Overall Cognitive Reserve
Lower CRI-Total score was associated with HAND, independent of HIV disease severity measures. As CRI-Total is a proxy CR measure, this supports our hypothesis that individuals with higher CR are less likely to meet HAND criteria.
The CRIq has only been used to evaluate CR in one other HAND study (25). This study also reported that lower CRI-Total was independently associated with HAND, though sample size was smaller (n=60) and predominantly male (51/60).
Our findings appear similar to other studies from HICs linking higher CR to lower HAND prevalence using other measures. In a US study, Foley et al. (24) used a CR composite score consisting of years of education and word-reading ability, and reported that higher CR appeared protective for HAND in older people with HIV (24).
We used years of formal education as a proxy CR measure. . In contrast, CRI-Education includes completed adult education or training. We found that years of formal education was independently associated with HAND, whereas CRI-Education was not. This suggests that childhood education may be more important in this setting although interestingly, measures of childhood disadvantage such as head circumference and femur length were not associated with HAND. Low education may not necessarily be a marker of childhood disadvantage in some LMIC settings and may instead reflect historical lack of access to schooling in rural areas(42).
We specifically found that ‘no education’ was independently associated with HAND despite use of a low-literacy battery and locally derived norms. This suggests that even elementary education may protect against HAND in this setting. The association between HAND and CR measured through educational level is widely reported, but most HIC studies have identified this relationship in settings with markedly higher educational background than we report here. The USA MACS(20) is a well-known study reporting that low education (<12 years versus >12 years) was a risk factor for HAND. Similarly, Milanini et al. (25) reported that CRI-Education was independently associated with HAND(25) in another US cohort with a median 12 years education. A similar relationship has been found in a higher-educated Zambian cohort (mn 10 years education, mn age 41 years), not assessed for HAND but where lower education was associated with deficits in executive functioning, learning and processing speed. Whilst education as a predictor of CR has consistently been shown to be a protective factor for HAND in well-educated populations, our findings suggest a similar association in a population with a much lower level of formal education. It may be that even a limited amount of early education may contribute to CR, possibly by strengthening neural pathways that are protective in later life (43).
Similarly, other HIC studies are difficult to compare as they include younger individuals with advanced HIV. For example, De Ronchi et al. also reported that lower educational level was associated with HAND(21) in a cohort where most participants were younger than 28 years and many were either symptomatic or had a CD4 count of less than 500.
In this setting, the median CRI-Total was 97.3, compared with 114 reported in the study by Milanini et al. that included participants of a similar age(25). Although we found a consistent association between education and HAND, we did not find the same association for other potential CR measures. CRI-Work and CRI-Leisure scores were higher in participants without HAND, but differences were not statistically significant. In contrast a number of previous studies include occupational level in CR composite measures, and demonstrate less cognitive impairment in those with higher CR(22, 23, 25, 26). It may be that work and leisure are less valid differentiators of CR in this setting. In Tanzania, it is usual to work into old age in farming and traditional occupations, with many farming alongside another occupation. It is unusual to be unemployed. The majority of our cohort were occupied in farming or low skilled manual work (61.8% and 68.2% respectively). In this setting where many had similar occupations, the effect on CR may be less. The original CRIq categories may therefore have some cultural limitations, and has not been validated in Africa.
Milanini et al. also failed to find an independent effect of CRI-Work on HAND, despite reporting independent associations with CRI-Total, CRI-Education, and CRI-Leisure. It is possible that occupational complexity has less impact than other CR factors in an older population.
Similarly, in our setting CRI-Leisure was not associated with HAND, in contrast to HIC data (25). Leisure activities are widely used as a proxy for CR in other neurological disorders (44, 45). It seems likely that occupation and leisure activities are poor measures of CR in this population, due to sociocultural and economic factors. We also evaluated other potentially more appropriate measures such as formal community roles but found no association. Overall, the effect of education and particularly formal childhood education appears more robust than other proxy CR measures in this setting.
HIV disease factors were not associated with HAND, which is potentially surprising, given the association between nadir CD4 count (the ‘legacy effect’) and HAND shown in multiple studies(12). Legacy effect and HIV disease control may be less important in HAND risk in older age groups who may have increased comorbidities, and risk factors and where HIV disease is well-controlled. Similar findings were reported by Patel et al, who found that CR was the most robust predictor for cognitive impairment, above HIV disease severity markers in a study of multiple HAND risk factors(27).
Previous studies in the Kilimanjaro region of Northern Tanzania, completed by our team and collaborators report similar findings. Although we did not investigate cognitive reserve per se, we found that using a similar (but not identical) methodology,. illiteracy and social isolation were independent predictors of symptomatic HAND and most HIV-disease factors were not (46). A major limitation of that study was that HIV viral load was not locally-available at the time of data collection (2016). In a subsequent 2019 study, lower education was associated with HAND, but HIV disease measures were not, including HIV viral load and nadir CD4 count. (32). These cohorts were educationally-similar, with the majority completing primary school but no further/higher education. Taken together, these findings suggest that CR may be a more important predictor of HAND than HIV disease severity in the older SSA population in the post-cART era.
Although associated with overall HAND, CRI-Total was not associated with symptomatic HAND (MND/HAD). Comparison with existing data is challenging. Most previous studies of CR and HAND have not assessed functional impairment (20-24, 27, 28). Only one HIC study to date has looked at CR in relation to symptomatic HAND (26) reporting that CR scores were lower in individuals with symptomatic HAND relative to those with asymptomatic HAND. This is a HIC study with a predominantly male sample and therefore difficult to compare directly to our cohort (47).
Functional impairment is challenging to measure and heavily reliant on subjective measures such as self and informant-report.
Consequently, symptomatic HAND may be under or over-estimated. It is possible that high CR does not prevent manifestation of functional deficits, but may delay or prevent cognitive impairment. Lack of formal education was significantly associated with symptomatic HAND. This finding supports evidence from Morgan et al. showing CR to be protective for syndromic HAND(26). They included years of education as part of a composite CR score. As CRI-Total showed no association with symptomatic HAND, this might suggest that education is a more powerful predictor of CR in this setting than the CRIq. Further studies are required as the relationship between CR and functional impairment is unclear.
Limitations
Our sample included only people actively attending clinic follow-up leading to potential bias towards individuals motivated to attend healthcare appointments.
There were also limitations in our operationalised HAND diagnostic process. Participants were not independently evaluated for cognitive and neurological impairment by a specialist. Although we excluded a small number of individuals where there was a clear alternative diagnosis such as severe depression, delirium or stroke, it is possible that other individuals were labelled as meeting HAND criteria where another diagnosis may have been made following detailed specialist evaluation. This could not be avoided due to both resource limitations and the restrictions posed by the COVID-19 pandemic. Similarly, determination of functional impairment as in other studies relied heavily on self and informant report and might therefore be susceptible to bias. We elected not to exclude individuals unable to provide an informant, to increase generalisability but in those cases identification of functional impairment relied on self report and clinician assessment only.
We selected the CRIq because it was being utilised by collaborators in Malaysia and we were not able to identify a validated measure for Africa. Although we discussed and minimally adapted the measure with local clinicians prior to use in the study, the CRIq questions may not be appropriate measures of CR in an African setting.
Background demographic and clinical data to compare to our systematically selected sample were only made available to us in MZRH as the lead site. Whilst we were able to demonstrate that this background population was demographically similar to our sample, we were not able to obtain equivalent data from other hospitals involved to ascertain whether samples were representative at all sites.