As previously mentioned, the best sources for the actual figures for the above calculations resides within the FDA but may, as I alluded, be inaccessible. In recent years, figures given by the Agency, e.g. in elderly drug-testing studies, appear to have been hand-tallied rather than garnered from composite computer access. However, the agency is now involved in a large effort to 'mine' data across therapeutic classes, some of which, with meta-analysis, will provide data which individual drug programs never could, nor were designed to show. In time, the ability to access data across drugs and across drug classes will grow as more firms put in computer-assisted NDAs (CANDAs) in appropriate and compatible programs and formats.
What is the risk of a fetus being damaged during an 'average' NDA drug development program? Obviously small. Clearly, toxic but 'life-saving' treatment will carry a heavy embryotoxic risk; anticancer, anti-AIDS drugs, and fetal intrauterine surgical procedures are obvious examples, but the clear-cut risks involved are usually deemed acceptable. A more subtle judgment call involves the development of anti-epileptic drugs. Let's look at two examples. It has been estimated that exposure of pregnant women to normal therapeutic doses of valproic acid may give rise to 1% fetal abnormality rate involving the neural tube (Lindhaut and Schmidt 1986)—10 times the natural incidence. Many of these defects are correctable with modern surgical techniques. Exposure to phenobarbitone also has a reported higher incidence of cleft lip and palate defects (Frederick, 1973): most are surgically correctible. If used in combination, the incidence of anticon-vulsant teratogenic effects are increased (Lindhaut et al 1984). Would either of these drugs be developed in today's litigious atmosphere? I doubt it. But both drugs are valuable in many circum stances; they may be the only drugs suitable for some patients and, indeed, frequently can be life-saving. Certainly status epilepticus is very injurious to the fetus, often resulting in miscarriage or premature birth.
The incidence of neonatal abnormalities in mothers taking anticonvulsant treatment is 70/ 1000 live births (Frederick, 1973). This is 2.4 times the 'spontaneous rate' in the general population (29 abnormalities/1000 live births). Thus, even using a known 'low-incidence' teratogen could cause 40 additional cases/1000 live births, but to determine that accurately would require many thousands of female patient exposures to be detectable against the 'spontaneous' background incidence.
So, back to the opening question. What is the likelihood ofdetecting low-incidence, drug-induced congenital effects in a drug development program? With our presumed database of 4000 patients, only 0.8-5 fetuses would be exposed to a background 'spontaneous' risk of 2.9%. Each program could carry a 1 in 33 to 1 in 6 chance of a single 'spontaneous' abnormality occurring. If the drug or procedure should have low teratogenic activity (at the level of an anticonvulsant), this risk rises to 1 in 14 to 1 in 2.5 that a child will be born with a congenital abnormality in any drug development program. Both 'spontaneous' or drug-induced abnormalities may occur, e.g. a neural tube defect. Thus, on a single-case basis, the abnormalities will be indistinguishable for drug causality. This in turn can lead to litigation, and certainly to a reference in the package label insert.
Wilson has estimated that both drugs and environmental chemical exposures only account for 23% of developmental defects in man (Wilson, 1972).Thus, a product-label reference of such an occurrence will be undeserved at least 97% of the time, but also may be the first signal of a terato-genic risk. It may now be appreciated why this 2-3% risk is termed the 'phantom fetus' and also why the difficulty in disproving liability dominates the mainstream concerns of research, regulatory authorities, and industry alike. This 'ghost risk' creates 'discrimination' against female patients in drug research. This 'ghost' must be exorcised and contained; possible solutions will be discussed later.
Was this article helpful?