The strides now being made in genetic medicine presage a revolution in health care. Already, it is possible to determine a patient’s susceptibility to certain hereditary diseases. Within a decade, genetic tests will indicate who will respond well to a drug, who will barely respond, and who will suffer adverse reactions. Physicians will then be able to tailor therapies to the needs of individual patients.

Significant scientific and technical challenges remain, however. To date, scientists have identified the function and location of only a small fraction of the genes that define the human race. And every gene displays a number of variations—a phenomenon called polymorphism. Much more genetic data will thus have to be collected and deciphered before genetic medicine can reach its full potential. Moreover, testing patients to identify their genes is still a slow and expensive process that costs from $1,000 to $2,000 for each test, though a few simple tests for infectious diseases cost as little as $20.¹ The cost of individualized drug therapy, too, is likely to be high. And there are challenges in the realm of policy: what are the privacy rights of people who give genetic information to scientists and clinicians, for example, and how should those rights be protected?

Nonetheless, the scientific community has already witnessed tremendous breakthroughs in genetic medicine and expects the pace of new discoveries to increase exponentially in coming years. The time is ripe for pharmaceutical and biotechnology companies to consider the implications of these advances for their own businesses and to develop appropriate strategies, for the ability to tailor drugs to the genetic makeup of individual patients—known in the trade as pharmacogenomics—represents a watershed for the industry. Although the strategic implications of this breakthrough will certainly become more clear as the technology evolves, some are clear already.

Enhancing (or sometimes reducing) the commercial value of drugs

Genetic testing could let pharmaceutical companies segment people into four groups, according to how they react to a given drug: high responders (those who respond especially well), low responders, nonresponders, and adverse reactants. Researchers at Bristol-Myers Squibb, for instance, have shown that different individuals have different degrees of responsiveness to the cholesterol-lowering drug Pravachol, depending on which of two variants of a polymorphism encodes the cholesterol ester-transfer protein (CETP).

In some cases, the issue isn’t degrees of responsiveness but the risk of an adverse reaction. To give only one example, patients with mutations in the CYP3A family of cytochrome P450 enzymes can have severe cardiac reactions to the antihistamine Seldane if they are concurrently taking the antibiotic erythromycin. Hoechst Marion Roussel (HMR) eventually had to remove Seldane from the market because of toxicity problems related to the drug.

The ability to categorize patients on the basis of likely responses to specific drugs can enhance a drug’s commercial value in two ways.

Saving drugs

First, the genetic approach can be used to screen out patients who might be put at risk by a drug already on the market. Time and again, drug companies have suffered when a small part of the population reacted badly to an otherwise effective drug. Warner-Lambert, for instance, temporarily pulled Rezulin (a new drug it had licensed from Sankyo and registered in Britain) from the UK market and suspended its regulatory filing in the rest of Europe to investigate why the drug proved seriously toxic to the livers of 1 diabetic patient in 60,000. (Sales of the drug have since resumed.) Novartis’s Clozaril became a drug of last resort after a small number of patients taking it for schizophrenia developed severe blood problems, including agranulocytosis (the complete loss of certain blood cells). And the heart valve problems encountered by a few users of the weight reduction drug Redux forced American Home Products and Interneuron to withdraw it and to focus on litigation that has cost millions in fees.

Rare polymorphisms (or combinations of them) may have produced many of these adverse reactions. If it had been possible to screen patients before they took the drugs in question, the drugs might have been saved and huge economic losses averted. A diagnostic for the CYP3A mutation might have helped HMR, the producer of Seldane, save a $600 million franchise. Warner-Lambert’s revenues from Rezulin would have grown much more quickly had it been possible to identify the patients susceptible to the liver problem described above, who could then have taken the drug provided their liver enzyme levels were monitored. Instead, the company lost about $200 million in potential sales while it investigated the problem.

The wisdom of rescuing a drug, however, will depend on the cost of conducting genetic tests for potential adverse reactants. Developing cost-effective diagnostic methods will clearly be more difficult for drugs, such as Rezulin, whose toxic effects are quite rare.

Improving treatment

Second, genetic diagnostics could increase the commercial value of some medications by identifying which class of drug would be most effective for specific groups of patients. Hypertension, for example, can be treated with angiotensin-converting enzyme (ACE) inhibitors and beta-blockers, as well as with calcium channel blockers. But not all patients respond to these medications in the same way: many doctors now tend to prescribe ACE inhibitors and beta-blockers to African-Americans, for instance, because these drugs seem to work more effectively on them. An unidentified genetic trait probably explains such differences, but without proof, physicians must arrive at a course of drug treatment by trial and error. Genetic diagnostics might provide scientific confirmation, thereby organizing and accelerating the process of matching drugs for all sorts of diseases with appropriate patients or groups of patients.

Genetic diagnostics could be equally valuable for treating serious diseases, such as cancer or cardiovascular complications from diabetes, that take a dangerously long time to respond to drugs. At present, for example, only after treatment is under way can a physician know whether a particular course of chemotherapy is suitable for a particular patient’s cancer. By predicting which class of drug would likely produce the best responses, the fewest side effects, or both, in different groups of patients, genetic diagnostics would reduce or eliminate the risk that the patient’s condition might deteriorate before their doctors found the right treatment. Yet it will take some time to master the genetics of disease progression, for the problem is complex, and large and long clinical trials are needed to prove any effect.

Furthermore, when a form of drug therapy, such as chemotherapy, is highly toxic to healthy and diseased cells alike, genetic diagnostics could help doctors match patients to the class of drug that was least harmful. Scientists conducting research on rheumatoid arthritis, for example, recently discovered that people with certain genetic characteristics responded better to a single class of drug than to a highly toxic combination of three classes, the more frequently prescribed therapy.

Genetic diagnostics can also reveal important differences among responses to drugs within the same class. Warner-Lambert and Pfizer collected clinical data that demonstrated the superior efficacy of Lipitor, a cholesterol-lowering medication (in the "statin" class of drugs) that Warner-Lambert produces and Pfizer markets. Often, however, a drug’s effectiveness in treating some groups of patients but not others remains unclear from clinical evidence alone. Antidepressants appear to be a case in point: patients who take a single class of these drugs—selective serotonin reuptake inhibitors (SSRIs), such as Eli Lilly’s Prozac—register a broad spectrum of currently unpredictable responses.

Such variations have consequences both for the profits of drug companies and for the treatment of patients: very clear clinical outcomes took Lipitor from fifth to second place in market share in less than a year. The use of genetic data to distinguish high from low responders would make such leaps in market share common; if drugs of generally mediocre effectiveness were prescribed for a segment of high responders, commercial also-rans would become blockbusters. Of course, the larger the group of high responders, the more profitable the drug. In all likelihood, companies that market latecomers to drug categories would be more likely than marketers of first-in-class drugs to use data on high responders as a competitive differentiator.

That said, the divergence in efficacy between drugs in the same class tends to be smaller and harder to measure than the difference between drugs in different classes. Even if tests to distinguish among responses to drugs in the same class became more sophisticated, the economic logic of continually refining techniques for matching or tailoring drugs to ever-smaller patient cohorts might therefore reach the point of diminishing returns.

The flip side

A more precise understanding of a drug’s effect on different people has dangers as well as advantages for pharmaceutical and biotechnology companies. Should physicians learn of scientific proof that a drug was relatively ineffective for entire groups of patients, the market for it could collapse. Of course, that might not happen if the genetic data also uncovered a large number of high responders—a fact that a pharmaceutical or biotechnology company would probably be keen to publicize, especially if it were coming late to market.

If regulatory authorities decided to collect genetic data on the performance of individual drugs and forced manufacturers to state which groups of patients were most likely to benefit from them, the industry would be spared the ethical dilemma of whether to disclose a drug’s ineffectiveness for certain groups. The result—government-mandated microsegmentation of the market—would give some companies substantial gains and other companies substantial losses.

How well any particular drug fared under this regime would still hang largely on the ratio of high responders to low responders and nonrespond-ers, but the drug’s commercial prospects would also depend on regulatory and reimbursement issues. These include how payors set the reimbursement levels of drugs shown to be best-in-class for certain population segments and how to discourage the use of inferior drugs, which some physicians might continue to prescribe out of habit. Even the screening of individual patients might be required, as it is for thalidomide, a potentially toxic drug administered to patients with AIDS-related ulcers only if they agree to have their response to it monitored strictly.

Clinical development

Besides influencing the commercial value of drugs already on the market, genetic diagnostics could remove much risk and uncertainty from the lengthy and expensive drug development pipeline. In the United States, companies spend an average of $240 million (including a very significant amount of money for failures) on the clinical trials required as a drug progresses from Food and Drug Administration (FDA) approval for initiating clinical trials in humans all the way to final approval (Exhibit 1). Pharmacogenomics would help save some of that money and might dramatically improve the quality of a company’s clinical development portfolio.

Phase I of a development program tests for toxicity in humans. The screening of individual compounds at this time would allow companies to pursue only compounds that had higher overall chances of success. At present, rare tox-icity problems, such as those bedeviling Seldane, often can’t be identified in Phase I, which typically involves small numbers of patients; problems are more likely to surface after drugs are marketed and large numbers of patients start taking them. If genetic tests helped identify patients likely to have adverse reactions, it would be easier to weed out drugs with toxicity problems during Phase I, thus reducing the risk that they might become evident only after the drugs were marketed. On average, pharmaceutical companies could save some $36 million for each compound by eliminating unpromising ones earlier: Phase I trials typically cost about $7 million; Phase III trials (when the efficacy of compounds is evaluated in large groups of patients), $43 million.

Moreover, if pharmacogenomic testing in clinical development eventually became widespread, fewer problem drugs would enter the market, and so fewer overall would require pharmacogenomics to rescue them. In addition, these techniques could reduce the number of patients in each trial and thus the associated costs. Preselecting people likely to have positive responses to drugs would have the greatest economic effect on small-scale trials and on very large trials, involving more than 10,000 patients.

The number of patients in small-scale trials, including Phase II trials (to gather preliminary data on efficacy), could be cut by as much as half, since including patients who respond more strongly to a drug doubles a test’s predictive power. If the difference in efficacy between one drug and another (or a placebo) is expected to be small, it is now usually necessary to conduct very large trials, such as the GUSTO² trial of tissue plasminogen activator (t-PA), used for treating heart attacks. But a smaller trial including preselected, more responsive patients could reveal statistically significant differences; the bigger the trial would have been without genetic preselection, the greater the potential saving. Of course, the FDA might not allow companies to use small trials.

Even if the FDA does not demur, conducting trials that include only high responders runs the risk that the FDA would insist on making the drug available only to that group. A company may therefore wish to use a smaller preselected clinical subject base to win rapid (though restricted) approval and then take advantage of broader clinical trials in Phase IV to expand indications.

Finally, pharmacogenomics might reduce the length of trials, something that would be particularly beneficial for developing drugs to treat degenerative diseases whose early stages are relatively asymptomatic. Some evidence suggests, for example, that genetics may be a factor in early-onset Alzheimer’s disease, which often takes a long time to manifest itself. If the gene (or genes) responsible for it could be identified and a genetic diagnostic developed, clinical trials involving patients who have that gene inheritance would reveal more quickly—and therefore more economically—whether a new drug for early-onset Alzheimer’s actually worked.

If the three levers of cost reduction—a reduction of the number of compounds, the number of patients, and the length of trials—were applied across all phases of clinical development, pharmacogenomics could eliminate $60 million to $85 million in costs for each approved drug (Exhibit 2). The savings take into account the higher expense of pharmacogenomic testing—perhaps $5 million to $10 million for every approved drug.³

The use of genetic diagnostics to cut the cost of trials by segmenting patients is already fairly advanced in the area of toxicology. We suspect that the biggest near-term impact of pharmacogenomics on drug manufacturers will be to reduce the number of drugs evaluated after Phase I.

Operational strategies

Pharmacogenomics will probably affect not only the cost of developing drugs and their commercial value but also the operational strategies of much of the pharmaceutical business.

In the basic-research stage—particularly in the identification of a drug’s molecular targets, such as a receptor or an enzyme—scientists will have to understand both the breadth of genetic variation among specific drug targets within various populations and how the drug would work on its molecular target. In the development stage, companies will have to plan additional exploratory clinical trials, to examine linkages among genetic polymorphisms and variations both in the presentation of diseases and the responses of patients to particular therapies. In the sales and marketing stage, pricing strategies will become increasingly important levers of profitability as markets become more segmented. Some drug companies will succeed in charging higher prices for certain drugs, but pharmacogenomic studies showing that particular patients receive unusually large benefits from these drugs will be needed to do so.

Only a few pharmaceutical concerns (including SmithKline Beecham and Roche) have as yet invested significantly in genomics. They tend to be the companies—like the biotechnology businesses Millennium Pharmaceuticals and Incyte Pharmaceuticals—that already have major genomics databases. Since leveraging them promises to transform the overall economics of developing and selling drugs, the question for the rest of the pharmaceutical and biotechnology industries isn’t whether to invest but how soon.