PREVALENCE CONTROVERSIES: ISSUES OF DIAGNOSTIC ACCURACY

Part of the validity of the Frascati HAND diagnostic criteria (Antinori et al., Reference Antinori, Arendt, Becker, Brew, Byrd, Cherner and Wojna2007) has been recently challenged (Gisslen, Price, & Nilsson, Reference Gisslen, Price and Nilsson2011; Nightingale et al., Reference Nightingale, Winston, Letendre, Michael, McArthur, Khoo and Solomon2014). While there is general agreement that NP testing is probably accurate in determining moderate to severe forms of HAND, researchers disagree over the criterion validity of NP testing in accurately identifying a clinically meaningful level of mild neurocognitive impairement, and in delineating asymptomatic neurocognitive impairment (ANI) from impairment resulting from confounding factors. Because of this, several studies have resorted to only including mild neurocognitive disorder (MND) and HAD cases, which can be useful at the level of primary care (McCombe, Vivithanaporn, Gill, & Power, Reference McCombe, Vivithanaporn, Gill and Power2013). But for biomarker discovery and prognostic value, the a priori exclusion of ANI is an issue, as this mild level of HAND is predictive of further cognitive decline (Grant et al., Reference Grant, Franklin, Deutsch, Woods, Vaida, Ellis and Heaton2014; Sacktor et al., Reference Sacktor, Skolasky, Seaberg, Munro, Becker, Martin and Miller2016).

There is empirical evidence that the Frascati criteria recommending at least a minus one standard deviation (−1 SD) on two cognitive domains, or 16% low NP performance threshold in an HIV- control sample optimally balances specificity and sensitivity for mild HAND detection (Taylor & Heaton, Reference Taylor and Heaton2001). Yet researchers who focus on this number at face value interpret that HAND rates are systematically over-estimated (Gisslen et al., Reference Gisslen, Price and Nilsson2011; Nightingale et al., Reference Nightingale, Winston, Letendre, Michael, McArthur, Khoo and Solomon2014). While some researchers contend that a -1 SD, as opposed to a proposed −1.5 or −2 SD, impairment threshold reflects the influence of confounding factors (e.g., psychiatric disorders that are common in HIV population) on rates of ANI, proponents of the −1 SD threshold point to data demonstrating relationships between objective markers of HIV brain involvement (e.g., neuropathology), and mild NP impairment at −1 SD (Cherner et al., 2002; Masliah et al., 1997; Moore et al., 2006).

Furthermore, there is substantial cumulative evidence that a cut-off at −1.5 SD or −2 SD will indisputably miss patients who have disease (Cherner et al., 2007; Grant et al., Reference Grant, Franklin, Deutsch, Woods, Vaida, Ellis and Heaton2014; Tierney et al., Reference Tierney, Sheppard, Kordovski, Faytell, Avci and Woods2017). Importantly, low NP performance in an HIV- sample is not the exact equivalent of impaired NP performance in a clinical sample. A recent review (Gates & Cysique, Reference Gates and Cysique2016) has attempted to clarify this issue where it was demonstrated that the recommended −1 SD in two cognitive domains low NP performance threshold in an HIV- control sample yielded a 5% misclassification rate for HAND in the demographically comparable HIV+ sample, and not a 16% rate. This result is associated with two factors: higher correlations between NP tests measures as a function of greater impairment, and the selection of test batteries that focus on detecting HIV-related cognitive impairment but not general cognitive impairment.

Adding to the misunderstanding is the fact that some researchers have computed the Frascati criteria incorrectly when conducting comparisons between the Frascati criteria and potential new ones (Su et al., Reference Su, Schouten, Geurtsen, Wit, Stolte, Prins and Schmand2015). Computational errors may also explain why some studies found unusual discrepancies between the cognitive rating and the global deficit score methods (McDonnell et al., Reference McDonnell, Haddow, Daskalopoulou, Lampe, Speakman, Gilson and Rodger2014), despite the virtual equivalence between the two methods (Blackstone, Moore, Franklin, et al., Reference Blackstone, Moore, Franklin, Clifford, Collier, Marra and Heaton2012). Furthermore, in several instances, researchers assessing the robustness of the Frascati criteria in small test batteries have not modified—as required in this context (Ingraham & Aiken, Reference Ingraham and Aiken1996; Kamminga et al., Reference Kamminga, Bloch, Vincent, Carberry, Brew and Cysique2017) —the expected low performance rate in their HIV- control sample to meet the recommended 16% low-performance threshold. Failure to modify the expected threshold for impairment in small batteries by either using a stricter cut-off (Kamminga et al., Reference Kamminga, Bloch, Vincent, Carberry, Brew and Cysique2017), or/and requiring impairment in only one task (Ingraham & Aiken, Reference Ingraham and Aiken1996) will produce unexpectedly high rates of low performance in HIV- (McDonnell et al., Reference McDonnell, Haddow, Daskalopoulou, Lampe, Speakman, Gilson and Rodger2014). Finally, some reports have not used appropriate normative data (Simioni et al., Reference Simioni, Cavassini, Annoni, Rimbault Abraham, Bourquin, Schiffer and Du Pasquier2010) or have used control samples that diverged substantially from clinical samples in terms of some socio-demographics (McDonnell et al., Reference McDonnell, Haddow, Daskalopoulou, Lampe, Speakman, Gilson and Rodger2014).

PREVALENCE CONTROVERSIES: THE ROLE OF NP CONFOUNDS ON HAND PREVALENCE

The HIV epidemic is characterized by situations that make the detection of HAND in very specific populations or demographic categories challenging. For example, the study by Crum-Cianflone et al. (Reference Crum-Cianflone, Moore, Letendre, Poehlman Roediger, Eberly, Weintrob and Hale2013) evaluated NP functions in 200 HIV+ and 50 demographically comparable HIV- military beneficiaries. The HIV+ sample was relatively healthy (HIV duration=5 years; 64% on cART initiated; median CD4 of 333 cells/mm³). The overall impairment rate was found to be 19% in the HIV+ group and surprisingly 30% in the HIV- group. While the level of impairment in the HIV+ group is akin to another study in patients with high CD4 cell counts and cART initiation always above CD4 of 350 cells/mm³ (despite different methods it was 20%) (Wright et al., Reference Wright, Grund, Cysique, Robertson, Brew, Collins and Price2015), the rate of impairment in the HIV- sample was very unusual and akin to a clinical population with a history of traumatic brain injury (Rabinowitz et al., Reference Rabinowitz, Li, McCauley, Wilde, Barnes, Hanten and Levin2015), condition that is prevalent in the U.S. military (Helmick et al., Reference Helmick, Spells, Malik, Davies, Marion and Hinds2015).

The largest ever study of HIV- and HIV+ women (Maki et al., Reference Maki, Rubin, Valcour, Martin, Crystal, Young and Anastos2015) showed an overall small effect of HIV on cognition (demographically adjusted Cohen’s d<.20 ) in a cohort composed of HIV+ women where 30% had AIDS, 53% were virally undetectable and 35% were sub-optimally adherent. In relative terms, the effect of reading achievement, age, years of education, and racial/ethnicity category had a much larger explanatory value with respect to NP performance. However, the effect of HIV disease biomarkers (viral detection, lower CD4 and AIDS) on cognition was detected yet masked at the group level due to a dramatic combination of socio-economic burdens (poverty, childhood trauma, domestic violence) that negatively and independently impacted cognition in both HIV- and HIV+ women, particularly for learning, memory, and psychomotor speed (Rubin et al., Reference Rubin, Pyra, Cook, Weber, Cohen, Martin and Maki2016). Interestingly, the longitudinal follow-up of this cohort can be found in Rubin et al., (Reference Rubin, Maki, Springer, Benning, Anastos, Gustafson and Valcour2017). An editorial highlighted the main finding which was that cognitive decline was greater in HIV+ women whether virally suppressed or not compared to their HIV- counterparts and above and beyond the effect of confounds (Cysique & Becker, Reference Cysique and Becker2017).

THEMES TO CONSIDER IN THE POTENTIAL UPDATE OF THE FRASCATI CRITERIA

New indexes may be considered for improving the detection of clinically meaningful ANI/MND, but they will need further study to assess their validity. Investigators of the POPPY studies have proposed other statistical indexes. However, this work is only based on the CogState battery (De Francesco et al., Reference De Francesco, Underwood, Post, Vera, Williams, Boffito and Sabin2016). Any CogState cognitive domain grouping would require some validation as this battery is different from standard NP testing (Cysique, Maruff, Darby, & Brew, Reference Cysique, Maruff, Darby and Brew2006; Kamminga et al., Reference Kamminga, Bloch, Vincent, Carberry, Brew and Cysique2017), while a global score as opposed to cognitive domain ratings is best to quantify impairment when using this battery (Kamminga et al., Reference Kamminga, Bloch, Vincent, Carberry, Brew and Cysique2017). One proposed method (also used by Su et al., Reference Su, Schouten, Geurtsen, Wit, Stolte, Prins and Schmand2015) is the multivariate normative comparison (MNC), which has been developed to control for family-wise error in large NP test batteries (Huizenga, Agelink van Rentergem, Grasman, Muslimovic, & Schmand, Reference Huizenga, Agelink van Rentergem, Grasman, Muslimovic and Schmand2016). Recent improvement in the MNC method shows that individual cases need to be compared to normal age, education and gender matched cases, and not the entire normative sample (Agelink van Rentergem, Murre, & Huizenga, Reference Agelink van Rentergem, Murre and Huizenga2017). To the best of our knowledge, this requirement has not been conducted in NeuroHIV studies which used this method. Also, the MNC validity to detect mild cognitive impairment in non-HIV populations has not been tested. The other method propsed by POPPY investigators (Underwood, Leech, Winston, Sabin, & De Francesco, Reference Underwood, Leech, Winston, Sabin and De Francesco2017) is the Malahanobis distance (Crawford & Allan, Reference Crawford and Allan1994), which a priori takes into consideration the test correlation within a test battery while the Frascati criteria do not, and the Global Deficit Score [GDS, (Carey et al., Reference Carey, Woods, Gonzalez, Conover, Marcotte, Grant and Heaton2004)] does, but not to the same extent (Gates & Cysique, Reference Gates and Cysique2016). Similar to the GDS, however, the Malanobis distance needs to be thresholded, so the issue of how to accurately detect clinically meaningful forms of mild cognitive impairment remains.

The exclusion of neuropsychiatric symptoms from the Frascati criteria is problematic because it is well known that alteration to striato-frontal circuits affect emotional regulation and motivation (Cole, Castellon, et al., Reference Cole, Castellon, Perkins, Ureno, Robinet, Reinhard and Hinkin2007). Therefore, some neuropsychiatric symptoms prevalent in HIV+ persons, including apathy, irritability (Watkins & Treisman, Reference Watkins and Treisman2012), anxiety (Brandt et al., Reference Brandt, Zvolensky, Woods, Gonzalez, Safren and O’Cleirigh2017), possible treatment-resistant depression (Cysique, Dermody, Carr, Brew, & Teesson, Reference Cysique, Dermody, Carr, Brew and Teesson2016), could represent progressing HIV-related brain injury (Cysique & Brew, in press). Furthermore, in women and ethnically diverse cohorts, it appears that major depressive disorder is more often associated with HAND (Fellows, Byrd, Morgello, & Manhattan, Reference Fellows, Byrd and Morgello2013), and is more often predictive of cognitive decline (Heaton et al., Reference Heaton, Franklin, Deutsch, Letendre, Ellis, Casaletto and Grant2015) than in less diverse male cohorts (Cysique et al., Reference Cysique, Deutsch, Atkinson, Young, Marcotte, Dawson and Heaton2007). Such subtle divergences in the expression of a neurological and neuropsychiatric disease should be expected and not systematically interpreted as evidence that “ethnic minorities and women are a problem” or that their disease expression is not the “real” one. Historically, symptoms that were later accepted as part of the CDC AIDS definitions were not initially recognized in African-American women back in the 70s because their symptoms did not correspond to the typical presentation of AIDS in men (The ACT UP/New York Women and AIDS Book Group, 1990). Further illustrating this complexity, a recent study from Nigeria showed that higher plasma levels of HIV driving activation of circulating monocytes may be more explanatory of HAND in women than in men (Royal et al., Reference Royal, Cherner, Burdo, Umlauf, Letendre, Jumare and Blattner2016).

In a potentially new formulation of the Frascati criteria, the question of impact on everyday functioning would have to be considered carefully. The assessment of current everyday functioning primarily relies on patients’ self-insight (Chiao et al., Reference Chiao, Rosen, Nicolas, Wendelken, Alcantar, Rankin and Valcour2013). This is less of an issue in high-functioning persons with mild HAND, who show correlation between insight into their cognitive difficulties and their actual level of cognitive impairment (Cysique et al., Reference Cysique, Heaton, Kamminga, Lane, Gates, Moore and Brew2014); although depressed HIV+ patients typically report more cognitive symptoms than non-depressed patients, and sometimes without any evidence of objective cognitive deficits (Carter, Rourke, Murji, Shore, & Rourke, Reference Carter, Rourke, Murji, Shore and Rourke2003).

However, as soon as the impairment is moderate, especially in a fronto-striatal disease such as HAND (Thames et al., Reference Thames, Becker, Marcotte, Hines, Foley, Ramezani and Hinkin2011), insight can be impaired enough that the self-assessment of everyday functioning is likely to be inaccurate (Blackstone, Moore, Heaton, et al., Reference Blackstone, Moore, Heaton, Franklin, Woods, Clifford and Grant2012). Therefore, modifying the diagnostic Frascati criteria to require impact on everyday functioning as in the current DSM-5 criteria (American Psychiatric Association, 2013) could have negative consequences by over-representing depressed, while those with lower self-insight and frontal symptoms would be systematically excluded. Recent advances have seen the development of several performance-based activities of daily living assessments which overcomes the issue of self-report (Blackstone, Moore, Heaton, et al., Reference Blackstone, Moore, Heaton, Franklin, Woods, Clifford and Grant2012; Sheppard, Woods, & Verduzco, Reference Sheppard, Woods and Verduzco2017), including some that are Internet based (Woods et al., Reference Woods, Iudicello, Morgan, Verduzco, Smith and Cushman2017). More work is needed to assess their construct validity as real-life indicators independent of cognitive functioning.

Further improvement in diagnostic validity will come not from any artificial reduction in the capacity of the criteria to detect baseline mild HAND, but by working toward the identification of biomarkers that will aid in determining if current HAND (including ANI) represent an active brain pathology or a static burnt out process (Brew, Reference Brew2004). This could be achieved by combining diagnostic criteria that determine if HIV+ patients have, within a clinically relevant period, stable, progressing, or incident cognitive decline (Gott, Gates, Dermody, Brew, & Cysique, Reference Gott, Gates, Dermody, Brew and Cysique2017; Tierney et al., Reference Tierney, Sheppard, Kordovski, Faytell, Avci and Woods2017) against selected biomarkers. The Frascati criteria could be improved by determining guidelines for cognitive decline (Gates & Cysique, Reference Gates and Cysique2016; Tierney et al., Reference Tierney, Sheppard, Kordovski, Faytell, Avci and Woods2017). These nomenclature changes would be similar to the current DSM-5 criteria for HAND, but with an improved formalization of cognitive impairment and decline (Tierney et al., Reference Tierney, Sheppard, Kordovski, Faytell, Avci and Woods2017). Such an update would reduce the weight of any diagnosis on a cross-sectional assessment and confirm its value longitudinally, enhancing the criteria for biomarker discovery, therapeutic and clinical management, as well as selection in trials.

CROSS-SECTIONAL NEUROPSYCHOLOGICAL PROFILE: GLOBAL FINDINGS IN NON-WESTERN COHORTS IN RELATION TO WESTERN COHORTS

The past decade has witnessed a notable expansion of comprehensive NP studies to Sub-Saharan Africa (Buch et al., Reference Buch, Chivero, Hoare, Jumare, Nakasujja, Mudenda and Joseph2016), Asia (Joseph et al., Reference Joseph, Achim, Boivin, Brew, Clifford, Colosi and Wood2013; Ku et al., Reference Ku, Lee, Ahn, Song, Kim, Kim and Choi2014), Eastern Europe (Joseph et al., Reference Joseph, Achim, Boivin, Brew, Clifford, Colosi and Wood2013) and South America (Joseph et al., Reference Joseph, Achim, Boivin, Brew, Clifford, Colosi and Wood2013). These pools of research strongly demonstrate the feasibility of NP studies in culturally and economically diverse non-Western settings using standard NP tests adapted for language and cultural variations. This development represents a remarkable achievement on the part of the global NeuroHIV research community, especially when we consider the paucity of studies that existed in 2009 as reviewed by (Robertson, Liner, & Heaton, Reference Robertson, Liner and Heaton2009). Key to the success of these studies has been the clinical expertise of local researchers, the contribution of collaborative training strategies for local testing personnel (Heaton et al., Reference Heaton, Cysique, Jin, Shi, Yu, Letendre and Wu2008; Pumpradit et al., Reference Pumpradit, Ananworanich, Lolak, Shikuma, Paul, Siangphoe and Valcour2010), and the commitment of local researchers to the recognition of HAND as a significant health issue.

These studies variously assess NP functions in antiretroviral naïve patients, patients starting treatment, and patients who are stable on treatment where the vast majority had experienced immunosuppression. In untreated cohorts, and as expected, HAD prevalence rate is elevated [e.g., 25.4% in (Joska et al., Reference Joska, Westgarth-Taylor, Myer, Hoare, Thomas, Combrinck and Flisher2011); 39% in (Sacktor et al., Reference Sacktor, Nakasujja, Okonkwo, Skolasky, Robertson, Musisi and Katabira2013]), similar to pre-cART Western cohorts. The profile of HAD in untreated non-Western samples is identical to that of original reports in the United States, with a subcortical profile dominated by slowing of information processing, motor deficits, working memory/attention deficits, learning and retrieval memory impairment, verbal fluency deficits, and executive dysfunctions (Gupta et al., Reference Gupta, Satishchandra, Gopukumar, Wilkie, Waldrop-Valverde, Ellis and Kumar2007; Robertson, Nakasujja, et al., Reference Robertson, Nakasujja, Wong, Musisi, Katabira, Parsons and Sacktor2007; Yepthomi et al., Reference Yepthomi, Paul, Vallabhaneni, Kumarasamy, Tate, Solomon and Flanigan2006). Although the effect of education on performance is often larger than in Western cohorts (Joska et al., Reference Joska, Westgarth-Taylor, Hoare, Thomas, Paul, Myer and Stein2012), this is probably due to a wider range of educational experience (Gupta et al., Reference Gupta, Vaida, Riggs, Jin, Grant, Cysique and Heaton2011). The use of appropriate controls, fairly comprehensive NP test batteries, and the rigorous application of the Frascati criteria have ensured an excellent detection of HAND (Gupta et al., Reference Gupta, Satishchandra, Gopukumar, Wilkie, Waldrop-Valverde, Ellis and Kumar2007; Heaton et al., Reference Heaton, Cysique, Jin, Shi, Yu, Letendre and Wu2008). However, tests requiring high levels of abstraction are probably best avoided among participants with very low levels of educations (Heaton et al., Reference Heaton, Cysique, Jin, Shi, Yu, Letendre and Wu2008), and reserved for participants with at least secondary school level (Gupta et al., Reference Gupta, Iudicello, Shi, Letendre, Knight, Li and Heaton2014). Finally, similar to North-American cohorts, patients with AIDS (vs. non-AIDS) show increased levels of cognitive dysfunction (Kanmogne et al., Reference Kanmogne, Kuate, Cysique, Fonsah, Eta, Doh and Njamnshi2010).

Several studies (Ghate et al., Reference Ghate, Mehendale, Meyer, Umlauf, Deutsch, Kamat and Marcotte2015; Joska et al., Reference Joska, Westgarth-Taylor, Hoare, Thomas, Paul, Myer and Stein2012; Robertson et al., Reference Robertson, Jiang, Kumwenda, Supparatpinyo, Evans, Campbell and Nair2012; Sacktor et al., Reference Sacktor, Nakasujja, Okonkwo, Skolasky, Robertson, Musisi and Katabira2013; Valcour et al., Reference Valcour, Shiramizu, Sithinamsuwan, Nidhinandana, Ratto-Kim, Ananworanich and Shikuma2009) demonstrated NP improvement after cART initiation. As a pool, these studies are more representative of an initial cART effect than Western studies because most patients were originally naïve participants, while in Western cohorts many were ART-experienced (Cysique & Brew, Reference Cysique and Brew2009). In studies using fairly comprehensive NP testing and a large number of patients with HAD, the domains that showed the most improvement were motor deficits, working memory/attention deficits, verbal fluency deficits, and executive dysfunctions, while improvement in learning/memory and psychomotor speed were either less pronounced or less consistent (Joska et al., Reference Joska, Westgarth-Taylor, Hoare, Thomas, Paul, Myer and Stein2012; Sacktor et al., Reference Sacktor, Nakasujja, Okonkwo, Skolasky, Robertson, Musisi and Katabira2013). In patients with mild to moderate or moderate levels of impairment (among which few were affected by HAD), most improvement occurred in learning, and to a lesser extent memory and motor functions (Ghate et al., Reference Ghate, Mehendale, Meyer, Umlauf, Deutsch, Kamat and Marcotte2015).

In long-term cART-treated cohorts including a majority of HIV+ patients who had initiated cART with a low CD4 count, a recent study in Zambia (Kabuba, Anitha Menon, Franklin, Heaton, & Hestad, Reference Kabuba, Anitha Menon, Franklin, Heaton and Hestad2017) detected 34.6% of HAND. This prevalence rate is not statistically different from North American studies, which have used the same NP methods [(Heaton, Clifford, et al., Reference Heaton, Clifford, Franklin, Woods, Ake, Vaida and Grant2010), including demographically corrected normative standard]. Also ANI is the most common HAND subtype (68%) showing that its a priori exclusion would be a major issue for the accurate assessment of HAND’s global burden. Moreover, the study demonstrated medium effect size impairment between controls and HIV+ participants in the domains of executive functioning, speed of information processing, working memory, learning, delayed recall, but not fine motor-coordination.

Another study of cART-treated patients in Botswana (Lawler et al., Reference Lawler, Jeremiah, Mosepele, Ratcliffe, Cherry, Seloilwe and Steenhoff2011) detected 37% with HAND. The study showed impairment in all domains tested (learning/memory; verbal fluency, executive functions, psychomotor speed) compared to demographically comparable controls. Of interest, fine motor-coordination was different at only p=.04 between groups. This relative insensitivity of the Grooved Pegboard to HIV effects has been found in several other African studies (Akolo et al., Reference Akolo, Royal, Cherner, Okwuasaba, Eyzaguirre, Adebiyi and Blattner2014; Clifford et al., Reference Clifford, Mitike, Mekonnen, Zhang, Zenebe, Melaku and Evans2007; Robertson, Nakasujja, et al., Reference Robertson, Nakasujja, Wong, Musisi, Katabira, Parsons and Sacktor2007). This could be consistent with the Western studies showing less motor-coordination deficits in cART compared to pre-cART cohorts (Cysique et al., Reference Cysique, Maruff and Brew2004; Heaton et al., Reference Heaton, Franklin, Ellis, McCutchan, Letendre, Leblanc and Grant2011), except that it has also been found in asymptomatic HIV+ patients in Nigeria that were not on treatment (Akolo et al., Reference Akolo, Royal, Cherner, Okwuasaba, Eyzaguirre, Adebiyi and Blattner2014).

Future studies dedicated to exploring this issue could perhaps use a more extensive fine-motor assessment, a thorough assessment of demographic factors that may affect manual dexterity (Gupta et al., Reference Gupta, Vaida, Riggs, Jin, Grant, Cysique and Heaton2011), and control for any peripheral neuropathy(Fellows et al., Reference Fellows, Byrd, Elliott, Robinson-Papp, Mindt and Morgello2012). In this regard, focusing assessment on motor functions should also probably be avoided when possible in non-Western cohorts, as it seems to lead to lower than expected rates of HAND (Robertson et al., Reference Robertson, Kumwenda, Supparatpinyo, Jiang, Evans, Campbell and Brouwers2011; Wright et al., Reference Wright, Brew, Arayawichanont, Robertson, Samintharapanya, Kongsaengdao and Wesselingh2008) (<15%). Finally, a meta-analysis by (Habib et al., Reference Habib, Yakasai, Owolabi, Ibrahim, Habib, Gudaji and Nashabaru2013) for sub-Saharan countries which included studies using screening tests and more comprehensive NP testing found that among those on ART for ≥6 months, the estimate for neurocognitive impairment was 30.4%. This is mostly consistent with what is detected in Western cohorts.

In other parts of the world, and in cohorts composed of mostly virally suppressed HIV+ persons tested with comprehensive NP test batteries, impairment rates were found to be around the 25–35% benchmark (Ene et al., Reference Ene, Franklin, Burlacu, Luca, Blaglosov, Ellis and Marcotte2014; Heaton, Cysique, et al., Reference Heaton, Cysique, Jin, Shi, Yu, Letendre and Wu2010; Kelly et al., Reference Kelly, van Oosterhout, Ngwalo, Stewart, Benjamin, Robertson and Solomon2014; Ku et al., Reference Ku, Lee, Ahn, Song, Kim, Kim and Choi2014; Pumpradit et al., Reference Pumpradit, Ananworanich, Lolak, Shikuma, Paul, Siangphoe and Valcour2010). In these studies, as in Western cohorts, ANI again accounts for the majority of diagnosis, MND for approximately a quarter, and HAD is typically present in less than 5%. Slight variation in HAND prevalence rates may be accounted for by differences in past experience of immunosuppression/viral replication, timeliness of cART initiation, duration of stability of current cART, and some portion of confounds if those were not excluded or controlled for.

LONGITUDINAL NEUROPSYCHOLOGICAL PROFILE: GLOBAL NATURALISTIC STUDIES

This overview of naturalistic longitudinal studies focuses on observational studies and their definitions of cognitive decline, thus exclusive of drug trials. We included 10 naturalistic longitudinal studies of cognitive change with cognitive decline investigated as a primary outcome as reported by August 2017, inclusive of a majority of HIV+ participants on stable cART. Seven were conducted in the United States, two in Australia and one in China, highlighting the need for further expansion. Several principal findings can be extracted from those studies: (1) The definition of cognitive decline as an absorbing state (wherein once an individual is defined as having declined this status is permanent) as compared to studies with multiple follow-up points (wherein an individual’s status can fluctuate across the duration of the project) yields apparent differences in cognitive decline rates that have to be interpreted carefully. Cysique, Maruff, and Brew (Reference Cysique, Maruff and Brew2006) defined cognitive decline as a unique endpoint and found 30% of HIV+ participants were defined as decliners at 6–12 month follow-up and 5% at 15–27 months.

In contrast, the large CNS HIV Antiretroviral Therapy Effects Research (CHARTER) study which defined cognitive decline as an absorbing state found an overall 23% cognitive decline rate over the 3-year study follow-up (Heaton et al., Reference Heaton, Franklin, Deutsch, Letendre, Ellis, Casaletto and Grant2015). A subsequent project by Cysique et al. (Reference Cysique, Letendre, Ake, Jin, Franklin, Gupta and Heaton2010) also used an absorbing rate definition and found an overall decline over the 27-month study period of 27%. (2) These three studies used standard global change scores (standard in the sense that scores are corrected for practice effect and regression toward the mean) and found fairly similar rates of cognitive decline ranging between ~30% across different cohorts context/nationalities. (3)

Some studies focused on incident HAND. One study (Sheppard et al., Reference Sheppard, Woods, Bondi, Gilbert, Massman and Doyle2015) found that HIV infection confers a nearly fivefold risk for developing a neurocognitive disorder over approximately 1 year. The sample was composed of a majority of participants on cART, of which 26% had detectable viral load. Of note, the authors applied a custom-made group practice effect correction based on the sub-sample of patients who had stable performance. They detected that 15.7% of the cohort had incident HAND. Another study (Gott et al., Reference Gott, Gates, Dermody, Brew and Cysique2017) in a sample of fully virally suppressed patients detected that incident HAND occurred in 7%, progressing HAND in 10%, while 35% has stable HAND over a 18-month period. The study used standard global change scores and combined this information with baseline HAND status. (4)

Studies that assessed deterioration within the various stages of HAND (thus not correcting for practice effect and regression toward the mean) typically found lower rates of cognitive decline. The Multicenter AIDS Cohort Study (MACS) found that across four years of follow-up, 10% of HIV+ individuals progressed to a worse stage of HAND stage (Sacktor et al., Reference Sacktor, Skolasky, Seaberg, Munro, Becker, Martin and Miller2016). Another study (Brouillette et al., Reference Brouillette, Yuen, Fellows, Cysique, Heaton and Mayo2016) reassessed the CHARTER sample using a relatively new group trajectory method but without practice effect corrections and found 15% cognitive decline.

Finally, in a more recent study, Vassallo et al. (Reference Vassallo, Fabre, Durant, Lebrun-Frenay, Joly, Ticchioni and Pradier2017) found 32% of cognitive decline in a group of patients of which approximately a quarter had detectable viral loads. While they used no correction for practice effect, they chose a fairly stringent criterion for degree decline (-2 SD) but only in one cognitive domain, potentially explaining the >30% decline rate. (5) Studies that used a restricted number of tests showed inconsistent results, suggesting that as with cross-sectional studies, a minimum of five cognitive domains should be assessed (Antinori et al., Reference Antinori, Arendt, Becker, Brew, Byrd, Cherner and Wojna2007). For example, when focusing on verbal memory, one study found worse performance over time in HIV+ individuals compared to HIV-individuals at 1 year follow-up (Seider et al., Reference Seider, Luo, Gongvatana, Devlin, de la Monte, Chasman and Cohen2014) and worse in the oldest participants, while another which focused only on two tasks of psychomotor speed and mental flexibility and a large selected group of long-term asymptomatic HIV+ adults reported stable cognitive performance across eight years (Cole, Margolick, et al., Reference Cole, Margolick, Cox, Li, Selnes, Martin and Miller2007). These results from the MACS cohort have never been replicated, probably because particularly healthy individuals were selected and many without any evidence of HAND in the first place. Overall, this overview demonstrates that standardization of the definition of cognitive decline is urgently needed, especially as the HIV population is aging with an increased prevalence of age-related comorbidities (Guaraldi et al., Reference Guaraldi, Orlando, Zona, Menozzi, Carli, Garlassi and Palella2011), and risk for brain premature aging (Brew & Cysique, Reference Brew and Cysique2017; Goodkin et al., Reference Goodkin, Miller, Cox, Reynolds, Becker, Martin and Sacktor2017).