Predictive models of multiple sclerosis-related cognitive performance using routine clinical practice predictors

BACKGROUND

The application of machine learning (ML) to predict cognitive evolution is exceptionally scarce. Computer-based self-administered cognitive tests provide the opportunity to set up large longitudinal datasets to aid in developing ML prediction models of risk for Multiple Sclerosis-related cognitive decline.

OBJECTIVE

to analyze to what extent clinically feasible models can be built with standard clinical practice features and subsequently used for reliable prediction of cognitive evolution.

METHODS

This prospective longitudinal study includes 1184 people with MS who received a Processing Speed (PS) evaluation at 12 months of follow-up measured by the iPad®-based Processing Speed Test (PST). Six of the most potent classification models built with routine clinical practice features were trained and tested to predict the 12-month patient class label (PST worsening (PSTw) versus PST stable). A rigorous scheme of all the preprocessing steps run to obtain reliable generalization performance is detailed.

RESULTS

Based on a 12-month reduction of 10% of the PST raw score, 187/1184 (15.8%) people with MS were classified as PSTw. The trees-based models (random forest and the eXtreme Gradient Boosting) achieved the best performance, with an area under the receiver operating characteristic curve (AUC) of 0.90 and 0.89, respectively. The timing of high-efficacy disease-modifying therapies (heDMTs) was identified as one of the top importance predictors in all the models evaluated.

CONCLUSION

Using trees-based machine learning models to predict individual future information processing speed deterioration in MS could become a reality in clinical practice.

The effect of timing of high-efficacy therapy on processing speed performance in multiple sclerosis

BACKGROUND

The potential influence of the timing of high-efficacy disease-modifying therapies (heDMTs) on processing speed (PS) performance is critically lacking in current literature.

OBJECTIVE

To assess the extent to which early commencement of heDMTs would be associated with a better PS evolution as compared to moderate efficacy disease-modifying therapies (meDMTs) and delayed commencement of heDMTs.

METHODS

In this ongoing prospective longitudinal study, the 695 MS patients that have received a PS evaluation at 12-month of follow-up measured by the iPad®-based Processing Speed Test (PST) were retained for the analysis. All patients who had ever been prescribed a high efficacy disease-modifying therapy (heDMT) were classified in tertiles according to the proportion of their disease duration that had been on heDMTs. Based on these tertiles and the time to the first heDMT from the disease onset, patients were divided into the early heDMT group and the delayed heDMT group. Between-group differences in mean PST standardized (Z-score) change from baseline were analyzed using a linear mixed model.

RESULTS

In the multivariable model, each year of delay in starting a heDMT was associated with increased odds of cognitive worsening at 12-month (OR = 1.0324, 95% CI = 1.014-1.062, p<0.05). MeDMT-treated patients were at a significantly higher risk for cognitive worsening than early heDMT patients (OR= 2.57, 95%CI = 1.02-6.17). Linear mixed model-based adjusted mean change in PST Z-score from baseline was significantly better in those patients with the longest proportion of their disease duration treated with heDMT (highest tertile) compared to the lowest tertile (difference 0.37 [95%CI 0.02-0.92;p=0.036) and medium tertile (difference 0.39 [95%CI 0.06-0.31;p=0.037).

CONCLUSION

Early he-DMT-treated patients are at significantly lower risk for cognitive worsening. Early administration of heDMTs is associated with greater cognitive functioning improvements than delayed commencement or meDMTs.

Comorbidity effect on processing speed test and MRI measures in multiple sclerosis patients

BACKGROUND

Comorbid conditions are known to affect the clinical course of multiple sclerosis (MS). Our objective was to determine the impact of comorbidities on the processing speed test (PST).

METHODS

We conducted a retrospective, longitudinal analysis of all patients who completed PST testing from June 2015 - August 2019 at our center. Our electronic medical record was queried to determine the presence of the following comorbidities: diabetes mellitus (DM), hypertension (HTN), hyperlipidemia (HLD), coronary artery disease, and depression. To help address baseline PST performance and practice effect, patients were also divided into four quartiles by baseline PST scores. Brain MRIs obtained within a 90-day window from the initial clinical assessment were quantitatively analyzed via fully-automated methods to calculate whole brain fraction (WBF), T2 lesion volume (T2LV), gray matter fraction (GMF), and thalamic volume (TV). Univariable and multivariable linear regression models were used to determine the relationship between the comorbidities, PST performance and MRI metrics over time.

RESULTS

A total of 4,344 patients (mean age 49.5 ± 12.4 years, 72.3% female, and 63.7% relapsing remitting MS) were included in the analysis with 13,375 individual patient encounters. Over half the cohort (52.4%) suffered from at least one comorbidity with the most common being depression (37.4%), HLD (20.9%), HTN (19.6%), and DM (6.4%). Patients with one or more comorbidity had lower baseline PST scores. Longitudinally, patients with two comorbidities lost 1.46 points on the PST per year relative to those with no comorbidities (95% CI -2.46 - -0.46, p = 0.004). Individuals with depression had lower PST scores than those without, and this difference persisted over time (β = -2.40, 95% CI -3.08 - -1.73, p < 0.001). At baseline, HLD patients had higher PST scores than non-HLD patients (β = 1.10, 95% CI 0.15 - 2.05, p = 0.022), but this difference did not remain over time. Individuals in the highest PST performance quartile were negatively impacted when diagnosed with depression, HTN, and DM relative to those without the comorbidities. There were no other correlations with PST scores and the remaining comorbidities. Depression was associated with lower baseline WBF (β = -0.0043, 95% CI -0.0084 - -0.0003, p = 0.033) and GMF (β = -0.0046, 95% CI -0.0078 - -0.0015, p = 0.004) along with larger T2LV (β = 0.1605, 95% CI 0.0082 - 0.3128, p = 0.039). HLD patients had more favorable baseline MRI measures, including higher WBF (β = 0.0076, 95% CI 0.0017 - 0.0135, p = 0.012) and TV (β = 0.0002, 95% CI 0.0000 - 0.0005, p = 0.041), with a lower T2LV (β = -0.2963, 95% CI -0.5219 - -0.0706, p = 0.010).

CONCLUSIONS

Comorbidities are common within a MS cohort and adversely impact processing speed. Depression adversely impacted PST scores with worse MRI outcomes. HLD was associated with lower longitudinal PST measures but favorable quantitative MRI metrics. MS patients with faster baseline processing speeds were most sensitive to comorbid conditions. Our findings suggest a complex interplay between cognition and comorbid conditions in MS patients.