Disease stages and therapeutic hypotheses in two decades of neurodegenerative disease clinical trials

Abstract

Neurodegenerative disease is increasingly prevalent and remains without disease-modifying therapies. Engaging the right target, at the right disease stage, could be an important determinant of success. We annotated targets and eligibility criteria for 3238 neurodegenerative disease trials registered at ClinicalTrials.gov from 2000 to 2020. Trials became more selective as the mean number of inclusion and exclusion criteria increased and eligible score ranges shrank. Despite a shift towards less impaired participants, only 2.7% of trials included pre-symptomatic individuals; these were depleted for drug trials and enriched for behavioral interventions. Sixteen novel, genetically supported therapeutic hypotheses tested in drug trials represent a small, non-increasing fraction of trials, and the mean lag from genetic association to first trial was 13 years. Though often linked to disease initiation, not progression, these targets were tested mostly at symptomatic disease stages. The potential for disease modification through early intervention against root molecular causes of disease remains largely unexplored.

Figure 1

Flowchart of trial inclusion and curation. Trials returned by our search strategy were sequentially excluded based on launch year, lack of therapeutic intervention, patient population, or other. See Methods for explanation.

Figure 2

Characteristics of neurodegenerative disease clinical trials. (A) Univariate count of trials by disease, sponsor, intervention class, phase, or placebo/SOC control status. (B) Total Univariate total of patient-years (see Methods) by the same variables. (C) Bivariate cross-tabulation of number of trials and patient-years. Intensity (patient-years per trial) is expressed as a color palette from transparent yellow to opaque blue; number of trials is expressed as the size of each circle. (D) Pie chart of patient-years of enrollment, with industry drug trials shown in 3 alternating shades of pink and all other trials in 3 alternating shades of green. Each wedge represents one trial, and trials are sorted by number of patient-years. (E) Barplot representation of contingency tables for whether trials are (purple) or are not (gray) placebo/SOC-controlled (top), completed (middle), or have a specified phase (bottom) depending on whether they are industry-sponsored drug trials (right) or other (left). (F–G) Mean total number of inclusion and exclusion criteria per trial as a function of (F) industry-sponsored drug trials versus all other, and (G) phase. Loess curves were fit on the raw individual values, but due to the large number of trials, individual values are not shown; instead the average for each year is shown as a semitransparent horizontal bar. Statistical significance was evaluated by linear regression, see Results text.

Figure 3

Disease stages eligible for trials. (A) Barplot of trial count by earliest disease stage (legend at left, see Methods for details) eligible to enroll. (B) Forest plot of odds ratios (Fisher’s exact test) for properties of preventive (stage 0–2) versus symptomatic (stage 3–4) trials. (C) Scatterplot of study duration (x axis) and enrollment (y axis) for preventive (stage 0–2, cyan) versus symptomatic (stages 3–4, gray) trials, with marginal histograms on both axes. Crosshairs represent mean and 95% confidence intervals of the mean on both dimensions. (D) Stacked area plot of the number of trials per year by earliest eligible disease stage. For 2020, only 3 months of data were included, so the raw number of trials was scaled by a factor of 4 to yield trials/year. (E) Eligible MMSE score ranges by year, N = 911. Each trial is displayed as a purple rectangle of 10% transparency stretching from the lowest to highest eligible score on the y axis and staggered by ± 0.5 years on the x axis, such that darker shades of purple indicate a greater density of trials recruiting patients in a given score range in a given year. Green lines represent best fits from linear regression models. Lower scores indicate greater impairment. (F) As in (E), but for Hoehn and Yahr, N = 757; note that on this scale, higher scores indicate greater impairment.

Figure 4

Therapeutic hypotheses tested in drug trials. (A) Area-scaled barplot of 7 categories of drug interventions tested. The length of each rectangle on the x axis is the number of trials; thickness on the y axis is proportional to the number of patient-years per trial. Thus the total area occupied by each rectangle is proportional to the total patient-years invested in each category, for which percentages of the total are overlaid in white text. (B) Barplot of number of trials per year for drugs FDA approved for the treatment of the indicated disease before (cyan) and after (gray) initial FDA approval. Trials of these same drugs in other neurodegenerative diseases, for which they are not yet labeled by FDA, are not included in this plot; trials where these drugs served as SOC arms are also excluded. Black triangles indicate years of first FDA approval; approvals prior to 2000 are left-aligned. (C) Proportional area Euler diagram of genes that encode targets of drugs approved for any indication (gray), targets tested in neurodegenerative disease clinical trials in this dataset (red), or targets supported by human genetic evidence (yellow). (D) Stacked area plot of cumulative number of genes associated to these 4 diseases by year. Association of ACE to AD is counted in 1999 acknowledging candidate gene studies which replicated by GWAS in 2018. (E) Stacked area plot of categories of trials as defined in (A) by year. (F) Scatterplot of genetically supported target-indication pairs pursued in drug trials, displayed as year of first reported genetic association (x axis) versus year of first trial in the genetically linked neurodegenerative disease (y axis), color-coded by disease. *Gamma secretase is represented here by PSEN1; members PSEN2 and APH1B also have genetic association to AD risk. (H) Barplot of number of trials for each genetically supported target, in its genetically supported indication, tested clinically.