| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
LETTER TO THE EDITOR |
Osteoporosis Research Center
Creighton University
601 North 30th Street
Omaha, NE 68131
E-mail: jmlappe{at}creighton.edu
Dear Sir:
We thank Schabas and Bolland and Reid for giving us the opportunity to clarify what seems to be a misunderstanding about the data presented in our article (1), to which these 2 correspondents refer. They call attention to a seemingly high cancer rate in our placebo group. To review, 50 cancers occurred in 1179 women over a 4-y period, for an overall incidence rate of 
1.1%/y. As stated in our article, the participants were allocated randomly to the 3 treatment groups. According to the null hypothesis, in effect, cancer incidence did not differ between the groups, and the apparent differences were due to random chance. However, differences as extreme as found in our study would have occurred by chance alone less than once in 50 such trials. That is the meaning of the published probability value (P < 0.02).
Can we prove that our trial was not that 1 out of 50? Of course we cannot. If this were the only information on this topic, we could not say much more. However, in fact, there is a large body of observational data indicating risk reductions of precisely the magnitude we observed, some of it cited in our article. Furthermore, since the submission of our manuscript, additional studies of improved vitamin D status and cancer risk have reported risk reductions ranging up to 70% or more (2, 3). Hence, there is a high degree of concordance across a variety of study designs and several different populations. Our results, rather than being surprising, are exactly what might have been expected.
Both correspondents commented that our cancer incidence rate in the placebo group was higher than expected. It is important to pay attention to the 95% CIs for our observed incidence rates. The binomial SD for our 50 cases of cancer overall is 4.0 cases, which means that our rate of 1.1% could have come from a population with an actual rate as low as 0.9% or as high as 1.3%. The rate for the placebo group alone (20 cases in 288 persons) was 1.7%/y, with a 95% CI of from 1.35%/y to 2.12%/y. These ranges of values are fully compatible with the values cited by both correspondents. Note that the rate cited by Bolland and Reid was 1.2%.
Despite the statements in the letter by Schabas, reevaluation of the Nebraska incidence data, which involved consultation with the professional staff of the Nebraska Cancer Registry, confirmed that our observed cancer incidence rate was about what would be expected for the urban fringe population that comprised the bulk of our sample.
Like Bolland and Reid, we did not find a statistically significant effect for calcium alone. Hence, there is no "contrast" (their term) to explain. Unlike vitamin D, which could plausibly operate in a number of different tissues, any effect that would be predicted for calcium would be confined largely to colon cancer (4, 5). There were only 3 colon cancers in our series; hence, the ability to see an effect was not there.
We thus conclude that, although a chance occurrence cannot be excluded, we take encouragement from the high degree of concordance between our results and those of the large body of observational studies.
ACKNOWLEDGMENTS
None of the authors had a conflict of interest.
REFERENCES
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
