At the International AIDS Conference in Bangkok, a generous handful of posters and presentations found promise in some of these approaches and put several nails in the coffin for total lymphocyte count as a CD4 surrogate.
A group of researchers loosely organized as the Afford CD4 Group, led by Professor George Jannosy of London, have sought to simplify cell sorting protocols and reduce the price per test by using generic, "home-brew" monoclonal antibodies instead of commercial reagents.
Researchers in Thailand performed a quality comparison of state-of-the-art two- and three-color cell sorting (using multiple monoclonal antibodies) with a simplified protocol called panleucogating that can use just one monoclonal antibody. However, this method only produces the relative percentage of CD4 T-cells in the blood; determining the familiar absolute CD4 count requires additional steps. They also compared panleucogating using generic reagents with the same technique using commercial reagents. Results from each of the methods correlated well with the others for determining CD4 percentage in 142 HIV-positive samples and 26 HIV-negative samples. The authors estimate that panleucogating with generic reagents could reduce the cost per test from $11.50 to $2.30 each. While these saving are significant, this method still relies on an initial investment in equipment that can run from $20,000 and up with expensive yearly maintenance contracts. (Pattanapanyasat, B3087)
Researchers in Barbados compared a panleucogating protocol to a sophisticated four-color method of cell counting using commercial monoclonal reagents. The coefficient of correlation was a respectable 0.97 over the entire range of counts (25 to 989 cells/mm3). The authors conclude that panleucogating is "an accurate method for enumerating CD4 T-cells and has major cost implications for the sustainability of the National HIV containment program in Barbados." (Sippy, B3108)
Researchers in Cambodia evaluated the precision of the Cyflow system by performing multiple repeated counts of the same samples. They also performed comparisons of results produced by Cyflow with those produce by a lab-based FACS. The coefficient of variance for the precision of the machine varied between 2.8% and 4.9%. The correlation with FACS was very high, with the Cyflow tending to undercount by about 20 cells on a mean CD4 count of 289. The system detected CD4 counts below 200 cells/mm3 with sensitivity of 100% and a positive predictive value of 97%. (Teav, B3089)
Researchers in Thailand compared Cyflow with two different FACS systems (3-color FACScan and 2-color FACScount) and reported high correlation between the results, with Cyflow producing mean counts 41.5 and 18.0 cells/mm3 lower than the two FACS systems in the range of 100-300 CD4 cells/mm3. The authors note that Cyflow offers advantages in cost and sample preparation, "but it requires technical expertise." (Pattanpanyasat, B3176)
In Malawi, investigators compared Cyflow to FACS on 311 blood samples. The mean difference in CD4 counts by Cyflow was -8.68 cells compared to FACS, with a correlation coefficient of 0.92. The authors note that "local district laboratory staff found the Partec machine easy to manipulate and robust under routine field conditions." (Fryland, B1149)
Researchers in Rwanda compared results from Cyflow (4-parameter, direct volumetric counting) with FACS (bead-based, FACScount) on samples from 73 HIV-positive pregnant women. Mean CD4 counts were 346 cells/mm3 with FACS and 377 cells/mm3 with Cyflow. (Jervais, B3172)
One limitation of performing CD4 counts in resource-limited settings is that there may be a significant time lag between when samples are collected and when they finally reach a testing facility. Researchers in Cambodia compared CD4 counts obtained by Cyflow on samples collected the same day and on EDTA-preserved samples kept at room temperature (<30C) for 4 days. In 27 samples tested, the correlation between fresh blood and aged blood was high, with aged blood samples tending to come in about 5 cells lower on a mean CD4 count of 241. (Teav, B3110)
CD4 counts were correlated with TLC for 747 participants in HIVNAT clinical trials in Thailand to determine the value of TLC for monitoring response to ART. Samples were collected at baseline and at weeks 12, 24 and 48 after beginning ART. Of 3578 paired samples, 29% had CD4 <200 cells/mm3. At baseline, TLC <1200 cells/mm3 had a sensitivity of 40% for predicting CD4 <200 cells/mm3 (specificity = 94%). Sensitivity was reduced to only 20% at week 48. "Thus, TLC is clearly not a good surrogate marker for monitoring HAART." (Ruxrungtham, B3154)
Researchers in Bahia, Brazil evaluated paired CD4 and TLC for 498 patients during May to December 2003. A TLC cutoff of 1000 cells/mm3 predicted CD4 count <200 cells/mm3 with sensitivity of 44% (specificity = 98.5%) and had a positive predictive value of 70.2%. Raising the cutoff to TLC <1500 cells/mm3 improved sensitivity to 76.9%, although the positive predictive value at this level was only 38.7%. The authors conclude that TLC estimates of CD4 counts for monitoring HAART are inaccurate. (Angelo, B3164)
Researchers in Jakarta, Indonesia evaluated 1062 paired CD4 and TLC results obtained between January 2002 and September 2003. Of 355 samples with TLC <1200 cells/mm3, 81% had CD4 counts <200 cells/mm3, while 20% of samples with TLC >1200 cells/mm3 had CD4 counts <200 cells/mm3. The authors conclude that if TLC alone were used to determine when to start ART, "then 39% of HIV infected Indonesians would be misclassified." (Donegan, B3177)
Researchers in Malvinas Argentinas, Argentina evaluated paired TLC and CD4 counts from 66 patients. While TLC <1500 cells/mm3 significantly predicted CD4 <200 cells/mm3 (p=0.01), sensitivity was 65% and specificity was 69%. With a cutoff of TLC <1200 cells/mm3, sensitivity was 50% (specificity = 89%) and with a cutoff of <1000 cells/mm3, sensitivity was 45% (specificity = 93.4%). The authors find that the sensitivity and specificity of TLC to predict CD4 cell counts <200 cells/mm3 is low, although lower cutoffs improve specificity. (Hojman, B7219)
Based on paired FACS and TLC data from 2419 patients, researchers in Pune, India derived an equation (CD4 = 0.24*TLC-5.97) to calculate CD4 counts using TLC values. They found that TLC was significantly correlated with CD4 (r=0.43; p<0.001). Using a TLC cutoff of <1500 cells/mm3 predicted CD4 counts <350 cells/mm3 with sensitivity of 72% and specificity of 78%. The positive predictive value was 79% and negative predictive value was 91%. However, for CD4 counts <200 cells/mm3, the equation was only 49% sensitive. (Thakar, B3105)
In Sagamu, Nigeria, investigators collected blood samples from 64 patients during a one year period ending in September 2003. CD4 counts were evaluated by FACS and by Dynakit, and TLC was calculated. While FACS and Dynakit were significantly correlated (r=0.831; p=0.001), TLC did not correlate with either method (FACS: r=0.061; p=.573). The authors conclude that TLC "is not a reliable substitute for CD4 countÉin this resource-limited setting." (Osho, B3096)
The one report to speak favorably of TLC also found fault with the CD4 count when it came to spotting opportunistic infections (OI).
Researchers at a major London, England hospital retroactively identified all patients with an AIDS-defining opportunistic infection (n=1097) to see if CD4 counts or TLC within the three months prior to the illness was more predictive of having an OI when compared to patients who did not develop OIs. TLC was significantly correlated with CD4 count (r=0.70; p<0.001) and the optimal cut-off for TLC was 1500 cells/mm3. While patients with TLC between 1000 and 1500 were 40% more likely to have an OI than those above 1500 (sensitivity = 68.6%; specificity = 75.6%), when using the CD4 count cutoff of 200 cells/mm3, those with CD4 between 150 and 200 were only 34% more likely to have an OI (sensitivity = 73.8%; specificity = 75.6%). The authors conclude that while CD4 <200 cells/mm3 is taken as the "gold standard for therapeutic intervention, this has relatively low specificity/sensitivity and results suggest that TLC is only moderately less reliable." (Jones, B3120)
There has been some interest in quantifying the success of making treatment decisions based on symptoms and clinical features, such as weight gain after starting ART. Yet the limits to this are apparent.
Researchers from the Centers for Disease Control (CDC) in Atlanta, Georgia investigated if weight gain could serve as a surrogate marker for response to ART by reviewing medical records from 709 patients with weight measurements at the time of and subsequent to beginning ART. Overall, the cumulative probability of having at least a 10% weight gain at 12 months was 0.15. The probability was increased in those starting ART with <200 CD4 cells/mm3 (0.31) and for those starting ART with BMI <20 (0.33). Although weight gain of 10% or more occurred in about one third of patients with low CD4 counts or low BMI, weight gain was not correlated with viral load reduction. The authors conclude that "weight gain following HAART initiation does not necessarily mean that there is virologic improvement." (Teshale, B3101)
Researchers in Uganda operating an ART program evaluated methods of screening for eligibility for therapy in 907 patients. Patient history, review of records, and physical exam to determine CDC Class C and B symptoms as well as other signs of HIV disease were performed. Following screening, CD4 counts were obtained and compared to clinical findings. Of 376 patients with CD4 counts 3, 39% met the clinical criteria for starting ART and 91% of those with CD4 counts >200 did not meet the criteria. The sensitivity of the clinical criteria was 39%, specificity was 91% and the positive and negative predictive values were 76% and 68%, respectively. The authors conclude that CD4 testing would be important to detect the two thirds of patients who qualify for starting ART. (Solberg, B2035) There seems to be no good way around having an absolute CD4 count in hand to make decisions about when to start therapy and to monitor response to therapy once it has begun. Some recent innovations seem to provide acceptable results with significant cost savings. Yet they fall short of an ideal solution to the need for low-cost, point-of-care monitoring. To get the most out of the expected scale up of treatment in the coming years, a breakthrough in diagnostic technology must be made a priority.
Back to the GMHC Treatment Issues July/August 2004 contents page.