Estimates on HIV called too high
- Now that surveys like the DHS are doing population-based HIV testing in
national sample surveys, previous national estimates of HIV prevalence should
continue to be revised downward- Scott
Estimates on HIV called too high
New data cut rates for many nations
By John Donnelly, Globe Staff | June 20, 2004
PRETORIA -- Estimates of the number of people with the AIDS virus have been
dramatically overstated in many countries because of errors in statistical
models and a possible undetected decline in the pandemic, according to new
data and specialists on the disease.
In many nations, analysts are cutting the estimates of HIV prevalence by
half or more.
Rwanda, for instance, a new United Nations estimate due out next month will
put HIV prevalence at about 5 percent, according to Rwandan officials, down
from more than 11 percent four years ago. In Haiti, a recent unpublished
study by the Centers for Disease Control and Prevention has found HIV prevalence
was less than 3 percent, compared with the UN's most recent estimate of
6 percent. And the numbers in India are coming under increasing scrutiny
because surveys in AIDS hot spots are indicating a prevalence rate that
is much lower than the national average.
Even with lower estimates, health specialists agree that AIDS remains the
most dangerous pandemic in the developing world. In particular, it threatens
to ravage societies in southern Africa, and throughout the continent the
disease has killed millions in the prime of their lives.
Several AIDS specialists said they think the current estimate of 40 million
people living with the AIDS virus worldwide is inflated by 25 percent to
50 percent, based on a wide spectrum of household surveys in nearly a dozen
countries. That would go against the grain of years of assertions by UNAIDS
that the disease is relentlessly on the rise.
A significant downward revision in AIDS and HIV numbers calls into question
many of the lessons on fighting AIDS that are based on prior estimates.
It also is likely to affect future budgets and cause many countries to consider
revising strategies on how to prevent and treat the disease.
"It is fundamental that we have accurate information of what we're up against,"
said Robert R. Redfield, cofounder of the University of Maryland's Institute
of Human Virology and a leading AIDS specialist. "If you are overestimating
the epidemic, you may attribute positive impacts to things that have nothing
to do with it."
UN epidemiologists and statisticians responsible for the current estimates
acknowledged in interviews that some country figures will be sharply cut
in a forthcoming report, but they played down any possibility now of a sizable
reduction in the overall number of infected people. They declined to disclose
their most recent estimates for 2003, set to be released next month before
the 15th International AIDS Conference in Bangkok.
The major error in the estimation of HIV and AIDS numbers has occurred primarily
because epidemiologists relied too heavily on HIV rates in urban areas and
failed to factor in much lower prevalence in rural areas, where surveys
are rare, AIDS specialists said.
Suggestions that prevalence may be overstated may not be embraced by some
who work to fight HIV or in countries hard hit by the epidemic. Billions
of dollars in aid is at stake in the coming years. If HIV seems less menacing,
more dollars may go to fighting other diseases or may be redirected away
from global health initiatives entirely.
Already, earlier this year, US officials told Rwandan AIDS administrators
that if HIV prevalence estimates were to drop to 5 percent, the country's
AIDS funding may be cut, according to both US and Rwandan officials, speaking
on condition of anonymity.
Another reason behind the plunging rates, some AIDS specialists say, is
that the pandemic may be slowing.
The upcoming UNAIDS report is expected to show that HIV prevalence is declining
in eastern Africa and leveling off or slightly slowing in West Africa, but
still maintaining a high rate in southern Africa.
Statisticians traditionally have had a difficult time estimating the size
of the pandemic. In 1986, Jim Chin, then a state epidemiologist in California
who later developed models for the World Health Organization to calculate
HIV prevalence, and several other US officials met in a West Virginia hotel
room to figure out how many Americans had HIV.
Chin recollected that the group arrived at a range of 1 million to 1.5 million
people; 18 years later, the number is at about 1 million Americans. "A lot
of it was guesswork, based on limited studies," Chin said. "It was the best
we could do."
The tools today are much more refined but still based on a long list of
More than a decade ago, AIDS researchers in sub-Saharan Africa found that
HIV tests on blood samples from pregnant women at prenatal clinics provided
a good indicator of HIV prevalence among adults aged 15 to 49 in countries
with high rates; early household surveys confirmed the finding.
But the surveys were limited at first to a few sites in countries. "We were
talking about four or five urban sites and one or two rural sites, and extrapolating
that to the whole country. You can see what potential inaccuracies there
can be with this crude methodology," said Chin, who now is an independent
AIDS analyst and criticizes UN estimates as overstated.
Other unknowns contribute to potential errors. One is estimating a country's
population; the estimates for Nigeria, for instance, range from 120 million
to 160 million people, but a census of the country has not been completed
in more than half a century. Another is that most countries do not collect
data on deaths.
Once officials from UNAIDS and the WHO arrive at an HIV prevalence estimate
for a country, they use that estimate to help determine AIDS deaths, AIDS
orphans, numbers of people needing antiretroviral treatment, and the average
life expectancy for people in countries. As HIV prevalence numbers are adjusted
downward, the numbers for the other AIDS-related categories also will be
To estimate the number of people dying from AIDS, for instance, epidemiologists
assume that on average a person will live for eight or nine years after
infection; they then plot the progression of a country's epidemic, determining
how many people were infected in each year. If HIV prevalence estimates
are significantly wrong in any year, estimates for AIDS deaths will be correspondingly
wrong eight or nine years later.
Some specialists raised questions about the estimates in mid-January, after
a report on a household survey in Kenya that estimated a 6.7 percent national
HIV prevalence rate, compared with the UN's 15 percent estimate in 2002.
Two weeks later in Addis Ababa, Ethiopia, at a conference on estimating
HIV and AIDS prevalence, epidemiologist Peter D. Ghys raised the question
of potentially inflated AIDS rates before several dozen epidemiologists.
Ghys estimates HIV and AIDS figures for UNAIDS. On the last slide of his
presentation, Ghys wrote, "Should UNAIDS/ WHO estimates be lowered by 25
Asked recently about his question, Ghys said he raised it in light of the
new Kenyan data: "We already had lowered estimates quite a bit in the previous
year, but since then, we've had additional information that came in somewhat
lower than our estimates. Maybe it will change again, but it's difficult
for us to say, yes, it should be 25 percent or something."
In 2003, the UN revised HIV prevalence in sub-Saharan Africa to 26.6 million
people, from 29.4 million the year before.
Ties Boerma, who is in charge of the WHO's estimates on the HIV virus, said
the UN was preparing to reduce prevalence estimates again this year, although
not by 25 percent. Yet Chin and several other specialists say new data from
several countries and from high-risk groups in countries suggest that the
UN should do so.
Chin said he thinks the global rate is inflated by 25 percent to 40 percent,
while two US health officials working on AIDS said they think the global
numbers may be 50 percent inflated. The two spoke on condition of anonymity.
Among their reasons:
In a slum of the Chennai district in India, a high-risk area for transmission
of HIV because of prostitutes and drug users, a recent household survey
found 0.2 percent HIV prevalence. The nationwide rate is 0.8 percent, or
4 million people. Ghys acknowledged: "It's difficult to say what is going
on in India. In India, there's always a great deal of uncertainty in those
estimates. Are we sure it's 0.8 percent? The honest answer is no, we are
A recent household survey in Burkina Faso put the national estimate at 1.9
percent; the UN's most recent estimate puts the rate at 7 percent. In Africa,
other recent surveys showing significantly lower rates have been in Zambia,
South Africa, Mali, and Ghana. Rates also were lowered by a third in Zimbabwe
because of significant numbers of faulty HIV tests.
Several years ago, UNAIDS estimated that up to 60 percent of the Angolan
military was HIV positive. Dr. Richard Shaffer, head of the US Department
of Defense's HIV/AIDS Prevention Program, said in an interview the estimate
was "nowhere near close to that. It's 6 to 7 percent. They based the earlier
number on a small sample, which included people outside the military, and
extrapolated that to the military as a whole."
In the late 1980s and early 1990s, HIV prevalence in adults aged 15 to 49
in Uganda was estimated as high as 30 percent; now HIV prevalence is estimated
at 5 percent. But now many no longer believe the 30 percent figure, raising
questions about the true impact of Uganda's much-touted prevention program.
Said Ghys: "If we recast our estimates, it wasn't 30 percent, it was maybe
22 or something."
Earlier this year, the US government announced its first substantial grants
in President Bush's multibillion-dollar plan to fight AIDS. The news release
cited a 15 percent HIV prevalence rate in Kenya and a 6 percent rate in
Haiti, even though US-funded surveys in both countries had recently concluded
that the rate was at least half those figures.
- Let me be the next to congratulate Steve Berry. I
believe this past weekend was when he got married.
Good luck Steve, and send us a photo when you can.
Do you Yahoo!?
Yahoo! Mail - 50x more storage than other providers!
- I don't know Steve, but this reminds me of a joke:
Two unbiased estimators are sitting at a bar having a drink. One asks "So
how do you like being married?" The other one says "It's okay, but you lose
a degree of freedom."
Well... at least it gets a chuckle in a room full of social scientists.
>-- Original Message --
>From: John Patten <jppatten98@...>
>Date: Wed, 7 Jul 2004 05:48:20 -0700 (PDT)
>Subject: [ujeni] Contratulations Steve Berry
>Let me be the next to congratulate Steve Berry. I
>believe this past weekend was when he got married.
>Good luck Steve, and send us a photo when you can.
>Do you Yahoo!?
>Yahoo! Mail - 50x more storage than other providers!
>Yahoo! Groups Links