The promise of many biotechnology drugs is that they are the body's own solutions. Human proteins like insulin and growth hormones, the same substances that the body uses every day, are made through genetic engineering and given to people who do not make enough of the proteins on their own.

But an outbreak of serious illnesses linked to the anemia drug Eprex shows that some patients do not react to genetically engineered proteins as if they were natural. The patients react as if the protein were a germ, and their immune systems try to destroy it.

In the case of Eprex, which is made by Johnson & Johnson and is sold only outside the United States, this immune response is widely believed to be responsible for 141 cases of pure red cell aplasia.

With this condition, the body is unable to produce red blood cells, making some patients dependent on transfusions to survive.

Although the Eprex case is the most serious, virtually all biotechnology drugs provoke immune responses in some patients, though usually just tiny fractions. The reactions are becoming of greater concern as the number of protein drugs increases.

"Sometimes there are miracle drugs, but they can still have severe side effects," said Dr. Huub Schellekens, a professor at Utrecht University in the Netherlands. "That has come as a surprise to us, really."

In some cases, patients experience allergic reactions or even potentially fatal shocks. In many cases, the body makes antibodies that attack the protein, rendering it less effective as a drug.

Up to one-third of hemophiliacs develop immune reactions to the blood-clotting protein Factor VIII. In desperation, some undergo yearlong treatments at a cost of $1 million to try to restore the drug's usefulness.

From 5 to 40 percent of multiple sclerosis patients develop resistance to beta interferon, according to Dr. Richard A. Rudick, director of the Edward J. and Louise E. Mellen Center for Multiple Sclerosis Treatment and Research at the Cleveland Clinic. Although there is debate about whether the presence of the antibodies is important, Dr. Rudick said, many such patients are unlikely to benefit from the drug.

What makes the illnesses associated with Eprex more alarming is that the antibodies produced by the patients do more than make the drug ineffective.

They also attack the patient's equivalent protein. So the patients can no longer produce any red blood cells, ending up with worse anemia than than they would have had without the drug.

Amgen saw the same problem a few years ago with patients in a clinical trial for a drug aimed at increasing the production of blood platelets. The company halted work on the drug, which never made it to market.

The reasons for the immune responses are not clear. In some cases, the drugs, which are made by putting the human gene for the protein into bacteria or animal cells, are subtly different from the natural protein. One such difference may be in the sugars that coat the protein. In other cases, the drug may have an impurity, or the proteins may clump together.

In the case of Eprex, the cause remains a mystery. The factory, in Puerto Rico, is under criminal investigation. Johnson & Johnson says it has done nothing wrong, and the plant passed two recent inspections.

The same protein, sold in the United States as Procrit by Johnson & Johnson and as Epogen by Amgen, is linked to just a handful of cases of red cell aplasia. Amgen, which developed the drug, makes both brands in a factory in Colorado.

Experts say that because living cells make biotech drugs, the output is not as predictable as with chemically made drugs. Even slight changes in manufacturing can affect the product, sometimes in ways that cannot even be detected.

"Despite best efforts to detect product differences and predict the impact of manufacturing changes, these surprises do continue to occur," Dr. Chris Joneckis, an official of the Food and Drug Administration, said in May at a workshop on the problem. The F.D.A. declined to make officials available for this article.

Immune responses also occur with protein drugs not made by genetic engineering like those purified from blood or derived from animals. The antivenin for rattlesnake bites, from horses' blood, can be used just once because after that a person's immune system is primed to destroy it immediately.

"When rattlesnake handlers get just a small bite, they don't get antivenin because they want to save it for a large bite," said Dr. Douglas J. Ringler, chief executive of TolerRx, a company in Cambridge, Mass., that is developing techniques to prevent immune reactions to protein drugs.

The 141 suspected cases of red cell aplasia pale in comparison to the three million people who have been treated with Eprex or its American counterparts for more than 10 years. Many of those patients are able to work or lead more normal lives because the drug eliminates the fatigue of anemia.

The drug is a form of erythropoietin, a protein produced by the body to spur the production of red blood cells, which ferry oxygen around the body. Before the advent of genetic engineering, scientists had tried to isolate the protein from human urine but could not obtain enough to use it as a drug.

In many cases, the genetically engineered proteins replace those taken from animals or from human blood or tissues. The biotech drugs are widely considered safer because they do not carry the risk of viral infection.

Human insulin made by genetic engineering supplanted insulin from animals. The human version generally provokes far fewer immune reactions and lets diabetics control their insulin better. Nevertheless, a small percentage of diabetics say they cannot tolerate the human version and are trying to keep cow and pig insulin available.

For hemophiliacs, Factor VIII made by genetic engineering is widely seen as safer than the same protein purified from human blood, which has a small risk of infections. But up to 35 percent of hemophiliacs develop antibodies to the genetically engineered factor, said Dr. Louis M. Aledort, a professor at the Mount Sinai School of Medicine. Fifteen percent developed antibodies to the blood-derived product, Dr. Aledort said, although the numbers may not be comparable because of differences in measuring.

Although the Eprex problem is something of a black eye for what has been the biotechnology industry's most successful product — Eprex and its American cousins exceed $5 billion in combined yearly sales — it could actually help in the long run, by eliminating the threat of generic competition. For drugs made chemically, generic manufacturers do not have to run expensive clinical trials. They need to show just that their drug is identical to the brand-name product.

But the biotechnology industry and the F.D.A. have said it is too difficult to show that two biotech drugs are equivalent without full clinical trials. Eprex is likely to become Exhibit A in such arguments.