On January 14, 1997, the National Institutes of Health (NIH) convened an independent panel to review data from two mouse studies, one of which suggested that very high daily doses of AZT may induce tumors in the offspring of pregnant mice given the drug during the last trimester of pregnancy. The panel, which consisted of basic and clinical researchers, epidemiologists, HIV-infected women, and a bioethicist, was asked to review the available data, to assist the NIH in developing a scientific summary of the technically complex information, and to offer recommendations for future research. The scientific summary will also be used by the NIH as the foundation for disseminating information to and communicating with the community of HIV-infected women and their health care providers, and in facilitating subsequent discussions on treatment recommendations and related public policy. The summary which follows is based upon the panel's review and discussion of the information from these and other relevant studies.
Administration of zidovudine (AZT) in the dose and schedule employed in AIDS Clinical Trials Group (ACTG) protocol 076 has been shown to reduce dramatically the likelihood of perinatally acquired HIV infection. A recent update of the data from this study continues to show a reduction in the rate of maternal-fetal transmission from 22.6 percent to 7.6 percent. This was the first major clinical trial of perinatal intervention to prevent HIV infection. To date, no other regimen of intervention has been studied as extensively or proven to be definitively effective. On the basis of this information, a number of professional organizations and the Public Health Service (PHS) have made recommendations for usage of AZT to prevent perinatally acquired HIV infection.
Subsequent population-based studies have demonstrated a substantial decrease in the incidence of perinatally acquired HIV infection. For example, a study of more than 500 infected pregnant women from five geographically separate sites in the U.S. (New York City, Boston, Chicago, Houston, and Puerto Rico) showed a decrease in the rate of mother-to-child transmission of HIV infection from 19 percent in infants born prior to March 1994 to 8 percent for infants born after March 1994, coinciding with implementation of the ACTG 076 AZT regimen in these sites. Another study from North Carolina has documented a decrease in transmission rate from 21 percent to 10 percent.
Scientists have known for a number of years that some nucleoside analogs, including AZT, are sometimes carcinogenic in animals. They have known since 1989 that AZT causes vaginal epithelial neoplasms when given to adult mice. This is believed to be a topical effect in mice, resulting from reflux of urine containing high concentrations of excreted AZT from the bladder into the vagina. No increase in the incidence of tumors in other organ sites has been seen in other studies conducted in adult mice and rats.
In humans, the short-term side effects of the regimen employed in ACTG 076 are limited largely to mild, reversible anemia in the infant. No tumors have been observed to date in studies of approximately 1000 AZT-exposed children followed for an average of 3 years. Current treatment guidelines stress that the long-term effects of this intervention in humans remain unknown, and that all children, whether uninfected or infected, born to mothers who receive AZT during pregnancy should have long-term follow-up which includes evaluation for late effects of this therapy, including development of tumors.
SUMMARY AND INTERPRETATION OF THE MOUSE STUDIES
New data from one of two studies in mice raise the concern that AZT may have a transplacental carcinogenic effect.
In an ongoing study being carried out by investigators at the National Cancer Institute (NCI), pregnant mice were given very high daily doses of AZT (close to the maximum dose allowing fetal viability) during the last trimester (days 12-18) of gestation. In offspring of the highest-dose group observed for twelve months, the investigators noted a several-fold increase in the incidence of liver and lung tumors, and the appearance of unusual reproductive organ tumors in 17 percent of the females. They also documented incorporation of AZT into DNA in a variety of newborn mouse tissues, although this did not clearly correlate with the presence of tumors. The significance of this AZT incorporation is not known, although it is not unexpected given the drug's mechanism of action.
A transplacental carcinogenic effect of AZT was not seen in a second study carried out by scientists at Glaxo-Wellcome, Inc. (G-W), the drug's manufacturer. In that study, pregnant mice were given one of several regimens of AZT, the doses based on pharmacokinetic data in mice and humans and intended to achieve blood levels somewhat higher (approximately three times) than those achieved in clinical practice. The daily doses received by the mice during gestation ranged from one-twelfth to one-fiftieth the daily doses in the NCI study. Some of the offspring received AZT over their entire lifespan. No increase in the incidence of tumors was observed in the offspring of these mice except in the lifetime-exposure group where vaginal tumors were once again noted.
It was the panel's assessment that both studies appeared to be well done. Furthermore, the panel noted that because the two studies differ substantially in design, it is not surprising that they come to different conclusions. The NCI study was designed to maximize the occurrence of transplacental carcinogenesis by giving large doses of the drug over a short period of time during pregnancy. The G-W study was designed to examine long-term effects of AZT administered in an approximation of clinical practice to prevent perinatal transmission of HIV.
The panel noted that, in general, very little is known regarding the sensitivity or reliability of this mouse model system in predicting transplacental carcinogenicity in humans. Demonstration of the utility of the mouse model in predicting transplacental carcinogenicity of drugs in humans would require comparison of long-term effects for a large number of different drugs given to both mice and women during pregnancy. However, only a few other drugs have been studied in the mouse model. Furthermore, long-term follow-up data in humans are available only for diethylstilbestrol (DES). Approximately 15 percent of the offspring of pregnant mice given DES within the range of doses received by humans develop vaginal tumors similar to those seen in the children of women given DES to prevent miscarriage. Approximately 1/1000 exposed daughters develop vaginal tumors. The mechanisms of carcinogenicity of DES and AZT are likely to differ substantially.
There are major differences between humans and mice in AZT distribution, metabolism and excretion, in duration of gestation and fetal development at birth, and in many facets of the maternal-fetal relationship. Therefore, the relationship between the regimens of AZT used in these two studies and those used in clinical practice in humans is complex, controversial, and of uncertain relevance. The implications of these findings will require further study.
SUMMARY OF MAJOR ISSUES DISCUSSED AND RECOMMENDATIONS
The panel was unanimous in concluding that the known benefits of AZT in preventing perinatal transmission appear to far outweigh the hypothetical concerns of transplacental carcinogenesis raised by the NCI mouse study.
At the same time, the panel emphasized the need for active and thorough discussion of this information about the theoretical risk of transplacental carcinogenesis of AZT with all HIV-infected pregnant women in the course of counseling them on the potential benefits and risks of any treatment interventions, whether in clinical practice or clinical trials.
The panel strongly emphasized the need for careful long-term follow-up of all children exposed in utero to antiretroviral therapy, including those who are not infected with HIV.
They also discussed a number of important research priorities that should be pursued. These include the following:
Recognizing the difficulties of maintaining contact and ensuring confidentiality, it is very important to intensify efforts to study the long-term effects of perinatal intervention in exposed children, whether infected or uninfected.
Improve awareness of the industry-sponsored Antiretroviral Pregnancy Registry among HIV-infected women and their health care providers.
Place a high priority on study of interventions that maximize safety and minimize the likelihood of long-term side effects. A number of such studies are underway.
Complete full 2-year follow-up of the remaining mice in the NCI study.
Conduct additional research on transplacental carcinogenesis of nucleoside analogs in the mouse model. This should include study of mechanisms, dose-dependency, and confirmatory studies in at least one other species.
Study the significance of incorporation of AZT into DNA of blood cells and solid tissues, and determine the relationship between AZT pharmacokinetics and incorporation.
IMPLICATIONS FOR PUBLIC HEALTH POLICY
The panel also heard and echoed concerns that the state of the art of antiretroviral therapy has changed substantially in the several years since the Public Health Service recommendations on the use of AZT to prevent perinatally acquired HIV infection were developed and published. Combination chemotherapy monitored by viral load measurements is now appropriate for many women who, until very recently, would have received the ACTG 076 regimen. A great deal of new information is also available regarding the pathogenesis of HIV disease, and the dynamics of perinatal transmission. These changes have a number of important implications for public health policy and for maternal and infant health. A thorough reassessment of the PHS guidelines is needed. The panel recommended that the PHS Task Force should be reconvened to carry out this important task. Any implications of the mouse carcinogenesis data should be considered in this broader and more immediately relevant context.