Towards early disease modification of Parkinson’s disease: a review of lessons learned in the Alzheimer field

Selecting at-risk populations is paramount for early intervention and prevention trials in PD. There are several options. First, at-risk populations can be defined based on clinical features that are supported by PD biomarkers, such as presence of a RBD or olfactory loss in combination with markers of advanced pathological evolution, e.g., dopaminergic transport imaging abnormalities. Without access to more advanced risk testing, a risk prediction can also be made on clinical observations, including complaints such as tremor, constipation, or dizziness [4]. A major advantage is that models based on such clinical risk markers have relatively high accuracy for predicting which individuals may subsequently be diagnosed with clinically overt PD [18]. However, recruiting people who are close to a PD diagnosis would leave limited room for actual early disease modification or prevention possibilities due to the generally advanced stage of neuronal loss at the moment in which first symptoms occur. This concern applies even more so to trials including recently diagnosed PD patients who averagely have a substantial reduction of dopaminergic neurons in the nigrostriatal pathway by the time of diagnosis [19]. Given the advanced level of neurodegeneration in those who (almost) meet PD diagnostic criteria, we speak of ‘early disease modification’ in this group rather than disease prevention. Even so, delaying the onset of overt PD, if only by a few years, would be a major achievement with clear clinical benefits. Evidently, the ideal timing of the trial also depends on the intervention target and the specific target population. PD is a heterogeneous disease, in which treatment benefits may differ per patient sub-group depending on risk profile. Hence, the current movement towards more personalized PD trials [2].

Second, it is also possible to define an earlier at-risk population that consists of asymptomatic individuals who are at increased risk to develop PD in a more remote future, the advantage being that putative disease-modifying treatments would theoretically have a greater effect here. Such prediction models would typically include genetic markers and midlife environmental risk factors, such as sex and non-use of caffeine. However, the accuracy of such models to predict clinical PD at the level of an individual is currently poor, especially within a limited timeframe [20, 21]. Some of the early disease modification AD trials, therefore, focused on recruiting only those with a autosomal dominant genetic predisposition for the disease [22]. However, for both AD and PD, autosomal dominant genetic predispositions are very rare—especially when narrowing it down to specific subtypes [23]. Such a focused recruitment strategy is thus challenging from a practical perspective and the trial results will not be representative for the vast majority of patients.

Third, combining clinical features and risk factors in trial recruitment criteria is currently considered the ideal option, being a compromise between predictive power and adequate room for prevention purposes, as reflected by the International Parkinson and Movement Disorder Society (MDS) research criteria for prodromal PD [3]. It should be noted that many risk factors for PD are also associated with Lewy Body dementia and multiple system atrophy, which increases the chance that participants with early symptoms are misdiagnosed as ‘prodromal PD’ in the recruitment process. This would negatively affect the reliability of clinical trial sample size calculations. To avoid a high screen failure in recruitment for these trials, previously set-up recruitment registries might be used to selectively invite those known to have certain PD-risk factors, although the true value of such registries as a basis for recruitment is yet to be shown [24].

Finally, for immunotherapy trials targeting α-synuclein in particular, ideally, only individuals with abnormal levels of α-synuclein would be recruited, because the intervention may only be effective in this population. However, the diagnostic accuracy of measures of elevated α-synuclein levels in CSF for PD pathology remains contentious [25, 26]. Positron emission tomography (PET) scans and serum and plasma biomarkers to measure α-synuclein levels are currently being developed, e.g., the real-time quaking-induced conversion (RT-QuIC) technique, but are not available for use in trial settings yet [27]. For serum and plasma biomarkers, the dynamics in fluid measures are unknown [28], but PD biomarker study results will be announced soon [29]. Given the many exceptions of α-synucleinopathy staging, according to the Braak hypothesis, it cannot be used as a tool to predict future symptoms [30]. Misdiagnosis in the recruitment process should be avoided to lower the noise in clinical trial power calculations, and thereby improve the chance of finding the potential effect of an intervention and limit the minimal number of participants.

According to current research criteria, someone has ‘prodromal’ PD when summing up the presence of (clinical) risk factors, such as olfactory loss and non-use of caffeine, indicates that someone’s overall risk to develop the disease is exceptionally high, up to a presumed risk > 80% if many risk factors are present [3, 5]. Since the word ‘prodromal’ refers to an early stage of a disease, i.e., before full manifestation of symptoms, the interpretation is that being at high risk for PD is the same as being in an early stage of the disease. Consequently, high-risk individuals recruited for trials can be labeled as having an early or ‘prodromal’ stage of a brain disease, while they would not have received such a diagnosis in a clinical setting—similar to the current situation in AD prevention trials.

We learned from the AD field that this scenario is vulnerable to miscommunication. Despite elaborate education sessions, some cognitively healthy research participants did not understand that having biomarkers of amyloid-β indicates an increased but uncertain risk for dementia [31]. Moreover, telling people they are in an early stage of a neurodegenerative disease, rather than being at increased risk, may have consequences for how they understand themselves and how they are perceived by others [32]. Some people may start thinking of themselves as “sick” and feel less able to take part in certain activities, for example, or colleagues will perceive them in that way, even if such an early diagnosis reflects, in fact, risk status. Since a diagnosis of a neurodegenerative disease may also provide a reason to “excuse” someone from societal or work obligations, expanding the disease label may create new challenges for employment and insurance policies, as well. If clinical practice will gradually adopt the expansion of PD diagnostic criteria applied in research, as occurred in the AD field, this may have severe social and psychological disadvantages [15, 33]. Moreover, substantial financial investments will then be needed for diagnostic tools and, potentially, further monitoring and follow-up [28]. We, therefore, propose to stick to terms of ‘risk’ rather than ‘prodromal’ (or ‘preclinical’) disease, also to prevent discrepancies between research and clinical practice.

For some, being informed of being at increased risk for a neurodegenerative disease may cause anxiety and stress [34]. For others, in the case of PD rather than AD, it could offer relief by providing an explanation for the psychological or physical difficulties that people experience in the years before receiving a diagnosis [35, 36]. Whether an early diagnosis is preferable, i.e., timely, will depend on the individual [37]. In the absence of an effective treatment, however, the majority of patients with PD are sceptical about early risk disclosure [38]. People who received an increased AD risk status contemplated a change to their health behavior, living situation, or future plans more compared to those who did not, even though the validity of this risk status for future symptom development is still uncertain [39].

Future trials including those at risk for PD can benefit from thought-out communication strategies from the AD research field to inform research participants of their risk status [40, 41], and should incorporate research on the impact of knowing to be at risk in the case of PD. A key challenge in communicating risks with patients is to adequately convey the limited diagnostic accuracy of currently available biomarkers and the precariousness of risk algorithms. Furthermore, participants should have legal safeguards against discrimination based on their PD-risk status and to protect their privacy. These are among the reasons why risk algorithms, for AD as well as PD, may be less suited for clinical practice, while being useful for recruitment purposes of upcoming clinical trials.

AD prevention trials have encountered major logistical challenges due to slow recruitment, high screen failure rates, and low retention. Immunotherapy trials that selected only those with elevated levels of amyloid-β have taken 3 years to complete enrollment, with screen failure rates up to 90% [60], partly due to the exclusion of persons with co-morbidities [61]. Setting up large transnational recruitment registries with clinical and biomarker information of potential research participants could be set up to tackle these issues [62]. However, the success of such approaches remains to be determined in the AD field, and it poses new ethical challenges if a small minority of those who registered for these trial-readiness cohorts will ever be eligible for a clinical trial. Moreover, designing such a registry and linking personal data on risks for future PD development requires careful consideration of ethical issues related to privacy, informed consent, and disease risk disclosure [63]. When informing people that they have an increased risk for a progressive and debilitating neurodegenerative condition and then invite them to participate in a trial of an agent that might mitigate that risk, people might feel more inclined to take part. Researchers should be warranted in these cases for false hope on the side of the potential participant.

For the upcoming trials that intend to investigate anti-α-synuclein immunotherapy in individuals with ‘prodromal’ PD [16], there is already some evidence for the short-term safety and tolerability for certain anti-α-synuclein antibodies [17, 47]. However, it remains essential to consider safety outcomes in the upcoming phase 1 and phase 2 trials. In AD trials, side-effects of immunotherapy included very serious adverse events, including meningoencephalitis and brain microhemorrhage [64]. Some were not detected until the phase 3 trial [65]. Future immunotherapy PD trials should, therefore, anticipate on these and other serious side-effects, and continue to monitor potential imaging abnormalities during phase 3 trials with the scrutiny level of a phase 1 or 2 trial.

The healthier the target population and the further away from a diagnosis of PD, the more difficult it is and the longer it will take to establish a clinical effect of the intervention. To have sufficient power to detect a clinical effect, there are two main possibilities: first, to design trials with a duration as long as 5–10 years, which carries a high risk of drop-outs and for which it is difficult to attract funding; second, to enroll a very high number of subjects, which could limit the trial duration to a shorter period, but at exponentially high costs. Modeling long-term effects based on small effect sizes for clinical or even biological outcomes in short-duration studies remains treacherous and may result in overestimating potential treatment effects, especially when outcome measures may leave room for a certain placebo effect [66].

The movement disorder society—unified Parkinson's disease rating scale (MDS-UPDRS) has long been the most common clinical outcome measure in PD clinical trials, but its scoring comes with several methodological challenges [67]. Importantly, this scale cannot detect changes in the prodromal stages of the disease due to floor effects.

Alternatively, documenting an effect on a biomarker targeted by the intervention, as measured either in body fluid or using neuroimaging as a proxy outcome measure, seems attractive, and has been applied. However, similarly to biomarkers of AD, PD biomarkers are not directly related to clinical parameters and are, therefore, no reliable indicator of clinical efficacy [26, 30, 46]. This can partly be explained by the fact that it remains a fundamental question whether PD’s range of clinical symptoms results from one common, or many pathophysiological processes [68, 69]. If the latter becomes more likely, PD research will presumably become increasingly individualized [57].

More sensitive outcome measures, e.g., changes in motor performance as detected with modern technologies including wearables, could increase the possibility to detect small potentially relevant effects of an intervention [70]. Still, any effect on an intermediate outcome which is not directly translatable into clinical relevance can at best only serve as proof of concept. Intermediate biomarker or digital monitoring outcomes will, therefore, not suffice in phase 2 trials when targeting the early disease stages of PD, i.e., with subtle clinical features, where an indication of a clinical effect at the level of relevant symptoms remains desirable [65].

Finally, similar to AD trials, before any drug is approved, an effect on a patient-oriented clinically meaningful endpoint is highly desirable, such as an effect on activities of daily living (ADL) or quality of life (QOL). This is challenging in its own right. For example, a recent trial of exercise—another intervention with potentially a disease-modifying and preventive potential—showed the stabilization of MDS-UPDRS motor scores in patients with the early stage PD, with an effect size that exceeded the minimal clinically important difference, yet without concurrent improvements in quality of life [71]. QOL outcome measures could be combined with other physical and psychological symptoms and life impact measures to better reflect the disease state [72].

As it would be difficult to establish a clinical consequence in a preclinical population, the Food and Drug Administration (FDA) is exploring alternative outcome measures for accelerated approval in the case of AD prevention trials [73]. These discussions may open the path for approval based on more subtle outcome measures in PD trials, although it also carries the risk of descending into the approval of therapies which have never shown to be effective in terms of patient benefit.

In AD research, anti-amyloid-β immunotherapy trials showed no improvement of cognition or functional ability in phase 2 after proof of concept was established. Even so, based on post hoc sub-group analyses, the tested antibodies advanced to phase 3 trials, which have all led to disappointing results [12]. Looking back, the misleading expectations resulted from an over-interpretation of subgroup analyses [65]. In light of these occurrences, the PD field should be warned against spending scarce resources on phase 3 trials in the absence of strong suggestions of clinical efficacy in phase 2 trials, and apply the appropriate sub-group analyses guidelines to avoid the over-interpretation of results.