In this prospective cohort study of 110,215 women, we found limited evidence that isothiocyanates and hydroxyl radical absorbance capacity are associated with reduced risk of DLBCL. However, we found no compelling evidence that isoflavones, lignans, isothiocyanates, or antioxidant compounds are associated with risk of other major NHL subtypes, overall B-cell NHL, MM, or HL. The major food sources of these compounds, including total fruits and vegetables, were not consistently associated with risk of lymphoid malignancies. The latter findings contradict those of previous case-control and cohort studies that found an inverse association between fruit and/or vegetable intake and risk of NHL or MM (1
). However, our results accord with those from the large, prospective European Prospective Investigation into Cancer and Nutrition (EPIC), which showed no association with risk of overall lymphomas, NHL subtypes, or HL, although the investigators found an inverse association between total fruit intake and MM risk (4
). In general, recent null findings from prospective cohort studies suggest that the strong inverse associations with fruit and vegetable intake previously detected in retrospective case-control studies may have been overstated, and may have resulted from selection, recall, or survival bias (25
Our results contrast with those of Thompson et al
., who found that intake of several antioxidants (e.g., dietary vitamin C and alpha-carotene), fruits, and vegetables was significantly inversely associated with risk of overall NHL and FL in particular in the Iowa Women’s Health Study (WHS) cohort. While our finding of an inverse association between moderate isoflavone intake and CLL/SLL risk somewhat accords with their observation of a similar association with overall NHL risk (5
), the lack of a dose-response trend in our data and the very low levels of intake at which statistically significant associations were observed lead us to believe that our findings are not biologically meaningful. Zhang et al
. also detected an inverse association between intake of fruits and vegetables, but not antioxidants, and overall NHL risk in the Nurses’ Health Study (NHS) cohort (3
). The difference in findings is unlikely to be due to lesser statistical power to detect these associations in our study, as the number of cases in the CTS (N
=518, including 145 DLBCL and 115 FL) was comparable to that in the WHS (N
=415, including 184 DLBCL and 90 FL) and the NHS (N
=199). One potential explanation for the discrepancy is that average consumption of fruits and vegetables (and, consequently, antioxidants) was higher in both other cohorts than in ours, and we may have lacked sufficient exposure variation to detect an association with the highest levels of fruit, vegetable, and antioxidant consumption. However, Rohrmann et al
. detected no such association in the EPIC cohort, which had levels of intake comparable to those in the WHS and NHS (4
), and we did not observe inverse associations with B-cell NHL or FL risk even when we categorized intake using the same cutpoints as Thompson et al
. or Zhang et al
. (data not shown). Other possible explanations for the different results include chance, residual confounding, effect modification by characteristics that varied between the cohorts (e.g., other dietary factors, physical activity), differences in the types of vegetables and fruits commonly consumed in each cohort, and perhaps varying effects of diet by age or time period, as the mean age of the NHS at the baseline dietary assessment in 1980 was 45 years, that of the WHS at baseline in 1986 was 62 years, and that of the CTS cohort at baseline in 1995-1996 was 52 years.
The weak inverse associations of DLBCL risk with isothiocyanates and hydroxyl radical absorbance capacity suggest that phytocompounds may have a minor protective effect against risk of certain B-cell NHL subtypes. Specifically, isothiocyanates may help to prevent DLBCL by inducing phase II cellular detoxification enzymes or by promoting apoptosis of malignant cells (9
). Antioxidant activity against hydroxyl radicals is highest in kale, brussels sprouts, alfalfa sprouts, beets, spinach, and broccoli flowers (19
)—plant foods that have some overlap with those high in isothiocyanates, but enough dissimilarity perhaps to indicate an independent effect of hydroxyl radicals—the most reactive of all free radicals (26
)—in DLBCL development. However, given the lack of dose-response trends and the large number of tests performed in our study, these and any of the other observed associations could also have been due to chance.
The limitations of our study are similar to those of previous cohort and case-control studies of diet and risk of lymphoid malignancies. We performed only one dietary assessment, preventing us from accounting for dietary changes over time. We lacked biological measures of dietary intake among all cohort members, although our questionnaire-based measures of isoflavone and lignan intake correlated well with excreted urinary levels in a small subset of the cohort [(17
) and unpublished data]. Future studies would be strengthened by including measures of phytocompounds in urine or plasma, which should be collected repeatedly over time, as these biomarkers reflect only recent dietary intake (27
). Finally, we lacked sufficient statistical power for detecting weak effects or associations with less common lymphoid malignancies, including most NHL subtypes, MM, and HL. A further limitation in our study, as well as many other studies of isoflavones in US populations, is that our baseline food frequency questionnaire included only limited soy-based foods and excluded soy milk, an important source of isoflavones in western populations. We did assess soy milk consumption in our follow-up questionnaire in 1997-98, but found no inverse associations with risk of any lymphoid malignancies (data not shown).
These limitations are countered by the strengths of our study, including its detailed dietary assessment, estimation of isoflavone intake based not only on soy-based foods, such as tofu, but also common foods in the US diet that contain soy flour and soy protein, investigation of novel measurements of antioxidants, complete ascertainment of incident cancer through linkage to the California Cancer Registry, and prospective design, which minimized the problems of selection, recall, and survival bias that often invalidate retrospective studies.
In summary, we found that a range of dietary phytocompounds were not related to risk of overall B-cell NHL, common NHL subtypes, MM, or HL, although we detected modest inverse associations of isothiocyanate intake and the antioxidant index measuring hydroxyl radical absorbance capacity with DLBCL risk. Despite our generally null results, we cannot exclude the possibility that isoflavones, lignans, isothiocyanates, or antioxidants exert a protective effect against lymphoid malignancies when consumed in early life, in greater amounts, or by persons with particular genetic or other host characteristics. Furthermore, our finding of no association between total fruits and vegetables and risk of B-cell NHL, MM, or HL does not rule out a beneficial effect of specific types of fruits, vegetables, or unmeasured components in some of these foods. Nevertheless, our findings do not support increasing dietary intake of isoflavones, lignans, isothiocyanates, antioxidants, and foods rich in these compounds as a promising strategy for decreasing risk of lymphoid malignancies overall.