This review is based on a literature search conducted primarily using the PubMed and Ovid databases in September 2014 with further searches including PsycINFO, CINAHL and Embase. Search terms used were “(Transcranial magnetic stimulation OR TMS OR rTMS OR brain stimulation) AND (geriatric OR elderly OR late life OR late onset OR age OR older) AND depress*.” Articles in English published prior to the search date were included.

RCTs evaluating the efficacy and safety of TMS (irrespective of pulse frequency) were included in this review. Due to paucity of research in this area we also included uncontrolled trials and retrospective reviews in the field. The primary outcome measure was chosen as a categorical response or a continuous change defined a priori in the individual trials on the basis of a standardised depression rating scale (e.g., Hamilton Depression scale). In addition we also studied the adverse events reported in the trials. Only studies with a mean age of patient sample > 60 years were included. An exception to this rule was applied when subgroup analysis was performed with mean age > 60 for at least one subgroup.

One reviewer (PS) assessed articles identified as a result of the search and where necessary, in cases of uncertainty, a second reviewer (LP) also considered the articles. Studies were initially screened for relevance in the title and abstract. Duplicate articles were removed and the remaining full text articles were reviewed. The final papers chosen for inclusion were also hand searched to ensure relevant references were not missed.

In addition to the systematic review focussing on efficacy and safety in geriatric depression, we also appraised the wider literature to investigate the known moderators of treatment efficacy in geriatric depression. For this purpose, a broader search was conducted using the above search terms, but without restricting studies to a specific mean age of the samples.

We identified 4 RCTs that specifically included older subjects, resulting in mean sample age > 60 years (Table 2). Of these 2 trials reported no benefit from rTMS when compared to sham[30,31] while 2 other trials (presented in a single manuscript) reported a substantial benefit[32]. Both Manes et al[30] and Mosimann et al[31] used stimulation intensity that was fixed at or lower than the motor threshold. In contrast, Jorge et al[32] employed a higher intensity. Jorge et al[32] have also shown that the delivery of a higher dose (larger number of pulses - 18000 instead of 12000) is more beneficial in the older than in the younger age group of patients with treatment resistant vascular depression[32].

Several uncontrolled trials that focus on the efficacy of rTMS in older samples (age > 60) have been published. We identified 5 uncontrolled trials with mean sample age > 60 (Table 3), and 2 trials that specifically studied the subgroup effect for patients aged > 60[33] or 65[34]. Response rates (defined as proportion showing a 50% or greater drop in HDRS score) greatly varied, ranging from 18% to 58.5%. Notable variation was also observed for the rTMS dose parameters, with some studies using 100% or less of Motor Threshold (MT: the intensity of magnetic pulse required to elicit an observable thumb movement)[34-36] and others using > 100% of MT[33,37]. One study employed an intensity adjustment for predicted frontal atrophy by measuring the MRI-based distance between the coil and the cortical surface[38]. Four studies explicitly studied the association between age and treatment response[33,34,37,38]. While 2 studies found a reducing treatment effect with increasing age[33,37], the other 2 studies failed to find a similar association[34,38]. Taken together, there is no consistent evidence negating the utility of rTMS in the elderly. The heterogeneity among studies suggests that several clinical and treatment-related factors may moderate the therapeutic effect of rTMS in the elderly.

A consistent observation supporting a high degree of tolerability and safety among the elderly patients emerged across the RCTs and the uncontrolled trials. In most of the studied cohorts, there were no dropouts due to adverse effects. Only 2 studies reported loss of subjects due to adverse effects[33,39]. Out of 102 patients of various ages reported by Pallanti et al[33], dropouts occurred due to anxiety (n = 4), insomnia (n = 5), mood elevation (n = 1), discomfort of the scalp (n = 5), and hospitalization during treatment (n = 4). The 18.6% dropout rate reported in this study is unusually high, when compared to the overall 4.5% observed from other larger multisite studies[40]. Pallanti et al[33] did not study the influence of age on the dropouts. In the sample reported by Abraham et al[39] one subject dropped out due to local discomfort. Jorge et al[32] reported local discomfort in around 33% of the randomized sample. But no subjects discontinued treatment due to this effect, and the distribution of side effects was not significantly different between the sham and the active treatment groups[32].

ECT has an important role in the clinical management of geriatric depression. A substantial proportion of all referrals received by ECT clinics fall into the age group > 65[41]. ECT appears to be generally safe and highly efficacious in geriatric depression with response rates close to 70% reported in some studies[42,43]. Interestingly, older age is associated with a more favourable response to ECT[44,45]. In this context it is important to consider direct comparisons ECT and rTMS in treating geriatric depression.

We located 4 meta-analyses that synthesize the evidence from studies that compare rTMS and ECT (Table 4). The sample mean age of these studies span a range of 31.0 to 63.6 years[46-49]. All 4 meta-analyses find that ECT is superior to rTMS in short-term when categorical response is considered, though Ren et al[46] indicate that change in continuous Hamilton Depression Rating Scale (HAMD) scores do not significantly differ between the two treatments. Furthermore, Berlim et al[47] noted baseline differences favouring ECT (shorter duration and more severe illness). There is insufficient data to comment on medium or long-term efficacy[46]. Interestingly, two of the three individual trials showing significant difference in favour of ECT[50,51] included participants with higher mean age than the rest. When comparing ECT and rTMS, Janicak et al[52] noted that older subjects required a higher number of rTMS sessions to experience a favourable treatment response when compared to younger subjects. Nevertheless, when considered as a moderator in the meta-analytic setting, mean sample age did not significantly predict the effect size of rTMS v ECT differences[47]. An important predictor appears to be the presence of psychotic symptoms[46,47]; in the presence of psychosis, ECT appears to be significantly superior to rTMS[51]. But in samples with no psychosis, rTMS performs as well as ECT[53]. Given that many patients with geriatric depression have psychotic symptoms, this may partly explain the superior efficacy of ECT in this setting, though more focused studies are required to provide conclusive support to this notion.

Rapidity of response is an important factor for which ECT is sought in the elderly. Difference in speed of response has not been studied directly in the 4 meta-analyses, but an important observation suggests that ECT may be superior to rTMS in terms of the rapidity of response. Xie et al[48] observed that when rTMS treatment period was less than 4 wk, rTMS was significantly inferior to ECT. When the treatment period was increased to 4 wk, the difference between rTMS and ECT began to decrease, suggesting that ECT results in far more cases of early response than rTMS.

With respect to cognitive side effects in the relatively younger samples included in meta-analyses, ECT was associated with more impairment in several cognitive domains[46], though the dropout rates (acceptability) did not differ between the two interventions.

Studies included in this review examined the role of several moderators on clinical response as described below.

Age: Age itself is not a consistent predictor of the antidepressant effects of rTMS as shown by conflicting evidence from adult samples[29]. A large naturalistic study (n = 307) that included a broad age range (18 to 90) of patients receiving rTMS for depression concluded that age had no notable influence on the improvement in quality of life noted in patients[54]. Furthermore the RCTs and uncontrolled trials reviewed here defy the notion that rTMS is ineffective in geriatric depression. Given the wide range of response rates reported among the elderly, it becomes important to understand the factors that may influence treatment response in this group. Some of the important factors that have been identified in this context are: (1) brain atrophy; (2) intensity and number of pulses (dose-response relationship); and (3) clinical profile of patients.

Brain atrophy: Manes et al[30] first observed that the volume of frontal cortex was inversely related to the antidepressant response to rTMS in the elderly. This observation was later confirmed by Fabre et al[35] and Jorge et al[32]. In an atrophic brain, the distance between scalp and cortex (Scalp-Cortex Distance or SCD) increases. When applying rTMS the intensity of the magnetic field decays exponentially as we move from the scalp surface to the cortical surface[55]. Herbsman et al[56] have shown that nearly 60% of inter-individual variation in Motor Threshold can be explained by the distance between scalp and motor cortex alone (motor SCD). Using a simple linear model, Stokes et al[57] estimated that for every 1 mm increase in motor cortex SCD, a 2.8% increase occurs in consistent with this observation, the degree of rTMS induced change in frontal perfusion corresponds inversely with increasing frontal cortex SCD in depressed subjects[58].

One may argue that as the intensity of stimuli used in the treatment of depression is generally based on an individual’s MT, the effect of brain atrophy is already taken into account by adjusting the intensity applied to frontal cortex on the basis of MT. But in fact, with age, a disproportionately higher volume reduction occurs in the frontal cortex compared to motor cortex[59]. As a result, the MT fails as a “benchmark” to calculate the intensity required for stimulating the frontal cortex. Several authors have proposed approaches to adjust the intensity of rTMS according to the measured SCD of the stimulated cortex[38,55,60,61].

But the utility of employing a SCD-adjusted intensity is questioned by other authors. Firstly, the relationship between SCD and treatment response is not a consistent one. Unlike Fabre et al[35], neither Kozel et al[62] nor Jorge et al[32] could find a relationship between frontal SCD and treatment response, though Jorge et al[32] noted that the volume of prefrontal cortex mediated the beneficial effects of treatment. Secondly, while higher SCD is a proxy for age-related (or pathological) atrophic process, the biological effect of brain atrophy is better reflected by a measure of cortical thickness or surface area. Interestingly, lower motor cortical thickness in older adults appears to be associated with lower (not higher) resting MT[63]. This indicates that brain atrophy could result in two contrasting changes 1. A reduction in the strength of the magnetic field that reaches the cortex upon the application of rTMS 2. An increase in the excitability of cortex, such that even smaller intensities are now sufficient to stimulate the brain. Taken together, there may not be any appreciable overall change in the MT with age. In fact, direct comparisons of the MT in the elderly and younger subjects have been so far inconclusive with some studies reporting age-related increase in MT, whereas others reporting no differences in MT between different age groups (reviewed by Rossini et al[64]). Most studies examining the effect of rTMS on cognition in dementia employ < 100% MT on DLPFC but still observed notable improvements in cognitive performance[65]. It is likely that physiological effects of rTMS are induced despite notable cortical atrophy when stimulating frontal cortex in the elderly. On the basis of simulated models of brain atrophy that included both volume shrinkage and sulcal widening, Wagner et al[66] argue that SCD based correction of pulse dose/intensity[38,55,60,61] is an inaccurate oversimplification that fails to consider the geometrical changes and altered excitability of the shrunken brain.

Despite the limitation of distance adjustment approaches, a superior therapeutic response is noted when dose-adjusted intensity is employed in treating depression[38,67], when compared to conventional doses delivered as a fixed proportion of MT. The intensity adjustment approaches employed by Nahas et al[38] 2004 and Mosimann et al[67] require an anatomical MRI to calculate the SCD. A more pragmatic solution was put forward by Johnson et al[68], on the basis of a multisite RCT that studied 185 adults of age 22-69 years. The authors reported that most of the effect of frontal atrophy could be overcome by using an intensity that is 120% of MT[68]. This resulted in distance-adjusted intensities that ranged from 93% to 156% of MT and was found to be both safe and efficacious in patients. So far, no studies in the elderly have evaluated the efficacy of 120% MT protocol, though most TMS clinics are now employing an intensity of 120% MT for left prefrontal stimulation working age adults with depression[54,69].

Number of pulses delivered: In rTMS studies of depressed working age adults, a relationship between dose and response has been previously reported. A review by Gershon et al[70] noted that rTMS studies delivering higher doses (pulses/day) for longer duration of treatment were more effective than shorter studies using lower doses. In this context, it is worth noting that over the years, the total number of pulses delivered in treatment of depression has increased in general[11,19,71], partly explaining the increase in effect size of rTMS in depression over the years[71].

An interesting observation emerges from the TMS trials (both RCTs and uncontrolled trials) in the depressed elderly reviewed here; almost all studies that conclude in favour of efficacy of rTMS in the elderly use a high number of pulses (18000[32,34,36,38]) while most of the trials that found age-related reduction in efficacy employed smaller number of pulses (2500 to 6300[30,33,37]). Jorge et al[32] examined this issue closely by designing 2 experiments within a single trial. They noted that the elderly subjects with vascular depression who received 18000 stimuli in total responded more robustly than those who received only 12000 stimuli a day. While there are a number of other factors that could have influenced the heterogeneity of individual study results, the observation of lack of age-related reduction in efficacy when using higher doses (number of pulses) merits further consideration in clinical practice.

Clinical profile: Apart from age, several other features differentiate geriatric depression form the depression seen in working-age adults. These include the higher prevalence of treatment resistance[3,4], the excess of somatic/melancholic[72] and psychotic features[73], higher degree of cognitive impairment/dementia and medical comorbidity[74,75] among the elderly. Evidence for the influence of these factors on rTMS treatment response in the elderly is considered below.

Treatment resistance: Meta-analytic evidence in depressed working age adults supports the use of rTMS in the treatment of refractory depression (no. of studies = 18 with at least 2 antidepressant failures; duration 1 to 6 wk; sample sizes 12 to 74; NNT = 5 for remission)[76]. Some studies in depressed working age adults have observed a relationship between the degree of treatment-refractoriness and higher response rates for rTMS[77], though other studies have observed a reversed relationship[27,69]. Allan et al[29] reviewed this issue and concluded that the presence of TRD in a sample does not influence its response to rTMS. In line with this, there is a lack of a predictable association in either direction among the studies considered in this review as well. Of the 12 rTMS trials (RCTs and uncontrolled studies) in the depressed elderly reviewed here, 5 include samples with at least 2 antidepressant failures while the rest include samples with minimum 1 antidepressant failure. Studies concluding in favour of rTMS in treating geriatric depression have included samples both with TRD (2 antidepressant failure)[33,34,36,78] and 1 antidepressant failure[32,38]. Taken together, the influence of the degree of treatment refractoriness on therapeutic response to rTMS in the elderly is still uncertain. No focused studies directly comparing subjects with TRD and without TRD in geriatric depression have been reported to date.

Psychosis: Several studies in working-age samples support the notion that ECT achieves a superior treatment response in psychotic than non-psychotic depression. In their meta-analysis comparing ECT and rTMS, Berlim et al[47] reported that ECT was significantly superior to rTMS in primary studies where the samples had psychotic symptoms. Ren et al[46] concluded that rTMS works as well as ECT in the absence of psychosis. The study that reported one of the highest response rates for rTMS in patients > age 60 (47.2% response after 3 wk) only included patients without psychotic symptoms[33]. The mediating influence of psychosis on the antidepressant effect of rTMS in the elderly has not been directly investigated to date.

Melancholia: Melancholic features are more common among the elderly with depression. rTMS does have an alleviating effect on core melancholic features such as psychomotor retardation in working-age adults[79]. A reduction in melancholic features occurs irrespective of overall clinical response, suggesting a symptom-specific effect of rTMS on melancholia[80]. In particular, psychomotor agitation, which is a common feature among the elderly, appears to respond to high-frequency rTMS[81]. While some rTMS studies in working-age samples suggest that the presence of melancholic features itself indicate poorer overall response to rTMS[82] other studies suggest the opposite[83]. Among the elderly, the specific effect of melancholic features on rTMS treatment response has not been examined so far.

Cognitive impairment: Patients with other comorbid psychiatric illnesses including dementia are often excluded from rTMS studies of depression. Geriatric depression is accompanied by a greater degree of cognitive impairment and a substantial number of patients, who are initially found to have depression with cognitive impairment, later develop dementia, suggesting that a subgroup of geriatric depression may indeed be a prodrome for dementia[84]. After an episode of depression, the risk of dementia increases by nearly 87% over a period of 25 years[85].

A number of studies indicate that rTMS may improve deficits and delay cognitive decline in Alzheimer’s dementia[86], with some studies suggesting a therapeutic effect that lasts for nearly 3 mo[87]. Rutherford et al[88] suggest that some of the improvement in cognitive performance seen in patients with dementia after rTMS treatment can indeed be attributed to the antidepressant effect of rTMS. Furthermore, as concluded by Ren et al[46] in their meta-analysis comparing rTMS and ECT, rTMS appears to be superior to ECT in terms of cognitive side effects. Taken together, this suggests that contrary to the prevailing practice in most ECT clinics, there is no need to exclude elderly subjects with notable cognitive impairment from rTMS studies.

Medical comorbidities: To our knowledge, the moderating influence of medical comorbidities on the therapeutic response to rTMS has not been studied either in the working-age samples or the elderly. Most rTMS trials in working-age adults have excluded patients with significant medical illnesses; as a result the moderating effect of this variable has not been studied in the extant meta-analytic literature on the efficacy of rTMS. Among studies considering geriatric depression, some authors excluded patients with uncontrolled medical illnesses[31,33,38,39,78] while others included subjects with medical comorbidities[32,37]. rTMS has been investigated as a treatment modality in several neurological disorders that often present with comorbid depression in the elderly (e.g., parkinsonism, stroke[89]), and has been found to be safe.

While the potential of rTMS in the treatment of depression is acknowledged widely, it has not entered the standard stepped-care approach recommended for the treatment of depression in the elderly. An appraisal of the rTMS literature relevant to geriatric depression highlights several deficiencies and offers insight on the pitfalls of recommending routine use of rTMS in geriatric depression. Firstly, the practice of excluding older adults from rTMS trials has resulted in a dearth of good quality RCT data in this age group. The available evidence does not provide an unequivocal support for age related reduction in the antidepressant effect of rTMS. In contrary, it hints at several possible mechanisms for the inconsistently observed differential treatment response. Secondly, there is a scarcity of experimental studies investigating the variations in rTMS parameters to improve response rates in the elderly. Third, despite the numerous phenomenological and neurobiological differences between working-age adults and elderly with depression, moderators other than age have not been systematically studied in rTMS studies of geriatric depression.

We acknowledge that despite being broadly inclusive, several grey literature such as conference abstracts reporting on TMS efficacy for geriatric depression might have been missed. In addition, publications in languages other than English were not included in this search.

Future studies are needed to replicate and confirm the hypotheses regarding the efficacy of higher number of pulses in the elderly. In addition, systematic exploration of the influence of various features (other than age) that differentiate geriatric depression from early-life depression is required to understand the moderating effect of these features.

Despite the limitations identified above, several clinical practice points emerge from this appraisal of rTMS focussed on older adults with depression. On the basis of the current evidence, the practice of excluding elderly patients from rTMS clinics and trials cannot be supported. Age-related reduction in antidepressant efficacy, even if present, is not specific to TMS[90]. Adjusted dosing schedules that deliver higher intensity and pulses appear to improve the therapeutic response in the elderly, and these dose variations must be made available for the elderly depressed seeking treatment. In contrast to working-age adult samples where TMS is considered as an alternative “in line” with ECT, for elderly depressed patients, given the indications for a superior efficacy of ECT, rTMS could be offered either after an unsuccessful or poorly tolerated trial of ECT. In some carefully selected cases of non-psychotic depression, rTMS could be a potential alternative to ECT when rapidity of response is not crucial but undesirable cognitive side effects to ECT are highly likely. While it is premature to recommend rTMS for regular use in geriatric depression, continued exclusion of this group of depressed patients from a well-tolerated and safe treatment option for resistant depression on the basis of their age appears to be clearly untenable.