To the Editors: Sinner et al. (1) report a lack of association between obesity and pancreatic cancer in a cohort study of elderly women with 209 cases of pancreatic cancer. These results are valuable given the prospective nature of the study. However, we disagree with the authors' conclusion that their findings provide “evidence” that obesity is not a risk factor for pancreatic cancer. Moreover, we find the authors' discussion of the existing literature to be misleading.
The authors indicate that “Since 2000, six studies of obesity and pancreatic cancer incidence reported no association whereas six others reported a positive association” (p. 1572). The existing literature, however, can be more accurately summarized as follows: Since 2000, positive associations between obesity and pancreatic cancer have been consistently observed in the largest studies (total of 2,348 female cases; refs. 2-5) whereas smaller studies have observed no association between obesity and pancreatic cancer (total of 367 female cases; refs. 6-11). Furthermore, among studies with null findings, one study provided no data on body mass index and pancreatic cancer and indicated that “in our Japanese population, body mass index is distributed in a lower range than in Western countries” (6); one study lacked an adequate number of obese participants (the highest body mass index cutpoint was 24.7 kg/m2; ref. 7); two studies found associations in men but not in women (case numbers were small; refs. 8, 9); one study included only five women (10); and one study observed a risk ratio of 1.21 (95% confidence interval, 0.70-2.06) among women (11).
Surprisingly, the authors downplayed their statistically significant findings of an association between obesity at age 40 and pancreatic cancer risk (hazard ratio, 1.72; 95% confidence interval, 1.12-2.63; ref. 1). This association is unlikely to be due to chance, given the existing literature, and should have been highlighted rather than buried in the text. The relation between obesity and pancreatic cancer is likely to be complex and the latency period, as with many cancers, is likely to be long. Therefore, it is quite conceivable that weight at age 40 is more important than weight later in life.
The literature to date, when evaluated critically, indicates that obesity plays an important role in the etiology of pancreatic cancer. Understanding the temporal relationship between body mass index and pancreatic cancer may provide new insights into the etiology of this disease and may well explain why studies in this area have not always been consistent.
In Response: We thank Drs. Michaud and Fuchs for the letter regarding our recent publication “Lack of association of physical activity and obesity with incident pancreatic cancer in elderly women” (1). We chose to present our findings from the Iowa Women's Health Study in the “null results in brief” format. However, this prevented a complete discussion of the literature, and this exchange of letters provides an excellent forum to expand on the discussion. We limit our review to studies of women, as this was our original focus.
Table 1 presents the evidence regarding obesity and incident pancreatic cancer or pancreatic cancer mortality among women. With this table, we can apply the methodology used by the World Cancer Research Fund/American Institute for Cancer Research Publication for synthesis of evidence from observational studies (2). In doing so, we find that the evidence falls into the low end of the category entitled “possible,” with 38% of the observational studies reporting a statistically significant positive association. The categories from this approach for synthesis of evidence from observational studies are as follows: convincing, >75% showed a statistically significant association; probable, 60% to 75% showed a statistically significant association; possible, 30% to 60% showed statistically significant association; insufficient, less than three total studies; evidence of no effect, <30% showed a statistically significant association.
Author and year . | No. (cases/deaths) . | No. (controls) . | No. (in cohort) . | Relative risks by body mass index groupings . | . | . | . | . | . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | . | . | . | Low . | 2nd . | 3rd . | 4th . | High . | P for trend . | |||||||||
Cohort studies | ||||||||||||||||||
Sinner et al., 2005 (1) | 209 | n/a | 38,002 | 1.0 | n/a | 0.94 | n/a | 1.14 | null* | |||||||||
Calle et al., 2003 (3) | 1,650 | n/a | 495,477 | 1.0 | 1.11 | 1.28 | 1.41 | 2.76 | <0.001 | |||||||||
Michaud et al., 2001 (4) | 210 | n/a | 117,041 | 1.0 | 1.10 | 1.18 | 1.28 | 1.70 | 0.05 | |||||||||
Nilsen and Vatten, 2000 (7) | 68 | n/a | 32,374 | 1.0 | n/a | n/a | n/a | 0.82 | null* | |||||||||
Gapstur et al., 2000 (8) | 43 | n/a | 15,183 | 1.0 | 0.48 | 1.09 | n/a | 0.73 | null* | |||||||||
Wolk et al., 2001 (9) | 34 | n/a | 19,964 | 1.0 | n/a | n/a | n/a | 1.1 | null* | |||||||||
Case control | ||||||||||||||||||
Pan et al., 2004 (5) | 275 | 2,492 | n/a | 1.0 | n/a | 0.85 | n/a | 1.63 | 0.06 | |||||||||
Silverman, 2001 (6) | 238 | 774 | n/a | 1.0 | 1.4 | 1.5 | n/a | 1.5 | 0.13 |
Author and year . | No. (cases/deaths) . | No. (controls) . | No. (in cohort) . | Relative risks by body mass index groupings . | . | . | . | . | . | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
. | . | . | . | Low . | 2nd . | 3rd . | 4th . | High . | P for trend . | |||||||||
Cohort studies | ||||||||||||||||||
Sinner et al., 2005 (1) | 209 | n/a | 38,002 | 1.0 | n/a | 0.94 | n/a | 1.14 | null* | |||||||||
Calle et al., 2003 (3) | 1,650 | n/a | 495,477 | 1.0 | 1.11 | 1.28 | 1.41 | 2.76 | <0.001 | |||||||||
Michaud et al., 2001 (4) | 210 | n/a | 117,041 | 1.0 | 1.10 | 1.18 | 1.28 | 1.70 | 0.05 | |||||||||
Nilsen and Vatten, 2000 (7) | 68 | n/a | 32,374 | 1.0 | n/a | n/a | n/a | 0.82 | null* | |||||||||
Gapstur et al., 2000 (8) | 43 | n/a | 15,183 | 1.0 | 0.48 | 1.09 | n/a | 0.73 | null* | |||||||||
Wolk et al., 2001 (9) | 34 | n/a | 19,964 | 1.0 | n/a | n/a | n/a | 1.1 | null* | |||||||||
Case control | ||||||||||||||||||
Pan et al., 2004 (5) | 275 | 2,492 | n/a | 1.0 | n/a | 0.85 | n/a | 1.63 | 0.06 | |||||||||
Silverman, 2001 (6) | 238 | 774 | n/a | 1.0 | 1.4 | 1.5 | n/a | 1.5 | 0.13 |
No P value for linear trend was provided, instead, confidence intervals were used to determine that results were null.
It is true that the two largest cohort studies, with 1,650 and 210 cases, respectively, show significant results in women (3, 4). However, the Iowa Women's Health Study (with 209 cases) had 80% power to detect a statistically significant relative risk of 1.5, far less than the 1.7 and 2.76 observed by these two larger studies, respectively. Furthermore, there are two large case control studies with more cases than were included in our report or in Michaud's article (5, 6). In one of these case control studies, the association was marginally significant (with 275 cases; ref. 5), in the other (with 238 cases), it was not (6).
Our own adequately powered results did not show evidence of an association. We disagree that the finding of an association of pancreatic cancer with obesity at age 40 (but not at ages 18, 30, and 50) is evidence for a causal association. If this association were causal, why would it disappear by age 50? Why would there be no association by age 30? Why would there be no trend across these time points?
Taken in context, as stated in our paper, the next step for this line of inquiry may be a meta-analysis across the cohort studies to address this hypothesis further and seek sources of heterogeneity. We agree that understanding changes in body mass index across the life span in relation to pancreatic cancer etiology would be of value.
Our conclusion, based on review of the literature in Table 1, and despite our own findings, is that it is possible that obesity is inversely associated with pancreatic cancer incidence or mortality among women. We did not observe the association in our data and stated so in our report. Perhaps additional study of this hypothesis will lead us to conclude, as have Drs. Michaud and Fuchs, that obesity plays an important role in the etiology of pancreatic cancer. However, given the current evidence, we feel that such a conclusion is premature.