Treatment switches on basis of CD4 declines often unnecessary, Uganda research shows

This article is more than 14 years old.

Switching people to second-line antiretroviral treatment on the basis of CD4 declines, without information from viral load tests, could result in a large numbers of unnecessary switches to more expensive second-line regimens in resource-limited settings, a study in Uganda has found.

Researchers from Johns Hopkins University Medical School and the Rakai Health Sciences Program reported their findings last week at the Sixteenth Conference on Retroviruses and Opportunistic Infections in Montreal.

Their study examined the utility of immunologic changes – identified by the World Health Organization (WHO) as predictive of antiretroviral treatment failure – to correctly predict virological failure. Changes in CD4 count are used in many resource-limited settings to detect treatment failure because it is not possible to monitor viral load, due to the need for well-equipped laboratories capable of conducting polymerase chain reaction testing (PCR), using kits provided by manufacturers such as Roche and Abbott, at an average cost of $40 to $50 a test.



A test used to measure something.


When using a diagnostic test, the probability that a person who does have a medical condition will receive the correct test result (i.e. positive). 


The fluid portion of the blood.

negative predictive value

When using a diagnostic test, the percentage of those testing negative who really don’t have the medical condition. This will vary according to the prevalence in the local population.


Short for logarithm, a scale of measurement often used when describing viral load. A one log change is a ten-fold change, such as from 100 to 10. A two-log change is a one hundred-fold change, such as from 1,000 to 10.

WHO recommends a switch in treatment if the CD4 count falls by more than 50% from its previous peak level, or if the CD4 count falls to its pre-therapy baseline (or below); or if it persistently remains below 100 cells/mm3.

It’s unclear how accurately CD4 count changes predict treatment failure, and there has been concern that, by the time the CD4 count begins to decline, viral load may have been rising for many months, leading to drug resistance. (The research on this subject up to mid-2007 was reviewed in HIV & AIDS Treatment in Practice in September 2007.)

The Rakai study was designed to investigate how accurately the WHO immunologic criteria predicted virologic failure in an African population receiving antiretroviral therapy.

The study enrolled 1133 people who were receiving antiretroviral therapy through the Rakai Health Sciences Program. During a median follow-up time of 20.2 months (interquartile range, 12.4 to 29.5 months), study participants were assessed in terms of both WHO immunologic failure criteria and a virologic failure endpoint that researchers defined as a viral load greater than 10,000 copies/ml on a single follow-up test. (It’s worth noting that the switch threshold of 10,000 copies/ml has not been validated in a clinical trial; it’s a number derived from observational analysis of resistance levels in cohort studies.)

CD4 cell counts were checked every six months for the first year of study participation, then every six months thereafter. Viral load testing was performed every six months.

One hundred and twenty-five study participants (11.0%) met WHO immunologic failure criteria and 80 (7.1%) experienced virologic failure. Only 18 people (1.6%) were identified by both monitoring methods as antiretroviral therapy non-responders.

Using virologic monitoring as their gold standard, the researchers determined the sensitivity, specificity and positive and negative predictive value of the WHO criteria for people at two different viral load levels.

Performance of immunologic criteria


Viral load level



Positive predictive value


Negative predictive value (NPV)

>10,000 copies/ml





>400 copies/ml





Presenting the results, Dr Stephen Reynolds of Johns Hopkins University said that the immunologic criteria had a very low sensitivity to true virologic failure, and that as well as switching patients unnecessarily, the use of the criteria would have missed a substantial number of virologic failures.

He asked: “How much money are we putting into ongoing CD4 monitoring? Should we be [concentrating on] monitoring viral load once individuals have shown a CD4 reponse on therapy?”

He noted that 107 patients would have switched treatment unnecessarily, at an extra cost of $75,000 a year for drugs alone.

He said that the high rate of immunologic failure without virologic failure was largely driven by 50% declines in CD4 count from peak levels, possibly a consequence of intercurrent infections – such as malaria – that were not being properly recorded.

The findings give added impetus to the call for the development of low-cost, point-of-care viral load tests that can be used outside sophisticated laboratory settings.

“All of us who work in clinics are very eager for a viral load assay, a qualitative assay, that can tell us if there is virus there or not,” he said.

A number of poster presentations at the conference provided encouraging information about validation of testing methods that might provide interim solutions for expanding use of viral load testing until true point-of-care tests have been developed.

Cavidi ExaVir

Cavidi’s ExaVir Load assay is designed for use in resource-limited settings and is based on the same ELISA technology used to carry out confirmatory antibody tests.

A comparison of viral load levels in 119 samples (86 non-B subtypes), using the Cavidi ExaVir Load version 3, Abbott RealTime and Roche Ampliprep/COBAS TaqMan assays found good agreement between the three assays, although the Cavidi assay tended to underestimate viral load by approximately half a log. In 16 cases the ExaVir assay failed to detect viral load when the Abbott and Roche assays found viral loads up to 2,000 copies/ml (range 41 to 2070 copies/ml by Abbott, 83 to 1523 copies/ml by Roche) (Garcia-Diaz).

A similar analysis carried out by the AIDS Clinical Trials Group, using 342 plasma samples from 108 patients (HIV subtype unspecified), showed a similarly high level of agreement between the ExaVir assay and the Roche Amplicor Monitor assay, with an average underestimate of only 0.15 log for the ExaVir assay. Less good agreement was shown by two versions of an ultrasensitive p24 antigen assay (Stewart). The ACTG study estimated that the viral load measurements using the ExaVir assay would cost around $15 to $20 each, and would take around three hours to complete.

Dried blood spots

Another approach to implementing viral load testing in the field is the use of dried blood spots to collect blood samples, which can then be sent to a laboratory equipped to carry out viral load testing. This method is already being used successfully in over 30 countries that are scaling up access to infant HIV diagnosis, using HIV DNA testing. However, until now the use of dried blood spots for capturing samples that yield measurable HIV RNA – the viral material measured by viral load assays – has not been validated in field conditions.

Several studies showed that dried blood spots are suitable for detecting HIV RNA at levels above 1000 – 3000 copies using the PCR-based viral load assays manufactured by Abbott and Roche.

A collaboration by Tanzanian and Norwegian researchers, evaluating performance of dried blood spots and normal sample-handling methods, using the Cobas Amplicor HIV-1 Monitor Test version 1.5 to measure HIV in plasma and the NucliSens EasyQ HIV-1 v1.1 assay to measure HIV in dried blood spots, found a mean difference of 0.03 log in measurement of viral load in 98 samples. Dried blood spots had a sensitivity of 82.1%, 88.9%, and 100% to detect viral loads greater than 400, 1000, and 3000 copies/ml, respectively (Johanssen).

A similar survey, evaluating the agreement between plasma and dried blood spot or dried plasma spot testing in 205 samples from 41 patients using the Abbott Real-Time assay, found poor detectability for spot methods below 6000 copies/ml, and only modest sensitivity under 60,000 copies/ml (Mbida).


Reynolds S et al. Evaluation of the WHO immunologic criteria for ART failure among adults in Rakai, Uganda. Sixteenth Conference on Retroviruses and Opportunistic Infections, Montreal, abstract 144, 2009.

Garcia-Diaz A et al. Comparative evaluation of the performance of the Cavidi ExaVir Load assay for the measurement of B and non-B subtype HIV-1 RNA load in plasma. Sixteenth Conference on Retroviruses and Opportunistic Infections, Montreal, abstract 1002, 2009.

Stewart P et al. Performance of the Cavidi ExaVir Load assay and the ultra-sensitive p24 antigen assays relative to the Roche Monitor HIV-1 RNA assay: ACTG New Works Concept Sheet 227. Sixteenth Conference on Retroviruses and Opportunistic Infections, Montreal, abstract 1004, 2009.

Johannessen A et al. Dried blood spots perform well in viral load monitoring of patients on ART in rural Africa. Sixteenth Conference on Retroviruses and Opportunistic Infections, Montreal, abstract 1007, 2009.

Mbida A et al. Measure of viral load by using the Abbott Real-Time HIV-1 assay on dried blood or plasma spot specimens collected in 2 rural dispensaries in Cameroon Sixteenth Conference on Retroviruses and Opportunistic Infections, Montreal, abstract 1007, 2009.