A Statistical Solution to Testing the Blood Supply for HIV
A Stanford GSB scholar devises an advanced model that allows inexpensive, accurate testing by pooling blood samples.
The blood supply has been a chief pathway for the spread of AIDS. To protect it, the American Red Cross long ago started testing the blood of all volunteer donors. But in poor regions of the world, such vigilant screening is out of the question. In parts of Africa, for example, nearly 30 percent of all blood donors may go untested in a given month because there is not enough money to procure test kits for everyone. Doctors test as many as they can. The rest is left to chance.
The human consequences are devastating. Testing for AIDS is a difficult problem in which Stefanos Zenios, assistant professor of operations, information, and technology, has taken a keen interest. Using advanced statistical analysis, Zenios and Lawrence Wein, professor of operations, information, and technology, have devised an accurate way to test more cheaply by pooling blood samples. If a batch of 10 samples produces a negative result, then the cost of nine tests has been saved. If the test is positive, samples can be subdivided and retested.
It sounds simple. In fact, the idea of pooling is not new. World War II draftees were screened for syphilis using pooling methods. But the nature of the laboratory process used to detect antibodies for HIV, the virus that causes AIDS, significantly complicates group testing. The chief bugaboo is a dilution effect that obscures the presence of the telltale antibodies.
Blood under scrutiny for HIV is not tested in a simple one-step process. Rather, a blood sample is incubated in a dish and then enzymes are added. That results in a color change, which is measured by an automated spectrometer to determine the concentration of antibodies for the deadly virus. If the reading exceeds a critical cutoff, the patient is declared HIV positive. The problem is that when samples are pooled, the sheer volume of blood can dilute the antibodies so they fade out of sight. “The test doesn’t have infinite accuracy,” says Zenios.
As a result, Zenios and Wein constructed a model that captures the dilution effect by selecting a cutoff point and pool size that do not compromise the sensitivity of the test. Their model predicts how much an infected sample must be diluted in order to produce a false negative result. By testing the parameters and factoring in the percentage of those infected in the population at large, the researchers were able to devise a safe cutoff and pool size. Currently, the World Health Organization condones pooled testing only when the pool size is less than six.
However, Zenios and Wein determined that in countries such as the Democratic Republic of the Congo (formerly Zaire), where 2.5 percent of blood donors are infected, as many as 10 samples could be accurately tested together. A minuscule risk would still remain, but the population would be far more protected than if testing continued as it has. In that country’s Kinshasa, for example, 28 percent of the 3,741 blood units transfused in February 1990 were not screened at all because of budget constraints. “They test individually until they run out of tests, then they stop screening until the next shipment,” says Zenios. Using the data from Kinshasa, Zenios and Wein calculated that pooling would reduce costs by 40 percent and lessen the expected number of infected units transfused from 27 to nearly zero.
In countries like the United States, where HIV-infected individuals represent only .04 percent of the low-risk blood donor population, the researchers predict as many as 40 samples could theoretically be pooled with accurate results. However, pooling is not used in the United States because it has the resources to test each individual.
Zenios and Wein’s model cannot account for “window period” donors who may be infected but have not produced enough antibodies to show up in the test, producing a false negative. Nevertheless, monitoring techniques, such as excluding donors who have had risky sexual contact within a 12-week period, would alleviate a large part of that risk in developing countries. While Zenios and Wein will conduct more research to validate their results, they conclude: “It is clear that pooled testing, if used properly, can save hundreds of lives worldwide.”
For media inquiries, visit the Newsroom.