• lactose;
  • ovarian cancer;
  • milk


  1. Top of page
  2. Abstract
  6. Acknowledgements

The milk sugar lactose is an hypothesized risk factor for epithelial ovarian cancer because of possible direct toxic effects of its metabolites on oocytes or by compensatory gonadotropin stimulation. Women are presently encouraged to consume dairy products as a source of calcium to prevent osteoporosis. The objective of our study was to prospectively assess lactose, milk and milk product consumption in relation to ovarian cancer risk among 80,326 participants in the Nurses' Health Study who had no history of cancer other than nonmelanoma skin cancer. Participants in the Nurses' Health Study reported on known and suspected ovarian cancer risk factors in questionnaires mailed biennially from 1976 to 1996. Food frequency questionnaires were included in the years 1980, 1984, 1986 and 1990. Newly reported ovarian cancer was documented by review of medical records. During 16 years of follow-up (1980–1996), 301 cases of invasive epithelial ovarian cancer were confirmed. Pooled logistic regression was used to control for age, body mass index (kg/m2), caffeine intake, oral contraceptive use, smoking history, parity and tubal ligation. For all subtypes of invasive ovarian cancer combined, we observed a nonsignificant 40% greater risk for women in the highest category of lactose consumption compared to the lowest (multivariate relative risk (RR) 1.40, 95% confidence interval (CI), 0.98–2.01). We observed a 2-fold higher risk of the serous ovarian cancer subtype among those in the highest category of lactose consumption compared to the lowest (RR 2.07, 95% CI, 1.27–3.40). For each 11-gram increase in lactose consumption (the approximate amount in one glass of milk), we observed a 20% increase in risk of serous cancers (RR 1.20, 95% CI, 1.04–1.39). Skim and low-fat milk were the largest contributors to dietary lactose. Women who consumed one or more servings of skim or low-fat milk daily had a 32% higher risk of any ovarian cancer (RR 1.32, 95% CI, 0.97–1.82) and a 69% higher risk of serous ovarian cancer (RR 1.69, 95% CI, 1.12–2.56) compared to women consuming 3 or less servings monthly. Controlling for fat intake did not change our findings. Our findings provide some support for the hypothesis that lactose intake increases risk of epithelial ovarian cancer. However, the observed excess risk appeared limited to the serous subtype of ovarian cancer in our study. © 2004 Wiley-Liss, Inc.

Ovarian cancer is the 5th most common malignancy among U.S. women, with 25,580 new diagnoses expected in 2004.1 Ovarian cancer etiology is poorly understood, though proposed models include cumulative effects of repeated ovulation2 and exposure of the ovary to high gonadotropin levels.3 Increasing parity, oral contraceptive use, lactation history and tubal ligation are the only established factors associated with decreased risk.4 Differences in ovarian cancer incidence between countries5 as well as migrant studies6 suggest lifestyle and dietary patterns as possible etiologic factors.

Lactose is hypothesized as a risk factor for ovarian cancer because of the possible effects of its metabolites on the ovary and on the ovarian-pituitary axis. Lactose is a disaccharide found only in milk and milk products and is cleaved by intestinal lactase to produce glucose and galactose. Galactose is further metabolized in the liver by a series of enzymes including galactose-1-phosphate uridyl transferase (GALT). In classic galactosemia, a rare mutation of the GALT gene results in the absence of enzymatic activity. Girls with galactosemia become hypogonadal, have elevated gonadotropins and subsequently develop ovarian failure.7 Although there are no reports of ovarian cancer in this rare disorder, little residual ovarian function remains.8 High levels of gonadotropins secondary to ovarian failure may also contribute indirectly to ovarian carcinogenesis.9, 10

There are 2 ecologic studies that suggest a relationship between per capita milk consumption and increased ovarian cancer incidence or mortality.11, 12 Case-control studies of ovarian cancer using recalled history of milk and dairy consumption have yielded conflicting results.10, 13, 14, 15, 16, 17, 18 In the only cohort study to examine these relationships, the Iowa Women's Health Study reported increased risk associated with skim milk intake based on 139 incident cases. A positive but statistically nonsignificant association between lactose and ovarian cancer risk also was observed.19

Since current dietary recommendations for women include increasing calcium intake20 and milk is widely promoted as a source of calcium, adverse health effects of milk consumption would have important public health implications. We, therefore, examined the relation of lactose consumption to ovarian cancer risk using 16 years of follow-up in the Nurses' Health Study.


  1. Top of page
  2. Abstract
  6. Acknowledgements

Study population

In 1976, 121,700 married, female registered nurses from 11 US states, aged 30–55 years, completed a mailed questionnaire on known or suspected risk factors for cancer and coronary heart disease. Follow-up questionnaires were mailed biennially to update information on risk factors and newly diagnosed diseases. We added a 61-item food frequency questionnaire (FFQ) in 1980 and expanded this to 126 items in 1984. Participants completed the expanded FFQ again in 1986 and 1990 to update information on diet. The follow-up rate through May 31, 1996 was 95% of potential person-years.

Lactose and dairy foods assessment

Details of the semiquantitative FFQ and documentation of its reproducibility and validity have been published elsewhere.21, 22, 23, 24 The 1980 questionnaire was derived by identifying 61 foods allowing maximal discrimination for intakes of specific fats, fiber and 12 other nutrients. For each food item, the questionnaire specified a common serving size and queried respondents on average intake of that food in the past year; responses in 9 categories ranged from almost never to 6 or more per day. In 1980, the dietary assessment included intake of skim or low-fat milk (8 oz glass, 240 mL), whole milk (8 oz glass, 240 mL), yogurt (1 cup, 240 mL), ice cream (1/2 cup, 120 mL), cottage cheese (1/2 cup,120 mL), hard cheeses (slice or 1 oz serving, 28 grams) and butter (number of pats). When the questionnaire was expanded in 1984, additional questions assessed intake of sour cream (1 tablespoon, 15 mL), cream cheese (1 oz, 28 grams) and other cheeses (1 oz, 28 grams). We also queried women in the 1980 questionnaire on whether intake of each dairy product was greatly increased or decreased over the prior 10 years. In a validation study using 2 questionnaires 1 year apart and 2 one-week diet records 6 months apart, we found Pearson correlation coefficients between the 2nd dietary questionnaire and the diet records of 0.81 for skim and low-fat milk, 0.62 for whole milk and 0.94 for yogurt.24

We calculated consumption of lactose and other nutrients by multiplying the frequency of intake by the nutrient content of specified portions. Each cup of milk (any type) contains approximately 11 grams of lactose.25 In this cohort in 1980, 55% of total dietary lactose came from skim or low-fat milk, 23% from whole milk, 5% from yogurt, 5% from ice cream, 4% from cottage cheese and 0.8% from hard cheese. By 1990, the relative contributions were: skim/low-fat milk, 68%; whole milk, 5%; yogurt, 6%; ice cream, 3%; cottage cheese, 2% and hard cheese, 0.4%. Mixed foods generally containing milk such as mashed potatoes and pudding also contributed a small amount to total lactose.

Ascertainment of other covariates

We assessed exposure status for multiple known or proposed risk factors for ovarian cancer from responses to the biennial questionnaires. Oral contraceptive use was assessed from 1976 to 1982, when use became rare due to the age of the participants. We used duration of oral contraception use (never, less than 3 years, 3 to 5 years or more than 5 years) as a covariate since this measure of exposure has been consistently associated with decreased risk in our study and others.26, 27, 28, 29 Parity, defined as number of pregnancies lasting at least 6 months, was queried from 1976 through 1984 and was categorized as none, 1 to 2, 3 to 4 or more than 4. Age at menopause was queried in 1976 and biennially thereafter and was categorized as premenopausal, less than 45, 45–49, 50–52 and 53 or greater. Tubal ligation history (ever, never) was assessed in each cycle from 1976 to 1982 and in 1994. Smoking history was assessed in each questionnaire and was categorized as never, past or current use. Women reported height and weight in 1976 and weight was updated biennially thereafter. We assessed other dietary factors including dietary fat, fiber, protein, caffeine, calcium and vitamin D from foods as described above. In addition, participants reported vitamin and supplement use including calcium and vitamin D. All analyses used updated covariates when data were available.

Ascertainment of ovarian cancer cases

We identified incident cases of epithelial ovarian cancer reported in each biennial questionnaire from 1980 to 1996. For women reporting such a history, we obtained discharge summaries and pertinent pathology reports. Trained physician reviewers unaware of exposure status extracted information on histologic type, subtype, morphology and stage. Deaths in the cohort were identified through family members, the US Postal Service and the National Death Index. We estimated that 98 percent of all deaths for the entire cohort were ascertained.30


We excluded women who did not complete the 1980 FFQ or had implausibly high or low dietary intakes (n = 29,233). Additional exclusions included reported diagnosis of cancer besides nonmelanoma skin cancer (n = 3,703) and history of bilateral oophorectomy, hysterectomy with unknown number of ovaries removed or history of pelvic irradiation (n = 8,438). We previously reported 97% agreement between self-reported surgical menopause and medical records.31 After exclusions, 80,326 women remained in the analysis at the start of follow-up. Exclusions were updated biennially. During 16 years of follow-up and 1,142,686 person-years, 527 cases of ovarian cancer were reported by participants. Of these cases, 42 were nonepithelial, we could not confirm 46 due to unavailable medical records or nurse refusal, 102 were rejected after reviewing records or contacting the nurse and 337 (301 of which were invasive) were confirmed by medical record review.

Statistical analysis

We calculated person-years for participants from the 1980 questionnaire return date to the date of ovarian cancer diagnosis, the date of death or May 31 1996, whichever was sooner. Incidence rates were calculated by dividing the number of cases by the number of person-years in each category of lactose consumption. We computed relative risks by dividing the rates for categories 2 through 5 by that for category 1. We used pooled logistic regression with 2-year time increments32 to control for age (in 5-year increments), body mass index (kg/m2), caffeine intake, duration of oral contraceptive use, parity, tubal ligation and smoking. We found no change in the association between lactose and ovarian cancer when we additionally adjusted for other factors including age at first birth, age at menarche and menopause, menopausal status, postmenopausal estrogen use and physical activity. Similarly, we found no evidence of confounding by dietary fat, fiber, other nutrients (including vitamin A, the carotenoids, vitamin E, vitamin C and folate) or vitamin supplements. Therefore, these variables were not included in models shown here. We reported our findings on the relationship between antioxidants and ovarian cancer,33 and dietary fat and ovarian cancer34 elsewhere. We used stratified analyses and the likelihood ratio test (comparing models with and without interaction terms) to assess for effect modification by each of the factors in the final model as well as postmenopausal estrogen use and dietary fat. We reported RRs and 95% CIs.

We adjusted lactose for total energy intake using the nutrient residual method.35 Additional control for total energy in multivariate models did not alter the findings when added to the models. Ovarian cancer risk associated with lactose and dairy food intake was calculated using diet reported in 1980, diet from 1980 to 1990 using simple updating and diet from 1980 to 1990 using a method of cumulative updating. For assessment of cumulative exposure, we used an updated average intake, for example, 1980 intake for the 1980–1984 time period, the average of 1980 and 1984 diet for 1984–1986 time period and so on.36 We also assessed effects of consistency of each dairy food by restricting the analysis to participants who, on the 1980 FFQ, did not report greatly increased or decreased consumption over the previous 10 years.

In additional analyses, we assessed the role of family history of ovarian cancer and breast feeding history using only the last 4 and 10 years of follow-up, respectively, because these exposures were not ascertained until later in follow-up. We also assessed exposure latency by conducting analyses using 1980 diet only, most recent diet and most recent diet with time lags of either 4–6 or 8–10 years. For example, in the analysis of a 4–6 year lag, 1980 diet was used to predict risk from 1984–1988, 1984 diet for 1988–1990, 1986 diet for 1990–1994 and 1990 diet for 1994–1996.


  1. Top of page
  2. Abstract
  6. Acknowledgements

We confirmed 301 incident cases of invasive ovarian cancer from 1980 to 1996 among women eligible for follow-up. Daily median lactose consumption for women in the lowest, middle and highest categories were 3.2, 11.1 and 26.0 grams, respectively. Higher lactose consumption was associated with increasing calcium and vitamin D intake and inversely correlated with caffeine and fat intake. Women with higher lactose consumption were more likely to be never-users of oral contraceptives and less likely to have had a tubal ligation (Table I).

Table I. Age-Standardized Baseline Characteristics of 80,326 Women According to Category of Energy-Adjusted Lactose consumption1
 Category of lactose Consumption
1 n = 16,0782 n = 16,2193 n = 16,1914 n = 15,8905 n = 15,948
  • 1

    All p-values < 0.01 except parity, postmenopausal hormone use, and family history of ovarian cancer.

  • 2

    Among women who used oral contraceptives.

  • 3

    Among parous women, lactation history ascertained in 1986.

  • 4

    Family history ascertained in 1992.

 Age at first birth24.724.724.824.924.9
 Duration oral contraceptive use (months)249.547.947.946.746.3
 Duration of lactation (months)
 Body mass index (kg/m2)24.124.324.224.324.4
Percent of study women     
 Never user of oral contraceptive52.051.049.850.449.6
 Current postmenopausal estrogen user9.59.410.09.59.8
 Current smoker36.228.827.025.526.7
 History of tubal ligation17.417.316.816.115.9
 Family history of ovarian cancer42.
Mean daily intake     
 Total energy (kilojoules)6,4816,4026,8956,4226,577
 Calcium, food and supplements (mg)4505576738181,165
 Total vitamin D (IU)176216266331467
 Dietary caffeine (mg)429421401385360
 Dietary fat (grams)7473706864

Lactose intake appeared positively associated with epithelial ovarian cancer, with a 40% increase in risk (all subtypes combined) for women in the highest category of cumulative average intake compared to the lowest, (RR 1.40; 95% CI, 0.98–2.01), although a test of trend was not statistically significant (Table II). For women with serous ovarian cancer, we observed a 2-fold increase in risk for those with the highest lactose intake compared to the lowest (RR 2.07; 95% CI, 1.27–3.40, p for trend 0.003), (Fig. 1). For each 11-gram increment of lactose per day (8 oz of milk), we observed a 20% increase in risk of serous cancers (RR 1.20, 95% CI, 1.04–1.39). Additional control for dietary fat or hormone replacement therapy in the models did not change the results. Findings were similar when evaluated within strata of each of the covariates in the model, history of postmenopausal estrogen use, oral contraceptive use, menopausal status or fat intake. In analyses limited to mucinous (n = 61), endometrioid (n = 45) or other subtypes (n = 31) of ovarian cancer, no substantial associations with dietary lactose were observed, although we had limited power to examine these relationships.

Table II. Multivariate Relative Risks of Ovarian Cancer Among 80,326 Women Between 1980 and 1996 According to Updated Energy-Adjusted Lactose Intake
 Category of lactose consumption (median and range of lactose intake, grams/day)
3.2 (0–4.9)7.0 (5.0–8.9)11.1 (9.0–13.3)16.1 (13.4–19.8)26.0 (19.9–110.1)p-value for trend
  • 1

    Model adjusted for age, body mass index (kg/m2), caffeine intake, duration of oral contraceptive use, parity, tubal ligation and smoking. Lactose is energy-adjusted and updated using the cumulative updating method.

All invasive tumors (n = 301)n = 51n = 66n = 46n = 65n = 73 
 Age adjusted RR (95% CI)1.001.35 (0.94–1.95)0.93 (0.63–1.39)1.31 (0.91–1.89)1.44 (1.00–2.06)0.07
 Multivariate RR (95% CI)11.001.36 (0.94–1.96)0.93 (0.62–1.39)1.30 (0.90–1.88)1.40 (0.98–2.01)0.10
Serous invasive tumors (n = 174)n = 24n = 35n = 27n = 40n = 48 
 Age adjusted RR (95% CI)1.001.53 (0.91–2.58)1.17 (0.68–2.03)1.72 (1.04–2.86)2.01 (1.23–3.29)0.005
 Multivariate RR (95% CI)*1.001.56 (0.93–2.63)1.19 (0.69–2.07)1.78 (1.07–2.96)2.07 (1.27–3.40)0.003
thumbnail image

Figure 1. Spline curve: risk of serous ovarian cancer according to lactose consumption. Multivariate relative risk of serous ovarian cancer according to lactose intake as a continuous variable. The 95% confidence intervals for the relative risk are shown as gray bands. Relative risks are adjusted for age, body mass index, caffeine intake, duration of oral contraceptive use, parity, tubal ligation and smoking.

Download figure to PowerPoint

Results using only the 1980 dietary assessment with person-time through 1996 were somewhat attenuated, with a 48% increase in risk in the highest category of lactose intake compared to the lowest for serous cancers (RR 1.48; 95% CI, 0.93–2.34, p for trend 0.02). For all subtypes combined, we observed little association with lactose assessed in 1980 (RR for highest vs. lowest intake 1.06, 95% CI, 0.75–1.52).

When we evaluated dairy foods, skim or low-fat milk consumption was modestly associated with overall ovarian cancer risk when cumulatively updated consumption was considered (RR 1.32, 95% CI, 0.97–1.82, p for trend 0.06) (Table III). For serous ovarian cancers, women consuming one or more servings of skim and low-fat milk daily had a 69% higher risk (RR 1.69, 95% CI, 1.12–2.56, p for trend 0.006). Women consuming 5 or more servings of yogurt weekly also appeared at increased risk for serous cancer (RR 2.35, 95% CI, 1.09–5.06), although this relationship was weaker for all subtypes combined (Table III). We noted no association between updated whole milk consumption and ovarian cancer risk. Of note, the median number of servings in the highest category of both whole and skim milk consumption was 2.0. Similarly, analyses for specific cheese products and butter showed no association with risk of ovarian cancer. Ice cream consumption was associated with a nonsignificant 63% greater risk of serous tumors among women consuming 5 or more servings weekly. In analyses, combining servings of all types of milk (skim, low-fat milk or whole milk), we found a 55% higher risk of serous cancers (RR 1.55 for one or more servings daily, 95% CI, 1.00–2.40, p for trend 0.01).

Table III. Multivariate Relative Risks of Ovarian Cancer Among 80,326 Women Between 1980 and 1996 According to Cumulative Updated Dairy Food Consumption1
 Frequency of consumption
Almost never, 1–3/month1/week2–4/week5–7/week1 or more/dayp-value for trend
  • 1

    Models in table are adjusted for age, body mass index (kg/m2), duration of oral contraceptive use, parity, tubal ligation and smoking history. Population median servings per week as follows: skim or low-fat milk, 4–5 servings per week; whole milk, 1–2 per week; yogurt, less than 1 per week; ice cream, 1–2 per week.

All invasive tumors (n = 301)      
 Skim or low-fat milk (240 mL)1.001.12 (0.77–1.62)0.82 (0.56–1.19)1.12 (0.80–1.57)1.32 (0.97–1.82)0.06
 n = 103n = 42n = 38n = 50n = 68 
 Whole milk (240 mL)1.001.21 (0.86–1.70)1.02 (0.67–1.55)0.93 (0.57–1.54)1.18 (0.68–2.03)0.71
 n = 204n = 41n = 25n = 17n = 14 
 Yogurt (240 mL)1.001.19 (0.87–1.62)0.80 (0.46–1.37)1.26 (0.59–2.67)0.69
 n = 230n = 50n = 14n = 7 
 Ice cream (120 mL)1.001.09 (0.84–1.41)0.85 (0.58–1.24)1.24 (0.70–2.18)0.65
 n = 173n = 84n = 31n = 13 
 Hard cheese (28 grams)1.000.86 (0.62–1.19)0.84 (0.61–1.16)0.65 (0.43–0.97)0.05
 n = 59n = 94n = 107n = 41 
 Cottage cheese (120 mL)1.000.86 (0.65–1.14)1.11 (0.79–1.55)0.82 (0.43–1.56)0.67
 n = 41n = 183n = 65n = 43 
Serous invasive tumors (n = 174)      
 Skim or low-fat milk (240 mL)1.001.04 (0.61–1.76)1.14 (0.71–1.82)1.35 (0.86–2.12)1.69 (1.12–2.56)0.006
 n = 53n = 20n = 27n = 30n = 44 
 Whole milk (240 mL)1.001.19 (0.77–1.85)0.89 (0.50–1.58)0.85 (0.43–1.68)0.98 (0.46–2.12)0.76
 n = 121n = 24n = 13n = 9n = 7 
 Yogurt (240 mL)1.001.33 (0.89–1.97)0.83 (0.41–1.70)2.35 (1.09–5.06)0.04
 n = 128n = 31n = 8n = 7 
 Ice cream (120 mL)1.001.19 (0.84–1.68)0.97 (0.59–1.60)1.63 (0.82–3.26)0.21
 n = 96n = 50n = 19n = 9 
 Hard cheese (28 grams)1.001.00 (0.64–1.56)0.96 (0.62–1.49)0.82 (0.49–1.40)0.41
 n = 30n = 56n = 62n = 26 
 Cottage cheese (120 mL)1.000.75 (0.51–1.11)1.27 (0.83–1.94)1.19 (0.58–2.48)0.45
 n = 105n = 33n = 28n = 8 

When only 1980 diet was considered for the dairy foods analysis, skim or low-fat milk consumption was less strongly associated with risk of serous tumors compared to analyses using cumulative updating (RR for one or more daily servings 1.35, 95% CI, 0.86–2.13), as was yogurt consumption (RR for 5 or more weekly servings 1.63, 95% CI, 0.83–3.20). However, whole milk consumption in 1980 appeared more strongly associated with risk of serous tumors than in analyses using updated diet, although we had limited power to assess this relationship (RR for one or more daily servings 1.53, 95% CI, 0.80–2.92).

Models using most recent updated diet to assess lactose intake revealed similar results to the main models with cumulative updating (for all tumors combined, RR 1.41 for highest category compared to the lowest, 95% CI, 0.97–2.05; for serous tumors, RR 1.80 for highest category, 95% CI, 1.08–3.00). We did additional analyses to examine 4–6 and 8–10 year latencies between the time of dietary assessment and start of follow-up. These analyses showed similar results to the main models for the 4–6 year lag and attenuation of the association with the 8–10 year exposure lag. However, there were only 155 total invasive and 93 serous invasive tumors remaining in the analysis for the 8–10 year exposure lag, limiting our statistical power. We found no change in the observed association between lactose and ovarian cancer after adjusting for breast feeding history in analyses restricted to 1986–1996 follow-up (after lactation history was ascertained) or family history of ovarian cancer (1992–1996 follow-up). In analyses restricted to women who, in 1980, reported no substantial increase or decrease in dairy foods over the past decade, we found similar associations between dairy foods and ovarian cancer risk. Since intake changed very little among most women, we were unable to evaluate associations among the small group of women who did report substantial changes in lactose consumption. We did not have adequate statistical power to adequately assess risk associated with consistently high or low lactose intake.

Milk and milk products provide the main source of calcium in the Western diet. We found a high correlation between lactose and calcium from food (r = 0.90). When both lactose and calcium from food were included in the model, the association for lactose remained similar (RR for serous cancers, category 5: 2.15, 95% CI, 0.90–5.16), while calcium from food did not appear to be associated with ovarian cancer risk (RR for category 5: 0.85; 95% CI, 0.36–2.00). When lactose was removed from the model, calcium from food was modestly associated with serous ovarian cancer (RR for category 5: 1.44, 95% CI, 0.91–2.29). Similarly, the RR for category 5 of total calcium (including food and supplements) was 1.47 (95% CI, 0.88–2.47) without lactose in the model.

We did not include tumors with borderline morphology in our main analyses. Including both borderline and invasive tumors in additional analyses revealed somewhat attenuated results compared to analyses including only invasive tumors. For all tumors combined (n = 337), we observed an RR of 1.27 for the highest category of lactose intake compared to the lowest, 95% CI, 0.91–1.78. For the serous subtype (n = 200), we observed an RR of 1.81, 95% CI, 1.15–2.84).


  1. Top of page
  2. Abstract
  6. Acknowledgements

In our prospective cohort study, we observed an elevated risk of ovarian cancer associated with high cumulative lactose consumption that was statistically significant for the serous subtype. We found significant associations between serous cancers and consumption of both skim and low-fat milk and yogurt. Lactose and calcium consumption are closely correlated, but we found no evidence that calcium in milk is responsible for our results. In contrast with other reports,13, 14, 15, 16, 17 dietary fat did not influence our observations. We observed no association for whole milk and ovarian cancer, but by 1990, whole milk contributed less than 5% of total dietary lactose. We did see a positive association with whole milk consumption in 1980 when whole milk contributed to 23% of dietary lactose.

The pertinent exposure period remains unclear. In the our study, we did not find differences in risk associated with lactose intake according to menopausal status. The latency period for ovarian cancer is also unknown. Analyses using only 1980 diet did demonstrate increased risk, albeit more modest than updated analyses. Introduction of lag periods showed results similar to cumulative updated diet for lags of 4–6 years between dietary assessment and disease follow-up, but attenuated results after 8–10 years. The consistent findings after introducing a 4–6 year lag makes it less likely that undetected preclinical disease accounts for our findings. The attenuation in results with longer lags could indicate that lactose is acting as a promoter, but we had considerably less cases in these analyses to assess such relationships. Only with further follow-up will we be able to assess, in detail, longer latencies such as 10 or more years. The strongest associations were observed in analyses of cumulative diet, which could suggest that longer term exposure is also important, or could result from minimizing misclassification by using repeated dietary intake measures.

Potential limitations of our study include the ability of FFQ to measure dairy intake. However, the FFQ has been used to document diet-disease associations,37 and an earlier validation study showed a high correlation between 4-week diet records and the FFQ for milk.24 Measurement error is expected to be random and would generally bias results towards the null. Although we did not observe a monotonic increase in risk across categories of lactose, we did observe a significant test for trend for serous tumors. This lack of a dose-response may be due to small numbers of cases in some categories. We did not have adequate power to examine these relationships in more detail. Alternatively, the lack of obvious dose-response or threshold effect may indicate a chance finding in the highest intake categories.

Another potential limitation is our inability to control for family history of ovarian cancer. However, we found little variation in the percentage of participants with a family history across categories of lactose. Since our data were collected prospectively, recall bias is unlikely to influence the findings. We have prospective measures of most known ovarian cancer risk factors. Our cohort was mainly Caucasian, potentially limiting generalizability. However, we expect our findings to be generally applied to populations able to digest lactose. We could not assess the impact of large changes in lactose intake. However, among women reporting little change in lactose consumption, our findings were similar. We used the residual method as described to adjust for total calories and found no evidence of additional confounding by calories when we examined this in our final models.

Our finding that the serous subtype was most strongly related to lactose consumption may represent a chance finding and was not an a priori hypothesis. We examined subtypes of ovarian cancer based on findings by others that ovarian cancer risk factors may differ by subtype.38, 39, 40 However, few studies have reported on risk factors by subtype, and our result may reflect differences in tumor biology that merit further study. Where possible, future studies should examine differences in dietary and other risk factors by subtype.

Exposure to bovine growth hormones (rBGH) used to augment milk production in dairy cows is also unlikely to explain our current findings. rBGH was not widely used until 1993, several years after our last dietary assessment. Also, the majority of our cases occurred before this point, and analyses excluding cases diagnosed after 1992 showed similar results.

Another possible explanation for the association between dairy consumption and ovarian cancer is that consumption of dairy products increases serum insulin-like growth factor I (IGF-I). IGF-I is positively associated with several malignancies, possibly including ovarian cancer.41 There are 3 studies that have now reported that increased milk consumption is associated with higher serum IGF levels.42, 43, 44 Since IGF-I is a protein hormone (subject to digestion by intestinal proteases), absorption from the diet is unlikely to contribute substantially to serum levels, although few data are available to assess this possibility. Although protein and energy restriction both decrease serum IGF-I,44, 45 it remains unclear how dairy products might influence endogenous IGF-I production.

Our results are similar to those from the Iowa Women's Health Study, the only previous prospective study to examine this association.19 That study, with 139 epithelial cancers, reported a RR of 1.60 for the highest quartile of lactose intake (95% CI, 0.95–2.70). Their findings for skim milk are similar to ours, with a RR of 1.73 in those women consuming more than one serving daily. Yogurt consumption was also positively associated with risk in previous studies.10, 17 A recent study also showed a trend towards worsened survival after ovarian cancer diagnosis with higher lactose intake.46 Several case control studies13, 14, 15, 16, 17 examining lactose and ovarian cancer relationships have not found an association, and 2 have shown decreased ovarian cancer risk associated with higher milk intake.47, 48 These conflicting findings may be due to the absence of a true association, or because of inability to measure a true association because of using a single dietary assessment, selection and recall bias or limited power to assess serous cancers.

There are 2 studies that have examined lactose absorption in relation to ovarian cancer risk, with differing results.16, 49 Although countries with higher proportions of the population able to absorb lactose may have increased ovarian cancer incidence,12 data from ecologic studies may be misleading.

Animal models suggest direct ovarian toxicity of galactose,50, 51 and girls with classic galactosemia develop ovarian failure unless lactose intake is restricted.7, 8 Although classic galactosemics have no functional GALT, several polymorphisms result in variable degrees of GALT activity.52, 53, 54 The N314D polymorphism is the most common polymorphism, present in 4–14% of the population.55 Heterozygotes for N314D have 69–75% of normal GALT activity, while homozygotes have 43–50%.55, 56 Cramer et al.10 found an increased risk associated with an increased ratio of lactose to GALT activity. More recent studies since GALT genotyping became available also have examined the possible association with ovarian cancer risk. Morland et al.57 showed an allele frequency for N314D of 0.167 among 63 serous cancers compared to 0.089 among 248 controls (p = 0.03), and another report found a higher prevalence of N314D homozygosity in women with endometrioid and clear cell ovarian cancers.18 Women with a family history of ovarian cancer also may have increased prevalence of the N314D polymorphism.58 In contrast, several recent studies showed no significant associations between GALT activity or N314D genotype and ovarian cancer risk.16, 17, 59, 60

Osteoporosis is an important public health problem, resulting in substantial morbidity61 and healthcare costs.20, 62 Current recommendations suggest postmenopausal women consume 1,000—1,500 mg of calcium per day.20 To reach this goal from dairy alone, a woman must drink approximately 3 glasses (8 ounce) of milk daily. This would result in 33 grams of daily lactose; more than the median in our highest risk group. Despite the hypothesized benefit of greater milk intake, several large studies have observed no association between dairy consumption and fracture risk in adult life.63, 64, 65, 66 Alternative dietary sources of calcium are described in a recent review.67

In summary, we did not observe a significant overall relationship between dietary lactose and ovarian cancer. Among the serous subtype only, we found a 2-fold increase in risk associated with dietary lactose at a level recommended for optimal calcium intake. These data are consistent with other prospective data and are biologically plausible given the potential ovarian toxicity of galactose and resultant elevations in gonadotropin stimulation.


  1. Top of page
  2. Abstract
  6. Acknowledgements

We thank the participants in the Nurses' Health Study for their continuing cooperation; and to Gary Chase, Karen Corsano, Barbara Egan, Diane Feskanich, Dorota Gertig, Frank Hu, Mary Louie, Bernard Rosner and Shumin Zhang. Dr. Fairfield received a Cancer Prevention Training Grant (CA57711-08) from the National Institutes of Health.


  1. Top of page
  2. Abstract
  6. Acknowledgements