METHODS

We undertook a systematic review of the published literature in an attempt to assess the efficacy of surgical intervention for PD. This entailed identifying studies of lesioning and DBS for PD published between January 1990 and March 2002; the choice of time period reflected the recent renewed interest in surgical procedures for the treatment of PD. The search did not include studies of transplantation for PD. We searched databases including the Cochrane Library, Medline, Embase, and PubMed and hand searched major journals in the field as well as conference proceedings to identify presentations made at meetings. Experts in the field were contacted in an attempt to find studies not identified by electronic and hand searching and to identify trials in progress. Research registers were searched to identify ongoing research and additional information was sought from scanning reference lists of retrieved papers, in particular review papers and web sites relating to PD. We also attempted formal meta-analyses of surgical randomised controlled trials (RCTs) that reported the Unified Parkinson’s Disease Rating Scale (UPDRS) scale or adverse event data as outcomes, or both.

RESULTS

This systematic review of the published literature identified 503 studies of functional stereotactic surgery for PD. Most of these were small non-randomised, non-controlled studies, and mainly case series. We found 10 700 patients had been included in 496 reported non-randomised studies, averaging about 21 patients per study. Only 2% of studies involved 100 or more patients and only 10% involved 50 or more. There have also been many single case reports. These statistics are based on published reports; the actual number of PD patients that have been included in studies worldwide is likely to be much higher.

We found considerable variation in follow up length across non-randomised studies ranging from days to years. For those studies with reported follow up duration, the maximum follow up period averaged 15 months with only 27% of studies reporting outcome beyond 12 months and only 15% of studies reporting two or more years. Very few studies (1%) followed up patients for five or more years. Moreover, not all patients were followed up for the maximum time period specified in the publications.

Reporting problems were also evident with many studies having poorly defined designs, for example, poorly described eligibility criteria, patient characteristics, follow up duration, and outcome measures. Identifying multiple publications of the same study also caused evaluation problems.

Strikingly, only seven RCTs of stereotactic surgery in PD were identified (table 1). Most were single centre trials and with small numbers of patients (196 patients in total, representing only 1.7% of the total number of patients identified and an average of 30 patients per trial). The maximum follow up in the randomised studies averaged only six months. Two of the randomised studies, including 73 patients, compared PD surgery with medical treatment.^1,2 Three trials (64 patients) compared electrical stimulation with surgical lesion,^3–5 and two (59 patients) compared GP with STN stimulation.^6,7 Attempts to perform meta-analysis of these trials were hampered by a paucity of consistently reported data. For example, standard deviations (SDs), confidence intervals, and p values were often not reported. Some data had to be assumed (for example median = mean; range = SD) or estimated from the available data (for example, using a p value to estimate the SD) and other data were estimated from plots or figures shown. Effect sizes were difficult to calculate because many trials reported percentage change in each arm but did not compare the difference between the two arms. Furthermore, adverse event data were only reported clearly in two RCTs. However, previous estimates of the occurrence of adverse events in surgery for PD indicate that transient adverse events occur in 5%–60% of cases and permanent effects in up to 40%.² In terms of efficacy, results from the surgery compared with medical management RCTs seem to suggest that surgery is superior to medical management in advanced PD patients, at least in the short-term. According to the NEPAS study group, pallidotomy reduces off phase motor impairments by 30% on the UPDRS motor subscale after six months, compared with a 4% worsening in the control arm (difference; p<0.001).² These results are consistent with those of the Atlanta study group: 34% improvement pallidotomy arm and 4% worsening medical arm. Unlike the NEPAS study though, changes over time were not analysed and they gave only statistics for total UPDRS.¹

The RCT comparing unilateral lesion with DBS in the gobal pallidus³ did not show significant differences in the efficacy or safety of the procedures. Similarly, no differences were found between thalamotomy and thalamic stimulation, although stimulation had fewer adverse events.⁵ In the trial investigating simultaneous bilateral pallidotomy versus simultaneous GP lesion combined with contralateral stimulation, the protocol was discontinued because of severe adverse events found in the bilateral pallidotomy arm.⁴ The two trials comparing DBS of the STN and GP demonstrated no significant differences in efficacy between treatment arms and no serious complications of either procedure.^6,7 The results from the existing RCTs should be interpreted with caution because of the small numbers, short follow up, and possible publication bias. Indeed, many of the surgical RCTs have indicated the need for larger, longer term, and hence more reliable trials.

Most studies of surgery in PD have relied on clinician based rating scales to measure motor impairments and disability, most frequently the UPDRS. Only three trials reported using the Core Assessment Programme for Intracerebral Transplantation.^1,3,4 Interestingly, only 5% of all studies (including three of the RCTs^2,5,7) reported quality of life measures. Only one non-randomised study of surgery for PD reported health economic data.⁸

DISCUSSION

While many of the studies identified in this review suggest that surgery is beneficial for at least some patients in the short-term, the reliable interpretation of these data is severely limited by methodological flaws that are inappropriate for generating unbiased evidence on the long term balance of benefits and risks of surgery for PD. For example, random errors attributable to the play of chance are likely to occur when numbers are small, as with most trials of surgery for PD. Furthermore, non-randomised studies are notoriously unreliable because of the potentially large biases from patient selection factors.⁹ Randomisation avoids selection bias by balancing measurable and unmeasurable prognostic factors between treatment and control groups. There are other potential sources of bias, even in the randomised trials, such as lack of or ineffective blinding, which is usual in PD surgery trials because of ethical and practical issues surrounding sham surgery¹⁰ but such biases are minor compared with those in non-randomised studies.

Publication bias is important to consider and is likely to exaggerate any real benefits from surgery with successes more likely to be reported than failures. This can be compounded by multiple publications of the same series of patients that are often difficult to identify. Problems in evaluating clinical trial evidence could be minimised by using registration systems to prospectively identify RCTs at initiation. Similarly reporting of trials could be improved by adopting the CONSORT guidelines.¹¹

Ideally, meta-analysis of randomised trials of surgery for PD for each question addressed would help to increase numbers and also give a more balanced view of the total evidence. However, no reliable conclusions about the balance of benefits and risks of surgery could be drawn at this stage because results were too inconsistently and incompletely reported. Such meta-analysis at present would be of limited value because of the inappropriately short follow up in most studies. PD is a progressive disease and it is therefore essential that future studies of surgery evaluate long term effectiveness and safety.

The choice of clinician rated assessments of motor impairments and disability (for example, UPDRS) as sole outcome measures in most previous surgical studies also limits their value. PD in its advanced stages is complicated by additional problems such as depression, dementia, hallucinations, and sleep disturbances. These may have a greater impact on quality of life than physical impairment. It is important that future trials should include reliable, validated, patient rated quality of life measures (for example, PDQ-39) to assess aspects of the patient’s life that are considered important by them rather than by clinicians.¹² Furthermore, in this age of limited healthcare resources, cost effectiveness needs to be assessed as well as clinical effectiveness.

To obtain more reliable evidence, there is an urgent need for larger scale randomised trials, with long term follow up, including patient rated quality of life and health economics as outcome measures. Participation in such trials (for example, the UK Medical Research Council’s PD SURG Trial) should be encouraged.