Viral Load: New Proof of Importance for Disease Staging and Prognosis

Research findings published May 24(1,2) have extended earlier work showing the importance of viral load in predicting HIV disease outcome. Measurement of plasma HIV RNA (indicating the concentration of virus in the blood) was found to be a much more accurate predictor of later illness than CD4 (T-helper) count, strongly suggesting that patients and physicians should consider viral load when making HIV treatment decisions. This data provides longer followup than for any previous study of viral load.

The research team, mostly from the University of Pittsburgh but including two members from Chiron Corporation, looked at disease outcomes in a cohort of 180 HIV-positive gay or bisexual men who were enrolled between April 1984 and March 1985 in the Pittsburgh site of the MACS study (Multicenter AIDS Cohort Study). The 180 include all who were HIV-positive when they enrolled at the Pittsburgh site, and for whom frozen plasma samples were available for testing. It is not known how long the men were infected before they enrolled.

Only 41% of the 180 volunteers received antiretroviral treatment at any time during the study. [It is important to note that the average survival times today will almost certainly be longer than those reported below, which indicate survival starting ten years ago when much less treatment was available.] [Note: Brackets are used to indicate our comments about the article -- JSJ.]

The data were looked at in several different ways:

* First, the researchers used viral load at entry into the study to divide the 180 volunteers into quartiles (four equal groups, of 45 each, with the highest, second-highest, second-lowest, and lowest viral load). They measured the viral load with a "second generation" bDNA test from Chiron Corporation; this test, not yet available commercially but only for research, can measure down to a cutoff of about 500 copies of HIV RNA per milliliter of plasma. (The bDNA test which has been available to practicing physicians until now has a cutoff of 10,000 copies, 20 times higher than the second-generation test.)

Of those in the lowest viral quartile, which was 4530 copies per ml. or less, only 8% progressed to AIDS within five years. In the next quartile, 4531-13,020 copies, 26% progressed. In the second-highest quartile, 13,021 - 36,270 copies, the proportion was 49%; and in the highest quartile, above 36,270, 62% developed AIDS in five years -- almost eight times the progression rate of the lowest quartile. [These particular viral-load levels might be lower than those measured today, since these samples were processed 10 years ago with procedures not optimized to preserve them for viral load testing; it is likely that damage to the RNA led to lower counts being recorded, meaning that the numbers above are too pessimistic. One of the researchers told us that the samples in this study are likely to have suffered a 1.6-fold to 2-fold decrease in viral load counts due to how they were preserved -- meaning that the counts above may actually have been 1.6 to 2 times higher than the numbers given above.]

Looking at survival, in the lowest quartile of viral load, 5% of the volunteers died within 5 years. For the next quartile it was 10%, the second-highest quartile 25%, and in the highest quartile, 49% died within 5 years.

The increased risk of progression to AIDS, and risk of death, was still evident even ten years after this single viral load measurement.

* Viral load was even more predictive when the average of the first two measurements was used, instead of a single viral load test (The MACS study drew blood samples every six months). And the predictive ability was further improved by eliminating the 29 patients (out of 172 for whom two tests were available) who had an 80% or more drop in number of copies between their first and second test. (This drop probably meant that those people were still recovering from their primary HIV infection, meaning that their viral-load tests may not have properly measured the "set point," or long-lasting, relatively stable viral level which remains after primary infection and seems to control the speed of disease progression.) Once these adjustments were made, those in the highest quartile (average viral load greater than 28,720, in this group) had a median survival time of only 2.5 years, with only about 5% surviving for ten years. The authors interpret this as "evidence that a persistently high viral load is almost always associated with more rapid disease progression."

On the average, after statistical adjustments including both HIV RNA (viral load) and CD4 counts, the relative risk of death was 1.57 (a 57% increase) for each threefold increase in viral load.

* CD4 counts were much less predictive of AIDS progression and survival. When the same volunteers were divided into quartiles by CD4 count (less than 322, 322-527, 528-787, and greater than 787) the three higher quartiles showed no difference in survival, nor in progression to AIDS. Only the lowest quartile showed a worse outcome. When the 29 volunteers with an 80% or more drop in viral load were excluded, the predictive ability of the CD4 test did not improve.

In another look at CD4 count, the researchers found that 50% of those with CD4 over 500 [ten years ago], if they also had a viral load greater than 10,900, died within six years -- compared with only 5% of those with the same CD4 counts but a viral load of less than 10,900. They suggested re-evaluating the practice of using CD4 level as a trigger to begin antiretroviral therapy. [This does not affect the use of CD4 counts for starting prophylaxis for pneumocystis or other opportunistic conditions; in this situation, the CD4 level does show who is at risk.]

Comment

This long-term followup study shows the predictive value of modern viral load testing. But in order to have the long-term progression and survival information available, the researchers necessarily had to use frozen samples collected years ago. Therefore, the particular numbers reported above should not be blindly applied today. But the conclusion that viral load testing is important for patient management -- probably considerably more important than CD4 testing for guiding antiviral therapy -- seems hard to dispute.

Even before these results were published, there was clear consensus among experts that viral load testing is valid for prognosis (predicting whether an individual is likely to do well or poorly in the future). But there has been controversy about its use for patient management, with some saying that the data to support such use is not yet available. The problem is that to get conclusive proof of the value of viral load testing for management requires randomly assigning some patients to have their physicians use viral load testing as part of their care, and randomly assigning others to not use viral load, and then looking for a statistically significant difference in the number of deaths or AIDS complications in the two groups. Such trials are now being done, but they will take time, since existing frozen samples cannot be used (as there is no way to know what a physician in the past would have done with the additional information, or how a patient would have responded to the resulting change in treatment).

We have argued elsewhere that there are fundamental problems with this kind of trial to prove the value of viral load testing. Aside from the ethical concerns, it will be hard to reach statistical proof, since antiviral therapies are still limited, and many patients in the trials will not have good access even to what does exist (as these trials do not provide or pay for treatment); if physicians cannot make effective use of the viral load information, how much difference could it make? A negative result would not show that viral load is not useful, but only that it was not useful within the limitations of the particular (and largely undefined) guidelines and therapeutic options existing at the time of the trial. How helpful is this information, when the therapies and guidelines surely will have changed considerably by the time the trial results are available? And we do not see how these trials will have the statistical power to add much to our knowledge about HOW to use viral load, when they are designed primarily to answer the go/no-go decision of whether viral load should be used in individual patient care.

There is already so much information consistently supporting the importance of viral load, that it is inconceivable that this test could be found to have no use in clinical care. Viral load testing has already become a de facto standard for those who can afford it. The problem today is to get better care out to everyone else. This is especially important now that many more treatment choices are available. Viral load will be essential for making these choices intelligently -- for example, to find out when an antiviral regimen is failing to work for a particular patient, so that different treatments can be tried instead.

References

1. Mellors JW, Rinaldo CR, Gupta P, White RM, Todd JA, and Kingsley LA. Prognosis in HIV-1 infection predicted by the quantity of virus in plasma. SCIENCE May 24, 1996; volume 272, pages 1167-1190.

2. Ho, DD. Viral counts count in HIV infection. SCIENCE May 24, 1996; volume 272, pages 1124-1125.