Sometimes research reports are as interesting to me for the story behind the story as much as the actual story itself. Such is the case with two abstracts presented at the annual meeting of the American Society of Clinical Oncology here in Orlando, which closes today.
The two reports deal with the impact of anti-depressants on the effectiveness of tamoxifen, a drug that has been used for decades in the adjuvant (preventive) treatment of breast cancer after surgery and radiation, as well as for the treatment of breast cancer that has spread to other parts of the body.
But the “story behind the story” is that the research was done through the analysis of large medical data banks, one of which is owned by a public company which happens to be a major prescription benefit manager here in the United States.
The relatively recent development of these megadatabases will, in my opinion, become something we will see used more and more in health care going forward, hopefully for good purposes.
First, the concept behind the research:
Tamoxifen is a drug which is commonly described as an “anti-estrogen”, and essentially blocks the effects of estrogen on breast cancer cells which are hormonally sensitive, or estrogen positive on biopsy. There is an enzyme in the body which converts tamoxifen into its active form. The levels of this breakdown product in the blood directly predict the effectiveness of the drug in treating the cancer. Some women do not have this enzyme, so they have lower levels of this “converted” form of the drug, and therefore do not get as much benefit.
There are also some drugs which are known to block this particular enzyme that breaks down tamoxifen. The enzyme is called CYP2D6. Among those drugs are anti-depressants. Women take anti-depressants not only for depression, but also because they are effective in treating the symptoms of menopause and may also block the menopausal-like symptoms associated with tamoxifen.
In one of the studies, researchers in the Unites States used the prescription records of their company to look at the question of whether women who took both tamoxifen and certain anti-depressants had poorer outcomes with their breast cancer compared to women who only took tamoxifen. They reported that women who took both drugs had a recurrence rate of 13.9% over two years, while women who took only tamoxifen had their breast cancer recur at a lower rate of 7.5% over the same time period.
In the other study, which was from the Netherlands, the researchers used their national health system records to determine that among 2000 women with breast cancer using tamoxifen from 1994 to 2006, 215 also used a drug that could inhibit the CYP2D enzyme. However, unlike the United States study, they did not find that the women fared poorly if they took an anti-depressant.
There are reasons why the studies may differ in their outcome, but the net result is that we still don’t know for certain whether or not it is safe to use the drugs together. The Dutch study clearly looked at women who used tamoxifen for the adjuvant therapy of breast cancer, and they also had computer based medical records which are likely more specific in defining when the breast cancer recurred.
The United States study examined the computer records of women with breast cancer who started taking tamoxifen between 2003 and 2005 and took the medicine for over two years. There was no independent confirmation of medical records, and no independent confirmation that this drug was used for adjuvant therapy. Recurrence was determined based on medical insurance claims that used certain codes to suggest that the disease had progressed.
The other interesting point about the United States study was that the database used in the research was developed as part of the actual daily business model of the pharmacy benefit manager. It contained the records of 10.7 million health plan members for whom the company administered the drug benefit portion of the client companies’ insurance program. The information was “deidentified,” so the researchers could not link the record to any specific patient, and could only look at the larger trends.
Using these large megadatabases for medical research and improving medical care has intrigued me for some time. What you may not know is that insurers, health plans, pharmacy benefit managers, health information technology companies and large medical group practices are now accumulating this data at an accelerating rate.
Although this information has long been used for business purposes and to some degree to manage the medical care of the patients included in these databases, the opportunity to mine this information for research purposes and looking at medical care trends is only now beginning to take hold.
The opportunities are—in my opinion—enormous. If we could get these databases to operate in “real time,” they could be used to look at potentially alarming trends in illness, such as the recent flu outbreak, to get an early warning signal about a rapidly spreading epidemic. They could be used to find unexpected side effects from drugs, as we have seen with medicines such as blood-thinners which can result in an increased number of emergency visits. The Food and Drug Administration is beginning a pilot program to test the theory that we may be able to use these large databases to screen for side effects from new and established drugs that are on the market, to find out if there are serious unanticipated problems.
For me, the dream has always been to establish seamless health care monitoring systems that help us improved the quality of care we offer our patients. For example, did an eligible woman have a mammogram? Was it positive or negative? Did she get a biopsy? How long did it take to get her surgery? What surgery did she have? Did she get the right follow-up treatment? And did she come back the next year for her follow-up?
We could also monitor physician performance, and get away from the crude measures we use today to really determine in some circumstances whether or not doctors meet certain standards.
For example, in the above scenario, how about measuring the performance of the radiologist doing the mammogram? How many mammograms does the doctor read a year (the number read is an indicator of quality. Too few is not good.)? How many mammograms does the doctor read as positive? How many biopsies are performed because of the doctor’s interpretation of the mammogram? How many were positive? How many were negative? Does the doctor miss too many lesions that subsequently are diagnosed as cancer?
So, here we have some research studies that unfortunately don’t give us a clear answer to the question about the use of anti-depressants in women with breast cancer who take tamoxifen.
But we do have a window on the future as to how we may get clues about the problems we face in health care, and perhaps develop a more rational, less intrusive way to spot early safety signals about our drugs and treatments, while at the same time be able to finally measure the quality of the health care we provide in real time.
I feel compelled to add that I don’t think we are not there yet. Even in the United States study described above the researchers had to rely on what we call “administrative claims data,” which is frequently not as accurate as we would like. It is also one very large step away from the actual “source” data which is the medical record.
But make no mistake: the next decade will see a substantial increase in the availability of this type of information. The real question is whether we have the skills to use that data wisely and effectively to improve the care we offer our patients.