Classification models for neurocognitive impairment in HIV infection based on demographic and clinical variables

Machine learning prediction models for clinical management of blood-borne viral infections: a systematic review of current applications and future impact

BACKGROUND

Machine learning (ML) prediction models to support clinical management of blood-borne viral infections are becoming increasingly abundant in medical literature, with a number of competing models being developed for the same outcome or target population. However, evidence on the quality of these ML prediction models are limited.

OBJECTIVE

This study aimed to evaluate the development and quality of reporting of ML prediction models that could facilitate timely clinical management of blood-borne viral infections.

METHODS

We conducted narrative evidence synthesis following the synthesis without meta-analysis guidelines. We searched PubMed and Cochrane Central Register of Controlled Trials for all studies applying ML models for predicting clinical outcomes associated with hepatitis B virus (HBV), human immunodeficiency virus (HIV), or hepatitis C virus (HCV).

RESULTS

We found 33 unique ML prediction models aiming to support clinical decision making. Overall, 12 (36.4%) focused on HBV, 10 (30.3%) on HCV, 10 on HIV (30.3%) and two (6.1%) on co-infection. Among these, six (18.2%) addressed the diagnosis of infection, 16 (48.5%) the prognosis of infection, eight (24.2%) the prediction of treatment response, two (6.1%) progression through a cascade of care, and one (3.03%) focused on the choice of antiretroviral therapy (ART). Nineteen prediction models (57.6%) were developed using data from high-income countries. Evaluation of prediction models was limited to measures of performance. Detailed information on software code accessibility was often missing. Independent validation on new datasets and/or in other institutions was rarely done.

CONCLUSION

Promising approaches for ML prediction models in blood-borne viral infections were identified, but the lack of robust validation, interpretability/explainability, and poor quality of reporting hampered their clinical relevance. Our findings highlight important considerations that can inform standard reporting guidelines for ML prediction models in the future (e.g., TRIPOD-AI), and provides critical data to inform robust evaluation of the models.

Development of a Nomogram-Based Tool to Predict Neurocognitive Impairment Among HIV-positive Charter Participants

Background:

Despite the widespread use of combination antiretroviral therapy (cART), HIV-associated neurocognitive impairment (NCI) persists in people living with HIV (PLWH). Studies have generated inconsistent results regarding etiological factors for NCI in PLWH. Furthermore, a user-friendly and readily available predictive tool is desirable in clinical practice to screen PLWH for NCI.

Objective:

This study aimed to identify factors associated with NCI using a large and diverse sample of PLWH and build a nomogram based on demographic, clinical, and behavioral variables.

Methods:

We performed Bayesian network analysis using a supervised learning technique with the Markov Blanket (MB) algorithm. Logistic regression was also conducted to obtain the adjusted regression coefficients to construct the nomogram.

Results:

Among 1,307 participants, 21.6% were neurocognitively impaired. During the MB analysis, age provided the highest amount of mutual information (0.0333). Logistic regression also showed that old age (>50 vs. ≤50 years) had the strongest association (OR=2.77, 95% CI=1.99-3.85) with NCI. The highest possible points on the nomogram were 626, translated to a nomogram-predicted probability of NCI to be approximately 0.95. The receiver operating characteristic (ROC) curve's concordance index was 0.75, and the nomogram's calibration plot exhibited an excellent agreement between observed and predicted probabilities.

Conclusion:

The nomogram used variables that can be easily measured in clinical settings and, thus, easy to implement within a clinic or web-interface platform. The nomogram may help clinicians screen for patients with a high probability of having NCI and thus needing a comprehensive neurocognitive assessment for early diagnosis and appropriate management.

Predicting occupational outcomes from neuropsychological test performance in older people with HIV

OBJECTIVE

The ability to work is amongst the top concerns of people living with well treated HIV. Cognitive impairment has been reported in many otherwise asymptomatic persons living with HIV and even mild impairment is associated with higher rates of occupational difficulties. There are several classification algorithms for HIV-associated neurocognitive disorder (HAND) as well as overall scoring methods available to summarize neuropsychological performance. We asked which method best explained work status and productivity.

DESIGN

Participants (N = 263) drawn from a longitudinal Canadian cohort underwent neuropsychological testing.

METHODS

: Several classification algorithms were applied to establish a HAND diagnosis and two summary measures (NPZ and Global Deficit Score) were computed. Self-reported work status and productivity was assessed at each study visit (four visits, 9 months apart). The association of work status with each diagnostic classification and summary measure was estimated using logistic regression. For those working, the value on the productivity scale was regressed within individuals over time, and the slopes were regressed on each neuropsychological outcome.

RESULTS

The application of different classification algorithms to the neuropsychological data resulted in rates of impairment that ranged from 28.5 to 78.7%. Being classified as impaired by any method was associated with a higher rate of unemployment. None of the diagnostic classifications or summary methods predicted productivity, at time of testing or over the following 36 months.

CONCLUSION

Neuropsychological diagnostic classifications and summary scores identified participants who were more likely to be unemployed, but none explained productivity. New methods of assessing cognition are required to inform optimal workforce engagement.

International HIV Dementia Scale for HIV-Associated Neurocognitive Disorders: A Systematic Review and Meta-Analysis

The present study aims to systematically review the evidence on the accuracy of the International HIV Dementia Scale (IHDS) test for diagnosing human immunodeficiency virus (HIV)-associated neurocognitive disorders (HAND) and outline the quality and quantity of research evidence available on the accuracy of IHDS in people living with HIV. We conducted a systematic literature review, searching five databases from inception until July 2020. We extracted dichotomized positive and negative test results at various thresholds and calculated the sensitivity and specificity of IHDS. Quality assessment was performed according to the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) criteria. Fifteen cross-sectional studies, published between 2011 and 2018, met the inclusion criteria for meta-analysis. Overall, 3760 patients were included, but most studies recruited small samples. We assessed most studies as being applicable to the review question, though we had concerns about the selection of participants in three studies. The accuracy of IHDS was investigated at thirteen cut-off points (scores 6-12). The threshold of 10 is the most useful for optimal HAND screening (including asymptomatic neurocognitive disorder, symptomatic HAND, and HIV-associated dementia) with fair diagnostic accuracy.

HIV-associated neurocognitive impairment in stable people living with HIV on ART in rural Tanzania

OBJECTIVES

Few studies have assessed cognitive impairment among healthy people living with HIV (PLWH) who are stable on antiretroviral treatment (ART) in sub-Saharan Africa.

METHODS

We conducted a cross-sectional study among a random sample of stable adult PLWH from rural Tanzania on ART for more than 1 year and without immunological failure or pre-existing neurological disease. We evaluated the prevalence and risk factors for neurocognitive impairment (NCI), assessed through neuropsychological tests, functional and depression questionnaires and defined as a mean Z-score ≤ -1 in two or more cognitive domains.

RESULTS

Among 243 participants [median age = 44.3 years (interquartile range: 36-52] and 71% female] we found a rate of NCI of 19.3% (95% confidence interval: 14.8-24.8%). Memory and psychomotor domains demonstrated the highest impairment. Independent predictors of NCI were age and self-reported alcohol use. Other classical risk factors were not associated with HIV-associated NCI.

CONCLUSION

Despite effective ART roll-out, NCI remained a prevalent condition in this healthy rural Tanzanian population of PLWH on ART. Age and alcohol use were key risk factors.

Significant differences between verbal and non-verbal intellectual scales on a perinatally HIV-infected cohort: from pediatrics to young adults

Delayed neurodevelopment is a common outcome in perinatally HIV-infected children. Our aim was to assess the intellectual profile of our cohort, considering both the infection and socio-environmental related variables. A cross-sectional cohort study was undertaken at seven major hospitals in Spain belonging to the CoRISpeS cohort (n = 97). Patients were followed up according to a standard protocol. Intellectual measures, psychosocial profile and HIV infection-related data have been analysed. The average patient age was 15 years. The median CD4 cell percentage was 35% (1,59). Viral load was undetectable in 80% of the patients and 27% were on AIDS category; 38% of whom had encephalopathy. The average composite score of both crystallized intelligence (CI) and intelligence quotient (IQ) for the cohort was lower than that of the general population (p < 0.001). Results revealed a significant difference of 38% between crystallized and fluid intelligence. There was a clear association between IQ and age of diagnosis (p = 0.022); CI and CDC classification (p = 0.035), CD4 count (p = 0.011) and CD4 nadir (p = 0.001). Higher parental education was associated with better performance across all intelligence scales (p < 0.002). A regression model showed that CI was influenced by the academic level of caregivers (p = 0.002), age at start of cART (p = 0.050) and primary language (p = 0.058). Findings revealed significant differences in verbal and non-verbal intellectual scales resulting in a misleading IQ Composite score. Crystallized intelligence demonstrated the highest level of impairment despite adequate treatment and good immunovirological status, while fluid intelligence results were average. Caregiver level of education was the strongest factor across all intelligence measures.

COMT Val158Met Polymorphism, Cardiometabolic Risk, and Nadir CD4 Synergistically Increase Risk of Neurocognitive Impairment in Men Living With HIV

OBJECTIVE

The Val allele of the Val158Met single-nucleotide polymorphism of the catechol-o-methyltransferase gene (COMT) results in faster metabolism and reduced bioavailability of dopamine (DA). Among persons living with HIV, Val carriers display neurocognitive deficits relative to Met carriers, presumably due to exacerbation of HIV-related depletion of DA. COMT may also impact neurocognition by modulating cardiometabolic function, which is often dysregulated among persons living with HIV. We examined the interaction of COMT, cardiometabolic risk, and nadir CD4 on neurocognitive impairment (NCI) among HIV+ men.

METHODS

Three hundred twenty-nine HIV+ men underwent COMT genotyping and neurocognitive and neuromedical assessments. Cohort-standardized z scores for body mass index, systolic blood pressure, glucose, triglycerides, and high-density lipoprotein cholesterol were averaged to derive a cardiometabolic risk score (CMRS). NCI was defined as demographically adjusted global deficit score of ≥0.5. Logistic regression modeled NCI as a function of COMT, CMRS, and their interaction, covarying for estimated premorbid function, race/ethnicity, and HIV-specific characteristics. Follow-up analysis included the 3-way interaction of COMT, CMRS, and nadir CD4.

RESULTS

Genotypes were 81 (24.6%) Met/Met, 147 (44.7%) Val/Met, and 101 (30.7%) Val/Val. COMT interacted with CMRS (P = 0.02) such that higher CMRS increased risk of NCI among Val/Val [odds ratio (OR) = 2.13, P < 0.01], but not Val/Met (OR = 0.93, P > 0.05) or Met/Met (OR = 0.92, P > 0.05) carriers. Among Val/Val, nadir CD4 moderated the effect of CMRS (P < 0.01) such that higher CMRS increased likelihood of NCI only when nadir CD4 <180.

DISCUSSION

Results suggest a tripartite model by which genetically driven low DA reserve, cardiometabolic dysfunction, and historical immunosuppression synergistically enhance risk of NCI among HIV+ men, possibly due to neuroinflammation and oxidative stress.

Neurocognitive and quality of life study in perinatally HIV-infected young people and their peers. NeuroCoRISpeS study

BACKGROUND

Assessing the role of HIV and non-HIV related factors is essential for a better understanding of the neurocognitive outcomes in perinatally HIV-infected (PHIV+) young people. The aim of our study was to assess cognition and quality of life (QoL) of a PHIV+ cohort of young people and to compare it with a control group.

METHODS

Thirty PHIV+ and 30 HIV(-) healthy young adults matched by age, sex and socioeconomic status completed a protocol that included neurocognitive tests, a psychosocial semi-structured interview and a QoL questionnaire (PedsQL). Neurocognitive domain-specific and domain-general (NPZ-5) Z-scores were calculated. CDC AIDS-defining category C or not C (PHIV+/C, PHIV+/noC) was considered to evaluate differences within the PHIV+ group. Univariate and multivariate analysis were performed.

RESULTS

Sixty patients were included; 67% were female; median age (IQR) 19 years (18-21). Regarding PHIV+ young people, 27% showed CDC C category (none encephalopathy), 93% were on ART and 77% had undetectable viral load. No differences regarding occupation were found, although the HIV(-) group repeated less grades (p=0.028) and had a higher education level (p=0.021). No differences were found between PHIV+/noC and HIV(-) participants. However, the PHIV+/C group showed poorer performance than PHIV+/noC (NPZ-5, p=0.037) and HIV(-) subjects (crystallised intelligence, p=0.025; intelligence quotient, p=0.016). Higher nadir CD4+ T-cell count was related to better Z-score in memory (p=0.007) and NPZ-5 (p=0.025). Earlier and longer exposure to ART resulted in better performance in memory (p=0.004) and executive functions (p=0.015), respectively.

CONCLUSIONS

No significant differences were found in the neurocognitive profile nor QoL between PHIV+/noC and HIV(-) adolescents; however, PHIV+/C participants obtained lower scores. The use of longer and earlier ART seems to have a beneficial effect.

Montreal Cognitive Assessment (MoCA) for HIV-Associated Neurocognitive Disorders

This study aims to systematically review the evidence on the accuracy of the Montreal Cognitive Assessment (MoCA) test for diagnosing HIV-associated neurocognitive disorders (HAND) and to outline the quality and quantity of research evidence available about the accuracy of MoCA in populations infected with HIV. We conducted a systematic literature review, searching five databases from inception until January 2019. We extracted dichotomized positive and negative test results at various thresholds and calculated the sensitivity and specificity of MoCA. Quality assessment was performed according to the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) criteria. Eight cross-sectional studies met the inclusion criteria for meta-analysis. Overall, 1014 patients were included but most studies recruited small samples. Recruitment period ranged from 2009 to 2015. We assessed most studies as being applicable to the review question though we had concerns about the selection of participants in three studies. The accuracy of MoCA for diagnosing HAND was reported at six cut-off points (scores 22-27). The MoCA test provides information about general cognitive functioning disturbances that contribute to a diagnosis of HAND. A lower threshold than the original cut-off of 26 is probably more useful for optimal screening of HAND, as it lowers false positive rates and improves diagnostic accuracy. Nonetheless, the choice of cut-off always comes with a sensitivity-specificity trade-off, the preferred cut point depending on whether sensitivity or specificity is more valuable in a given context.

A survey of machine learning applications in HIV clinical research and care

A wealth of genetic, demographic, clinical and biomarker data is collected from routine clinical care of HIV patients and exists in the form of medical records available among the medical care and research communities. Machine learning (ML) methods have the ability to identify and discover patterns in complex datasets and predict future outcomes of HIV treatment. We survey published studies that make use of ML techniques in HIV clinical research and care. An advanced search relevant to the use of ML in HIV research was conducted in the PubMed biomedical database. The survey outcomes of interest include data sources, ML techniques, ML tasks and ML application paradigms. A growing trend in application of ML in HIV research was observed. The application paradigm has diversified to include practical clinical application, but statistical analysis remains the most dominant application. There is an increase in the use of genomic sources of data and high performance non-parametric ML methods with a focus on combating resistance to antiretroviral therapy (ART). There is need for improvement in collection of health records data and increased training in ML so as to translate ML research into clinical application in HIV management.

HIV-Associated Neurocognitive Disorders: A Global Perspective

The present review on HIV-associated neurocognitive disorders (HAND) provides a worldwide overview of studies that have investigated the rate and neuropsychological (NP) profile of HAND research since the inception of the 2007 HAND diagnostic nomenclature. In the first part, the review highlights some of the current controversies around HAND prevalence rates. In the second part, the review critically assesses some solutions to move the field forward. In the third part, we present the cross-sectional NP profile in non-Western HIV+ cohorts and in relation to Western cohorts' findings. The adopted global perspective highlights the successful expansion of NP studies in HIV infection to culturally diverse low- to medium-income countries with high HIV burden. These studies have produced interestingly similar rates of HAND whether patients were naïve or treated and/or virally suppressed compared to the rich income countries where the NP research in NeuroHIV has originated. The perspective also demonstrates that globally, the group which is the most representative of the HIV epidemic, and thus at risk for HAND are persons with chronic HIV infection and survivors of past immunosuppression, while in relative terms, those who have been treated early with long-term viral suppression represent a minority. In the last part, we present a review of the naturalistic longitudinal NP global studies in HIV+cohorts, discuss the role of longitudinal design in solving issues around the question of asymptomatic neurocognitive impairment, and the question of biomarker discovery. Finally, we conclude by calling for greater methods and data harmonization at a global level. (JINS, 2017, 23, 860-869).