Risk Tool Helps Discern When Complaints about Memory Matter

by Kristina Fiore for MedPage Today:A biomarker-based personalized risk estimate may help predict which patients with subjective cognitive decline (SCD) will progress to dementia, Dutch researchers reported here.A model that takes into account age, Mini Mental State Exam (MMSE) scores, MRI brain volume, and cerebrospinal fluid (CSF) biomarkers had good prognostic performance (C-statistic of 0.82) in a cohort of patients who visited a memory clinic in the Netherlands, reported Ingrid van Maurik, PhD, of VU Medical Center in Amsterdam, at the Alzheimer's Association International Conference."About 25% of the patients at our memory clinic are diagnosed with subjective cognitive decline. But what do those results mean for our patients? They could be a precursor to future clinical progression," van Maurik said. "Our aim is to use biomarkers to make an individualized risk prediction."SCD has been an evolving concept over the last few years, as researchers have struggled with how best to classify patients who complain that their memory -- or some other cognitive aspect -- isn't what it used to be. The idea has also taken hold as researchers need to recruit patients into clinical trials at earlier stages of the disease.Frequently, these complaints will not develop into anything serious. But predicting which patients will experience problems has been a challenge. van Maurik's group focused on the ABIDE project to better understand which of their memory clinic patients with SCD would progress to dementia. For the current study, they included 481 patients who had at least 6 months of follow-up.Their prognostic model included age, MMSE scores, whole-brain and hippocampal volume on MRI, and CSF levels of amyloid beta and total tau. They calculated the probability of progression for each patient within 1, 3, and 5 years.A total of 70 patients progressed to mild cognitive impairment (MCI) or dementia during follow-up.Van Maurik reported that age and MMSE scores alone were a decent predictor of progression, with a C-statistic of 0.70. But when adding CSF levels of amyloid and tau, that figure rose to 0.82."Age and MMSE had good prognostic performance ... but when you add CSF amyloid and tau, the C-stat is way better," she said.MRI whole-brain and hippocampal volumes, however, did not offer anything valuable in terms of risk prediction, indicating "that what you see on MRI may be more age-driven than disease driven," van Maurik said.CSF tau appeared to be particularly useful in predicting progression in younger patients (ages≤ 65), she said: "For young patients, when you have low tau, you have a low risk of progression even despite an abnormal amyloid beta. For older patients, the risk of progression is driven by amyloid beta."Van Maurik cited three examples of the predictive model at work in their clinic. The first was a 61-year-old man who presented in 2008. He had an MMSE of 27 and a normal MRI, but an abnormal CSF with low amyloid and high tau. The model determined him to be at high risk and during follow-up he was diagnosed with probable Alzheimer's disease.In a second case, a 57-year-old man in 2009 had an MMSE of 29 and a family history of dementia. He had mild hippocampal atrophy on MRI but normal CSF readings. He was determined to be at low risk for progression and he still has not progressed.Finally, a 52-year-old male with an MMSE of 25, normal MRI, and normal CSF levels was predicted to be at low risk and has not progressed over 4 years.Van Maurik noted that the majority of patients could be assessed to be at high or low risk, but 12% of the cohort was inconclusive because their risk was between 30% and 60%. And overall, the majority of cases were at very low risk of progressing.She said her group is building an app similar to one that they already have in place for predicting which patients with MCI will progress to full dementia. The idea is to develop something that can easily be used by "local professionals," so they used the easy-to-administer MMSE rather than a stronger battery of cognitive tests.Frank Jessen, MD, of University Hospital Cologne in Germany, who has been a driving force in defining and establishing criteria for SCD, noted during the session that the risks must be validated in an external cohort. van Maurik agreed, given that the study cohort was a younger and more highly educated sample.When asked by MedPage Today whether amyloid and tau PET imaging should be incorporated into such a predictive model, Jessen noted many patients in the sample were enrolled before such imaging was more widely available. Also, some work suggests CSF and PET imaging may tell similar stories about risk, he said.Nonetheless, "I think you could do the same thing with PET imaging, and I imagine this could even improve the prediction," said Jessen, who was not involved in the study.Source: http://bit.ly/2uhYTSK