Parkinson’s patients can develop more than motor difficulties—learning and memory can fail as well, but not always. That creates a dilemma for trials of drugs that might stave off cognitive loss in PD—researchers want to enroll only those patients at the highest risk for dementia. A June 16 Lancet Neurology paper proposes a way. Scientists led by Clemens Scherzer, Harvard Medical School, created a cognitive risk score (CRS) that combines seven individual factors previously proposed to predict cognitive decline. In nine cohorts, the CRS picked out who would lose cognitive function and who would stay in the normal range within 10 years of PD diagnosis. The score could trim sample sizes for trials, thereby cutting down on time and cost, claim the researchers.
“Some patients have slow cognitive decline and add a lot of noise to the data,” Scherzer told Alzforum. “If we target trials to patients with a high CRS, then we have much stronger power to see whether a novel drug is effective.”
“A single cognitive risk score that incorporates multiple factors could be of great practical value in quickly determining those PD patients on the path to future cognitive decline,” Daniel Weintraub, University of Pennsylvania, Philadelphia, wrote to Alzforum. He praised both the large sample size and inclusion of both clinical and genetic data.
Downward Trend. Cognition as measured by the MMSE (left) and more sensitive MoCA (right) falters more quickly in PD patients with a high cognitive risk score (red) than in people with low CRS (blue). [Liu et al., Lancet Neurol.]
Past studies proposed a number of factors that predict decline in PD patients, including older age at disease onset, depression, more severe disease at baseline, and fewer years of education (see Zhu et al., 2014; Aarsland et al., 2001). Others have linked mutations in β-glucocerebrosidase (GBA) to cognitive decline in PD (Liu et al., 2016; Cilia et al., 2016). However, these studies were mostly short and/or based on relatively small populations. Scherzer and colleagues wondered if they could combine those risk factors in thousands of patients to better predict who would develop cognitive impairment and dementia.
First author Ganqiang Liu and colleagues compiled a discovery population of 1,350 PD patients to derive the score. These patients came from six cohorts: the Harvard Biomarker Study (HBS); Cambridgeshire Parkinson’s Incidence from GP to Neurologist (CamPaIGN); Parkinsonism: Incidence, Cognition, and Non-motor heterogeneity in Cambridgeshire (PICNICS); Drug Interaction with Genes in Parkinson’s Disease (DIGPD); the PROfiling PARKinson’s disease (PROPARK) study; and the Parkinson’s Disease Biomarkers Program (PDBP).
The authors retrospectively analyzed longitudinal data that had been collected for up to 12 years for each patient. All had been cognitively normal at diagnosis. The researchers took note of whether a person developed cognitive decline (defined as a score of 25 or below on the MMSE) or dementia within 10 years of their diagnosis. They then calculated which combination of the following factors best predicted that loss: age at onset, years of education, baseline MMSE, MDS-UPDRS II and MDS-UPDRS III motor scores, Hoehn and Yahr scale of motor symptoms, GBA mutation status, depression, and gender.
Seven of the nine factors were predictive to varying degrees. Hoehn and Yahr stage and MDS-UPDRS II gave no added predictive value and were excluded. The group calculated the CRS in part by assigning a score to each risk variable and then multiplying that by a coefficient derived from Cox proportional hazards analysis to account for the relative strength of each risk factor. They incorporated the sum of these products into their final calculation, deriving a CRS scale that ranged from zero to 1, with 1 being the likeliest to develop dementia within 10 years.
In the discovery cohort, 75.3 percent remained cognitively intact and 24.7 percent declined. A CRS of 0.196 seemed to be the optimal threshold; most with a score of 0.196 or below stayed normal, while those above 0.196 tended to deteriorate. This cutoff predicted who would decline with a specificity and sensitivity of 72 and 87 percent, respectively.
The CRS worked similarly in a replication cohort consisting of 1,123 patients from three studies: the Deprenyl and Tocopherol Antioxidative Therapy of Parkinsonism (DATATOP); the Parkinson Research Examination of CEP-1347 Trial/A Longitudinal Follow-up of the PRECEPT Study Cohort (PreCEPT/PostCEPT); and the Parkinson’s Progression Marker Initiative (PPMI). Of the combined cohort, 26 percent declined on the MMSE, and the 0.196 cutoff predicted this with 85 percent accuracy. Specificity and sensitivity were 74 and 73 percent, respectively.
According to the authors’ power analysis, focusing on the high CRS scores would reduce by sixfold the sample size needed for a three-year clinical trial of a drug that prevents cognitive decline in Parkinson’s (see image at right). Scherzer noted that the CRS predicted global cognitive decline even for time periods of one year. If more sensitive readouts than the MoCA were used to track cognition in trials, those targeted to CRS-high patients could become even faster, he said.
The full CRS scale may not work for everyone, since GBA genotypes are not always available, but a clinical-only score option worked almost as well. Weintraub wrote that he would have liked to see more biomarkers included in the score, namely those from cerebrospinal fluid and functional or structural imaging. The authors pointed out that while Aβ42 in the cerebrospinal fluid or brain can strengthen predictions, they require either an invasive lumbar puncture or an expensive and less widely available and PET scan.
The CRS will be useful, possibly immediately, in clinical trials, Scherzer said. In fact, pharmaceutical companies and academic investigators are already working on incorporating this score into their trial designs, he noted. However, it’s not yet ready for use in routine clinical care, he added, because the data needs to be replicated in a large prospective trial.
His group plans to refine the score, adding other predictive genetic variables besides GBA variants. They are combing PD patient genomes and epigenomes for such factors. Meanwhile, they have made available a beta version of a CRS calculator for researchers who may want to select high-risk patients for research studies.
“Anything we can do to find a population more likely to decline so that we could better see the benefit of a drug would be valuable,” said Warren Olanow, Mount Sinai Hospital, New York. This is a great start, he continued. “Maybe we can find ways to explore this data further and predict who would decline at an even faster rate.”—Gwyneth Dickey Zakaib
- Zhu K, van Hilten JJ, Marinus J. Predictors of dementia in Parkinson's disease; findings from a 5-year prospective study using the SCOPA-COG. Parkinsonism Relat Disord. 2014 Sep;20(9):980-5. Epub 2014 Jun 26 PubMed.
- Aarsland D, Andersen K, Larsen JP, Lolk A, Nielsen H, Kragh-Sørensen P. Risk of dementia in Parkinson's disease: a community-based, prospective study. Neurology. 2001 Mar 27;56(6):730-6. PubMed.
- Liu G, Boot B, Locascio JJ, Jansen IE, Winder-Rhodes S, Eberly S, Elbaz A, Brice A, Ravina B, van Hilten JJ, Cormier-Dequaire F, Corvol JC, Barker RA, Heutink P, Marinus J, Williams-Gray CH, Scherzer CR, International Genetics of Parkinson Disease Progression (IGPP) Consortium. Specifically neuropathic Gaucher's mutations accelerate cognitive decline in Parkinson's. Ann Neurol. 2016 Nov;80(5):674-685. PubMed.
- Cilia R, Tunesi S, Marotta G, Cereda E, Siri C, Tesei S, Zecchinelli AL, Canesi M, Mariani CB, Meucci N, Sacilotto G, Zini M, Barichella M, Magnani C, Duga S, Asselta R, Soldà G, Seresini A, Seia M, Pezzoli G, Goldwurm S. Survival and dementia in GBA-associated Parkinson's disease: The mutation matters. Ann Neurol. 2016 Nov;80(5):662-673. Epub 2016 Oct 3 PubMed.
- Liu G, Locascio JJ, Corvol JC, Boot B, Liao Z, Page K, Franco D, Burke K, Jansen IE, Trisini-Lipsanopoulos A, Winder-Rhodes S, Tanner CM, Lang AE, Eberly S, Elbaz A, Brice A, Mangone G, Ravina B, Shoulson I, Cormier-Dequaire F, Heutink P, van Hilten JJ, Barker RA, Williams-Gray CH, Marinus J, Scherzer CR. Prediction of cognition in Parkinson's disease with a clinical-genetic score: a longitudinal analysis of nine cohorts. Lancet Neurol. 2017 Jun 16; PubMed.