Incidence and Nature of Cognitive Decline Over 1 Year Among HIV-infected Former Plasma Donors in China Lucette A. Cysique; Scott L. Letendre; Christopher Ake; Hua Jin; Donald R. Franklin; Saurabh Gupta; Chuan Shi; Xin Yu; Zunyou Wu; Ian S. Abramson; Igor Grant; Robert K. Heaton
Authors and Disclosures
Posted: 05/12/2010; AIDS. 2010;24(7):983-990. ? 2010 Lippincott Williams & Wilkins
Objective: To quantify and characterize the nature of cognitive change over 1 year in a cohort of HIV-positive former plasma donors in rural China.
Design: The present study is an observational cohort study.
Methods: One hundred and ninety-two HIV-positive and 101 demographically comparable HIV-negative individuals, all former plasma donors, who lived in a rural part of China, received comprehensive medical and neuropsychological examinations. At study entry, 56% of HIV-positive group was on combination antiretroviral treatment and 60.9% at follow-up. Multiple regression change score approach was used with the HIV-negative sample to develop norms for change that would be then applied to the HIV-positive participants. Follow-up test scores adjusted for the control group practice effect.
Results: Fifty-three HIV-positive individuals (27%) developed significant cognitive decline as compared with five (5%) HIV-negative individuals. Cognitive decline was predicted at baseline by AIDS status, lower nadir CD4, and worse processing speed; at follow-up, it was associated with lower current CD4 cell count and failure of viral suppression on combination antiretroviral treatment. Neuropsychological decline also was associated with decreased independence in activities of daily living. Using neuropsychological impairment scores that were corrected for 'practice' on repeated testing, we found that among the decliners, 41.5% (N = 22) had incident impairment, whereas 38% (N
= 20) declined within the impaired range and another 20.7% (N = 11) declined within the normal range.
Conclusion: The present study demonstrates that despite ongoing combination antiretroviral treatment, cognitive decline in HIV-positive people is common over a 1-year follow-up. Regression-based norms for change on western neuropsychological tests can be used to detect disease-related cognitive decline in a developing country.
Several recent studies have provided prevalence estimates of HIV-associated neurocognitive [2,3]disorders (HAND) in 'nonwestern' regions of the world such as sub-Saharan Africa, India, [5,6]China, and south-east Asia. Taken together, the findings of these studies have demonstrated that HAND can be validly assessed in resource-limited countries, when using appropriate control samples. HAND prevalence appears to be similar in these areas of the world to the estimates reported with western HIV-positive cohorts (i.e., 20–50%, depending upon disease stages and comorbid conditions).
Combination antiretroviral therapy (cART) has greatly decreased HIV-associated mortality and medical morbidity, including HIV-associated dementia (HAD), but the prevalence of milder forms of HAND remains high. Several longitudinal cohort studies in western countries have assessed cognitive [9–13]decline and its potential contributing factors in the cART era. Published studies, after varying
follow-up periods, have found that between 22 and 63% of participants on cART demonstrated
[12,13]continuing impairment and at least 21% had incident impairment. Variability in these estimates of
cognitive impairment and change probably is related to methodological factors such as inclusion of appropriate normative standards for detecting these outcomes.
Just as it is important to define cross-sectional neuropsychological impairment in individual cases, rather than relying on group mean comparisons, longitudinal methods for classifying meaningful neuropsychological change in individuals also are needed. Advantages of reliably detecting cognitive change at the individual level include understanding factors responsible for variable manifestations and course of HAND, and alerting clinicians to consider changes in antiretroviral regimens early (e.g., to include agents with better central nervous system (CNS) penetration or adjunctive treatments that may specifically benefit the CNS).
To our knowledge, the current study is the first to assess long-term neurocognitive outcomes in a large cohort of HIV-positive persons in China and in the cART era. Our aim was to detect and quantify neuropsychological decline over a 1-year period in 192 individual HIV-positive former plasma donors (FPDs) in the Chinese province of Anhui. For this, we developed norms for change using longitudinal data of a fairly large sample (N = 101) of demographically comparable HIV-negative individuals. The [17,18]norms for change were derived using a regression change score approach. Here we propose an
extension of this method by introducing a battery approach to derive a normed summary change score [19,20]for each individual. In addition, we propose an adaptation of the Global Deficit Score method to
estimate follow-up impairment rate corrected for practice effect.
Details regarding participant recruitment procedures and baseline results were published in study by Heaton et al. Briefly, at baseline, 203 HIV-positive and 198 HIV-negative participants, virtually all of whom were farmers in the rural area of Anhui province, were enrolled into the study at a local hospital in Fuyang city. By design, at 12 months postbaseline, half of the baseline HIV-negative participants (N = 101) were reassessed. In addition, all available HIV-
positive participants were reassessed (N = 192), yielding a 5.4% (11/203) attrition rate. The
causes for dropout were one case of cerebral infarct; two cases of blindness and an additional case of substantial vision decline precluding valid neuropsychological testing; four moved away for work; and three deaths. The 101 HIV-negative participants had been randomly selected in order to compose a sample representative of the total baseline group with respect to demographic and baseline neuropsychological characteristics. This sample of HIV-negative participants was followed up at 1 year in order to develop normative standards for neuropsychological change (see also e-data analysis, http://links.lww.com/QAD/A35). At
baseline, 106 (55.2%) of the 192 HIV-positive participants met CDC-1993 criteria for AIDS. Virtually, the same subgroup (N = 107; 55.7%) was being prescribed cART at baseline
consistent with prevailing antiretroviral initiation guidelines in China. HIV duration estimated from self-report averaged 154 months (SD = 42). When they were retested after 12 months, 20 additional HIV-positive participants had transitioned to AIDS and at that time, 60.9% of the total group was receiving cART.
As in the baseline assessment, participants underwent comprehensive neurocognitive and neuromedical evaluations, and a structured psychiatric examination, as well as a self-report
assessment of daily functioning. Examinations were performed by the same Chinese psychiatric staff who had conducted the baseline assessments the previous year. They had been trained and certified in the standardized testing procedures by the US research team (R.K.H., D.F., and H.J.). Prior to starting the second annual testing phase, the Chinese examiners participated in another training session in order to review and practice test administration and test scoring procedures.
Details of the neuropsychological battery selection and adaptation for use in China are provided in studies by Cysique et al. and Heaton et al. (see also Table e-1 & e-2). Clinical
significance of any neuropsychological impairment or change was examined with standard instruments measuring participants' experience of cognitive difficulties in everyday life (Patient's Assessment of Own Functioning Inventory – PAOFI) and decreases in degree of
independence in activities of daily living [modified version of the Lawton & Brody Instrumental Activities of Daily Living (IADL) scale]. Participants also completed the Beck Depression [23,24]Inventory-II (BDI-II).
To identify individuals who presented with overall neuropsychological change, we used a [17,18]statistical methodology based on the multivariate regression change score approach (see
for review of this change score approach). The advantage of the regression-based change score approach is that it accounts for practice effect, regression toward the mean, and other factors that may influence normal test–retest variability in neurologically stable people (e.g., test–retest interval, demographics, and overall baseline neuropsychological competence).
Our detailed method is available in the supplemental file e-Data analysis, http://links.lww.com/QAD/A35. Briefly, to define neuropsychological decline in the HIV-positive sample, the 101 HIV-negative individuals were used as a reference sample to develop normative regression formulas. The final regression formulas were then applied to the HIV-positive sample providing a Z-score for each of 17 neuropsychological variables. These Z-
scores reflect how well or poorly the person performed at follow-up, relative to normal expectation for someone with the same baseline neuropsychological and other relevant characteristics. The Z-scores were then summed to provide a summary regression change score (sRCS). We determined a 90% confidence interval (CI) on the sRCS to define 'no change' on the test battery. That is, the cut-off for the top 5% of the sRCS distribution of the HIV-negative controls defined the 'improved' range and the cut-off for the bottom 5% defined the 'decliners' range. This was applied to the HIV-positive sample (Fig. 1).
Percentage of neuropsychological change as defined by the summary of
regression change score in the HIV-positive and HIV-negative samples.
Secondary analyses were conducted to clarify the nature of both baseline predictors and follow-up correlates of cognitive decline in this population. The HIV-positive decliners and
nondecliners (as defined by the sRCS) were compared on baseline and follow-up demographic, HIV disease-related laboratory measures, AIDS status, Global Deficit Score (GDS), and ability domain summary scores on the neuropsychological test battery, treatment-related variables, 2cognitive complaints, IADL, and BDI-II using t-test or χ test as appropriate. We also conducted
standard multivariate analyses to define which combination of baseline factors provided the most robust prediction of neuropsychological decline (defined by the sRCS). The GDS and ability domain T-scores at follow-up were corrected for the HIV-negative sample [6,19,20]median practice effect (see also e-Data analysis, http://links.lww.com/QAD/A35 ).
The baseline demographic characteristics of the HIV-negative and HIV-positive persons who participated in the 12-month follow-up assessment are presented in Table 1 . Test–retest
correlations on the neuropsychological battery were robust for both groups [HIV-negative group's mean scaled score r = 0.85 and HIV-positive group's mean scaled score r = 0.84]. tttt
Incidence of neuropsychological decline in the HIV-positive group (27.6%) was significantly greater than that in our reference HIV-negative sample (5%) as defined by the sRCS approach 2[χ = 21.4; P < 0.0001]. We found that 53 HIV-positive individuals were classified as having (1)
neuropsychological decline at the 12-month follow-up (i.e., they performed below the 5% cut-off reference range that defined the 'decliners' in the HIV-negative sample; CI of 95% one-tailed; see Fig. 1).
When comparing the 53 HIV-positive participants who were classified as decliners at the 12-month follow-up with the 139 HIV-positive nondecliners, we found that decliners tended to be older (P < 0.08), were more likely to have AIDS at baseline (P < 0.03), and had lower nadir CD4
count (P < 0.05). At follow-up, they had lower current CD4 cell count (P < 0.03) and also were
more likely to have detectable virus in plasma when on cART (P < 0.01; Table 2 ). To explore
further the CD4 cell count finding at follow-up and take into account any cART effect, a two-way analysis of variance (ANOVA) was performed with the CD4 cell count at follow-up as the outcome variable, and the decliner status as well as the treatment status at baseline (whether receiving cART or not at baseline) as predictors, and their interaction. We found that the follow-up CD4 cell count still tended to differ between the decliners and nondecliners (P = 0.06). No
other findings approached statistical significance. Importantly, the decliner and nondecliner groups did not differ in education and sex characteristics, or for prevalence of hepatitis C virus (HCV) infection or cART duration. They also did not differ in overall baseline neuropsychological performance (GDS) or prevalence of baseline neuropsychological impairment.
About one quarter of both the decliners and nondecliners evidenced clinically significant level of depressive symptoms (BDI-II ? 17) at baseline (26.4 versus 25.2%). At follow-up, neither
group showed an increase in rates of depression, but the nondecliners improved somewhat more (see Table 2 ). Decliners and nondecliners did not differ in their numbers of reported cognitive problems either at baseline (P = 0.22) or follow-up (P = 0.60), nor in the difference
between baseline and follow-up (P = 0.13). Results remain comparable when the BDI-II was
entered as a covariate. However, decliners were more likely to have significant decrease in IADL independence at follow-up (9.4 versus 1.4%; P < 0.007; significant decrease in IADL was defined as decrease in dependence in at least two everyday functioning areas).
Although neuropsychological impairment rates for the decliners and nondecliners were almost identical at baseline (37.7 versus 36.7%, respectively), not surprisingly, a much higher prevalence of impairment was seen in the decliners at follow-up (79.25 versus 28.78%; P <
0.0001). Overall impairment rate in the HIV-positive group at follow-up was 42.71% (82/192), which was just slightly higher than that at baseline (36.9%; 71/192).
When using the follow-up scores that were uncorrected for practice effect, 51% of decliners and 15.1% of nondecliners would be classified neuropsychologically impaired at follow-up (and only 25% in the total HIV-positive group – see also Table e-2 e-Data analysis,
http://links.lww.com/QAD/A35). Importantly, without correction for practice effect, impairment prevalence estimates in the HIV-negative controls also dropped from an expected 14% to only 2% (P = 0.0004). There is no known biological reason for such change; this result should, therefore, be considered as evidence for error due to failure to correct for practice. Although overall neuropsychological performance at baseline did not differ between decliners and nondecliners, we did find that speed of information processing (SIP) at baseline was significantly lower in the HIV-positive who would go on to decline as compared with the nondecliners (medium effect size, d = 0.51 – see Table 3 ). At follow-up and using the practice
effect corrected neuropsychological scores to derive demographically corrected ability domain T-scores, we found that the decliners demonstrated significantly lower performance in all domains when compared with the nondecliners, with the greatest difference found in learning, memory, and SIP (all large effect sizes; d > 0.8).
We found that among the 53 HIV-positive decliners, 22 (41.5%) showed incident impairment on the corrected GDS at follow-up. Eleven (20.7%) declined within the normal range and another 20 (38%) declined within the impaired range. These subgroups did not statistically differ on the sRCS.
Finally, we tested predictors of cognitive change using a stagewise regression approach. First, in a multivariate model with the sRCS as the outcome and including the predictors that were found (P < 0.10) to be different between HIV-positive decliners and nondecliners at baseline (see Table 2 ), we found that SIP (P = 0.004) and AIDS at baseline (trending at P = 0.06) 2remained unique contributors to the model (R = 0.10; P = 0.007; baseline age did not uniquely
predict change). Second, in a final two-covariate logistic regression model that predicted decliners' status, baseline SIP mean T-score had an odds ratio of 1.07 (CI 1.03–1.13) for a one-
unit decrease and baseline AIDS had an odds ratio of 1.94 (CI 0.98–3.84) relative to baseline
The present study demonstrated that performance on the neuropsychological tests originally developed in western regions of the world was quite reliable over time in both the infected and uninfected rural Chinese participants. Together with the earlier finding of greater baseline impairment associated with HIV and HCV infections and associations of neuropsychological impairment with both HIV disease severity and indications of decreases in everyday functioning, this suggests that neuropsychological tests standardized and widely used in western regions of the world (here historically and geographically defined as Europe, Australia, and North America) generalize to people having very different cultural, linguistic, and
educational backgrounds. It may be, therefore, that the abilities assessed by most or all of these tests capture some universal features of cognition.
Moreover, using the regression change score approach that has been validated mainly in non-HIV populations in the United States, we found that we could detect cognitive decline in 27.6% of the HIV-positive participants after a 1-year interval. This greatly exceeded the expected 5% normative cut-off of the HIV-negative sample. Even though our infected sample remained relatively healthy from a general medical point of view (average immune status improved and few participants had new AIDS-defining illnesses) and reported excellent adherence to their cART ( Table 2 ), many evidenced worsening of their neurocognitive status and slightly fewer than expected showed any improvement. Thus, though cART has dramatically improved the mortality and medical morbidity associated with HIV disease, such treatment appears less effective in alleviating or reversing CNS compromise that has been observed since the beginning of the HIV epidemic.
The minimal 12-month dropout rate (5%) for our HIV-positive FPDs is remarkable and reflects high levels of cooperation and dedication of the Chinese participants and examiners. Of the 11 baseline HIV-positive patients who could not be retested, seven were presumed to be due to HIV disease phenomena (three deaths, one stroke, three with recent severe visual impairment), so their loss to follow-up may have slightly underestimated rates of HIV-related neuropsychological decline.
[26–29]Similar to prior longitudinal neuropsychological findings in western studies, test–retest
interval in our rural Chinese population was not a significant predictor of variability in cognitive performance. Still, the variation in retest interval was restricted by design (actual range 237–
494 days) and it is possible that more discrepant test–retest intervals may have yielded
different results. Also similar to results in western studies, we found that the most robust predictors of follow-up neuropsychological scores were baseline scores on the same tests, as [18,27,28,30,31]well as overall neuropsychological competence at baseline, and to a lesser extent age
(see also e-Data analysis, http://links.lww.com/QAD/A35).
Supporting the current method for classifying change is the fact that neuropsychological decline was associated with markers of advanced HIV disease such as AIDS at baseline and lower nadir CD4 count. The same disease-related predictors were associated with neuropsychological impairment at baseline. Worse immune recovery, as reflected by lower
current CD4 cell counts at follow-up, also was associated with cognitive decline. Even though both groups' CD4 cell counts improved at follow-up, immune recovery was slightly worse for the decliners. Although cART use and duration did not differ between the groups, the cART of the decliners was less likely to be effective (they were more likely to have detectable viral loads at follow-up, suggesting more treatment failure in the decliner group).
Altogether, our findings would imply that current virological response to cART and past immune injury are important factors for HAND prognosis. Lower nadir CD4 cell counts may reflect prior HIV-related brain injury, representing a neuropathogenic process that may persist despite partial immune recovery, whereas current viral load detection would reflect ineffective cART. In any event, our results regarding long-lasting effects of earlier immunosuppression are [10,13,32–34]consistent with several reports in western regions as are findings suggesting that successful cART is an important factor for maintaining stable cognitive functions. Together,
these findings strengthen the hypothesis that effective cART initiated earlier in HIV disease may be more efficient at preventing HAND.
As has been noted in relation to studies of NeuroAIDS in the west, the precise timing and nature of the neuropathological mechanisms occurring in Chinese HIV-positive persons with stable or incident HAND are not well understood. Coinfection with HCV, though associated
with somewhat increased risk of neuropsychological impairment in this population at baseline, was unrelated to further cognitive decline over the next year. One explanation for this observation is that any brain injury that was previously linked to HCV had already occurred and was static and nonprogressive. Another is that some individuals who tested positive for HCV antibodies at baseline no longer had active HCV viral replication, though this hypothesis was not tested in this study. Future longitudinal research with coinfected (and HCV-monoinfected) samples should determine whether detectable HCV in plasma confers an increased risk for neuropsychological decline over time.
Another confirmatory finding with the previous literature in NeuroAIDS is that lower SIP performance was predictive of future neuropsychological decline. Impairment of this ability
appears to be a common and possibly early feature of both HAND and HAND progression. If so, the presence of even fairly isolated deficiencies in information processing speed may suggest the possibility of an evolving process. Studies in other countries should be undertaken to see whether this is a generalizable feature of HAND, independently of the cultural context. We showed that our estimate of cognitive decline was associated with decreased independence in activities of daily living, but not change in cognitive complaints or depressive symptoms. Independence in everyday life serves as an important benchmark to distinguish asymptomatic from symptomatic forms of HAND. We had previously demonstrated that
neuropsychological criteria for HAND could be effectively adapted to the Chinese rural context cross-sectionally and that neuropsychological impairment was associated with reported decrease in performance of IADL. This longitudinal finding is the first to be reported in a Chinese HIV-positive cohort and supports our definition of neuropsychological change and its clinical meaningfulness.
In conclusion, this study demonstrates that despite ongoing cART, cognitive decline in HIV-positive people is common over a 1-year follow-up. Regression-based norms for change on western neuropsychological tests can be used to detect disease-related cognitive decline in a developing country.