 (95% CI = 1.09–1.68) for cohort studies.
Seven case-control and 2 cohort studies reported RR estimates for whole milk consumption and risk of ovarian cancer. There was significant heterogeneity between all studies (Q = 16.55, p = 0.04; I2 = 51.7%), but not between the 2 cohort studies (Fig. 4). The summary RR for the highest vs. the lowest category of whole milk consumption was 1.27 (95% CI = 0.97–1.68) in case-control and 1.17 (95% CI = 0.81–1.68) in cohort studies. The heterogeneity between all studies was largely explained by the studies by Mettlin et al.24 and Cramer et al.23 After excluding these 2 studies, the overall RR for case-control and cohort studies combined was 1.20 (95% CI = 1.03–1.41), without significant heterogeneity between studies (Q = 2.49, p = 0.87, I2 = 0%).
The meta-analysis of yogurt consumption included 7 case-control and 2 cohort studies. One case-control study26 provided estimates for consumption of full-fat and low-fat yogurt separately; both estimates were included in the meta-analysis. There was no evidence for the presence of heterogeneity between the 10 RR estimates from case-control and cohort studies (Fig. 5). The overall RR for high vs. low yogurt consumption was 1.13 (95% CI = 0.96–1.33).
Seven case-control and 3 cohort studies on cheese consumption and risk of ovarian cancer were identified. Two of the 7 case-control studies provided results for cottage cheese. There was no significant heterogeneity between the 11 studies (Fig. 6). Combining the data from case-control and cohort studies yielded an overall RR of 0.95 (95% CI = 0.80–1.12). The findings did not change materially when the 2 studies that reported results for cottage cheese were excluded (RRoverall = 0.92; 95% CI = 0.75–1.13).
The association between lactose intake and risk of ovarian cancer was examined in 9 case-control and 3 cohort studies. Overall, significant heterogeneity was present among results of all studies (Q = 24.23, p = 0.01; I2 = 54.6%), but not within each study design (Fig. 7). The summary RR for high vs. low intake of lactose was 0.88 (95% CI = 0.73–1.06) in case-control and 1.47 (95% CI = 1.17–1.84) in cohort studies.
Dose-response meta-analysis of lactose
Six case-control and 3 cohort studies (including the Swedish Mammography Cohort) were included in the dose-response analysis of the association between lactose intake and risk of ovarian cancer. There was significant heterogeneity among results of all studies but not within each study design (Table I). The summary RR for an increment of 10 g/day of lactose (the approximate amount in 1 glass of milk) was 0.96 (95% CI = 0.89–1.03) in case-control and 1.13 (95% CI = 1.05–1.22) in cohort studies.
|No. of studies||RR (95%CI)*||Q||p-value||I2|
|All studies||911, 12, 13, 16, 17, 19, 25, 26||1.02 (0.95–1.09)||16.54||0.04||51.6%|
|Case-control||613, 16, 17, 19, 25, 26||0.96 (0.89–1.03)||6.76||0.24||26.0%|
|Prospective cohort†||311, 12||1.13 (1.05–1.22)||0.09||0.96||0%|
We evaluated publication bias by means of visual inspection of funnel plots and formal statistical tests.8 There was no evidence of publication bias with regard to consumption of total dairy foods, milk, yogurt, cheese or lactose and ovarian cancer risk (data not shown).
In our meta-analysis we have observed that there are discrepancies between results of case-control and cohort studies concerning the relationship between milk consumption and lactose intake and ovarian cancer risk. In case-control studies, low-fat milk consumption is inversely and whole milk consumption positively associated with risk of ovarian cancer. Case-control studies provided no support for a role of lactose in the etiology of ovarian cancer. Prospective cohort studies indicated that high intakes of milk and lactose may increase the risk of ovarian cancer. In case-control and cohort studies combined, yogurt consumption was associated with a modest nonsignificant increased risk of ovarian cancer, whereas cheese consumption was not associated with risk. The amounts of free galactose and lactose (that is cleaved by intestinal lactase to produce galactose and glucose) in cheese are lower than in milk.31 As a comparison, 1 glass (200 ml) of milk contains about 450 mg free galactose and 10 g lactose, and one serving (28 g) of hard cheese about 25 mg free galactose and no lactose. In the cohort studies, a significant 13% increase in the risk of ovarian cancer was observed for a daily increment of 10 g of lactose.
There are several possible explanations why case-control and cohort studies provide conflicting results pertaining to the relation between intakes of low-fat milk (but not whole milk) and lactose and ovarian cancer. In case-control studies, dietary intake is assessed after the cancer diagnosis. Consequently, these studies may be subject to recall bias and inaccurate or biased measurements of dietary intake attributed to changes in the dietary practices among cases after their cancer diagnosis. Even though the cases are inquired about their diet-before-disease, their recall may well be influenced by their current diet.
Case-control studies are also prone to selection bias, i.e., differential participation with respect to exposure or severity of the disease. Participation rates in studies that provided this information were relatively high in some (≥70% response rate)15, 16, 19, 26, 28, 29 but low in others (55–70%).17, 20, 22, 23, 27, 30 Nevertheless, we could not discern any pattern regarding the findings from studies with high compared to low participation rates. One exception was whole milk consumption, which was positively associated with risk of ovarian cancer in all but one study. Participation rate in the study that failed to found an association with whole milk was only 55% among cases.23 In the case-control studies, 5–27% of the ascertained cases died before they were contacted.13, 15, 19, 23, 26, 27, 30 This could introduce selection bias if a case's likelihood of surviving until interview is related to the consumption of dairy foods. In fact, intakes of dairy foods and lactose have been associated with poorer survival from ovarian cancer.32 Hence, associations between dairy food and lactose intakes and ovarian cancer risk may have been underestimated in case-control studies if cases with high dairy food and lactose intakes died before interview.
Another difference between case-control and cohort studies relates to the time interval between the period covered by the questionnaire and the cancer diagnosis. This time interval is usually 1 or 2 years (recent diet) in case-control studies, whereas in cohort studies it could be as large as 10–20 years (recent diet at baseline). If the diet 10–20 years before the diagnosis of ovarian cancer may be most relevant and if changes in dietary patterns have occurred, then an association with dairy food consumption may not be found in case-control studies in which recent diet was assessed. It is also likely that long-term dietary intake may be most relevant to chronic diseases like cancer with a long duration of development. Two cohort studies, the Nurses' Health Study11 and the Swedish Mammography Cohort,4 used repeated measures of dietary intake, which took into account changes in diet over time and thus better reflect long-term diet as well as reduce random within-person measurement error due to one single dietary assessment.
Only the Nurses' Health Study11 and the Swedish Mammography Cohort4 examined milk and lactose intakes in relation to histological subtypes of ovarian cancer. In both studies the associations with milk and lactose intakes were confined to serous ovarian cancer.
Besides lactose, dairy foods are a source of cholesterol, which has been associated with an increased risk of ovarian cancer in some12, 30 although not all studies.33, 34 Two studies that examined cholesterol from eggs and cholesterol from other sources separately observed that only cholesterol from eggs was positively associated with ovarian cancer risk.35, 36 It is unlikely that the apparent positive association between milk consumption and risk of ovarian cancer is attributable to cholesterol.
It is possible that high intake of dairy foods, and consequently of lactose, increases the risk of ovarian cancer only in certain subgroups of the population, such as those with specific genetic or biochemical features of galactose (lactose metabolite) metabolism. Animal models have shown that high dietary galactose causes ovarian toxicity.37, 38 Moreover, hypogonadism or ovarian failure occurs frequently among women with galactosemia,39, 40 which arises from an autosomal recessive defect in the galactose-1-phosphate uridyl transferase (GALT) gene. Impairment of the GALT gene can lead to an accumulation of galactose and other metabolites in the body, including the ovary. Several polymorphisms of the GALT gene result in variable degrees of GALT activity. Cramer et al. found that women with a high ratio of lactose intake to GALT activity had a significant over 2-fold elevated risk of ovarian cancer. Some23, 41 but not all26, 42 studies have also found a higher prevalence of a polymorphism of the GALT gene, N314D among ovarian cancer cases than among control subjects. The N314D polymorphism can be associated with either high or low GALT enzyme activity depending on whether an associated polymorphism 1721CT is in cis with it as well as another polymorphism in the translation start site.43, 44 This might explain the discrepant results pertaining to the N314D polymorphism and ovarian cancer risk in different populations.
Several potential limitations of our study are worth considering. First, this meta-analysis is based on observational studies, and the inherent limitations of such studies may affect our findings. The possibility of bias and uncontrolled or residual confounding cannot be excluded. Although confounding cannot be ruled out, it should be noted that many studies included in the meta-analyses controlled for other risk factors and the associations with dairy foods and lactose, when present, persisted in multivariate analyses. Second, our findings were likely to be affected by misclassification of dairy food consumption due to imprecise measurement of diet. This type of random measurement error would generally bias results toward the null, and is thus unlikely to explain an observed association. Finally, we cannot discard publication bias, which may distort the representativeness of available results. Nevertheless, we found no evidence of publication bias involved in our findings.
In conclusion, results of this meta-analysis provide some support for the hypothesis that high intakes of milk and lactose may be associated with an increased risk of ovarian cancer. This evidence is largely limited to cohort studies, whereas case-control studies have produced conflicting results. Future studies should consider specific subtypes of ovarian cancer, and the interrelations between intakes of dairy foods and lactose, genetic polymorphisms and ovarian cancer risk.
- 8Systematic reviews in health care: meta-analysis in context. London: BMJ Publishing Group, 2001., , .
|Study and location (follow-up)||Cases (cohort size)||Measure of exposure and range of exposure||RR (95%CI)||p1||Covariates|
|Kushi et al., 199912 United States (1986–1995||139 EOC, including borderline tumors (28,944)||Lactose: >25.2 g/day vs. <7.7 g/day||1.60 (0.95–2.70)||0.12||Age, education, parity, age at menopause, family history of breast or ovarian cancer, hysterectomy, smoking, waist-hip ratio, physical activity, energy|
|Total dairy foods: >23 serving/week vs. <9 servings/week||1.76 (0.99–3.13)||0.03|
|Skim milk: >1 serving/day vs. <1 servings/week||1.73 (0.99–3.02)||0.04|
|Cheese: >4 serving/week vs. <1 servings/week||1.56 (0.85–2.86)||0.14|
|Fairfield et al., 200411 United States (1980–1996||301 invasive EOC (80,326)||All epithelial ovarian tumors:||Age, OC use, parity, tubal ligation, smoking, BMI, caffeine|
|Lactose: >19.8 g/day vs. <5.0 g/day||1.40 (0.98–2.01)||0.10|
|Skim/low-fat milk: ≥1 serving/day vs. never/seldom||1.32 (0.97–1.82)||0.06|
|Whole milk: ≥1 serving/day vs. never/seldom||1.18 (0.68–2.03)||0.71|
|Yoghurt: >4 servings/week vs. never/seldom||1.26 (0.59–2.67)||0.69|
|Hard cheese: >4 servings/week vs. never/seldom||0.65 (0.43–0.97)||0.05|
|Serous epithelial ovarian tumors (174)|
|Lactose: >19.8 g/day vs. <5.0 g/day||2.07 (1.27–3.40)||0.003|
|Skim/low-fat milk: >1 serving/day vs. never/seldom||1.69 (1.12–2.56)||0.006|
|Whole milk: ≥1 serving/day vs. never/seldom||0.98 (0.46–2.12)||0.76|
|Yoghurt: >4 servings/week vs. never/seldom||2.35 (1.09–5.06)||0.04|
|Hard cheese: >4 servings/week vs. never/seldom||0.82 (0.49–1.40)||0.41|
|Larsson and Wolk, 2005 Sweden (1987–2004)||288 invasive EOC (61,057)||All epithelial ovarian tumors:||Age, education, OC use, parity, BMI, fruits, vegetables, energy|
|Lactose: ≥21.9 g/d vs. <9.9 g/day||1.48 (1.50–2.09)||0.04|
|Total dairy foods: ≥4 servings/day vs. <servings/day||1.61 (1.07–2.41)||0.02|
|Total milk: ≥2 servings/day vs. never/seldom||1.40 (0.99–1.98)||0.07|
|Low-fat milk: >1 serving/day vs. never/seldom||1.27 (0.90–1.79)||0.13|
|Whole milk: >1 serving/day vs. never/seldom||1.16 (0.71–1.88)||0.83|
|Yoghurt: ≥1 serving/day vs. never/seldom||0.89 (0.63–1.26)||0.54|
|Hard cheese: ≥3 servings/day vs. never/seldom||1.24 (0.72–2.16)||0.51|
|Serous epithelial ovarian tumors (135 cases):|
|Lactose: ≥21.9 g/d vs. <9.9 g/day||1.96 (1.15–3.35)||0.009|
|Total dairy foods: ≥4 servings/day vs. <servings/day||2.15 (1.73–3.97)||0.02|
|Total milk: ≥2 servings/day vs. never/seldom||1.82 (1.08–3.09)||0.04|
|Low-fat milk: >1 serving/day vs. never/seldom||1.69 (1.01–2.84)||0.04|
|Whole milk: >1 serving/day vs. never/seldom||1.66 (0.86–3.22)||0.56|
|Yoghurt: ≥1 serving/day vs. never/seldom||0.90 (0.54–1.50)||0.68|
|Hard cheese: ≥3 servings/day vs. never/seldom||1.17 (0.51–2.67)||0.73|
|Study and location (years of data collection)||Cases (Participation rate)||Controls (Participation rate)||Measure of exposure and range of exposure||RR (95%CI)||P1||Covariates|
|Cramer et al.,1984 United States (1978–1981)||215 EOC (NA) interviewed within ∼3 months from diagnosis||215 age-, race, and area-matched population controls (NA)||Skim milk: regular use vs. no regular use||0.58 (0.39–0.86)||—||Age, race, and study area|
|Whole milk: regular use vs. no regular use||1.51 (1.03–2.21)||<0.05|
|Cheese: regular use vs. no regular use||0.66 (0.19–2.36)||—|
|Mori et al., 198814 Japan (1981–1981, 1985–1986)||110 EOC (NA)||110 age- matched hospital controls (NA)||Drinks milk daily||0.6 (0.4–1.0)||—||Age|
|Cramer et al., 198922 United States (1983–1987||235 EOC [including borderline tumors (69%)]||239 age-, race, and area-matched population controls (51%)]||Skim milk: ≥1 serving/month vs. <1 serving/month||1.0 (0.7–1.5)||—||Age, education, religion OC use, parity, energy|
|Whole milk: ≥1 serving/month vs. <1 serving/month||1.1 (0.7–1.6)||—|
|Yoghurt: ≥1 serving/month vs. <1 serving/month||1.7 (1.1–2.4)||0.01|
|Cottage cheese: ≥1 serving/month vs. <1 serving/month||1.4 (1.0–2.1)||0.08|
|Mettlin et al., 199024 United States (1982–1988||303 EOC (NA)||606 age-matched hosptial controls (NA)||Skim milk: >1 glass/day vs. never/seldom||0.7 (0.3–1.2)||0.06||Age|
|Whole milk: >1 glass/day vs. never/seldom||3.1 (1.8–5.5)||<0.001|
|Engel et al., 199125 United States (1984–1989||71 EOC (NA)||141 age- and ethnicity-matched hosptial controls (NA)||Lactose: ≥14.7 g/day vs. ≤2.8 g/day||0.9 (0.4–2.0)||—||Smoking, BMI, β-carotene|
|Risch et al., 199419 Canada (1989–1992||450 EOC [including borderline tumors (71%)], 9% of cases died before contacted, interviewed within ˜3 months from diagnosis||564 age-matched population controls (65%)||Lactose: ≥20 g/day vs. <6 g/day||1.07 (0.72–1.59)||—||Age, parity, OC use, fat, fiber, energy|
|Herrinton et al., 199520 United States (1989–1991||108 stage 1 EOC (60% of eligible)||108 [70 friends (74%) and 98 population (53%)] controls matched by age||Lactose: Highest vs. lowest quartile||0.80 (0.52–1.20)||—||Age, race|
|Webb et al., 199826 Australia (1990–1993||˜8002 EOC [90% of eligible cases, ˜74% of ascertained cases, (including borderline tumors)], 6% of cases died before contacted||˜8002 age- and area-matched population controls (73%)||Lactose: >31 g/day vs. ≤8 g/day||0.97 (0.67–1.41)||0.5||Age, education, smoking, parity, OC use, BMI, energy|
|Total milk: ≥2.5 glasses/day vs. <0.5 glass/day||1.03 (0.73–1.46)||0.97|
|Skim milk: >1 glass/day vs. never/seldom||0.57 (0.34–0.95)||0.001|
|Low-fat milk: >1 glass/day vs. never/seldom||0.91 (0.59–1.39)||0.2|
|Whole milk: >1 glass/day vs. never/seldom||1.35 (0.88–2.07)||0.03|
|Low-fat yoghurt: >4 servings/week vs. never/seldom||0.87 (0.51–1.46)||0.3|
|Full-fat yoghurt: >1 serving/week vs. never/seldom||1.23 (0.82–1.85)||0.004|
|Hard cheese: >1 serving/day vs. <1 serving/week||1.38 (0.81–2.36)||0.2|
|Cramer et al., 200023 United States (1992–1997||563 EOC [including borderline tumors (55%)], 9% died before contacted||523 age-matched population controls (˜60%)||Latcose: Highest vs. lowest quartile||0.92 (0.62–1.37)||0.69||Age, study area, education, OC use, parity, family history of breast or ovarian cancer|
|Skim/low-fat milk: > glass/day vs. never/seldom||1.10 (0.74–1.64)||—|
|Whole milk: >1 glass/day vs. < glass/month||0.64 (0.32–1.26)||—|
|Yoghurt: > glass/day vs. never/seldom||0.85 (0.18–4.00)||—|
|Cottage cheese: > servings/week vs. never/seldom||1.04 (0.65–1.66)||—|
|Bertone et al., 200127 United States (1991–1994)||327 EOC (65%), 23% died before contacted||3129 age—matched population controls (75%)||Total milk:≤7.0 servings/week vs never/seldom||0.84(0.63–1.1)||0.20||Age, state, parity, tubal ligation, family history of ovarian cancer|
|Skim/low-fat milk:≤8.0 servings/week vs never/seldom||0.67(0.46–1.0)||0.09|
|Whole milk:≤4.0 servings/week vs never/seldom||1.1(0.75–1.6)||0.72|
|Yoghurt:≤2.0 servings/week vs never/seldom||1.2(0.86–1.6)||0.80|
|Cottage cheese:≤2.0 servings/week vs ≥1.0 serving/week||1.1(0.84–1.5)||0.27|
|Bosetti et al,. 200128 Italy (1992–1999||1031 EOC (96%)||2411 hospitals controls (96%)||Total milk: Highest vs lowest quintile||0.97(0.73–1.29)||0.73||Age, study area, education, parity, OC use, energy|
|Cheese: Highest vs lowest quintile||0.74(0.55–0.98)||0.04|
|Zhang et al,. 200229 China (1999–2000||254 EOC (100%), 75% interviewed within 12 months from diagnosis||652 (601 hospital and 51 population) controls matched by age and area (96%)||Milk and products: > g/day vs <27 g/day||1.18(0.5–2.6)||NS||Age, education, study area, income, marital status, menopausal status, parity, tubal ligation, OC use, smoking, family history of ovarian cancer, physical activity, BMI, tea, alcohol, energy|
|Salazar-Martinez et al., 200216 Mexico (1995–1997)||84 EOC (100%)||629 age-matched hospital controls (98%)||Lactose: ≥33 g/day vs ≤23 g/day||0.54 (0.29–0.99)||0.04||Age, parity, weight changes, physical activity, diabetes, energy|
|Total dairy foods: Highest vs. lowest tertile||1.09 (0.50–2.36)||0.58|
|Goodman et al., 200217 United States (1993–1999)||558 EOC [431 invasive and 127 borderline tumors (62%)]||607 age-, ethnicity-, and area-matched population controls (67%)||Lactose: >15.9 g/day vs. <3.8 g/day||0.61 (0.42–0.89)||0.007||Age, ethnicity, study area, education, OC use, parity, tubal ligation, energy|
|Total dairy foods: >372 g/day vs. <94 g/day||0.59 (0.41–0.87)||0.003|
|Total milk: >262 g/day vs. <34 g/day||0.64 (0.45–0.91)||0.009|
|Low-fat milk: >196 g/day vs. <7.9 g/day||0.59 (0.41–0.83)||0.002|
|Whole milk: >37 g/day vs. <4.4 g/day||1.06 (0.72–1.54)||0.80|
|Yoghurt: >43 g/day vs. 0 g/day||0.01 (0.72–1.43)||0.68|
|Cheese: >44 g/ vs. <6.4 g/day||0.89 (0.57–1.39)||0.88|
|Cozen et al., 200313 United States (1992–1998)||290 invasive EOC (48% of ascertained, 78% of eligible cases) 27% died) or were too ill) before contacted||645 age-, ethnicity-, and area-matched population controls (NA)||Lactose: >19.7 g/day vs. <5.2 g/day||1.0 (0.66–1.52)3||0.53||Age, ethnicity, study area, education, socioeconomic status|
|McCann et al., 200318 United States (1993–1991)||124 EOC (NA), 93% interviewed within 6 months from diagnosis||696 age–matched population controls(NA)||Total dairy foods: >348 g/day vs <61 g/day||1.29(0.70–2.39)||—||Age, education, menstrual history, difficult becoming pregnant, OC use,, menopausal status, energy|
|Yen et al., 200415 Taiwan(1993–1998)||86 invasive EOC (96%),interviewed within ˜3–4 months from diagnosis, 5% died before contaced||396 hospital controls matched by age, hospital, and date of admission (93%)||Drinks milk: yes vs no||0.45 (0.28–0.74)||–||Age, income,during marriage, education|
|Pan et al., 200430 Canada (1994–)||442 EOC (57% of ascertained, 69% of eligible cases), 11.4% died before contacted||2135 age-matched population controls (% of contacted, 65% of ascertained controls)||Total dairy foods: Highest vs. lowest quintile||1.20 (0.89–1.62)||0.25||Age, residence, education, parity, menstruation years, menopausal status, smoking, BMI, physical activity, alcohol, energy|
|Total milk: Highest vs. lowest quintile||1.25 (0.91–1.69)||0.11|
|Low-fat milk: Highest vs. lowest quintile||1.19 (0.86–1.67)||0.27|
|Cheese: Highest vs. lowest quintile||0.86 (0.63–1.17)||0.75|