HIV Prevalence Estimates: Fact or Fiction

Better data have led to a drop in the estimated global HIV prevalence in some countries, but the epidemic still continues to outpace international responses

By Kristen Jill Kresge

Science and politics often clash—consider embryonic stem cell research or the even more quotidian debate over global warming and its consequences. These and many other issues are hotly contested in both political and scientific circles.

Politics has always been at the forefront of the HIV/AIDS pandemic. Even before it had a name, HIV was a political issue. In the days when it first started spreading in the US, rapidly killing those who became infected, the people who would soon be branded AIDS activists implored the US government to openly discuss and actively confront this new disease. As a result there is more legislation singly devoted to HIV/AIDS than any other disease.

Now some are suggesting that science and politics may be colliding again—this time in the fundamental way the scope of the global HIV/AIDS epidemic is measured. Some epidemiologists have called into question the accuracy of global HIV prevalence estimates, which represent the total number of people who are thought to be infected with the virus in a region or country at a specific point in time. Prevalence figures are used by governments, public-health agencies, and donor organizations to gauge the severity of the pandemic and this, in turn, drives decisions about how and where money is spent on both HIV prevention and treatment. These estimates are regularly updated by The Joint United Nations Programme on HIV/AIDS (UNAIDS), in partnership with the World Health Organization (WHO), but the responsibility of collecting the data falls to the individual countries.

In recent years many of the HIV prevalence estimates have been revised based on improved data and in almost all cases the new estimates are lower than previously thought, sometimes dramatically so. As a result the total number of people in the world thought to be infected with HIV keeps going down. A few years ago UNAIDS estimated that 42 million people were HIV infected. Now the number stands just below 40 million, according to the2006 Report on the Global AIDS Epidemic. The question about the accuracy of the estimates was pushed to the forefront recently when India cut its HIV prevalence numbers by half. UNAIDS had previously estimated that India had over five million HIV-infected individuals, meaning it had more people living with HIV than any other nation.

This has led some to accuse UNAIDS of crying wolf about the severity of the global pandemic—some of the harshest critics even suggest UNAIDS is purposely exaggerating prevalence figures to sustain political momentum and funding for HIV prevention and treatment programs. But even if a more accurate picture of the global HIV/AIDS pandemic shows that prevalence is lower than originally thought, everyone agrees it still warrants urgent attention. "Even if you cut the [HIV prevalence] numbers in sub-Saharan Africa in half, it's still a huge problem," says James Chin, a retired epidemiologist and faculty member at the University of California, Berkeley who is one of the most outspoken critics.

Getting better data

HIV prevalence estimates are generated by epidemiologists using HIV infection data from small subsets of the population that can be extrapolated using mathematical models. These models combine national population estimates and epidemiological data collected in a country and then churn out estimates of national HIV prevalence, based on a series of assumptions. In South Africa, where there is the largest number of HIV-infected individuals, the national HIV prevalence among adults between the ages of 15 and 49 is estimated by UNAIDS to be nearly 19%. The number of HIV infections is not evenly distributed within the population—many countries have epidemics that are still mainly contained within certain regions or in groups that are at especially high risk, such as injection-drug users (IDUs) or commercial-sex workers (CSWs). In some regions of South Africa or in high-risk populations, the prevalence estimates can be twice as high as the national estimate.

Since its inception in 1995, UNAIDS, along with WHO, has released annual estimates of regional HIV prevalence and biannual estimates of national HIV prevalence that serve as the standard measure of the extent of the pandemic. These numbers are one of the primary drivers behind decisions about funding for AIDS-related prevention and treatment programs worldwide and therefore receive a great deal of international attention.

Over recent years and months, many of the UNAIDS estimates have been revised. Most of these revisions reflect substantial downgrades in the numbers, indicating that there are fewer HIV-infected individuals than previous estimates suggested (see Figure 1). Several factors contribute to this revision of HIV prevalence, including the increased or improved surveillance of HIV infection in many countries, better population estimates, and more accurate models for estimating prevalence on a population basis. The positive influence of HIV prevention campaigns also plays a role, though it is often difficult to directly pinpoint. In Uganda, many epidemiologists have suggested that the dramatic drop in HIV prevalence in the mid-1990s was at least partly due to the high death toll of individuals who were infected early on in the epidemic.

Figure 1. Adjusted HIV Prevalence Estimates.

View larger image

But in most cases the revisions to the UNAIDS estimates are based on the collection of better data that more accurately represents the burden of HIV infection in individual countries. In its 2006 update on the pandemic, UNAIDS said that "new systems, including greatly improved surveillance, tell us with increasing accuracy where and how the epidemic is moving." Many countries are conducting more rigorous surveillance of their epidemics, both in the general population and in high-risk groups, by either increasing access to voluntary counseling and testing services or conducting household surveys that are part of the broader demographic and health surveys (DHS). These population-based surveys allow researchers to track the spread of several diseases in developing countries and monitor trends in overall health. In DHS surveys, researchers randomly visit households in a community and collect medical information from the available family members. Recently this survey was altered to include collection of a saliva sample that could later be used to conduct an HIV test.

Previous prevalence estimates have been based primarily on sentinel surveillance data collected from pregnant women who visited antenatal clinics, one of the few settings where there is mandatory HIV testing. The original method of projecting prevalence based on data from antenatal clinics was established in the 1980s by Chin when he was working at the Global Program on AIDS at WHO, years before the job of tracking the pandemic came under the purview of UNAIDS. He thought HIV prevalence data collected from sexually-active women would be a good surrogate for national prevalence.

But in most cases this data was not representative of HIV infection for the entire population. Most antenatal clinics are located in urban areas, where HIV prevalence is generally much higher, and the pregnant women who would take advantage of healthcare generally have a higher income, which introduced another bias. When Zambia conducted the country's first population-based health study it found that estimates for HIV prevalence based on antenatal clinic data correlated very closely in urban areas, but that neglecting rural populations led to a gross overestimation of national HIV prevalence.

"Data from antenatal clinics help monitor trends over time," says Karen Stanecki, a senior advisor at UNAIDS in Switzerland. "The intent [with data from pregnant women] is to monitor changes, not to predict the actual number of people who are infected," says Prabhat Jha, professor of epidemiology at the Center for Global Health Research at the University of Toronto.

Watch out for falling estimates

Following pressure from donor organizations to come up with more accurate prevalence estimates, more countries began conducting population-based surveys, often leading to a drop, sometimes precipitous. Kenya reduced its estimated HIV prevalence in 2003 after conducting a population-based survey, from 2.3 million HIV-infected individuals to 1.2 million. "That was a huge reduction," Chin says.

Following that, more than a dozen other countries conducted population-based surveys that led to revisions in the UNAIDS prevalence estimates. In Ethiopia the number of HIV-infected individuals was cut by half to one million. Cambodia also lowered its national prevalence estimate, from 1.8% of the population to less than 1%. India is one of the latest countries to release new figures indicating that the estimated national HIV prevalence is only half that previously projected by UNAIDS.

Chin suspects that prevalence figures might also have been overestimated in the US. In 1986 it was estimated that there were between 1 and 1.5 million people infected with HIV nationally, and the fact that the numbers have not changed since then suggests to Chin that the original figures were an overestimate.

Now 30 countries have conducted population-based surveys to help better gauge the extent of their HIV/AIDS epidemics. In Benin, Mali, and Niger the results from these surveys were very similar to the figures estimated using sentinel surveillance data from antenatal clinics, but in the majority of cases the new figures were lower. "There are still plenty of countries that haven't done these types of studies," says Chin, which is why he thinks that the total UNAIDS figures are still an overestimate. But in countries that still use antenatal data as a basis for determining HIV prevalence, UNAIDS now adjusts these figures to account for rural populations.

Population-based surveys have several advantages—they reach more individuals in rural areas and include men. But they have disadvantages as well. "The other side of the coin is that people may refuse HIV testing," says Stanecki. "This introduces a bias."

These household surveys are also limited to countries where there is a well-developed HIV/AIDS epidemic. "We don't recommend that they be conducted in countries with low-level prevalence," Stanecki adds. Population-based surveys are only applicable in countries where 1% or more of the population is HIV infected, which excludes many Asian countries where the HIV epidemic has not progressed as rapidly as in sub-Saharan Africa.

These surveys also tend to exclude marginalized individuals who are often at the highest risk of HIV infection, including IDUs, CSWs, or transient workers, so countries where the HIV epidemic is still confined within high-risk groups could underestimate prevalence. To adjust for these discrepancies epidemiologists collect further data within these populations and then adjust accordingly. But the models are still imperfect. "There's always going to be a lot of bias," says Seth Berkley, president of IAVI, who was involved in tracking the HIV epidemic in Uganda when epidemiologists first started estimating prevalence there. But, he adds, "The numbers for HIV are probably better than for any other disease ever. It's AIDS that has been the big controversy."

The onus of collecting better data falls on the individual countries that have to finance and conduct population-based surveys. "We [UNAIDS] don't do any surveys," says Stanecki. "Surveillance is done by the countries themselves." UNAIDS and WHO just work with countries, holding regional training workshops on the modeling tools, and assist with calculating national HIV prevalence estimates.

Figure 2. Global HIV Prevalence Estimates.

View larger image

Politics at play

There are obvious political reasons both for and against individual nations collecting better data on the scope of the HIV/AIDS epidemic. Some countries are motivated to conduct household surveys to show that the epidemics are not as bad as estimates suggest and to prove to the international community that the government is handling the epidemic. Other countries may be leery of showing that there is less of an HIV/AIDS problem because it could result in funding cuts for the country's AIDS-related programs. This controversy was reignited when India's National AIDS Control Organization (NACO) released new prevalence estimates in July, in cooperation with UNAIDS and WHO.

NACO reported that the new estimates were the result of a considerable increase in the number of HIV testing sites in both rural and urban areas and in low-prevalence Indian states, as well as the conduct of comprehensive household surveys. Most agree that these new estimates are more accurate than before. Jha refers to the previous prevalence estimates in India as "guesstimates" and says that the "sources for the new data are better, but still not perfect." He also points to two supporting pieces of evidence that corroborate that the Indian epidemic is not as extensive as originally thought. A study published in 2006 showed that HIV prevalence in Southern Indian states, where two-thirds of the HIV-infections are located, were declining and other surveys indicated that AIDS mortality rates were less than original UNAIDS/WHO estimates (Lancet 367, 1164, 2006). But there is still a risk that basing the new prevalence estimates on household surveys, which limit access to high-risk individuals, may underestimate the scope of the problem.

Stanecki defends the new estimates for India. "The estimates that were done in the past were based on limited data and we now have better information," she says. "We recognize uncertainty in the estimates with ranges and [for India] it's now a much smaller range." But Jha says it is still a pretty wide confidence interval for the Indian prevalence estimates.

Strange bedfellows

As HIV prevalence estimates continue to be downgraded, some epidemiologists are questioning whether politics might be interfering with the science of tracking the pandemic. "Each year we get numbers from UNAIDS, but we don't have easy access to the supporting analyses and calculations," says David Ho, director of the Aaron Diamond AIDS Research Center in New York City. "Those [analyses] should be put out there for the entire scientific community to comment, along with the conclusions and projections."

Stanecki says this process is already in place. UNAIDS appoints a reference group, including outside scientists and experts, to review the models and publishes all of the findings from this group, she says. But the exact methodology that was used to establish the new prevalence figures for India has not yet been released publicly. Jha says that if anything the Indian experience should argue for making the prevalence numbers "completely transparent in the future."

Chin argues that UNAIDS is reticent to lower the estimates even further because it will only make it more difficult to sustain political momentum and funding for HIV prevention and treatment programs. This suggestion is controversial. "We've been continuously lowering our numbers over the past years and we do this in collaboration with countries," says Stanecki. "We don't have any agenda." Chin isn't convinced. "Regardless of whom they've been working with, they are making gross overestimates," he says. "They're an advocacy organization but they shouldn't ignore the science."

Whether or not the numbers are too high, funding and expanding HIV prevention and treatment programs remains critical—only a minority of HIV-infected individuals in developing countries currently receives life-saving antiretrovirals (ARVs) and last year alone four million people were newly infected with the virus. In a recently published book, Chin provocatively accuses UNAIDS of misapplying mathematical models to produce exaggerated estimates and then giving credit to the agency's prevention programs for the declining prevalence. Jha emphasizes the fact that the new lower prevalence estimates in India are not a direct result of HIV prevention programs. "It's not due to a control program, but to a computer program," he says.

Stephen Lewis, the former UN Special Envoy for AIDS in Africa, and Paula Donovan, who was Lewis's senior advisor at the UN, wrote in a review of Chin's book that it poses an "open challenge to the UN's role in the most eviscerating plague in human history," (Nature 447, 531, 2007).

Others also view some of the dire predictions and projections previously issued by UNAIDS on the potential expansion of the HIV/AIDS epidemic in Asia as exaggerated.

In a 2002 publication called "HIV/AIDS: China's Titanic Peril", UNAIDS projected that there would be 10 to 20 million HIV infections in the country by the end of this decade. "Although we still have two to three years to go, I do not think China will come close to that figure," says Ho. China already lowered its prevalence to 650,000 from around a million and, according to Chin, it could be cut in half again as more data is collected.

Many researchers and epidemiologists now agree that the Asian HIV epidemic is unlikely to bring about the devastation that was initially predicted. Yet the 2006 UNAIDS update report contained another dire warning about the spread of HIV in Asia, referring to "rapidly growing epidemics in regions such as Eastern Europe and South-East Asia that may come to rival that of sub-Saharan Africa in scope." Stanecki says that UNAIDS doesn't like issuing projections. "We don't really project to say where the epidemic is going," she says.

Mind the gap

While better surveillance has allowed epidemiologists to collect data that most agree are more representative of the scope of the epidemic, there is still an enormous gap between what is needed to control and eventually end the HIV/AIDS pandemic and what is currently being done. "The numbers are lower, but there's still the possibility of explosive growth," says Jha. There is an overwhelming need for improving the availability of ARVs to HIV-infected individuals in developing countries and new prevention methods, including AIDS vaccines, to help prevent the millions of new HIV infections that still occur each year.

"It's great to get better estimates, but from an epidemic point of view it isn't over," says Berkley, who points out that even in countries with a stable epidemic like the US, there are still a substantial number of new HIV infections each year. "What India, and the rest of the world, should do is focus on prevention, especially for high-risk populations, and continue accelerating vaccine research," says Jha.