Abstract

Introduction. The CD4 count and CD4 percentage (CD4%) are both strong predictors of clinical disease progression in human immunodeficiency virus (HIV). Although individuals may show discordancy between their CD4 count and CD4%, the clinical relevance of this is unclear.

Methods. Discordancy was defined where the CD4% was ≤10th percentile for a selected CD4 count range (referred to as low discordancy), within the central 80% range (concordant), or ≥90th percentile (high discordancy). Regression methods identified factors associated with low and high discordancy in untreated individuals and assessed the impact of discordancy on treatment responses to highly active antiretroviral therapy (HAART).

Results. High discordancy was associated with female sex, low viral load, and white ethnicity; low discordancy was associated with black or nonwhite ethnicity, older age, and injection drug use. Clinical event rates were higher in individuals with high discordancy starting HAART, but there was no association with subsequent HIV progression by 6 months after starting HAART. CD4 count increases remained lower, by 20 cells/mm3, in individuals with low discordancy, and higher, by 27 cells/mm3, in those with high discordancy.

Conclusions. Overall discrepancies between the CD4/CD4% are small, confirming the use of absolute CD4 counts as a monitoring tool.

Individually, the CD4 count and CD4 percentage (CD4%) are both strong predictors of clinical disease progression in untreated and treated human immunodeficiency virus (HIV)–positive individuals [1, 2]. Although the relative contribution of each to the determination of prognosis is still a matter for debate [1–7], the CD4 count remains the most frequently used of the 2 measurements [1, 8]. Although there is generally a strong positive correlation between the 2 markers, this correlation is not complete and some individuals have been reported to exhibit discordancy between their CD4 count and CD4% [9]. However, the clinical relevance of this discordancy is unclear, particularly in terms of clinical outcomes or immunological or virological responses to highly active antiretroviral therapy (HAART). Furthermore, much of the existing data on which this evidence is based is anecdotal and there has been no formal definition of what constitutes a discordant CD4/CD4% relationship in individuals infected with HIV.

Using information on the distribution of CD4% values in a large cohort of untreated HIV-positive individuals, we identified a group of individuals whose CD4/CD4% values may be viewed as discordant. Using this information, we then identified risk factors for a discordant CD4/CD4% value, assessed whether discordancy was associated with differential virological and immunological responses to HAART, and described the clinical implications of a discordant CD4/CD4% after 6 months of HAART.

METHODS

Patients and Methods

The data used in our analyses derive from the UK Collaborative HIV Cohort (UK CHIC) Study, a collaboration of some of the largest HIV treatment centers in the United Kingdom (UK). The current dataset utilizes information from 35 377 individuals seen at 13 clinical centers (see Appendix) since 1 January 1996 [10]. Participating centers provide information on demographics, AIDS diagnoses and deaths, antiretroviral treatment, and laboratory tests (including historic data where available). The UK CHIC Study was approved by a multicenter research ethics committee and by local ethics committees.

Statistical Methods

Assessment of Discordancy in Untreated HIV-Positive Individuals

As there is no standard definition for a discordant CD4/CD4% value, we used data from untreated participants to describe the distribution of CD4% values among individuals with absolute CD4 counts in various ranges (<50, 50–99, 100–149, 150–199, …, 700–749, and ≥750 cells/mm3). The unit for these analyses was the CD4/CD4% “pair”—an absolute CD4 count and CD4% measured on the same day in the same individual. Each patient could contribute multiple CD4/CD4% pairs to the analysis as long as the patient remained untreated at the time. Each CD4/CD4% pair was classified into 1 of 3 categories depending on whether the CD4% was ≤10th percentile for that CD4 range (referred to as low discordancy), within the central 80% range (concordant), or ≥90th percentile (high discordancy). Thus, approximately 10% of CD4/CD4% pairs would be expected to demonstrate low discordancy, 10% high discordancy, and the remaining 80% concordancy.

Multiple logistic regression methods, using generalized estimating equations to take account of the multiple measurements contributed by participants, were used to identify factors associated with low and high discordancy. The factors considered in these analyses were age, sex, HIV transmission risk group (men who have sex with men [MSM], heterosexual men and women, injection drug users [IDUs], and other/unknown), ethnicity (white, black African, other, unknown), calendar year, and viral load. Because many of the CD4/CD4% pairs were collected in the pre-HAART era, at a time when HIV viral loads were not monitored, these regression analyses were based on the subset of CD4/CD4% pairs where a viral load was also measured on the same day.

Relationship Between CD4/CD4% Discordancy Pre-HAART and Response to HAART

Included patients were antiretroviral-naive at the time of starting HAART (any regimen including at least 1 protease inhibitor [PI], nonnucleoside reverse transcriptase inhibitor [NNRTI], or abacavir, regardless of the number of drugs in the regimen), had started HAART from 1998 onward, had a CD4/CD4% pair measured in the 6 months prior to starting HAART, and had at least 1 CD4 count and viral load measured after starting HAART. Discordancy was defined using the cutoff values determined from the first part of the analysis.

Using Cox proportional hazard regression models, we assessed whether a discordant CD4/CD4% relationship prior to starting HAART was associated with the following outcomes on HAART: initial virological response (the date of the first viral load <50 copies/mL); viral rebound (a confirmed viral load >500 copies/mL in those with an initial response); and clinical response (the development of a new AIDS event or death). Patient follow-up started on the date of initiation of HAART. For analyses of viral load outcomes, patient follow-up was right-censored on the date of the last viral load measurement, death, or 1 January 2009, whichever occurred first; for analyses of clinical outcomes, patient follow-up was right-censored 3 months after the patient’s last clinic attendance, death, or 1 January 2009, whichever occurred first. We took an “intent-to-continue-treatment” approach and thus ignored any treatment changes in these analyses. All regression models also controlled for the pre-HAART CD4 count (as a continuous covariate) and the following potential confounders: age at start of HAART, sex, transmission risk group, ethnicity, type of HAART regimen (PI, NNRTI, or other), and calendar year.

In order to consider the impact of a discordant CD4/CD4% on the immunological response, we considered the change in CD4 count at 6 months; the CD4 count used for these analyses was the nearest measurement to the midpoint of the window from 20 to 32 weeks after starting HAART (this window was chosen as it reflects a 3-month time interval during which most patients would be expected to have at least 1 CD4 measurement). Multivariable linear regression models, adjusting for the same covariates as mentioned above, were used to identify the factors associated with the change in CD4 count over this interval.

Although our a priori hypothesis was that the effect of low or high discordancy on outcomes would be independent of the pre-HAART CD4 count, we investigated whether this was the case by formally testing for interactions between the pre-HAART CD4 count and low and high discordancy.

Finally, we assessed whether a discordant CD4/CD4% pair at 6 months after initiation of HAART was associated with a poorer subsequent clinical outcome. This timepoint was chosen as previous studies [11] have shown that the CD4 count at 6 months after initiation of HAART is a strong predictor of clinical outcome. For these analyses, we considered the time to a new AIDS event or death with patient follow-up beginning 6 months after initiation of HAART. Multivariable Cox regression models, adjusting for the factors listed above as well as the CD4 count attained at this time, were used to identify whether a discordant CD4/CD4% response was associated with a higher risk of clinical progression.

To ensure that our results remain valid in the modern HAART era, analyses were repeated on the subset of patients starting HAART from 2004 onward.

RESULTS

Assessment of Discordancy in Untreated HIV-Positive Individuals

A total of 141 121 CD4/CD4% pairs were available from 22 476 untreated individuals. The median (interquartile range [IQR]) CD4 count was 384 (263–539) cells/mm3 and the median (IQR) CD4% was 22 (15–28). As expected, the majority of CD4/CD4% pairs were contributed by white MSM aged 30–39 years (Table 1). Of the pairs, 74 546 were among individuals with a measured HIV viral load on the same day; as expected, given the requirement for these patients to be untreated, viral loads were high. As expected, there was a strong correlation between the 2 values (Figure), with CD4% values increasing as the absolute CD4 count increased. Overall, 12.3% of CD4/CD4% pairs met our definition of low discordancy and 11.4% met our definition of high discordancy (observed percentages were >10% due to the presence of a large number of tied CD4% values on the threshold). The proportions of patients exhibiting low and high discordancy in each subgroup are shown in Table 1.

Table 1.

Demographic Breakdown of Each CD4/CD4% Pair in Untreated Patients in the UK CHIC Study and Number (%) of Pairs Exhibiting Low and High Discordancy

Factor  No. (%) Low discordancy, No. (%) High discordancy, No. (%) 
Total number of CD4/CD4% pairs  141 121 (100.0) 17 395 (12.3) 16 026 (11.4) 
Sex Male 117 542 (83.3) 14 571 (12.4) 12 088 (10.3) 
 Female 23 577 (16.7) 2824 (12.0) 3938 (16.7) 
Age (years) <30 33 713 (23.9) 3243 (9.6) 4281 (12.7) 
 30–39 66 893 (47.4) 7921 (11.8) 7568 (11.3) 
 40–49 31 143 (22.1) 4504 (14.5) 3267 (10.5) 
 ≥50 9372 (6.6) 1727 (18.4) 910 (9.7) 
HIV risk group MSM 97 012 (68.7) 11 336 (11.7) 9663 (10.0) 
 IDU 6448 (4.6) 503 (7.8) 1512 (23.5) 
 Heterosexual 29 577 (21.0) 4636 (15.7) 3743 (12.7) 
 Other/not known 8084 (5.7) 920 (11.4) 1108 (13.7) 
Ethnicity White 96 381 (68.3) 11 071 (11.5) 11 575 (12.0) 
 Black 21 864 (15.5) 3566 (16.3) 2122 (9.7) 
 Other 15 544 (11.0) 1900 (12.2) 1576 (10.1) 
 Not known 7332 (5.2) 858 (11.7) 753 (10.3) 
Calendar year ≤1993 16 573 (11.7) 1551 (9.4) 1956 (11.8) 
 1994–1996 18 024 (12.8) 2467 (13.7) 1933 (10.7) 
 1997–1999 17 362 (12.3) 2294 (13.2) 1954 (11.3) 
 2000–2002 23 018 (16.3) 2956 (12.8) 2561 (11.1) 
 2003–2005 32 014 (22.7) 4258 (13.3) 3567 (11.1) 
 2006–2009 34 130 (24.2) 3869 (11.3) 4055 (11.9) 
Viral load <1000 9084 (12.2) 628 (6.9) 1841 (20.3) 
(copies/mL) 1000−9999 17 676 (23.7) 1355 (7.7) 2287 (12.9) 
(n = 74 546) 10 000−99 999 33 376 (44.8) 3788 (11.4) 3286 (9.9) 
 ≥100 000 14 410 (19.3) 2538 (17.6) 1140 (7.9) 
Factor  No. (%) Low discordancy, No. (%) High discordancy, No. (%) 
Total number of CD4/CD4% pairs  141 121 (100.0) 17 395 (12.3) 16 026 (11.4) 
Sex Male 117 542 (83.3) 14 571 (12.4) 12 088 (10.3) 
 Female 23 577 (16.7) 2824 (12.0) 3938 (16.7) 
Age (years) <30 33 713 (23.9) 3243 (9.6) 4281 (12.7) 
 30–39 66 893 (47.4) 7921 (11.8) 7568 (11.3) 
 40–49 31 143 (22.1) 4504 (14.5) 3267 (10.5) 
 ≥50 9372 (6.6) 1727 (18.4) 910 (9.7) 
HIV risk group MSM 97 012 (68.7) 11 336 (11.7) 9663 (10.0) 
 IDU 6448 (4.6) 503 (7.8) 1512 (23.5) 
 Heterosexual 29 577 (21.0) 4636 (15.7) 3743 (12.7) 
 Other/not known 8084 (5.7) 920 (11.4) 1108 (13.7) 
Ethnicity White 96 381 (68.3) 11 071 (11.5) 11 575 (12.0) 
 Black 21 864 (15.5) 3566 (16.3) 2122 (9.7) 
 Other 15 544 (11.0) 1900 (12.2) 1576 (10.1) 
 Not known 7332 (5.2) 858 (11.7) 753 (10.3) 
Calendar year ≤1993 16 573 (11.7) 1551 (9.4) 1956 (11.8) 
 1994–1996 18 024 (12.8) 2467 (13.7) 1933 (10.7) 
 1997–1999 17 362 (12.3) 2294 (13.2) 1954 (11.3) 
 2000–2002 23 018 (16.3) 2956 (12.8) 2561 (11.1) 
 2003–2005 32 014 (22.7) 4258 (13.3) 3567 (11.1) 
 2006–2009 34 130 (24.2) 3869 (11.3) 4055 (11.9) 
Viral load <1000 9084 (12.2) 628 (6.9) 1841 (20.3) 
(copies/mL) 1000−9999 17 676 (23.7) 1355 (7.7) 2287 (12.9) 
(n = 74 546) 10 000−99 999 33 376 (44.8) 3788 (11.4) 3286 (9.9) 
 ≥100 000 14 410 (19.3) 2538 (17.6) 1140 (7.9) 

Abbreviations: CD4, CD4 count; CD4%, CD4 percentage; HIV, human immunodeficiency virus; IDU, intravenous drug user; MSM, men who have sex with men; UK CHIC, UK Collaborative HIV Cohort.

Figure 1.

Distribution of CD4 percentage value in individuals with absolute CD4 counts in specified ranges; values shown are median, 10th, and 90th percentiles.

Figure 1.

Distribution of CD4 percentage value in individuals with absolute CD4 counts in specified ranges; values shown are median, 10th, and 90th percentiles.

Results from the multivariable logistic regression models (Table 2), based on 74 546 observations (8309 low discordancy, 8554 high discordancy) with viral load data, suggested that older age was associated with a higher risk of low discordancy (P = .0001) and a slightly lower risk of high discordancy (P = .03). Women were more likely to exhibit high discordancy and less likely to exhibit low discordancy (P = .0001 for each comparison) compared with men. Compared to MSM, heterosexuals and those with an unknown or other risk for HIV infection were more likely to exhibit high discordancy (global P = .0001) whereas IDUs were more likely to exhibit low discordancy (global P = .0001). Individuals of white ethnicity were more likely to exhibit high discordancy but less likely to exhibit low discordancy compared with those from other ethnic groups (global P = .0001). Finally, higher viral loads were associated with a reduced risk of high discordancy (P = .0001) but an increased risk of low discordancy (P = .0001). There were no significant trends with calendar year for either low or high discordancy.

Table 2.

Adjusted Odds Ratios (and 95% Confidence Intervals) From Multivariable Logistic Regression Models of the Factors Associated With Low and High Discordancy

  Low discordancy vs concordant
 
High discordancy vs concordant
 
  OR (95% CI) P value OR (95% CI) P value 
Age (per 5 years older) 1.10 (1.07–1.12) .0001 0.97 (.95–1.00) .03 
Sex Male .0001 .0001 
 Female 0.64 (.55–.75)  1.92 (1.61–2.28)  
HIV risk group MSM .0001 .0001 
 Heterosexual 0.91 (.66–1.25)  1.68 (1.27–2.21)  
 IDU 1.60 (1.38–1.86)  0.99 (.83–1.18)  
 Other/not known 1.06 (.90–1.27)  1.29 (1.12–1.50)  
Ethnicity White .0001 .0001 
 Black 1.70 (1.47–1.98)  0.48 (.41–.56)  
 Other 1.31 (1.13–1.51)  0.67 (.58–.78)  
 Not known 1.19 (.94–1.52)  0.75 (.61–.92)  
Year (per later year) 1.00 (.99–1.01) .73 1.00 (.99–1.01) .60 
Viral load (per log10 higher) 1.42 (1.35–1.49) .0001 0.76 (.73–.79) .0001 
  Low discordancy vs concordant
 
High discordancy vs concordant
 
  OR (95% CI) P value OR (95% CI) P value 
Age (per 5 years older) 1.10 (1.07–1.12) .0001 0.97 (.95–1.00) .03 
Sex Male .0001 .0001 
 Female 0.64 (.55–.75)  1.92 (1.61–2.28)  
HIV risk group MSM .0001 .0001 
 Heterosexual 0.91 (.66–1.25)  1.68 (1.27–2.21)  
 IDU 1.60 (1.38–1.86)  0.99 (.83–1.18)  
 Other/not known 1.06 (.90–1.27)  1.29 (1.12–1.50)  
Ethnicity White .0001 .0001 
 Black 1.70 (1.47–1.98)  0.48 (.41–.56)  
 Other 1.31 (1.13–1.51)  0.67 (.58–.78)  
 Not known 1.19 (.94–1.52)  0.75 (.61–.92)  
Year (per later year) 1.00 (.99–1.01) .73 1.00 (.99–1.01) .60 
Viral load (per log10 higher) 1.42 (1.35–1.49) .0001 0.76 (.73–.79) .0001 

Abbreviations: CI, confidence interval; HIV, human immunodeficiency virus; MSM, men who have sex with men; OR, odds ratio.

Outcomes on HAART

Of the patients starting HAART, 10 614 met the criteria for inclusion; 18.1% (n = 1924) and 10.4% (n = 1108) of these patients exhibited low and high discordancy, respectively, prior to starting HAART (Table 3).

Table 3.

Characteristics of 5829 Antiretroviral-Naive Patients Who Started HAART With Pre-HAART CD4/CD4% Measurements and at Least 1 CD4/CD4% Pair, With HIV RNA Measured After HAART

Characteristic  No. 
Number of patients  10 614 100.0 
Sex Male 7841 73.9 
 Female 2773 26.1 
HIV risk group MSM 5573 52.5 
 IDU 298 2.8 
 Heterosexual 3854 36.3 
 Other/not known 889 8.4 
Ethnicity White 5898 55.6 
 Black 3085 29.1 
 Other 1261 11.9 
 Not known 370 3.5 
Regimen PI-based 2958 27.9 
 NNRTI-based 6958 65.6 
 PI + NNRTI 200 1.9 
 NRTIs only 327 3.1 
 Other 171 1.6 
Year ≤2000 1882 17.7 
 2001–2002 2617 24.7 
 2003–2004 3469 32.7 
 ≥2005 2646 24.9 
Age group (years) <30 1953 18.4 
 30–39 4977 46.9 
 40–49 2709 25.5 
 ≥50 974 9.2 
Pre-HAART CD4 count
(cells/mm3
Median (IQR) 203 111–297 
Pre-HAART CD4% Median (IQR) 14 8–20 
Pre-HAART HIV RNA
(log10 copies/mL)
(n=10 078) 
Median (IQR) 4.8 4.1–5.3 
Characteristic  No. 
Number of patients  10 614 100.0 
Sex Male 7841 73.9 
 Female 2773 26.1 
HIV risk group MSM 5573 52.5 
 IDU 298 2.8 
 Heterosexual 3854 36.3 
 Other/not known 889 8.4 
Ethnicity White 5898 55.6 
 Black 3085 29.1 
 Other 1261 11.9 
 Not known 370 3.5 
Regimen PI-based 2958 27.9 
 NNRTI-based 6958 65.6 
 PI + NNRTI 200 1.9 
 NRTIs only 327 3.1 
 Other 171 1.6 
Year ≤2000 1882 17.7 
 2001–2002 2617 24.7 
 2003–2004 3469 32.7 
 ≥2005 2646 24.9 
Age group (years) <30 1953 18.4 
 30–39 4977 46.9 
 40–49 2709 25.5 
 ≥50 974 9.2 
Pre-HAART CD4 count
(cells/mm3
Median (IQR) 203 111–297 
Pre-HAART CD4% Median (IQR) 14 8–20 
Pre-HAART HIV RNA
(log10 copies/mL)
(n=10 078) 
Median (IQR) 4.8 4.1–5.3 

Abbreviations: CD4, CD4 count; CD4%, CD4 percentage; HAART, highly active retroviral therapy; HIV, human immunodeficiency virus; IDU, intravenous drug user; IQR, interquartile range; MSM, men who have sex with men; NRTI, nucleoside reverse transcriptase inhibitor; NNRTI, nonnucleoside reverse transcriptase inhibitor; PI, protease inhibitor.

Table 4.

Summary of Results From Multivariable Regression Analyses Assessing the Relationship Between Pre-HAART Discordancy Status and Response to HAART

  Unadjusted
 
Adjusted
 
 Median (IQR) change in CD4 Effect on mean (95% CI) P value Effect on mean (95% CI) P value 
Immunological response      
    Low discordancy 104 (38–180) −21.4 (−29.8 to −13.0) .0001 −19.6 (−27.7 to −11.5) .0001 
    Concordant 122 (56–205) … … 
    High discordancy 146 (71–243) 24.4 (13.7–35.0) .0001 27.4 (17.0–37.8) .0001 
 N (%) with outcome OR (95% CI) P value OR (95% CI) P value 
Initial virological response      
    Low discordancy 1758 (91.4) 0.99 (.94–1.04) .58 0.98 (.93–1.04) .56 
    Concordant 6835 (90.2) … … 
    High discordancy 973 (87.2) 0.95 (.94–1.08) .82 1.03 (.97–1.10) .36 
Viral rebound      
    Low discordancy 255 (14.5) 0.88 (.77–1.01) .07 0.86 (.75–.99) .03 
    Concordant 1072 (15.7) … … 
    High discordancy 189 (19.4) 1.29 (1.11–1.51) .001 1.17 (1.00–1.37) .05 
Clinical outcome (from HAART)      
    Low discordancy 217 (11.3) 1.10 (.94–1.28) .23 1.02 (.88–1.19) .77 
    Concordant 766 (10.1) … … 
    High discordancy 151 (13.6) 1.39 (1.17–1.66) .0002 1.33 (1.12–1.59) .001 
Clinical outcome (from 6 mos onward)      
    Low discordancya 74 (7.6) 1.01 (.79–1.29) .95 0.97 (.75–1.24) .78 
    Concordanta 403 (7.2)   
    High discordancya 85 (9.0) 1.30 (1.03–1.64) .03 1.37 (1.08–1.73) .01 
  Unadjusted
 
Adjusted
 
 Median (IQR) change in CD4 Effect on mean (95% CI) P value Effect on mean (95% CI) P value 
Immunological response      
    Low discordancy 104 (38–180) −21.4 (−29.8 to −13.0) .0001 −19.6 (−27.7 to −11.5) .0001 
    Concordant 122 (56–205) … … 
    High discordancy 146 (71–243) 24.4 (13.7–35.0) .0001 27.4 (17.0–37.8) .0001 
 N (%) with outcome OR (95% CI) P value OR (95% CI) P value 
Initial virological response      
    Low discordancy 1758 (91.4) 0.99 (.94–1.04) .58 0.98 (.93–1.04) .56 
    Concordant 6835 (90.2) … … 
    High discordancy 973 (87.2) 0.95 (.94–1.08) .82 1.03 (.97–1.10) .36 
Viral rebound      
    Low discordancy 255 (14.5) 0.88 (.77–1.01) .07 0.86 (.75–.99) .03 
    Concordant 1072 (15.7) … … 
    High discordancy 189 (19.4) 1.29 (1.11–1.51) .001 1.17 (1.00–1.37) .05 
Clinical outcome (from HAART)      
    Low discordancy 217 (11.3) 1.10 (.94–1.28) .23 1.02 (.88–1.19) .77 
    Concordant 766 (10.1) … … 
    High discordancy 151 (13.6) 1.39 (1.17–1.66) .0002 1.33 (1.12–1.59) .001 
Clinical outcome (from 6 mos onward)      
    Low discordancya 74 (7.6) 1.01 (.79–1.29) .95 0.97 (.75–1.24) .78 
    Concordanta 403 (7.2)   
    High discordancya 85 (9.0) 1.30 (1.03–1.64) .03 1.37 (1.08–1.73) .01 

Abbreviations: CI, confidence interval; HAART, antiretroviral therapy; IQR, interquartile range; OR, odds ratio.

a

Defined at 6 months after initiation of HAART.

The patients were followed for a median of 3.2 (IQR, 1.3–6.0) years over which time 9566 patients (90.1%) experienced an initial virological response to HAART, a median of 115 days (95% confidence interval [CI], 112–117) after starting HAART. There was no evidence that discordancy was associated with the likelihood of an initial virological response either before or after adjusting for other confounders (Table 4). A total of 1516 (15.9%) of the 9566 individuals with an initial response to HAART experienced viral rebound; the proportion with viral rebound by 1 year after initial virological response was 9.1%. There was some suggestion of an increased risk of viral rebound in individuals exhibiting high discordancy and a reduced risk in those exhibiting low discordancy; these effects remained significant after controlling for the pre-HAART CD4 count and other potential confounders. During follow-up, 907 (8.5%) patients developed a new AIDS event and 430 (4.1%) patients died (1134 patients either developed a new AIDS event or died). Clinical event rates were higher in those exhibiting high discordancy at the time of initiation of HAART (adjusted odds ratio, 1.39; P = .0002) than among those with a concordant CD4/CD4%; this finding persisted after adjustment for other known confounders, including the pre-HAART CD4 count.

A total of 7516 individuals (70.8%) had a CD4/CD4% pair 6 months after starting HAART (measured at a median of 6.0 [IQR], 5.5–6.6] months). By this time, the median CD4 count had risen to 325 (IQR, 213–460) cells/mm3 (median increase of 120 [IQR, 53–205] cells/mm3) and the median CD4% was 20 [IQR, 13-26] (median increase of 6% [IQR, 2%–9%]). Among this subgroup, median (IQR) baseline CD4 counts were 200 (76–306), 208 (120–290), and 190 (98–310) cells/mm3 in those exhibiting low discordancy, concordance, and high discordancy, respectively. Median (IQR) increases in CD4 count in these 3 groups were 104 (38–180), 122 (56–205), and 146 (71–243) cells/mm3, respectively (P = .0001). After adjustment for the pre-HAART CD4 count and other potential confounders, CD4 increases remained lower (by 20 cells/mm3; P = .0001) in those with low discordancy, and higher (by 27 cells/mm3; P = .0001) in those with high discordancy (Table 4). As we were concerned that these analyses might reflect regression to the mean (which may occur if the CD4 count was temporarily low as a result of random fluctuations), we repeated the analyses after redefining the baseline CD4 and CD4% as the average of the last 2 measurements prior to HAART where these were available, with similar results.

Our conclusions were unchanged when the analyses were repeated in the subset of individuals who started HAART from 2004 onward (data not shown). For time to both initial virological response and clinical progression there was no evidence of any statistical interaction between the pre-HAART CD4 count and either low (P value for interaction terms: .10 and .40, respectively) or high (P = .34 and .57) discordancy. Similarly, there was no evidence of any statistical interaction between the pre-HAART CD4 count and high discordancy for time to virological rebound (P = .22) or for predicting the change in CD4 count (P = .30). However, we did identify a significant interaction between low discordancy status and the pre-HAART CD4 count with time to virological rebound (P = .0003) and the 6-month change in CD4 count (P = .0001). In particular, the impact of low discordancy on virological rebound was reduced in individuals with higher pre-HAART CD4 counts whereas the detrimental impact on 6-month CD4 increases was increased.

At the 6-month timepoint, 971 (12.9%) individuals exhibited low discordancy and 945 (12.6%) high discordancy. Over a median follow-up of 3.1 (range, 1.3–5.7) years after the measurement of the 6-month CD4/CD4%, 256 (3.4%) individuals died, 414 (5.5%) developed an AIDS event, and 562 (7.5%) experienced clinical progression. Those who exhibited high discordancy at the 6-month timepoint were at higher risk of subsequent clinical progression, both before (odds ratio, 1.30; P = .03) and after (adjusted odds ratio, 1.37; P = .01), adjustment for potential confounding factors (Table 4).

DISCUSSION

In a large cohort of HIV-positive individuals, we found that for any given CD4 count, CD4% tended to be higher in women but was lower in older individuals, in those of nonwhite ethnicity, and in those with higher viral loads. While there appeared to be no significant effect of a discordant CD4/CD4% on virological responses to HAART, those exhibiting high discordancy were at greater risk of clinical progression after starting HAART than those with concordant values or those exhibiting low discordancy. At the same time, CD4 increases were greater in this subgroup. As those exhibiting high discordancy have low absolute CD4 counts for any given CD4%, this finding might be expected; however, these associations remained significant after adjusting for the pre-HAART CD4 count itself. Furthermore, the association with immunological response did not appear to be a consequence of regression to the mean, which may occur if patients were identified with high discordancy on the basis of a temporary random low CD4 count.

There is limited information on the rate of CD4/CD4% discordancy in other published studies, and comparison of results is complicated by the varying definitions of discordancy. For example, Le et al [9] reported that 8% of patients in the Military Medical Consortium for Applied Retroviral Research had CD4/CD4% discordancy, defined as a CD4% ≤20% among individuals with an absolute CD4 count of 500–700 cells/mm3. Among patients in the US CHORUS study [2], 25% of patients with CD4 counts ≥350 cells/mm3 had a CD4% that was <21% (low discordancy), whereas 17% of patients with a CD4 count <350 cells/mm3 had a CD4% ≥21% (high discordancy). Previous analyses from the same author [3] had revealed that 16% of patients with a CD4 count >350 cells/mm3 also had a CD4% ≤20%, whereas 25% of patients with a CD4 count <200 cells/mm3 had a CD4% ≥14%. In the study from Gebo et al [1], extreme discordance was rare: of CD4/CD4% pairs where the absolute CD4 count was 201–350 cells/mm3, the CD4% was <7% in only 0.3%, and 7%–14% in an additional 10.4%. Of CD4/CD4% pairs where the absolute CD4 count was <50 cells/mm3, 0.1% and 0.2% were associated with CD4 percentages of >21% and 15%–21%, respectively.

Several studies have considered whether the CD4 count and CD4% provide independent prognostic information among HIV-positive persons. Among patients in the CHORUS study starting HAART [2], the CD4% remained a significant predictor of clinical progression, even after adjusting for the absolute CD4 count. Among patients initiating HAART with a CD4 count ≥200 cells/mm3 [3], progression to a new AIDS-defining event or death tended to occur more frequently among those whose CD4% was <17%, although this association was only of borderline significance. Among those with a CD4 count >350 cells/mm3, however, clinical progression was significantly more rapid among patients with a CD4% <17%. Among untreated patients in the French Hospital Database on HIV [12], patients with lower CD4% values experienced more rapid progression to a new AIDS event, to a new serious AIDS event, and to death than those with higher CD4% values, even after adjusting for the CD4 count. In contrast, Gebo et al [1] reported that the CD4% was not associated with the development of AIDS after adjusting for the CD4 count and other demographic variables.

While these studies have investigated whether the CD4 count and CD4% provide independent prognostic information for clinical progression, to our knowledge only one of these studies has focused on the more direct question of whether discordancy itself is important [9]. In Le et al’s study, patients with CD4/CD4% discordancy did not appear to experience more rapid progression to several clinical outcomes, including anergy, Pneumocystis carinii pneumonia, thrush, a CD4 count ≤200 cells/mm3, a new opportunistic infection, or death. However, as all patients had a high CD4 count (500–700 cells/mm3), the number of patients experiencing an event was small with resulting lack of power.

A number of factors have been shown to influence the CD4 count in uninfected subjects, including the time of sampling [13], smoking [14], exercise, the presence of concomitant infections (including tuberculosis), and the use of other medications, including anticancer treatments [15]. While these factors increase the variability of the CD4 count, making its interpretation more difficult, they generally have a weaker impact on the CD4% and, as a result, this marker generally shows less variability. In contrast, the CD4% measurement reflects not only the number of CD4 cells but also the total lymphocyte count, which incorporates information on the number of CD8 and other CD3-positive cells. While our findings are of clinical interest, the question arises as to what any discordancy actually represents. If all patients had the same total lymphocyte count, then the CD4 count and CD4% would be perfectly correlated. Thus, any variation in the CD4% between individuals with the same CD4 count must reflect variations in the total lymphocyte count (and/or the CD8 count). Although studies have considered whether the total lymphocyte count, the CD8 count, and/or the CD4/CD8 ratio provide prognostic information in addition to that provided by the CD4 count, the results have often been inconsistent [16–18]. Unfortunately, we do not collect information on the total lymphocyte count, and information on CD8 and CD3 counts may be unavailable. Thus, we are unable to distinguish the relative components of any discordancy.

Among HIV-negative individuals, CD4 cell counts are higher in women than in men [19–21], and studies of HIV seroconverters have confirmed that these differences persist after infection, with women seroconverting to HIV, developing AIDS, and dying at higher CD4 counts [22]. Of note, most early findings highlighting the important role of the CD4 count and CD4%, as well as the development of algorithms for initiation of HAART, were based on data from largely male, white cohorts. While our findings might suggest that it may be appropriate to initiate HAART at a slightly different CD4 count in women than in men, these differences are unlikely to translate into a significant difference in clinical responses to HAART. While our data should not be used to make recommendations for the timing of HAART, when coupled with recent findings of lower mortality rates in women than in men in the post-HAART era [23], our findings do provide reassurance that existing treatment guidelines are likely to be sufficient, independent of gender.

There are also well-described differences in CD4 count between individuals from different ethnic groups. In one study [24], West African men had higher lymphocyte counts and CD4 counts than French men with similar CD4% and CD8 counts, resulting in higher CD4/CD8 ratios in the African men. In a study of US veterans [14], those of black ethnicity had higher CD8 counts than those who were white. Differences in lymphocyte subsets have also been reported in patients of Asian background, with studies [25, 26] reporting both lower CD4 counts and CD4% in Asians than in non-Asians. As with gender, it is unlikely that these differences will have a major impact on the outcomes of treatment, with any ethnicity-related differences in treatment outcome possibly being more likely to be a function of socioeconomic and behavioral differences [27] than any real biological effect.

Our finding that older individuals were more likely to have lower CD4% values for a given CD4 count is, however, somewhat surprising, and seems to contrast with other studies. CD4 counts decrease at older ages [20, 21] and our findings suggest that this might be exacerbated by an additional decline in CD4%. Patients who are older do suffer from accelerated HIV progression and this may be part of the explanation. Possible explanations for the age-related decline in CD4% values could include thymic involution, predominant memory response, and age-related immunodegeneration.

In conclusion, despite finding some associations between CD4/CD4% discordancy and immunological and clinical response to HAART, these effects were, on the whole, relatively small, suggesting that it is sufficient to simply monitor (and act upon) the CD4 count when assessing patient prognosis.

Notes

Acknowledgments.

The views expressed in this manuscript are those of the researchers and not necessarily those of the MRC.

Financial support.

Funding for this work was provided by the Medical Research Council, UK (G0000199 and G0600337).

Potential conflicts of interest.

All authors: No reported conflicts.

All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

APPENDIX

UK CHIC Steering Committee: Jonathan Ainsworth, Jane Anderson, Abdel Babiker, Loveleen Bansi, Valerie Delpech, David Dunn, Martin Fisher, Brian Gazzard, Richard Gilson, Mark Gompels, Teresa Hill, Margaret Johnson, Clifford Leen, Mark Nelson, Chloe Orkin, Adrian Palfreeman, Deenan Pillay, Andrew N. Phillips, Frank Post, Caroline Sabin, Memory Sachikonye, Achim Schwenk, John Walsh. Central Coordination: UCL Medical School, London (Loveleen Bansi, Teresa Hill, Andrew Phillips, Caroline Sabin); Medical Research Council Clinical Trials Unit (MRC CTU), London (Abdel Babiker, David Dunn, Stephen Sheehan). Participating Centers: King’s College Hospital, London (Anele Waters, Dorian Crates, Siti Mohamed-Saad, Frank Post); Brighton and Sussex University Hospitals NHS Trust (Martin Fisher, Nicky Perry, Anthony Pullin, Duncan Churchill, Wendy Harris); Chelsea and Westminster NHS Trust, London (Brian Gazzard, Steve Bulbeck, Sundhiya Mandalia, Jemima Clarke); Mortimer Market Centre, UCL Medical School (RFUCMS), London (Richard Gilson, Julie Dodds, Andy Rider, Ian Williams); Health Protection Agency—Centre for Infections, London (Valerie Delpech); Royal Free NHS Trust and RFUCMS, London (Margaret Johnson, Mike Youle, Fiona Lampe, Colette Smith, Helen Gumley, Clinton Chaloner, Dewi Ismajani Puradiredja); St Mary’s Hospital, London (John Walsh, Jonathan Weber, Shane Cashin, Christian Kemble, Nicky Mackie, Alan Winston); Barts and The London NHS Trust, London (Chloe Orkin, Rachel Thomas, Kevin Jones); Homerton Hospital, London (Jane Anderson, Selina Gann, Kevin Jones); Edinburgh (Clifford Leen, Alan Wilson); North Middlesex (Achim Schwenk, Jonathan Ainsworth); University Hospitals of Leicester, Leicester (Adrian Palfreeman, Anne Moore).

References

1.
Gebo
KA
Gallant
JE
Keruly
JC
Moore
RD
Absolute CD4 vs. CD4 percentage for predicting the risk of opportunistic illness in HIV infection
J Acquir Immune Defic Syndr
 , 
2004
, vol. 
36
 (pg. 
1028
-
33
)
2.
Hulgan
T
Shepherd
BE
Raffanti
SP
, et al.  . 
Absolute count and percentage of CD4+ lymphocytes are independent predictors of disease progression in HIV-infected persons initiating highly active antiretroviral therapy
J Infect Dis
 , 
2007
, vol. 
195
 (pg. 
425
-
31
)
3.
Hulgan
T
Raffanti
S
Kheshti
A
, et al.  . 
CD4 lymphocyte percentage predicts disease progression in HIV-infected patients initiating highly active antiretroviral therapy with CD4 lymphocyte counts >350 lymphocytes/mm3
J Infect Dis
 , 
2005
, vol. 
192
 (pg. 
950
-
7
)
4.
Taylor
JM
Fahey
JL
Detels
R
Giorgi
JV
CD4 percentage, CD4 number, and CD4:CD8 ratio in HIV infection: which to choose and how to use
J Acquir Immune Defic Syndr
 , 
1989
, vol. 
2
 (pg. 
114
-
24
)
5.
Burcham
J
Marmor
M
Dubin
N
, et al.  . 
CD4% is the best predictor of development of AIDS in a cohort of HIV-infected homosexual men
AIDS
 , 
1991
, vol. 
5
 (pg. 
365
-
72
)
6.
Moore
D
Hogg
R
Yip
B
Craib
K
Wood
E
Montaner
J
CD4 percentage is an independent predictor of survival in patients starting antiretroviral therapy with absolute CD4 cell counts between 200 and 350 cells/ml
HIV Med
 , 
2006
, vol. 
7
 (pg. 
383
-
8
)
7.
Goicoechea
M
Haubrich
R
CD4 lymphocyte percentage versus absolute CD4 lymphocyte count in predicting HIV disease progression: an old debate revisited
J Infect Dis
 , 
2005
, vol. 
192
 (pg. 
945
-
7
)
8.
Hammer
SM
Eron
JJ
Reiss
P
, et al.  . 
Antiretroviral treatment of adult HIV infection: 2008 recommendations of the International AIDS Society-USA Panel
JAMA
 , 
2008
, vol. 
300
 (pg. 
555
-
70
)
9.
Le
TP
Tribble
DR
Zhou
SYJ
, et al.  . 
Clinical significance of discordant CD4 count and CD4 percentage in HIV-infected individuals
AIDS
 , 
1997
, vol. 
11
 (pg. 
1395
-
6
)
10.
The UK Collaborative HIV Cohort Steering Committee
The creation of a large UK-based multicentre cohort of HIV-infected individuals: The UK Collaborative HIV Cohort (UK CHIC) Study
HIV Med
 , 
2004
, vol. 
5
 (pg. 
115
-
24
)
11.
Chene
G
Sterne
JA
May
M
, et al.  . 
Prognostic importance of initial response in HIV-1 infected patients starting potent antiretroviral therapy: analysis of prospective studies
Lancet
 , 
2003
, vol. 
362
 (pg. 
679
-
86
)
12.
Guiguet
M
Kendjo
E
Carcelain
G
, et al.  . 
CD4+ T-cell percentage is an independent predictor of clinical progression in AIDS-free antiretroviral-naive patients with CD4+ T-cell counts >200 cells/mm3
Antivir Ther
 , 
2009
, vol. 
14
 (pg. 
451
-
7
)
13.
Bofill
M
Janossy
G
Lee
CA
, et al.  . 
Laboratory control values for CD4 and CD8 T lymphocytes. Implications for HIV-1 diagnosis
Clin Exp Immunol
 , 
1992
, vol. 
88
 (pg. 
243
-
52
)
14.
Mili
F
Flanders
WD
Boring
JR
Annest
JL
Destefano
F
The associations of age, cigarette smoking, and smoking cessation to measures of the immune system in middle-aged men
Clin Immunol Immunopathol
 , 
1991
, vol. 
59
 (pg. 
187
-
200
)
15.
Vedel
SJ
Tholstrup
D
Kolte
L
, et al.  . 
Limited impact of the thymus on immunological recovery during and after chemotherapy in patients with diffuse large B-cell lymphoma
Scand J Immunol
 , 
2009
, vol. 
69
 (pg. 
547
-
54
)
16.
Chevret
S
Roquin
H
Ganne
P
Lefrere
JJ
Prognostic value of an elevated CD8 lymphocyte count in HIV infection. Results of a prospective study of 152 asymptomatic HIV-positive individuals
AIDS
 , 
1992
, vol. 
6
 (pg. 
1349
-
52
)
17.
Margolick
JB
Gange
SJ
Detels
R
O’Gorman
MR
RInaldo
CR
Jr
Lai
S
Impact of inversion of the CD4/CD8 ratio on the natural history of HIV-1 infection
J Acquir Immune Defic Syndr
 , 
2006
, vol. 
42
 (pg. 
620
-
6
)
18.
Stebbing
J
Sawleshwarkar
S
Michailidis
C
, et al.  . 
Assessment of the efficacy of total lymphocyte counts as predictors of AIDS defining infections in HIV-1 infected people
Postgrad Med
 , 
2005
, vol. 
81
 (pg. 
586
-
8
)
19.
Maini
MK
Gilson
RJ
Chavda
N
, et al.  . 
Reference ranges and sources of variability of CD4 counts in HIV-seronegative women and men
Genitourin Med
 , 
1996
, vol. 
72
 (pg. 
27
-
31
)
20.
Jentsch-Ullrich
K
Koenigsmann
M
Mohren
M
Franke
A
Lymphocyte subsets' reference ranges in an age- and gender-balanced population of 100 healthy adults --A monocentric German study. Clin Immunol
 , 
2005
, vol. 
116
 (pg. 
192
-
7
)
21.
Reichert
T
DeBruyère
M
Deneys
V
, et al.  . 
Lymphocyte subsets' reference ranges in an age- and gender-balanced population of 100 healthy adults–a monocentric German study
Clin Immunol Immunopathol
 , 
1991
, vol. 
60
 (pg. 
190
-
208
)
22.
Prins
M
Robertson
JR
Brettle
RP
, et al.  . 
Do gender differences in CD4 cell counts matter?
AIDS
 , 
1999
, vol. 
13
 (pg. 
2361
-
4
)
23.
Jarrin
I
Geskus
R
Bhaskaran
K
, et al.  . 
Gender differences in HIV progression to AIDS and death in industrialized countries: slower disease progression following HIV seroconversion in women
Am J Epidemiol
 , 
2008
, vol. 
168
 (pg. 
532
-
40
)
24.
Anglaret
X
Diabouga
S
Mortier
E
, et al.  . 
CD4+ T-lymphocyte counts in HIV infection: are European standards applicable to African patients
J Aquir Immune Defic Syndr Hum Retrovirol
 , 
1997
, vol. 
14
 (pg. 
361
-
7
)
25.
Howard
RR
Fasano
CS
Frey
L
Miller
CH
Reference intervals of CD3, CD4, CD8, CD4/CD8 and absolute CD4 values in Asian and non-Asian populations
Cytometry
 , 
1996
, vol. 
26
 (pg. 
231
-
2
)
26.
Webster
HK
Pattanapanyasat
K
Phanupak
P
, et al.  . 
Lymphocyte immunophenotype reference ranges in healthy Thai adults: implications of management of HIV/AIDS in Thailand
Southeast Asian J Trop Med Pub Health
 , 
1996
, vol. 
27
 (pg. 
418
-
29
)
27.
Silverberg
MJ
Leyden
W
Quesenberry
CP
Horberg
MA
Race/ethnicity and risk of AIDS and death among HIV-infected patients with access to care
J Gen Int Med
 , 
2009
, vol. 
24
 
9
(pg. 
1065
-
72
)