Amantadine treatment is an independent predictor of improved survival in Parkinson's disease

Amantadine has been used for more than 20 years in the treatment of Parkinson's disease (PD). Although most patients experience symptomatic improvement with amantadine, the mechanism for this response is uncertain. Possible mechanisms include increased synthesis or release or diminished reuptake of dopamine, as well as anticholinergic effects, driving striatal D2 receptors toward high-affinity states or influencing other neurotransmitter systems. ^[1] Amantadine and related drugs are also potent antagonists at the N-methyl-D-aspartate (NMDA) receptor. ^[2]

Several discoveries suggest that amantadine may have a neuroprotective effect in PD. First, dopamine uptake blockade prevents parkinsonism induced by 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP). ^[3] Amantadine's ability to reduce dopamine uptake may be useful therapeutically if an MPTP-like toxin is responsible for or important in the progression of PD. Second, excitotoxicity may have a role in the pathogenesis of PD (idiopathic parkinsonism) ^[4] and NMDA antagonists protect dopaminergic substantia nigra neurons from excitotoxicity produced by methamphetamine and MPTP. ^[5] In this regard, amantadine derivatives, which are potent NMDA antagonists, are protective of neurons in experimental models of excitotoxicity. ^[2]

In a previous study, designed to determine the effect of levodopa on survival, we evaluated longevity in a population of PD patients who lived before and after the introduction of levodopa and found that levodopa use was an independent predictor of improved survival. ^[6] In the present study, we chose to study patient survival in the levodopa era and to determine the effect of amantadine on this endpoint. If a neuroprotective effect is produced by amantadine, one might expect that use of the drug may be associated with improved survival.

Methods.

Information concerning all patients with parkinsonism attending the Movement Disorder Clinic (MDC) at Royal University Hospital, Saskatoon, Saskatchewan from 1968 to 1990 were entered into a computerized database. All patients were evaluated by the same neurologist (A.H.R). Information compiled for each patient included date of birth, sex, age of onset of parkinsonian symptoms, date of first clinic visit, Hoehn and Yahr ^[7] stage at first clinic visit, type of parkinsonism (PD or non-PD--e.g., progressive supranuclear palsy, multiple system atrophy, etc.), and date of death or most recent follow-up visit. Information concerning drug therapy (amantadine, anticholinergics, levodopa/levodopa-carbidopa, dopamine agonists) was recorded at each clinic visit. A diagnosis of PD was made if a patient demonstrated at least two of the three cardinal signs (resting tremor, bradykinesia, and rigidity) without identifiable cause. The diagnosis was updated after each follow-up visit and, in cases with autopsy, final diagnoses were appropriately recorded. A complete antiparkinsonian drug history (including start/end dates and dosages of amantadine, levodopa, dopamine agonists, and anticholinergic agents) was used to determine patients who received amantadine. We employed all of this drug information in subsequent analyses concerning amantadine-treated patients. Further details regarding this cohort are published elsewhere. ^[8] As a rule, follow-up was maintained at 6 to 12-month intervals. ^[9]

In patients not treated with amantadine, survival was determined from time zero (being the date of initial visit to the MDC) to the last follow-up or death. In patients treated with amantadine, time zero was the first known date of amantadine use or the initial visit to the MDC, whichever was later. Cumulative survival probabilities were estimated using the Kaplan-Meier method ^[10] and compared with the general population life expectancy for the West North Central region of the U.S., ^[11] for persons of like age, calendar year of birth, and gender. The one-sample log rank test was used to compare observed with expected survival. ^[12] The log rank test was employed to compare survival of groups of patients. ^[13] In all instances, p < 0.05 was used to denote statistical significance.

A Cox regression ^[14] model was utilized to identify a set of independent predictors of survival and to address the issue of treatment selection bias. Variables considered in this analysis were amantadine use, age, gender, 10-year expected survival probability (based on age, gender, and calendar year of birth using West North Central general population life tables), type of parkinsonism, presence or absence of resting tremor at initial visit, Hoehn and Yahr stage (defined as I to II versus III to V), year of onset of symptoms defined as before 1974 versus after, and year first seen in the MDC defined as before 1974 versus after. Backward elimination of nonsignificant variables was employed with p values <or=to0.01 used as evidence of a finding not due to chance. Interactions were considered in the Cox regression model after the initial backward elimination process was completed.

A separate survival analysis restricted to patients receiving amantadine was performed to assess whether the timing of initiation of amantadine, with respect to initiation of levodopa/carbidopa-levodopa, was a significant factor influencing survival. For this analysis, patients were partitioned into four groups: A, amantadine-treated with no levodopa; B, amantadine started first with levodopa started >2 months later; C, amantadine and levodopa started concurrently (both within <or=to2 months or both started prior to the initial visit to the MDC); and D, levodopa started first with amantadine started >2 months later. To eliminate spurious survival benefit in patients who received both drugs (amantadine and levodopa), time zero in these two groups (B and D) was defined as the latest of either the first visit to the MDC, the earliest known date of levodopa use, or the earliest known date of amantadine use. To assess the significance of the timing of amantadine initiation, two Cox proportional hazards regression models were compared, one that included the factors (other than amantadine use) found in the earlier analysis to be independent predictors of survival for parkinsonism patients and one that included these factors along with three dichotomous indicator variables that defined the four groups listed above. The p-value associated with the three degrees of freedom chi-square statistic (determined by the difference between the score statistics for the two models) was used to assess the overall significance of the timing of amantadine initiation.

Data analyses were performed using SPSS-X Release 3.0, installed on a VAX/VMS system at the University of Saskatchewan, Saskatoon, SK and using SAS at Mayo Clinic, Rochester, MN.

Results.

We studied survival in 836 parkinsonism patients with complete drug histories, seen at the MDC over a 23-year period (1968 to 1990); 92% of patients had PD. Further information regarding the cohort has been published elsewhere. ^[8]

From the multivariate survival analysis, five factors were identified as being independent predictors of improved survival: higher expected 10-year survival (based on age, sex, and calendar year of birth), absence of dementia at first MDC visit, use of amantadine, type of parkinsonism = PD, and low Hoehn and Yahr stage (I to II) at first MDC visit Table 1. Two-way interactions of amantadine use and the other factors found to be multivariately significant were assessed by adding the four interaction terms to the model and using a stepwise backward elimination process. From this analysis no interactions were found to be significant.

Of the 836 patients, 250 received treatment with amantadine for at least 2 months. The mean duration of treatment (100 mg b.i.d.) in this group was 37 months (median 24 months), with 99 patients still receiving amantadine at the time of their last visit to the MDC. None of the independent factors that predicted survival differed significantly between patients treated versus untreated with amantadine. The mean age at t₀ (the date of initial visit to the MDC for those not treated with amantadine and the later of the date first seen at the MDC and earliest known date of amantadine use for those treated with the drug) was 67.2 +/- 10.0 years in those not treated and 67.0 +/- 8.6 in those treated with amantadine. Of those not treated with amantadine, 61% were male, 89% were not demented at first visit, 92% had PD, and 51% had low Hoehn and Yahr stage (I or II) at first visit. For those treated with amantadine, 62% were male, 86% were not demented at first visit, 94% had PD, and 46% had low Hoehn and Yahr stage.

Of the 250 patients treated with amantadine there were 59 reported deaths with median time to death of 4.2 years. Of the 586 patients not treated with amantadine there were 154 reported deaths with median time to death of 4.6 years. For the amantadine-treated patients alive at last follow-up (n = 191) the median length of follow-up was 4.5 years (range 0 to 18.9 years). For the non-amantadine-treated patients alive at last follow-up (n = 432) the median length of follow-up was 1.0 years (range 0 to 19.0 years). The latest possible follow-up date was June 1992. Overall, follow-up for amantadine-treated patients was 57% of the maximum possible and 29% for the non-amantadine-treated patients. Patients without follow-up were censored from the survival analysis but were included in the data cited above regarding amantadine-treated versus untreated groups. We performed an analysis considering the independent predictors of survival to determine whether there were any significant differences between amantadine versus non-amantadine-treated patients when the analysis was restricted to patients with follow-up. When only patients with follow-up were included we found no significant differences between the amantadine versus non-amantadine-treated patients with respect to the independent predictors of survival.

The presence or absence of resting tremor at the initial visit was not identified as a significant predictive variable for survival. Additionally, there were no significant differences between amantadine-treated and non-amantadine-treated groups (92.2% of treated and 93.2% of non-amantadine-treated patients had resting tremor at their initial MDC visit).

Survival in the amantadine-treated group was better in comparison with the untreated group (p < 0.001). Figure 1 shows the Kaplan-Meier ^[10] survival curves for the amantadine-treated and untreated patients compared with the expected survival for the general population for the West North Central region of the U.S. and with each other. Survival for patients treated and untreated with amantadine was found to be poorer than the general population (treated: p = 0.046; untreated: p < 0.001).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Observed and expected survival for 250 patients treated with amantadine and 586 patients not treated with amantadine. For those not treated with amantadine time zero is the date of the first visit to the movement disorder clinic (MDC). For amantadine-treated patients time zero is the later date of first visit to the MDC or their initiating use of amantadine. Expected survival is based on all-cause mortality for persons in the general population of the West North Central region of the U.S. of like age, sex, and calendar year of birth. Survival was significantly worse than expected for both amantadine-treated (p = 0.046) and not-amantadine-treated (p < 0.001, one-sample log rank test). Survival was significantly better for amantadine-treated versus not (p < 0.001, two-sample log rank test).

Of the 250 patients who received amantadine, 231 also received levodopa/carbidopa-levodopa at some point during their treatment. Both drugs were started concurrently (within <or=to2 months) or prior to the first MDC visit in 61 patients, 81 started amantadine first and added levodopa >2 months later, and 89 started levodopa first and added amantadine >2 months later. Table 2 provides separate Kaplan-Meier survival estimates for each of these groups. The timing of the initiation of amantadine, with respect to the initiation of levodopa, was not found to be a significant factor influencing survival in patients receiving amantadine.

Discussion.

Of the many hypotheses regarding the pathogenesis of PD, some have implications for influencing the natural progression of this disorder. ^[15] The oxidative stress hypothesis led to concern that standard therapy with levodopa may be contributing to dopaminergic cell death in PD, ^[16] but our previous study showed that levodopa was associated with increased survival. ^[6] The present study evaluates the impact of amantadine on survival. This issue is of interest because (1) parkinsonian states may result from overactivity of glutamatergic activity arising from the subthalamic nucleus acting upon the medial globus pallidus ^[17] and (2) NMDA receptor antagonists prevent NMDA receptor-mediated glutamate excitotoxicity and MPTP-induced parkinsonism. ^[2,5]

Patients treated with amantadine had improved survival compared with those not treated. Although the amantadine-treated and untreated groups were not formed by random assignment, they appeared to be similar in relevant aspects, including all other independent predictor variables for survival. Our method addresses two major problems in the design of many survival studies of PD. We used the later of the dates at which amantadine was initiated or the time of first visit (in non-amantadine-treated patients) as "time zero'' for survival calculations. (Studies using the date of onset of symptoms as time zero give an erroneous survival advantage to all patient groups. ^[6]) In these ways, we minimized our chances of identifying a beneficial survival effect (spurious or otherwise) associated with treatment. Additionally, by statistical adjustment, employing a Cox regression model, we attempted to address the potential problem of treatment selection bias. (We hypothesized that amantadine-treated patients might tend to be those with either mild, early PD or those with longstanding, severe disease who were not having a good response to levodopa.)

Potential biases related to socioeconomic status may be difficult to identify unless patient income/occupation are available. However, in the present study of residents of Saskatchewan, we believe that the availability of neurologic care and treatment were independent of socioeconomic status. In 1962, Saskatchewan became the first Canadian province to implement a general provincial tax-supported universal health care system. ^[18] In 1972, the premiums for registration in the provincial insurance plan were eliminated, entitling all residents to complete physician services with no out of pocket expense. Therefore, cost was no deterrent for Saskatchewan residents to obtain consultation with a neurologist. Additionally, in 1975 a provincial drug plan, which included all drug costs (except dispensing fees), was instituted. In 1987, the drug plan was changed so that the patient paid for 20% of the drug cost. Consultation at the MDC and treatment with amantadine or any other therapy therefore would not appear to be strongly influenced by socioeconomic factors.

PD patients, in the aggregate, had poorer survival than the general non-parkinsonian population. This corresponds with our findings concerning survival in another population. ^[6] This includes the patients treated with amantadine, who had slightly but significantly poorer survival than the general population. The Cox model identified several variables, including amantadine use, as being independent predictors for improved survival. Not surprisingly, we did not identify levodopa use as a statistically significant predictor of improved survival in the amantadine-treated subset, as the vast majority (>85%) received levodopa.

Amantadine (1-amino-adamantane) and memantine (1-amino-3,5-dimethyladamantane) are antiparkinsonian drugs with similar structures. Both compounds are NMDA receptor antagonists and have been used (or suggested) as therapeutic agents for a variety of neurologic conditions including PD, ^[5,19] depression with parkinsonism, ^[20] epilepsy, ^[21] dementia, ^[22] AIDS-related dementia, ^[23] heredo-degenerative ataxias, ^[24] multiple sclerosis, ^[25] methyl parathion toxicity, ^[26] neuroleptic malignant syndrome, ^[27] dystonia, ^[28] and spasticity. ^[29] In PD, NMDA receptor antagonism may account for symptomatic and possibly neuroprotective effects. Recent studies have suggested that symptomatic improvement observed with amantadine treatment of PD may be mediated by antagonizing NMDA receptor-mediated acetylcholine release ^[30,31] rather than enhanced dopaminergic transmission. ^[32,33] As NMDA antagonists, these agents are capable of preventing the death of central neurons both in vitro and in vivo when given coincident with an excitatory amino acid insult. ^[27,34] Furthermore, they prevent both MPTP and methamphetamine-induced parkinsonism in animal models. ^[5] A recent study reported that remacemide, a glutamate antagonist with neuroprotective action, potentiates the antiparkinsonian effect of levodopa in animal parkinsonian models. ^[35]

This study represents the first report of an association between amantadine use and improved survival in human PD. ^[36] We believe our definition of time zero for amantadine-treated patients represents the appropriate definition for survival analyses. If anything, our definition should give the survival advantage to those not treated with amantadine. Nevertheless, our findings supported the conclusion that those receiving amantadine had better survival when assessed univariately and multivariately. This association may stem from an ongoing symptomatic benefit provided by the drug that leads to less disability and ultimately better survival. Additionally, amantadine may be directly influencing disease progression at the cellular level, providing neuroprotection. We speculate that such action may be mediated through NMDA receptor antagonism, although other mechanisms, such as dopamine uptake inhibition, could be playing a role. For these reasons, amantadine and related NMDA antagonists may have previously unappreciated therapeutic potential for the treatment of PD. Although our study disclosed a statistically significant survival treatment benefit with amantadine, this issue can only be definitively addressed by a randomized trial.