SEARCH

SEARCH BY CITATION

Keywords:

  • breast cancer;
  • diet;
  • vegetarianism;
  • migrants;
  • South Asians

Abstract

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

To investigate the role of lifelong vegetarianism on the aetiology of female breast cancer, we conducted a population-based case-control study among South Asian migrant women from the Indian subcontinent resident in England. A total of 240 South Asian breast cancer cases were identified from 2 cancer registries during 1995–1999. For each case, 2 age-matched South Asian controls were randomly selected from the age-sex register of the case practice. Lifelong vegetarians had a slight reduction, although not statistically significant, in the odds of breast cancer relative to lifelong meat-eaters, which persisted after adjustment for socio-demographic and reproductive variables [odds ratio (OR)=0.77; 95% confidence interval (CI)=0.50–1.18]. Analysis by food group revealed no linear trend in the odds of breast cancer with increasing consumption of meat (p=0.10) but the odds were higher for women in the top 75%. In contrast, there were strong inverse trends in the odds of breast cancer with increasing intake of vegetables (p=0.005), pulses (p=0.007) and fibre [non-starch polysaccharides, NSP (p=0.02)], with women in the highest 25% of intake of these foods having about 50% of the odds of those in the lowest ones. Adjustment for intake of vegetables and pulses reverted the odds of breast cancer in lifelong vegetarians relative to lifelong meat-eaters (OR=1.04; 95% CI=0.65–1.68) and attenuated the quartile-specific estimates for meat intake, whereas the inverse trends in the odds of breast cancer with intake of vegetables and pulses remained after adjustment for type of diet or meat intake. These findings suggest that lifelong vegetarianism may be associated with a reduction in the risk of breast cancer through its association with a higher intake of vegetables and pulses. Although it is not possible to exclude the possibility that lifelong meat abstention may also play a role, the findings provide evidence that a diet rich in vegetables and pulses, such as those typically found in South Asian diets, may be protective against this cancer. © 2002 Wiley-Liss, Inc.

Migrant1, 2 and ecological3 studies have supported the hypothesis that diet may play a role in the aetiology of female breast cancer. However, results from cohort and case-control studies have been inconsistent. Deficiencies in study design or inaccuracies of dietary assessment methods that introduce non-differential misclassification of exposure leading to relative risk estimates biased towards unity could have contributed to the conflicting results. In addition, a common difficulty in investigating relationships between dietary factors and breast cancer risks with case-control or cohort studies, especially if they are likely to be weak associations, is the relative homogeneity of the diet of Western populations where most such epidemiological studies have been carried out. South Asian populations such as those in India, Pakistan and Bangladesh are characterised by a large variability in dietary intakes with most of them influenced by strong religious and regional variations. Most Hindus do not eat any type of meat nor eggs but consume milk and its products, whereas Muslims are predominantly meat-eaters, although they do not consume pork. However, in contrast to Western vegetarians who usually stop meat consumption only in adult life, Hindus are vegetarians from early life. We report here findings on the role of lifelong vegetarianism in the aetiology of breast cancer from a population-based case-control study conducted among women of South Asian ethnicity who have migrated to England.

METHODS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The study was carried out in the catchment areas of the Thames and West Midlands population-based cancer registries, which cover some of the areas in England with the largest numbers of residents of South Asian origin.4 Within each registry, we restricted, for logistic reasons, the study to health authorities with reasonable numbers of South Asian patients, that is, essentially those in Greater London, Birmingham, Coventry and Wolverhampton. The study began initially as a pilot in Greater London and was later extended to the other study areas. Thus, cases were identified through the Thames registry from December 1995 to March 1999 and through the West Midlands registry from July 1997 to May 1999. The study was approved by all relevant ethics committees.

Selection of cases and controls

Cases were women of South Asian ethnic origin, under 75 years, with a newly diagnosed breast cancer, who were born in the Indian subcontinent or in East Africa, resided in 1 of the study areas, were reported to the Thames and West Midlands cancer registries during the study period and were still alive at the time of reporting. Cases registered more than 2 years after diagnosis were excluded. For each case, 2 population-based controls, individually matched to the case on year of birth (within 5 years), were randomly selected from all South Asian female patients born in the Indian subcontinent or East Africa who were registered with the same general practitioner (GP) as the case at the time of diagnosis. Occasionally, a neighbouring GP practice was used if no eligible controls were available from the case GP's list. Cases and controls with a previous diagnosis of any cancer, suffering from conditions requiring strict diets (e.g., insulin-dependent diabetes mellitus) or from mental or psychiatric disorders that would reduce the accuracy of the information elicited in an interview were excluded.

Since information on ethnicity and country of birth in cancer registry data or in GPs' lists was very incomplete, potential South Asian cases and controls were identified on the basis of their names and subsequent confirmation of ethnicity and country of birth was obtained from the GPs and the patients themselves. The validity of name analyses to identify people of South Asian ethnicity in Britain has been widely demonstrated.5, 6 After obtaining permission from her GP, a letter (in English and in the most likely language spoken by the subject) was sent to the woman explaining the aim of the study. To avoid motivation and recall biases, the study was described as a general survey on women's health and no mention was made of breast cancer. Failure in tracing a woman was only accepted after at least 2 home visits.

Cases and controls were interviewed at their home in English (48% of cases and 52% of controls) or in their mother tongue by trained interviewers. A structured questionnaire was developed to obtain data on socio-demographic, reproductive and general health variables. Dietary data were collected through a food-frequency questionnaire (FFQ) specifically designed for South Asian populations resident in England (see below). Although it was not always possible to keep the interviewers blind to the case-control status of the subjects throughout the whole interview, dietary questions were always completed before the section on general health. Measurements of weight, standing and sitting heights and hip and waist circumferences were taken at the end of the interview using standard procedures. The median time between case diagnosis and interview was 15 months (inter-quartile range: 12–19 months). For each matched triplet, events that occurred after the date on which the corresponding case was diagnosed were excluded from the analysis.

Food-frequency questionnaire (FFQ)

The interviewer-administered FFQ was developed to assess the habitual dietary intake of the participants 2–3 years prior to an interview. The development of this FFQ has been described in detail elsewhere.7 Briefly, the FFQ was organised in sections according to food groups. The foods/dishes included were those that covered the range consumed by the various South Asian ethnic subgroups and were likely to contribute to inter-individual variation in energy and macronutrient intake.7 The FFQ included 24 questions on meat dishes, 40 questions on vegetables, 21 on pulses, lentils and dhals and 23 on fruits. Frequency of consumption was ascertained by asking women how often they ate the particular quantity of food per day/week/month or never. For seasonal foods, such as mangoes, women were asked to estimate their average intake when the food was in season. Portion sizes were estimated on the basis of “natural” units (e.g., 1 egg), or by asking the women to select from previously validated sets of 8 colour photographs that represent different portion sizes8, 9 or by using average portion sizes derived from published sources.10–13 Additional questions ascertained the main eating habits (e.g., consumption of alcohol, meat and fish and type of fat used in cooking) and changes throughout life. Nutrient intakes were computed on the basis of the frequency of consumption of each unit of food and nutrient content of the specified portions. Data on the nutrient composition of traditional South Asian foods/dishes were available from recent publications.14–16 Nutrient analyses were conducted in COMP-EAT.17

This FFQ was validated against the average of 12 monthly 24 hr dietary recalls collected from each of a subset of 100 controls who participated in the present case-control study (Sevak et al., personal communication). The agreement between the 2 methods was 100% for type of diet (vegetarian/meat-eater) and high for food and nutrient intake. The percentage of women classified in the same ± 1 quartile by the 2 methods was 74% for total energy and 85%, 85%, 96% and 90% for energy-adjusted fat, carbohydrates, proteins and non-starch polysaccharides (NSP), respectively. Spearman correlation coefficients for total energy and energy-adjusted fat, carbohydrates, protein and NSP, were 0.45, 0.59, 0.56, 0.76 and 0.71, respectively. Similar results were observed for vegetarians and meat-eaters. The percentage of vegetarians who were classified in the same ±1 quartile by the 2 methods was 69%, 90%, 90%, 96% and 87% for total energy, total fat, carbohydrates, proteins and NSP, respectively. The equivalent figures for meat-eaters were 79%, 79%, 79%, 96% and 94%, respectively.

Statistical methods

Conditional logistic regression models,18 with each case and her 2 age- and GP-matched controls being identified in a common set, were used to analyse the data, taking into account potential confounding variables. Nutrient intake, independent of total energy intake, was estimated by using the residuals of the regression of the nutrient on total energy intake (both on a logarithmic scale due to heteroscedasticity of residuals on an untransformed scale).19 Quartiles of food consumption and nutrient residuals were defined according to the distribution in the controls. The cut-off points that define the energy-adjusted nutrient quartiles shown in the tables correspond to the nutrient range at 1,826 kcal, the median energy intake in controls. Tests for trend in the odds of breast cancer with food/nutrient quartiles were based on the likelihood-ratio test between the models with and without a linear term for food/nutrient's quartiles. To correct attenuation of the odds ratios due to non-differential errors in the measurement of nutrient intakes by the FFQ, corrected odds ratios were calculated using the linear approximation method of regressing 24 hr recall quartile scores on FFQ quartile scores in the validation study.19, 20

RESULTS

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

After excluding potential South Asian female breast cancer cases who were found not to be South Asian or not to have been born in the Indian subcontinent or East Africa (77), those who had died (51) or those who did not fulfil the eligibility criteria (23), there were 352 potentially eligible cases. Forty-nine were excluded because the patients and/or their GPs could not be traced. Of those traced, 62 were excluded because the patients or their GPs declined to participate and for 1, it was not possible to obtain appropriate controls. Thus, a total of 240 cases were included in the study, a response rate of 79% among South Asian cases who could be traced. Of the 480 eligible controls initially identified, 365 agreed to participate, which was a response rate of 76%. Second selected controls were interviewed to replace those initially selected who refused to participate. For 3 cases it was only possible to obtain 1 matched control.

Cases and controls had, as expected, similar ages at the time of diagnosis of the cases (matching variable) (Table I). There was no difference between cases and controls in time since migration to the United Kingdom but cases were more likely to have come from Pakistan. Relative to the controls cases had, on average, a higher educational level and a higher current social class, had a younger age at menarche, were older when their first child was born, were less likely to have ever breastfed but more likely to have a positive family history of breast cancer and parents who were first-degree relatives. Only 8.3% of cases and 6.5% of controls reported any intake of alcohol. There were no differences between cases and controls in relation to adult height, current body mass index (BMI), current waist:hip ratio and use of oral contraceptives or hormone replacement therapy.

Table I. Distribution of Cases and Controls by Socio-Demographic and Reproductive Characteristics
Cases n = 240Controls n = 477p-value for heterogeneity
Numbers (%) unless stated otherwise
  • 1

    Includes Gujerati Muslims, East African Ismaili Muslims and Bengali Hindus.

  • 2

    Measured as either the social class of the woman or of her partner, whichever was the highest.

  • 3

    Includes work in the armed forces, unemployed and missing information on occupation.

  • 4

    Parous women only (n = 672).

  • 5

    Post-menopausal women only.

Socio-demographic characteristics
 Age at case's diagnosis(mean (SD))51.45 (9.0)51.87 (9.2)
 Region/religion of originPakistani Muslim57 (23.8)83 (17.4)0.01
Gujerati Hindu73 (30.4)177 (37.1)
Punjabi Sikh58 (24.2)105 (22.0)
Bangladeshi Muslim6 (2.5)30 (6.3)
Punjabi Hindu18 (7.5)54 (11.3)
Other128 (11.7)28 (5.9)
 Current household social class2Non-manual147 (61.3)252 (52.8)0.01
Manual86 (35.8)214 (44.9)
Unclassified37 (2.9)11 (2.3)
 Years of formal educationMedian (inter-quartile range)10 (7–13)10 (5–13)0.04
 Years in the UKMean (SD)23.7 (8.4)23.2 (8.6)0.64
Reproductive-related characteristics
 Age at menarcheMean (SD)13.6 (1.8)13.8 (1.7)0.10
 Age at first full-term birth4Mean (SD)24.1 (4.9)23.5 (4.5)0.11
 ParityMean (SD)3.0 (1.8)3.2 (2.0)0.13
 Family history of breast cancer%6.33.80.08
 % post-menopausal%57.952.00.11
 Years since menopause5Median (inter-quartile range)6.7 (3.1–14.7)9.2 (4.5–14.5)0.14
 % who breastfed4%74.180.40.04
 Parents who were first-degree relatives%9.68.00.37

Lifelong vegetarianism was associated with a slight, although not statistically significant, reduction in the odds of breast cancer relative to lifelong meat-eaters, which persisted after adjusting for socio-demographic and reproductive variables (Table II). Further adjustment for height, current BMI, and use of oral contraceptives and hormone replacement therapy did not affect these results. The point estimates for women who changed from being vegetarians to being meat-eaters, or vice-versa, were based on small numbers. There was no evidence of a linear trend in the odds of breast cancer with increasing intake of meat dishes among current meat-eaters (p=0.23), although the odds were higher for the top 75% (Table III). Meat dishes contain meat and varying proportions of vegetables and staples but repetition of the analyses using total meat intake, as estimated from recipe information, gave similar results (p=0.18 for linear trend).

Table II. Odds Ratios (and 95% Confidence Intervals) for Breast Cancer by Dietary Group
Adjusted for age at case's diagnosis (through matching) and the following potential confounders1Life-long meat-eaterMeat-eater in childhood, vegetarian in adulthoodVegetarian during childhood, meat-eater in adulthoodLife-long vegetarians
OR (95% CI)OR (95% CI)OR (95% CI)
  • 1

    Categorisation of the potential confounders are as follows: age at menarche: <13, 13, 14, ≥15 years; age at first birth: <20, 20–24, 25–29, ≥30 years; parity: 0, 1–2, 3–4, ≥5; family history of breast cancer: yes/no; menopausal status: pre/peri/post; time since menopause: <5 years, 5–9, 10–14, ≥15 years; ever breast-fed: yes/no; adult socio-economic group: measured as either the social class of the woman or of her partner, whichever was the highest; years of formal education: 0, 1–7, 8–13, ≥14; years in the UK: continuous.

Number of cases/controls134/24716/3210/1779/179
10.86 (0.43, 1.73)0.89 (0.36, 2.23)0.77 (0.53, 1.13)
Age at menarche10.86 (0.43, 1.75)0.92 (0.37, 2.32)0.81 (0.55, 1.19)
Age at first full-term birth10.82 (0.41, 1.66)0.83 (0.33, 2.10)0.76 (0.51, 1.13)
Parity10.84 (0.42, 1.68)0.93 (0.37, 2.31)0.74 (0.50, 1.09)
Family history of breast cancer10.83 (0.41, 1.68)0.90 (0.36, 2.27)0.77 (0.52, 1.13)
Menopausal status and years since menopause10.85 (0.42, 1.72)0.96 (0.38, 2.43)0.76 (0.52, 1.12)
Ever breast-fed10.80 (0.40, 1.62)0.81 (0.32, 2.03)0.74 (0.50, 1.09)
Adult socio-economic group10.79 (0.39, 1.61)0.96 (0.38, 2.46)0.83 (0.56, 1.24)
Years of education10.91 (0.45, 1.83)0.92 (0.37, 2.30)0.77 (0.52, 1.13)
Years in the UK10.83 (0.41, 1.69)0.87 (0.35, 2.19)0.77 (0.53, 1.13)
Adjusted for all of the above variables10.66 (0.31, 1.44)0.84 (0.31, 2.26)0.77 (0.50, 1.19)
Table III. Odds Ratio for Breast Cancer by Intake of Selected Foods
Food groupsQuartileRangeOR195% CILRT for linear trend: p-value
  • 1

    Odds ratio (OR) adjusted for total energy intake, age at menarche (<13, 13, 14, ≥15 years), age at first birth (<20, 20–24, 25–29, ≥30 years), parous (yes/no), parity (1–2, 3–4, ≥5), breast feeding (never/ever), family history of breast cancer (yes/no), menopausal status (pre/peri/post), time since menopause (<5 years, 5–9, 10–14, ≥15 years) and years of formal education (0, 1–7, 8–13, ≥14).

  • 2

    p-value for heterogeneity.

Meat and meat dishes (servings/week)Vegetarian00.870.48, 1.590.102
10.1–2.51
22.6–4.81.450.75, 2.81
34.9–6.91.520.78, 2.95
4≥7.01.590.79, 3.21
Breads (g/day)1<13210.41
2132–1840.750.45, 1.22
3185–2641.140.70, 1.87
4≥2650.620.38, 1.14
Vegetable dishes (g/day)1<21010.005
2210–2990.710.44, 1.16
3300–4050.520.31, 0.87
4≥4060.480.27, 0.85
Pulses, lentils and dhals (g/day)1<35.010.007
235.0–67.50.960.61, 1.52
367.6–107.30.580.35, 0.97
4≥107.40.540.31, 0.94
Fruit (servings/day)1<0.8910.45
20.89–1.611.170.72, 1.92
31.62–2.490.970.51, 1.47
4≥2.500.890.50, 1.57

Lifelong vegetarians differed from meat-eaters not only in terms of meat consumption but also in their intake of vegetables and fruits. Among controls lifelong vegetarians had a much higher median daily intake of vegetables (332 g vs. 249 g) and pulses, lentils and dhals (92 g vs. 46 gr.) than meat-eaters but similar consumption of fruit (1.6 servings vs. 1.6 servings) and bread (184 g vs. 175 g). Analysis by food groups (Table III) exhibited strong inverse linear trends in the odds of breast cancer with increasing intake of vegetables (p=0.005) and pulses (p=0.007) but no clear trend with increasing intake of breads or fruits.

There were no associations between intakes of total energy, fat, protein and carbohydrates and the odds of breast cancer (Table IV). In contrast, there was a strong inverse association between the odds of this cancer and intake of NSP (p=0.02 for linear trend) with women in the highest 25% having only 61% of the odds of those in the lowest 1 (Table IV). This linear trend corresponded to an odds ratio (OR) of 0.83 [95% confidence interval (CI)=0.70–0.98; p=0.02] per unit increase in NSP quartile. After correcting for non-differential exposure measurement errors associated with the FFQ, the OR became stronger (corrected OR per unit increase in NSP quartile=0.75; 95% CI= 0.59–0.97; p=0.03). There was a marginal statistically significant inverse linear trend in the odds of breast cancer with intake of NSP from vegetables and pulses, but not from cereals or fruits, after adjusting for total energy intake and mutually for each other (Table IV). Among controls, the mean daily intake of NSP was 23.3 g/day (SD=6.8) in vegetarians and 19.3 g/day (6.6) in meat-eaters; these values corresponded to a mean of 11.9 g/1,000 kcal (1.9) and 10.4 g/1,000 kcal (1.9), respectively. The effect of NSP on breast cancer was, however, similar among vegetarians and meat-eaters (test for interaction p=0.47), with vegetarians in the top 25% of NSP intake having only 62% (95% CI=0.26–1.49) of the odds of those in the bottom 25%. Similarly, meat-eaters in the top 25% of NSP intake had only 47% (95% CI=0.24–0.95) of the odds of those in the lowest 25%.

Table IV. Odds Ratios for Breast Cancer by Intake of Macronutrients and Intake of NSP from Different Sources
MacronutrientsQuartileRange1OR (p-value for linear trend)95% CIAdjusted OR2 (p-value for linear trend)95% CI
  • 1

    The nutrient range for each energy-adjusted nutrient quartile corresponds to the appropriate range at the median energy intake in controls of 1826 kcal.

  • 2

    Odds ratio (OR) adjusted for age at menarche (<13, 13, 14, ≥15 years), age at first birth (<20, 20–24, 25–29, ≥30 years), parous (yes/no), parity (1–2, 3–4, ≥5), breast feeding (never/ever), family history of breast cancer (yes/no), menopausal status (pre/peri/post), time since menopause (<5 years, 5–9, 10–14, ≥15 years) and years of formal education (0, 1–7, 8–13, ≥14).

  • 3

    Quartiles of nutrient residuals from regression of nutrient on total energy (both on a log scale).

  • 4

    Each quartile of residual of NSP source is adjusted for total energy and the other 3 NSP sources.

Energy (kcal/day)1<155711
21557, 18901.040.67, 1.591.120.71, 1.77
31891, 22241.030.65, 1.620.960.59, 1.56
4≥22250.990.62, 1.580.840.51, 1.40
(p = 0.97)(p = 0.94)
Fat3 (g/day)1<7111
271–780.900.58, 1.400.870.55, 1.38
379–861.020.66, 1.581.050.65, 1.68
4≥870.760.47, 1.240.720.42, 1.21
(p = 0.41)(p = 0.36)
Protein3 (g/day)1<5011
250–562.071.25, 3.411.901.10, 3.28
357–631.590.96, 2.631.610.94, 2.76
4≥641.070.62, 1.851.170.64, 2.13
(p = 0.85)(p = 0.79)
Carbohydrates3 (g/day)1<22011
2220–2381.250.80, 1.951.150.72, 1.83
3239–2541.100.69, 1.771.090.65, 1.82
4≥2551.260.78, 2.021.280.77, 2.10
(p = 0.46)(p = 0.39)
NSP3 (g/day)1<17.611
217.6–20.40.840.55, 1.290.900.56, 1.43
320.5–23.20.610.38, 0.970.600.36, 1.00
4≥23.30.540.34, 0.880.610.36, 1.02
(p = 0.005)(p = 0.02)
NSP from different sources, adjusted for the other 3 NSP sources
 NSP from cereals34 (g/day)1<6.711
26.7–9.31.160.76, 1.761.150.72, 1.85
39.4–11.70.480.29, 0.780.480.28, 0.83
4≥11.80.660.41, 1.070.870.51, 1.49
(p = 0.006)(p = 0.10)
 NSP from pulses34 (g/day)1<0.911
20.9–1.61.390.88, 2.181.180.73, 1.91
31.7–2.50.910.56, 1.480.780.46, 1.32
4≥2.60.740.45, 1.250.660.38, 1.15
(p = 0.10)(p = 0.06)
 NSP from vegetables34 (g/day)1<4.211
24.2–5.70.910.59, 1.390.780.49, 1.23
35.8–7.40.820.52, 1.290.670.41, 1.09
4≥7.50.670.41, 1.110.660.39, 1.11
(p = 0.11)(p = 0.09)
 NSP from fruit34 (g/day)1<1.411
21.4–2.30.840.44, 1.630.800.48, 1.33
32.4–3.50.640.27, 1.470.610.36, 1.02
4≥3.60.970.32, 2.921.050.62, 1.78
(p = 0.98)(p = 0.80)

To assess whether the reduction in risk among vegetarians relative to meat-eaters was due to abstinence from meat or to higher intake of vegetables and pulses, 2 separate models were fitted. First, when both type of diet and intake of vegetables and pulses were included in the same model, the odds of breast cancer in lifelong vegetarians relative to lifelong meat-eaters was attenuated (from 0.77 to 1.04), whereas the magnitude of the quartile-specific estimates for vegetables and pulses was little affected and the tests for trend remained of borderline statistical significance (Table V, model A). Second, when both meat intake and intake of vegetables and pulses were included in the same model, the quartile-specific estimates for meat were attenuated, whereas the trend in the odds of breast cancer with intake of vegetables and pulses persisted, although they were no longer statistically significant (Table V, model B).

Table V. Odds Ratios for Breast Cancer by Dietary Group and Servings of Meat Per Week, Adjusted for Vegetable and Pulse Consumption
 Odds ratio195% CI Odds ratio195% CI
  • 1

    Odds ratio (OR) adjusted for total energy intake, age at menarche (<13, 13, 14, ≥15 years), age at first birth (<20, 20–24, 25–29, ≥30 years), parous (yes/no), parity (1–2, 3–4, ≥5), breast feeding (never/ever), family history of breast cancer (yes/no), menopausal status (pre-/peri-/post-), time since menopause (<5 years, 5–9, 10–14, ≥15 years) and years of formal education (0, 1–7, 8–13, ≥14).

  • 2

    p-value for heterogeneity.

  • 3

    p-value for linear trend.

Adjusted for breast cancer risk factors only
Dietary groupMeat and meat dishes (servings/week)
 Life-long meat-eater100.870.48, 1.59
 Was meat eater, now vegetarian0.680.32, 1.450.1–2.51
 Was vegetarian, now meat-eater0.860.32, 2.302.6–4.81.450.75, 2.81
 Life-long vegetarian0.770.50, 1.184.9–6.91.520.78, 2.95
(p = 0.572)≥7.01.590.79, 3.21
(p = 0.102)
Adjusted for breast cancer risk factors and for the other two dietary variables
Model AModel B
Dietary groupMeat and meat dishes (servings/week)
 Life-long meat-eater100.950.51, 1.77
 Was meat eater, now vegetarian0.850.39, 1.880.1–2.51
 Was vegetarian, now meat-eater0.940.34, 2.622.6–4.81.280.65, 2.52
 Life-long vegetarian1.040.65, 1.684.9–6.91.350.68, 2.66
(p = 0.972)≥7.01.250.60, 2.60
(p = 0.722)
Vegetable dishes (g/day)Vegetable dishes (g/day)
 <2101<2101
 210–2990.800.49, 1.30210–2990.770.47, 1.27
 300–4050.600.35, 1.04300–4050.620.36, 1.08
 ≥4060.640.36, 1.12≥4060.610.33, 2.60
(p = 0.073)(p = 0.083)
Pulses, lentils, dhals (g/day)Pulses, lentils, dhals (g/day)
 <35.01<35.01
 35.0–67.50.980.61, 1.5835.0–67.50.980.61, 1.57
 67.6–107.30.620.36, 1.0767.6–107.30.680.40, 1.17
 ≥107.40.640.36, 1.15≥107.40.680.37, 1.24
(p = 0.063)(p = 0.113)

Because a lifelong type of diet may be a marker of acculturation or other lifestyle habits that are causally associated with breast cancer, we examined the association between type of diet and breast cancer stratified by whether or not they migrated to Britain via East Africa, age at migration and main language spoken at home. The effect of lifelong vegetarianism on breast cancer was slightly stronger among women who migrated to the UK via East Africa (OR=0.50; 95% CI=0.11–2.22) than in those who came directly to the UK (OR=0.85; 0.37–1.95), although the test for interaction was not significant (p=0.51) but was similar in women who migrated to the UK when they were under 25 years of age (the median age at migration) and in those who migrated later in life (p for interaction=0.59) and in women who predominantly spoke English at home relative to those who used their mother tongue (test for interaction p=0.52). There was also no evidence that the effect of lifelong vegetarianism was modified by menopausal status (p=0.17), but the power was limited, or by time from diagnosis to interview of the cases (p=0.61).

DISCUSSION

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

This is the first study of breast cancer among South Asian first-generation migrant women resident in England and 1 of the few to examine the role of lifelong vegetarianism. One of the strengths of the study derives from the use of population-based controls. Despite this there is still potential for selection bias for various reasons. First, cases were identified from population-based cancer registries to ensure reasonable numbers but case ascertainment by registries was probably incomplete. To minimise possible bias, cases and controls were matched on GP to ensure as much as possible that had a control been diagnosed with breast cancer, she would have had a similar probability as her matched case of being entered into the cancer registry files. In addition, there were delays between diagnosis and registration, with some cases dying before registration or before it was logistically possible to organise a home interview. Examination of breast cancer risks by time since diagnosis did not suggest, however, that the observed associations between dietary factors and breast cancer were due to an underlying association between diet and survival. Second, selection bias might have been introduced even though the response rate among both cases and controls can be regarded as high for a population-based study, but examination of the few variables available from the cancer registration data indicated that untraceable and non-responders were similar to the participants in terms of their age, area of residence and stage. Third, as migrants with similar religious and social origins tend to concentrate in certain geographical areas and controls were selected from the same GP practices as the cases there is a possibility of overmatching, that is, of both groups having similar religious and socio-economic backgrounds and, hence, similar diets. This effect is unlikely to be strong and, if present, would bias odds ratios towards unity. It is reassuring, however, that the distribution of the controls by ethnic origin was similar to that that would have been expected on the basis of the 1991 Census.4 The observed/expected percentages were 77%/78%, 18%/14% and 6%/9% for Indians, Pakistanis and Bangladeshis, respectively.

Diet plays an important social role in South Asian populations and is largely determined by strong religious and regional influences. Diet and its recall may, therefore, be less influenced by a recent diagnosis of breast cancer than in other populations. Although FFQs are the only feasible method of assessing past dietary intake in case-control studies, there have been concerns with regard to their validity. In particular, it has been argued that the protective effect of fibre observed in some studies could be due to the fact that FFQs usually include a long list of vegetables and fruits that lead to over-reporting of consumption. The findings in the present study did not change when non-differential exposure measurement errors introduced by the FFQ were corrected in relation to an average of 12 monthly 24 hr recalls collected from a subset of the controls. Twenty-four hour recalls agree surprisingly well with weighed records and food diaries21, 22 and multiple 24 hr recalls have been widely used as the reference method in FFQ validation studies23 but their sources of error may not be totally independent since both methods rely on memory. The mean intake of vegetables and fruits among South Asians in our study were, however, similar to those reported by others that used different dietary assessment methods.13, 24 After controlling for known breast cancer risk factors, the results were virtually identical to the age-adjusted relative risks, suggesting that residual confounding by known lifestyle factors is unlikely to explain the findings. We cannot exclude unknown lifestyle factors, but they would need to be predictors of breast cancer and also associated with type of diet and intake of vegetables and pulses.

The present study found no clear association between meat intake and breast cancer risk. This finding is consistent with results from a hospital-based case-control study carried out in Bombay (India).25 It is also consistent with findings from large follow-up studies of Western vegetarians and religious groups who tend to abstain from meat.26–30 In these studies, however, most subjects cease meat consumption in adult life and therefore it was not possible to exclude the possibility that meat intake might play a role in the causation of breast cancer if pre-adult life is a more critical period. In contrast, the dietary habits of South Asian populations are adopted very early in life and therefore our findings seem to exclude that possibility.

Our study showed that high intakes of vegetables and pulses were associated with protection against breast cancer and that this could potentially account for the reduced breast cancer risk observed among vegetarians. A similar protective effect for vegetable consumption was found in a meta-analysis of 14 case-control and 3 cohort studies.31 No protective effect was found in a pooled analysis of data from 7 cohorts32 but, in some, ascertainment of fruit and vegetable intake33, 34 was poor. Moreover, 2 of the cohorts included in this analysis have since reported protective effects for vegetable intake. The inverse trend between NSP intake (a marker of fruit and vegetable consumption) and the odds of breast cancer observed in the present study is also consistent with that from a meta-analysis of case-control studies,35 which reported a weak, but statistically significant, protective association. Several recent studies, not included in the meta-analysis, have found similar statistically significant protective associations with various components of fibre,36–38 crude fibre39 and fibre from vegetables and fruit, but not cereals.40 No association between total dietary fibre intake and subsequent incidence of breast cancer was reported by 2 prospective studies conducted in the US,41, 42 but a weak protective effect was found in a prospective study conducted in Canada.43 Intake of vegetables in Western countries is lower than in South Asian populations. The median intake of total vegetables, including pulses, among the controls in our study was 355 g/day, much higher than in the cohorts included in the pooled analysis (range: 77–226 g/day32). Thus, the lower intake of vegetables and dietary fibre might have reduced the likelihood of finding associations in studies conducted among Western populations. The type of vegetables typically found in South Asian diets is also different from those found in Western diets with a greater preponderance of pulses.

In short, the findings from the present study seem to suggest that lifelong vegetarianism may be associated with a reduction in the risk of breast cancer through its association with a higher intake of vegetables and pulses. Although it is not possible to exclude the possibility that meat abstention may also play a role, the findings provide evidence that a lifelong diet rich in vegetables, such as those typically found in South Asians diets, may be protective against this cancer. Further studies in these populations will help to identify the precise foods and nutrients associated with this protective effect. In the meantime, these findings suggest that women with a diet rich in vegetables may be protecting themselves not only in relation to cardiovascular diseases44 and other (mainly digestive) cancers45 but also in relation to breast cancer.

Acknowledgements

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

We thank Mrs. N. Khurana and Mrs. R. Kabawala for their clerical support throughout the study; Dr. T. Kassam-Khamis for allowing us access to her South Asian nutrient database; Ms. B. Peachey and Mrs. H. Nguyen for their help with data entry; and Mrs. D. Lathiga, Mr. J. Gaffney and the late Ms. T. Pierpoint for their help with the home interviews. We are also grateful to staff at the Thames and West Midlands Cancer Registries for allowing us access to their data; the general practitioners who allowed us to interview their patients; the health authorities for their help with GP's patient lists; and all the patients and controls who helped us with the study.

REFERENCES

  1. Top of page
  2. Abstract
  3. METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES
  • 1
    Shimizu H, Ross RK, Bernstein L, et al. Cancers of the prostate and breast among Japanese and white immigrants in Los Angeles County. Br J Cancer 1991;63: 9636.
  • 2
    Ziegler RG, Hoover RN, Pike MC, et al. Migration patterns and breast cancer risks in Asian-American women. J Natl Cancer Inst 1993;85: 181927.
  • 3
    Armstrong B, Doll R. Environmental factors and cancer incidence and mortality in different countries, with special reference to dietary practices. Int J Cancer 1975;15: 61731.
  • 4
    Peach C. Ethnicity in the 1991 census, vol 2: The ethnic minority populations of Great Britain. London: HMSO, 1996.
  • 5
    Nicoll A, Bassett K, Ulijaszek SJ. What's in a name? Accuracy of using surnames and forenames in ascribing Asian ethnic identity in English populations. J Epidemiol Community Health 1986;40: 3648.
  • 6
    Nanchahal K, Mangtani P, Alston M, et al. Development and validation of a computerised South Asian Names and Group Recognition (SANGRA) for use in British health-related studies. J Publ Health Med 2001;23: 27885.
  • 7
    Kassam-Khamis T, Nanchahal K, Mangtani P, et al. Development of an interview-administered food-frequency questionnaire for use amongst women of South Asian ethnic origin in Britain. J Hum Nutr Dietet 1999;12: 719.
  • 8
    Nelson M, Atkinson M, Darbyshire S. 1994. Food photography. I. The perception of food portion size from photographs. Br J Nutr, 1994;72: 64963.
  • 9
    Nelson M, Atkinson M, Darbyshire S. 1996. Food photography. II. Use of food photographs for estimating portion size and the nutrient content of meals. Br J Nutr 1996;76: 3149.
  • 10
    MAFF (Ministry of Agriculture, Fisheries and Food). Food portion sizes, 2nd ed. London: HMSO, 1994.
  • 11
    Gregory J, Foster K, Tyler H, et al. The dietary and nutritional survey of British adults. London: HMSO, 1990.
  • 12
    Henderson JB, Dunnigan MG, McIntosh WB, et al. Asian osteomalacia is determined by dietary factors when exposure to ultraviolet radiation is restricted: a risk factor model. Q J Med 1990;76: 92333.
  • 13
    Anderson AS, Lean ME. Healthy changes? Observations on a decade of dietary change in a sample of Glaswegian South Asian migrant women. J Hum Nutr Dietet 1995;8: 12936.
  • 14
    Judd PA, Kassam-Khamis T, Thomas JE. The composition and nutrient content of foods commonly consumed by South Asians in the UK. London: Aga Khan Health Board for the United Kingdom, 2000.
  • 15
    Holland B, Welch AA, Unwin ID, et al. McCance and Widdwson's the composition of foods, 5th ed. Cambridge: Royal Society of Chemistry, 1995.
  • 16
    Holland B, Welch AA, Buss DH. Vegetable dishes: The second supplement to McCance and Widdowson's the composition of foods, 5th edn. Cambridge: Royal Society of Chemistry, 1996.
  • 17
    Nutrition Systems. COMP-EAT, version 5. London: Carlson Bengston Consultants Limited.
  • 18
    Breslow NE, Day NE. Statistical methods in cancer research, vol 1: The analysis of case-controls studies. Lyon: International Agency for Research on Cancer, 1980.
  • 19
    WillettW, ed. Nutritional epidemiology, 2nd ed. Monographs in Epidemiology and Biostatistics, vol 30. Oxford: Oxford University Press, 1998.
  • 20
    Rosner B, Willett WC, Spiegelman D. Correction of logistic regression relative risk estimates and confidence intervals for systematic within-person measurement errors. Stats Med 1989;8: 105169
  • 21
    Bingham SA, Gill C, Welch A, et al. Comparison of dietary assessment methods in nutritional epidemiology: weighed records vs. 24 hour recalls, food-frequency questionnaires and estimated-diet records. Br J Nutr 1994;72: 61943.
  • 22
    Block G, Thompson FE, Hartman AM, Larkin FA, Guire KE. Comparison of two dietary questionnaires validated against multiple dietary records collected during a 1-year period. J Am Diet Assoc 1992; 92: 68693.
  • 23
    Gnardellis C, Trichopoulou A, Katsouyanni K, Polychronopoulos E, Rimm EB, Trichopoulos D. Reproducibility and validity of an extensive semiquantitative food frequency questionnaire among Greek school teachers. Epidemiol 1995;6: 747.
  • 24
    Sevak L, McKeigue PM, Marmot MG. Relationship of hyperinsulinaemia to dietary intake in South Asian and European men. Am J Clin Nutr 1994;59: 106974.
  • 25
    Rao DN, Ganesh B, Desai PB. Role of reproductive risk factors in breast cancer in a low-risk area: a case-control study. Br J Cancer 1994;70: 12932.
  • 26
    Key TJ, Fraser GE, Thorogood M, et al. Mortality in vegetarians and nonvegetarians: detailed findings from a collaborative analysis of 5 prospective studies. Am J Clin Nutr 1999;70: 516S524S.
  • 27
    Kinlen LJ. Meat and fat consumption and cancer mortality: a study of strict religious orders in Britain. Lancet 1982, I: 9469.
  • 28
    Snowdon DA. Animal product consumption and mortality because of all causes combined, coronary heart disease, stroke, diabetes and cancer in Seventh-day Adventists. Am J Clin Nutr 1988;48: 73948.
  • 29
    Mills PK, Beeson WL, Phillips RL, et al. Dietary habits and breast cancer incidence among Seventh-day Adventists. Cancer 1989; 64, 58290.
  • 30
    Chang-Claude J, Frentzel-Beyme R, Eilber U. Mortality patterns of German vegetarians after 11 years of follow-up. Epidemiology 1992;3: 395401.
  • 31
    Gandini S, Merzenich H, Robertson C, Boyle P. Meta-analysis of studies on breast cancer risk and diet: the role of fruit and vegetable consumption and the intake of associated micronutrients. Eur J Cancer 2000;36: 63646.
  • 32
    Smith-Warner SA, Speigleman D, Yaun S-S, Adami H-O, Lawrence Beeson W, van den Brandt PA, et al. Intake of fruits and vegetables and risk of breast cancer. A pooled analysis of cohort studies. JAMA 2000;285: 76976.
  • 33
    Zhang S, Hunter DJ, Forman MR, Rosner BA, Speizer FE, Colditz GA, et al. Dietary carotenoids and vitamins A, C and E and risk of breast cancer. J Natl Cancer Inst 1999;91: 54756.
  • 34
    Verhoehen DT, Assen N, Goldbohm RA, Dorant E, van't Peer P, Sturn Hermus RJ, et al. 1997. Vitamins C and E, retinol, beta-carotene and dietary fibre in relation to breast cancer risk: a prospective cohort study. Br J Cancer 1997;75: 14955.
  • 35
    Howe GR, Hirohata T, Hislop TG, et al. Dietary factors and risk of breast cancer: combined analysis of 12 case-control studies. J Nat Cancer Inst 1990;82: 5619.
  • 36
    Baghurst PA, Rohan TE. High-fiber diets and reduced risk of breast cancer. Int J Cancer 1994;56: 1736.
  • 37
    Zaridze D, Lifanova Y, Maximovitch D, et al. Diet, alcohol consumption and reproductive factors in a case-control study of breast cancer in Moscow. Int J Cancer 1991;48: 493501.
  • 38
    De Stefani E, Correa P, Ronco A, et al. Dietary fiber and risk of breast cancer: a case-control study in Uruguay. Nutr Cancer 1997;28: 1419.
  • 39
    Yuan J-M, Wang Q-S, Ross RK, et al. Diet and breast cancer in Shanghai and Tianjin, China. Br J Cancer 1995;71: 13538.
  • 40
    Freudenheim JL, Marshall JR, Vena JE, et al. Premenopausal breast cancer risk and intake of vegetable fruits and related nutrients. J Natl Cancer Inst 1996;88: 3408.
  • 41
    Willett WC, Hunter DJ, Stampfer MJ, et al. Dietary fat and fiber in relation to risk of breast cancer: an 8-year follow-up. J Am Med Assoc 1992;268: 203744.
  • 42
    Graham S, Zielezny M, Marshall J, et al. Diet in the epidemiology of postmenopausal breast cancer in a New York State cohort. Am J Epidemiol 1992;136: 132737.
  • 43
    Rohan TE, Howe GR, Fridenreich CM, et al. Dietary fiber, vitamins A, C and E and risk of breast cancer: a cohort study. Cancer Causes Control 1993;4: 2937.
  • 44
    Liu S, Manson JE, Lee IM, Cole SR, Henneckens CH, Willett WC, et al. Fruit and vegetable intake and risk of cardiovascular disease: the Women's Health Study. Am J Clin Nutr 2000;72: 9228.
  • 45
    World Cancer Research Fund, American Institute for Cancer Research Expert Panel. Food, nutrition and the prevention of cancer: a global perspective. Washington, DC: American Institute for Cancer Research, 1997.