Long-term follow-up of subthalamic nucleus deep brain stimulation in patients with Parkinson’s disease: an analysis of survival and disability milestones

Design: a longitudinal retrospective study

Primary objective: to assess mortality of PD patients submitted to bilateral STN-DBS after a follow-up of 8 years.

Secondary objective: to determine the frequency and the determinants of disability milestones (falls, FOG, hallucinations, dementia, and institutionalization) in the same population of patients.

Patients: All PD patients undergoing bilateral STN-DBS from 2006 to 2012 at Hospital Santa Maria were included in the study. Patients implanted in the globus pallidus internus were excluded (n=2). PD was diagnosed according to the UK Brain Bank criteria²³ and the criteria for DBS were: i) presence of clinically significant levodopa-induced MC not optimally controlled with medication, ii)< 70 years-old, iii) a ≥ 33% reduction in the Unified Parkinson’s Disease Rating Scale (UPDRS)²⁴ motor score after a supra-threshold levodopa dose, iv) no dementia, v) no major/refractory psychiatric illness, and vi) no postural instability or FOG in the best ON state. Patients who had the leads or the implanted pulse generator definitively removed during the follow-up were identified and censored at the time of DBS system removal.

Study Outcomes: Primary outcome - mortality during the 8 years after DBS. Secondary outcomes - occurrence of any of the following disability milestones: falls; FOG; hallucinations; dementia; and institutionalization.

All patient data was retrospectively retrieved and analysed from baseline and up to 8 years post-surgery. Data was extracted from every single hospital visit of patients. Patients were followed-up until the time of death or the end of the follow-up period (8 years), whichever occurred first.

Assessment tools

Falls: defined as a fall report by the patient or caregivers on two consecutive visits.

FOG: defined by the presence of FOG observed by the clinician on two consecutive visits. Both falls and FOG were considered only after 6 months into the post-surgical period, as we considered 6 months the time required for treatment stabilization.

Dementia: defined according to the diagnostic criteria of the DSM-V.

Institutionalization: the date of placement in a nursing home was either the specific date recorded in files or the date of the visit when institutionalization was first mentioned.

Additional baseline pre-surgery data was collected: age and duration of disease at surgery, gender, scores of parts II (activities of daily living) and III (motor) of the UPDRS in both OFF and ON states, scores of part I (mental) and IV (motor complications) of UPDRS, Hoehn and Yahr (H&Y) stage²⁵ and Schwab and England score in both ON and OFF state, percentage of response to a supra-threshold dose of levodopa. The percentage of patients with a significant gait impairment (score ≥2 on item 29 of the UPDRS-III) or postural instability (score ≥2 on item 30 of the UPDRS-III) in the OFF state were calculate. Clinical phenotype, i.e. postural instability/gait disorder (PIGD) and tremor dominant (TD), and levodopa-equivalent daily doses (LEDD) (mg/day) were calculated as previously described.^{26, 27} Data from formal neuropsychological evaluation including the MMSE score were collected.²⁸ As our institution started using the MDS-UPDRS²⁹ only from 2012 onwards, the MDS-UPDRS II and III scores were converted from the corresponding UPDRS ones using the available conversion formula,³⁰ allowing for the standardization of the entire cohort.

Data availability statement

Anonymized data of this study will be available from the corresponding author on reasonable request from any qualified researcher, following the EU General Data Protection Regulation. analysis pipeline can be found at [https://github.com/Le-bruit-de-nos-pas/PD_Milestones_STN-DBS].

Clinical and demographic baseline characteristics

109 patients underwent STN-DBS during the study period. Clinical and demographic data are presented in Table 1. During follow-up, 8 patients were censored: 7 due to system removal because of infection (at months 5, 35, 45, 57, 61, and 2x 95) and 1 was referred to another institution and lost to follow up (at month 72). 84 patients reached the end of the 8-year follow-up: mean age of 68.6 ± 7.6, mean disease duration of 21.38 ± 4.38 years, mean H&Y 2.8 ± 1.1, mean LEDD of 600.90 ± 370.14 mg (vs 1252.66 ± 521.17 mg at baseline; average reduction of 49.64 ± 29.52 %, p < 0.001). Stimulation parameters at the end of follow-up in Table 2.

Mortality

Seventeen patients died during the study, accounting for a mortality rate at the end of follow-up of 16%. On average, patients died 62.1 ± 21.3 months after surgery (Fig. 1), at a mean age of 69.9 ± 6.0 years. The presence of disability milestones in the follow-up did not significantly alter the mortality rate (Fig. 1).

• Download figure
• Open in new tab

Figure 1.

Survival plots for the entire cohort (a) and for the same cohort split between patients developing a given milestone and those failing to do so (b-f). Comparisons performed using Log-rank (Mantel-Cox) tests.

In the patients who died during follow-up, falls preceded death by 34.6 ± 13.7 months, FOG by 35.8 ± 21.4, dementia and hallucinations by 18.5 ± 14.5 and 18.7 ± 14.4, respectively. None of the patients who died were institutionalized.

Non-adjusted, univariate analysis showed that a preoperative higher UPDRS part II score in OFF (HR 1.012, C.I. 1.005-1.020, p = 0.001) and older age at surgery (HR 1.108, C.I. 1.008-1.218, p = 0.034) significantly increased mortality rates (Sup. Fig. S2). On multivariate analysis, age at surgery (HR 1.121, C.I. 1.035-1.463, p = 0.019) and the preoperative UPDRS part III score in OFF (HR 0.942, C.I. 0.890 - 0.996, p = 0.037) were significant drivers of overall mortality (Fig. 3).

Disability Milestones

Falls were the most likely event during follow-up (Fig. 2), with a competing risk-adjusted probability of 73% at 8 years post-DBS (95% C.I., 0.636 - 0.897). The average time from surgery to first fall was 40.4 ± 25.4 months. FOG was observed in 47% (95% C.I., 0.377 - 0.566) of the cohort (Fig. 2). The average time from surgery to FOG was 39.6 ± 28.4 months. In turn, the competing risk-adjusted probability of developing hallucinations was 32% (95% C.I., 0.409 - 0.231) (Fig. 2), with an average time from surgery to hallucinations of 60.0 ± 20.7 months. 34% of patients developed dementia (95%C.I., 0.254 - 0.436) (Fig. 2), with an average time from surgery to dementia diagnosis of 56.2 ± 21.2 months. Only 6% of the cases were institutionalized (95% C.I., 0.024 - 0.117) (Fig. 2), after 62.3 ± 22.0 months of surgery.

• Download figure
• Open in new tab

Figure 2.

Estimated cumulative incidence curves (a-e) for each milestone event with mortality treated as competing risk. Shaded areas represent (upper and lower) 95% confidence intervals.

Determinants of axial milestones

Fallers had worse baseline scores in UPDRS II (28.0 ± 38.3 vs 19.9 ± 6.7, p=0.015) and UPDRS III item 30 (postural instability) (1.6 ± 1.0 vs 1.0 ± 1.1, p= 0.022) in OFF than non-fallers (Sup. Table S1). The baseline variables that in univariate analysis were significantly associated with falls are show in Sup. Fig. S2. When adjusting for confounders, older age at surgery (HR 1.184, 95% C.I., 0.779 - 1.086, p = 0.004) remained significant associated with falls (Fig. 3). The cumulative incidence of falls was significantly higher among patients who also developed FOG during follow-up (86%, 95% C.I., 0.718 – 0.932) compared to those who did not (62%, 95% C.I., 0.478 – 0.735, p = 0.010) and in those who developed hallucinations (85% %, 95% C.I., 0.67-0.94 vs 67%, 95% C.I., 0.55-0.77, p = 0.039) (Fig. 4).

• Download figure
• Open in new tab

Figure 3.

Forest plots depicting the Hazard Ratios from a Cox proportional-Hazards regression model (a, mortality) and several sub-distribution hazard models (b-f, milestone events). Each and every single variable is adjusted for all other variables in the respective plot. Multivariable regression analysis in the presence of competing risks was performed using the (semiparametric proportional hazards model) sub-distribution hazards of cumulative incidence functions based on the Gray’s test. A backward elimination method for selection of variables was using a cut-off of 0.2.

• Download figure
• Open in new tab

Figure 4.

Estimated cumulative incidence plots (a-f) of falls and freezing, with equality between plots based on the Gray’s test (i.e. comparison of the weighted averages for the sub-distribution hazards of each event of interest)

Freezers had significantly higher baseline scores in UPDRS III item 29 (gait) in OFF than non-freezers (2.0 ± 0.9 vs 1.6 ± 1.5, p = 0.029) (Sup. Table S2). Univariate analysis showed that longer disease duration at surgery (HR 1.070, 95% C.I., 1.010 - 1.140, p = 0.021), higher preoperative UPDRS part III score in ON (HR 1.080, 95% C.I. 1.020 - 1.140, p = 0.012) and higher preoperative UPDRS part II score in OFF (HR 1.010, 95% C.I. 1.010 - 1.020, p = 0.000) significantly modulated the rate FOG (Sup. Fig. S2). When adjusting for confounders, older age at surgery (HR 1.186, 95% C.I. 1.050 – 1.340, p = 0.006) and higher UPDRS part II score in OFF (HR 1.024, 95% C.I., 1.012 - 1.040, p = 0.000) were significantly associated with freezing. The cumulative incidence of freezing was significantly higher in the fallers (56%, 95% C.I., 0.440 – 0.659) when compared to the non-fallers (25%, 95% C.I., 0.106 – 0.416, p = 0.007) (Fig. 4).

Determinants of hallucinations, dementia and institutionalization

The preoperative UPDRS part II score in OFF (HR 1.010, C.I. 1.010 - 1.020, p = 0.000) significantly modulate the rate of hallucinations, dementia and institutionalization in univariate, non-adjusted analysis (Sup. Fig. S2). After adjustment for confounders, no variable remained significantly associated with hallucinations. As for dementia, both older age at surgery (HR 1.029, C.I. 1.000 - 1.374, p = 0.05) and preoperatory UPDRS part II score in OFF (HR 1.026, C.I. 1.014 - 1.037, p = 0.000) and ON (HR 0.996, C.I. 0.711 - 0.962, p = 0.014) significantly modulated the rate of event development. In the case of institutionalization, multivariate analysis showed that both preoperatory higher UPDRS part II score in OFF (HR 1.010, 95% C.I., 1.010 - 1.020, p = 0.000) and older age at surgery (HR 1.190, C.I. 1.070 - 1.320, p = 0.001) were significant drivers of institutionalisation(Fig. 3).

Mortality

Across different studies, highly variable mortality rates have been reported (from 17 to 61%) in long-term DBS cohorts, ^{10, 13, 33–36,13} probably reflecting high heterogeneity of study populations regarding disease and follow-up duration. We have herein observed figures for cumulative mortality and cumulative incidence of falls, dementia, and hallucinations similar to those recently reported by Mahlknecht et al.³⁶ for STN-DBS, who found a mortality rate of 17% after 7 years of follow-up. Age and disease duration at surgery were similar in both studies.³⁶ Likewise, a similar survival was found after 7 years of follow-up in a cohort with a similar age at surgery.³⁵ The importance of comparing cohorts with similar demographic characteristics is reinforced by the increase in mortality rates with older age at surgery.^{35, 37} Older age at PD onset has been associated with increased mortality,^{38, 39} however, findings from Kempster et al. have instead suggested that death (and disability milestones) occurs at around the same age regardless of the age at PD onset.^{19, 40} In our cohort, the mean age of death is around 70 years old, similar to what has been previously observed despite the longer disease duration.^{19, 40} Thus, mortality would be related to biological age more than to disease duration, or age of disease-onset.

Interestingly, regarding the milestone-to-death interval, we have found a similar temporal relationship as previous studies,¹⁹ reinforcing the idea that the late phase of PD (independently of age-of-onset, disease duration and DBS) follow a stereotyped pattern, with the appearance of milestones preceding death.

Falls and FOG

Falls and FOG were both the earliest and the most frequent disability milestones and tended to occur at around the same time after surgery. The incidence of falls surpassed that of freezing in the long run. In previous studies, falls were present in 61% of a DBS cohort followed-up during > 8 years³⁶ and in 64% of patients 7 years after surgery.³⁷ In the same study, 64% of the patients also developed FOG.³⁷ Different sample sizes and methods used to assess FOG and falls could explain the different rates between studies. A time-depended deterioration of axials signs after surgery^{10, 11, 14, 16, 37, 41} has been suggested and can explain the lower incidence of falls (32% of 260 patients) in a study with shorter disease duration at surgery (6 ± 3.8 years) and shorter follow-up time (median 3.1 years). ⁴²

Older age at STN-DBS was a strong driver of increase rates of both falls and freezing. Several studies have shown an association between age and the occurrence of falls and freezing^{33, 34, 42, 43}. Besides age at DBS, only a worse baseline UPDRS II OFF score was an independent predictor of freezing, but not falls. Our results reinforce the previously highlighted importance of baseline disease severity, namely in OFF, for future incidence of axial disability milestones, irrespectively of levodopa responsiveness.^6,11 Interesting, though, is the finding that the impairment on ADL in the OFF state is a predictor of axial disability milestones. When selecting patients for DBS, special attention should be given to the performance of patients in ADL in the OFF state, besides their objective motor disability, since it can help to identify patients with a higher risk of developing disability milestones.

Even though FOG is not often assessed when studying disability milestones in advanced disease,^{17, 19, 40} we have decided to evaluate freezing because it is a symptom associated with significant disability and deterioration in QoL, and most often unresponsive to levodopa.⁴⁴ The response of FOG to DBS-STN has been controversial⁴⁵ and as such, there is a need for future studies to accurately characterize its response to DBS.

Hallucinations and Dementia

Cognitive/behavioral milestones had a lower incidence than axial symptoms in our study. The rates of hallucinations and dementia after STN-DBS have been highly variable across studies, ranging from 18 to 61% for hallucinations,^{15, 33, 34, 36, 37, 46} and from 5 to 61% for dementia.^{13, 16, 33, 36, 46} An association between hallucinations and dementia (i.e., similar incidence and close temporal relationship) has been previously reported,^{17, 19} pointing to a common pathophysiology. On the one hand, a higher cortical burden of Lewy body pathology as well as a higher load of Alzheimer’s disease neuropathological changes have been associated with both dementia and hallucinations.^{19, 40} On the other hand, the presence of cortical Lewy bodies appears not to be correlated with motor milestones, indicating that cortical pathology is specifically implicated in the development of cognitive milestones.⁴⁶

Likewise, older age is a risk factor for dementia in PD,^{33, 34, 42, 43} which might explain the higher risk of dementia in those patients older at STN-DBS.

Nursing home placement

We found a much lower rate of admission to nursing home facilities compared to other post-DBS^{16, 34, 36, 37} and non-DBS studies^{47, 48} Nonetheless, a previous study of Portuguese late-stage non-DBS patients also found a lower frequency of nursing home admissions than that reported in the Swedish and Dutch counterparts.⁴⁸ Sociocultural stigma together with inadequate social support systems are likely to explain such lower institutionalization rates.^{36, 49}

The present study lacks a control group, but comparing the observed numbers with comparable size cohorts,^{21, 50} one is lead to conclude that STN-DBS patients developed the same disability milestones than non-DBS PD patients, but at a lower rate despite similar disease duration.^{21, 49, 50} One possible explanation is the difference in age at PD onset, with DBS patients being relatively younger at disease onset (47 vs 54, 56, and 61 years old at disease onset) and at the end of follow-up.^{21, 49, 50} Accordingly, a negative correlation between age at disease onset and time-to-development of the first disease milestone has been previously reported.^{17, 19} When compared to the Sydney cohort, a delay of 10 to 15 years in achieving the same level of disability milestones was found in the Toronto study.³⁷ A younger age of onset in the Toronto study was the explanation for the difference in findings, suggesting that the disease progresses relatively slowly in younger patients but disability milestones are reached at similar ages independently of age at disease onset.^{19, 37, 40} Our data supports this hypothesis.

Interestingly, even though STN-DBS patients are a particular sample of PD patients, considering that more frequently includes patients with younger age at onset, highly responsive to levodopa and no dopaminergic resistant axial symptoms before surgery, they still end up developing the same disability milestones (Fig. 5). Additionally, MC have been well-controlled with STN-DBS, this study shows that surgery does not prevent the development of such milestones.

• Download figure
• Open in new tab

Figure 5.

Summary incidence plot across all events under study adjusted for competing risks (falls, freezing, hallucinations, dementia, mortality, and institutionalization). For the sake of visibility, values on the X-axis correspond to the base 10 logarithmic of the number of elapsed months to event (1 -> 10 months, 1.5 -> 32 months, 2 -> 100 months).

In contrast to the usual sequence of events reported in late-stage non-DBS PD patients,^{17, 19} we have observed the development of falls antedating that of hallucinations. The decrease in dopaminergic medication after surgery and the lack of cognitive impairment before DBS may explain such a difference.³⁶