December 9th, 2010

Meta-Analysis Finds Aromatase Inhibitors Increase Risk for Heart Disease

A new meta-analysis has found that postmenopausal women taking aromatase inhibitors (AIs) instead of tamoxifen to treat breast cancer appear to be at increased risk for cardiovascular complications. The results were presented at the San Antonio Breast Cancer Symposium.

Previous studies suggested that women taking AIs were more likely to develop cardiovascular disease, while women taking tamoxifen were found to be at increased risk for endometrial cancer and venous thromboembolism. In 2008, the FDA warned about an increased risk for heart disease with one AI, anastrozole.

In the new analysis of data from 7 large randomized trials comparing AIs to tamoxifen, the researchers found a similar effect with all AIs. Any exposure to AIs was associated with a 20% increase in the risk for cardiovascular disease (p=0.01; NNT 143) and a 48% increase in the risk for bone fractures (p<0.00001; NNT 34). By contrast, AIs reduced the risk for venous thrombosis (OR 0.53, p<0.00001; NNT 67) and endometrial carcinoma (OR 0.32, p<0.00001; NNT 200).

9 Responses to “Meta-Analysis Finds Aromatase Inhibitors Increase Risk for Heart Disease”

  1. Robin Motz, M.D., Ph.D. says:

    Once again, meta-analysis is mathematically flawed, and should just be used to generate a hypothesis to then be formally tested by a prospective study.

  2. The myth of the randomized controlled prospective study.

    I agree that meta-analyses are intrinsically flawed but so are RCTs. A well done meta-analysis should do more than generate a hypothesis in some cases.

    RCTs have shown estrogen to cause cardiovascular events while other good science shows estrogen to prevent atherosclerosis and only oral prem-pro is associated with increased disease. Despite this, all estrogen has been tarred with the prem-pro brush.

    My point? Meta-analyses should not be so easily discarded and neither should RCTs be so blindly embraced.

    • Amen to both the point and to the example!
      RCTs are of necessity limited in scope. The results can easily be over-extended in their application. The current HRT guidelines, which essentially say that short-term estrogen is ok for early symptom releif but it MUST be discontinued soon after starting, are only very loosely based on the WHI results. Paradoxically, recent, more careful analysis of the WHI has shown that (1) unopposed estrogen is generally much more favorable than estrogen plus progestin, (2) HRT started before age 60 is favorable, while that started after age 65 is harmful, and (3) the LONGER one takes HRT, generally the more favorable the result. NONE of this has made a dent in the guidelines or in popular scientific or lay opinion, with the notable exception of the Endocrine Society statement of June 2010, which finally makes these critical points.

      Competing interests pertaining specifically to this post, comment, or both:

  3. Saurav Chatterjee, MD says:

    And essentially a well done meta-analysis with proper reporting of bias and usually a more robust sample size will often have a more valid outcome…but then I am a bit biased……..

    Competing interests pertaining specifically to this post, comment, or both:
    Member of the Cochrane Review Group on PVD.

  4. H Robert Silverstein, MD says:

    patients lipids/AIC/Lp(a)/blood pressure and DEXA need to be reviewed routinely in patients treated with AI’s AS SHOULD all patients. We as MD’s need to control as many risks for all diseasesa as possible, not focus on the “one.”

  5. Saurav Chatterjee, MD says:

    I agree with Dr Silverstein-hence the scope of possibly more studies/metaanalyses to look at these other outcomes maybe??…………..

  6. Leon Hyman, Ms M.D. says:

    One should explore the psychological impact of presenting the pro’s and con’s of AIS and Tamoxifen to women and then asking them what chemotherapy risk do you want to take, especially if you also include you can also do nothing and see what the cancer can do unopposed.

    Competing interests pertaining specifically to this post, comment, or both:

  7. Robin Motz, M.D., Ph.D. says:

    I can list three potential problems with meta-analyses off-hand:

    1) We are never told which studies were rejected and NOT included in the meta-analysis, so we cannot make our own independent judgment.

    2) There are two mathematical models: standard effects, or random effects, and therefore there are two possible results to the meta-analysis.

    3) Do you measure the effect using the standardized mean difference, or by correlations?

    As one can see, we already have a 2×2 matrix with 4 possible conclusions, and, in fairness, all four possible conclusions/interpretations should be stated.

    I have not even mentioned the possibility of the Simpson Effect creeping in.

  8. Saurav Chatterjee, MD says:

    Valid queries Dr Motz,however….
    1)If one is so inclined to dispute a study-most good journals require authors to outline their search methodology-which can be repeated ,if so deemed, to refute inclusion/exclusion of studies-and have their own publication in the process.Case in point being the ACEI and Cancer controversy….
    2)Usually the Standard or fixed effect is used for a meta-analysis with small sample size and low heterogeneity-while random effect model is used for a metanalysis with significantly larger sample size and heterogeneity.
    Ref-Cochrane Handbook of Systematic Reviews
    3)If one wants to do a regression meta-analysis(as the one by freemantle et al in the landmark BMJ paper on beta blockers)-one should use correlation analysis-for other systematic reviews,standardised mean difference should suffice.In any case-the methodology should be specified in the publication.

    Competing interests pertaining specifically to this post, comment, or both:
    Member of CRG on PVD.