February 10, 2009
My name is Michael Campsmith, and I'm an epidemiologist with the Centers for Disease Control and Prevention [CDC] in Atlanta. My group is the HIV incidence and case surveillance branch. I'm presenting at the CROI conference on an analysis of the prevalence of HIV in the United States at the end of 2006, and we're focusing on the percent of persons who are undiagnosed -- who have HIV infection but are undiagnosed.1
If we don't know they're diagnosed, what's the process the CDC uses to arrive at these numbers?
All areas [in the U.S.] are funded to collect surveillance data. They do that at the local level and send that data without personal identifiers to the CDC. We compile national data, and then there's a statistical modeling procedure.
Michael Campsmith, D.D.S., M.P.H.
When HIV and AIDS were first identified, we knew that HIV caused the infection, but AIDS was the end result. So there was interest in how many HIV cases would you need to result in the number of AIDS cases that you were seeing later? That was called a back calculation model because you just looked at the number of AIDS cases and estimated backwards -- how many HIV infections would have been needed years ago to result in that many AIDS cases.
That was good enough until highly active antiretroviral therapies [HAART] came about in the mid-1990s, which extended the period from HIV infection to AIDS in people who were taking therapy.
So a back calculation strictly from AIDS data is not sufficient, because it's not accurate enough. So now our technique is a revision, or an improvement, of that back calculation to be an extended back calculation.
We modeled that using HIV data from the states where it's available. In our analysis it was 40 out of the 50 states, D.C. -- District of Columbia -- also was not included for HIV data.
By the way, did you include New York, California and Florida -- three states with a lot of HIV cases?
California no; New York and Florida, yes. One of the limitations of the model is that we don't have HIV data from all areas. There is some uncertainty; people need to understand that this is an estimate. We can't directly observe this.
And compared to other states, Missouri or something, California's pretty densely infected.
California has a large population and a large number of cases. But again, we use this statistical estimation procedure to come up with reasonable estimates of this.
Tell me a little bit about your results.
The results are that overall, about 1.1 million people in the United States are estimated to be living with HIV infection, both diagnosed and undiagnosed. Of that 1.1 million, about 21%, or 232,000, don't know that they're infected, they don't know because they haven't been diagnosed yet. It's about 20, 21%.
So 1.1. million is the new number for overall infections in the U.S. Previously, for many years the CDC has reported that it was 1 million.
This is a revision, it's an update. As we gain more states contributing HIV data to our estimates, we get revisions of this. Also, over time, the number of new infections -- the incidence estimates which came out last year -- show that more people are getting infected than are dying from the disease, so there is that natural escalation in prevalence, or the number of people living with HIV.
And it reflects that people live a lot longer with HIV.
Yes, there are new HIV infections every year, but people are not dying as quickly as they were earlier in the epidemic, the goal is to get people who are infected to get diagnosed, get into care, and have a prolonged life. While they do know that they're infected, you want to get them into effective prevention programs as well, so they don't further transmit HIV.
Can you tell me how these numbers break down in terms of ethnicity and gender?
The highest percentage of cases, prevalent cases, is among African Americans. There is a differential percent for undiagnosed, where minorities, racial and ethnic minorities, tend to have higher percentages among undiagnosed than whites.
By risk, men who are infected heterosexually had the highest percentage of undiagnosed, followed by men who have sex with men. By age, the youngest age group, 13 to 24, had a higher percentage of undiagnosed compared to other age groups.
Is there overlap between heterosexual and IDU?
There's a hierarchy of presumed transmission. So if people exhibit all the risk characteristics, we have a hierarchy that says, "this is what we think this is the most likely one." We can't say exactly for sure. If someone had injection drug-use behavior and was a commercial sex worker, what was the exact mode of transmission? But the hierarchy goes, if they have injection drug use and they don't also have male-to-male sexual contact, they're counted as injection drug users. Heterosexual contact is specifically identifying heterosexual risk for someone known to have, or be at high risk for, HIV infection.
So for a heterosexual woman, she would identify risk as, "I had sex with an injection drug user," or "I had sex with a bisexual male."
For a heterosexual man, it would be, "I had sex with an injection drug user," or "I had sex with a commercial sex worker." Those sort of risks. So when we say heterosexual, it's the identification of the risk.
Injection drug users are classified, their transmission risk was injection drug use.
Previously the CDC had announced that 250,000 people were thought to be living with HIV and not know it, so the number has changed a little bit.
You have an increase in the overall prevalence, so if nothing else changed, that percentage undiagnosed goes down as the overall prevalence goes up. But there also has been more HIV testing done through the testing initiatives.
So if you take a look at the previous estimate of 250,000, our new estimate is just under 233,000. So there is both a decrease in the percentage and a decrease in the number of undiagnosed persons.
So the small amount of extra HIV testing going on across the country has contributed to changing these numbers? Because -- even though the CDC made the recommendations in 2006 -- so many states have struggled to get these testing recommendations implemented because of laws, and because of lack of ability ...
It is a challenge, it is a challenge. Part of it is changing the, sort of, mindset of HIV still being an issue. This is 30 years into the epidemic, there's burnout; it's not on the front page every day. So there's been some sort of complacency, if you will. But we all recognize that this is a really serious problem.
If you still have 21% undiagnosed, the prevalence keeps increasing, you have groups that are differentially impacted, particularly African Americans and men who have sex with men [MSM]. These are areas that need focus for prevention and treatment.
I'm looking at your abstract. It says the majority of undiagnosed cases are among men who have sex with men. This is really surprising, because you would think -- given this history, you would think this community would be the most likely to get tested and aware of HIV risks.
Early in the epidemic, there was a lot of very, very forceful advocacy for change in behaviors in this group. But over time, some of that has decreased, not only among MSM, but among other risk categories as well. It is a group that the CDC is focused to provide testing campaigns, initiatives, that sort of thing.
Which I think, over the last eight years, might have not happened particularly for MSM, so things might change now, I guess.
The new administration has expressed that they want to have the science drive the policy. So if the data show that certain groups need focus, the logical follow up would be that those groups would be addressed.
But it's true that that wasn't addressed for eight years, was it?
I would have to say that those decisions were made at a higher level. We acknowledge at CDC that the data show that there are differences based on race, based on risk, based on age, and those need to be properly addressed.
I don't know if you broke this down even further, but do you have data for the percentage of MSM who are African American?
Those data are not in this poster; the poster has a limited display. We are working on a manuscript that goes into more depth, has more breakdowns of race-by-risk and so those would be addressed in that follow on.
Based on my understanding of the issue, I would think that the greatest percentage of MSM who do not get tested are African-American MSM.
We'd have to look at the data to see that. There is a large percentage of white MSM as well.
There are a lot of issues for all MSM, but particularly for racial/ethnic minorities. Stigma, lack of access --
Homophobia, stigma, lack of access to resources for testing, for treatment, for basic health care services, that sort of thing.
So, yes: When the data come out, when the analysis comes out, we'll see what that is. But I'll have to have that come out, I can't tell you right off the top of my head which group would be differentially impacted more than others.
It's also striking, just the sentence here in the abstract that whites had the lowest percentage undiagnosed, 18.8% compared to Hispanics, African Americans, etc. Can you talk to that at all?
Again, stigma, lack of access to care and treatment services, lack of, sort of, acknowledgement of risk. There are a number of factors that are impacting that.
You have Asian and Pacific Islanders as actually the highest group of people.
Could you talk about that a little bit?
We don't get into the reasons why people don't get tested. This is off of surveillance data, which does not really interact with the client, ask them questions about why you did or didn't do anything. The racial/ethnic minority groups, again, often have difficulty with access. For Asian/Pacific Island groups, you may have cultural and linguistic isolation. It may be that they're just not acknowledging risk, or they don't have access to care and treatment.
For that group, if you have one percent of a million, that's quite large. But if you take a look at the rates, their rates of undiagnosed infections are actually lower than whites'. Their percentage is higher, but the population differences -- again, that's a smaller population group, so their rates are actually the lowest by race/ethnicity.
Are there going to be policy decisions based on these new overall data?
I'm not a policy person, so again, we would hope that the data would drive the response. As we've seen already with the heightened national response for African Americans, that's a group that is severely impacted and so they are getting more focus.
This is a snapshot. CDC is committed to analyzing these data going forward, not sure if it's going to be every year or every other year. You'd really need to have time pass so you can see changes that are meaningful, that we can make some sense out of.
Did you also look at where in the United States there are more people undiagnosed?
These data are not broken down by region. That's a little problematic. As of April 2008, all areas are reporting, but the data have not been matured. We have not been able to do some statistical adjustments that, again, account for that big influx of cases -- prevalent cases -- once a state goes to name-based reporting. It's going to be awhile before we can do any kind of regional analyses that include any kind of meaningful HIV data.
Is there anything that was really surprising about this, compared to years before?
I think this reinforces things that we've seen for awhile, in that MSM and racial/ethnic minorities have a differential impact compared to other population groups, and that's why CDC has been committed to focusing on that, through the heightened national response and some of the testing initiatives.
So how did women do, in this?
The percent undiagnosed is very close to the male percentage. If you start breaking that down by race by sex, we see that racial/ethnic minority women have greater disparities compared to white women. Again, that's something that's reinforcing, we've seen that in our other data.
What's the next step?
Again, this reinforces things that have come out before. It's another piece of the puzzle to say here are groups that need focus, get the initiatives, the testing in there. People who are then diagnosed get linked to care so that we can, over time, have some affect on these data and get the numbers for all groups going in a better direction, than increasing.
So the people at the CDC who would initiate particularly interventions and expand and fund HIV testing, that's a totally different department?
Correct, correct. What we know is that if people are diagnosed, they get into care and prevention services -- most of them do reduce risk behaviors. So it follows on that, as more people become diagnosed and aware of their infection, they then adapt protective behaviors and over time, the number of new cases can be impacted. Then, over more time, the number of prevalent cases can be impacted.
Do you estimate the undiagnosed HIV prevalence every couple of years? I haven't really seen this for a long time.
The last estimate was data from 2003. These are data from 2006. So, again, CDC is committed to doing this going forward; I can't say if it'll be every year, every two years, every three years, but we will be doing this as sort of updates, as time goes on.
Great, thank you very much.
Thank you, Bonnie.