Age or age of onset: which is the best criterion to classify late-life depression?

In late-life depression (LLD), several differences between patients whose first episode is reported after age 65 (late-onset depression, LOD) and those with early-onset depression (EOD) might reflect the effects of brain ageing. To test this hypothesis, we analysed the impact of current age and age at illness onset on a number of clinical and cognitive manifestations in 438 outpatients with major depressive disorder aged >60 years, treated with venlafaxine for 12 weeks. When compared to the EOD group, patients with LOD were older (P < 0.00001) and associated with lower depression severity (P = 0.0029), lower global cognitive functioning [Mini-Mental State Examination (MMSE): P = 0.0001; Repeatable Battery for the Assessment of Neuropsychological Status: immediate memory, P = 0.0009, and delayed memory, P < 0.00001; Delis-Kaplan Executive Function System measuring executive functions: Trail-Making Test (TMT) – P = 0.0004 and Colour-Word Interference Test, Inhibition – P = 0.0063], and more dyskinesias (Abnormal Involuntary Movement Scale: P = 0.0006). After controlling for its interactions with age of onset, current age was inversely correlated with Montgomery Åsberg Depression Rating Scale scores at baseline (P < 0.00001) and week 12 (P = 0.0066), MMSE (P < 0.00001), delayed memory (P < 0.00001), and TMT (P = 0.0021). Age of onset predicted impairment in immediate (P = 0.023) and delayed memory (P = 0.0181), and dyskinesias (P = 0.0006). Although most features of LLD are related to ageing rather than to late-onset, LOD is a possible separate diagnostic entity characterised by memory dysfunction and increased liability to movement disorders.


Introduction
Depression is a common and serious health concern in older people. In this group, the prevalence of any depressive disorder ranges from less than 10% to over 40% depending on diagnostic thresholds and settings (Beekman et al., 1999;Djernes, 2006). The occurrence of a depressive disorder after the age of 60 is called late-life depression (LLD). At the clinical level, LLD has been characterised by chronic course (Comijs et al., 2015) and increased morbidity and mortality (Unützer, 2007;Wei et al., 2022). At the diagnostic level, LLD is a broad term that does not differentiate between patients who are depressed for the first time in late life (late-onset depression, LOD) and those who experience late-life relapses of depressive disorders already present during their youth or adulthood (early-onset depression, EOD). In psychogeriatrics, a large amount of literature has shown that LOD and EOD are associated with distinct clinical (Holroyd and Duryee, 1997;Benazzi, 2004;Xiao et al., 2020), neurocognitive (Sachs-Ericsson et al., 2013;Mackin et al., 2014), and morphological (Ballmaier et al., 2008) features. In addition, several potential peripheral biomarkers for depression in the elderly, possibly related to the specific symptom profile and disease phase, have been suggested (Vogelzangs et al., 2014;Fanelli et al., 2019Fanelli et al., , 2020. Recent studies have also reported differences between early and late-onset suicide attempters (Kenneally et al., 2019;Szücs et al., 2020;Perry et al., 2021). All these findings apparently support age at onset of the first depressive episode as a stem diagnostic criterion for LLD. On the other hand, ageing process is itself characterised by physiological changes that involve the central nervous system, with consequences for neurological and cognitive functions (Amarya et al., 2018). In the geriatric population increasing age has been associated with greater cognitive impairment, loss of autonomy and diminished quality of life (Cohen-Mansfield et al., 2013;Kim, 2021). Such age-related differences could be reproduced in LLD. Thus, although most findings point to a clear-cut distinction between LLD and adult depression, their interpretation is actually a conundrum because some of them could be specific characteristics of LOD, whereas others could simply mirror the effects of ageing. However, this bias is seldom considered in LLD research.
The general purpose of this study was to disentangle the role of current age and age of onset as classifiers for LLD. In detail, we used data from a large sample of depressed outpatients aged 60 years or older with the following goals: (1) to test the association of current age and age of onset with a number of clinical and neurocognitive variables related to LLD severity; (2) to investigate the impact of age and age of onset on treatment adherence and tolerability; and (3) to analyse interactions between current age and age of onset as moderators of LLD.

Study sample
We performed a secondary analysis of the IRL GREY (Incomplete Response in Late-Life Depression: Getting to Remission) sample (NCT00892047). IRL GREY is a multi-site study including subjects aged 60 and older, with a Diagnostic and Statistical Manual of mental disorders (DSM)-IV diagnosis of major depressive disorder (confirmed by the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID-I) and a Montgomery Åsberg Depression Rating Scale (MADRS) score of at least 15). Exclusion criteria were dementia [based upon DSM-IV criteria, as well as a Mini-Mental State Examination (MMSE) score less than 24], lifetime diagnoses of bipolar or psychotic disorders, alcohol or substance abuse/dependence within the past 3 months and high risk for suicide (e.g. current/recent intent or plan) (Joel et al., 2014).
All subjects were included in the IRL GREY trial after obtaining their written informed consent. This research group certifies that data collected from the IRL GREY trial were exclusively used for scientific investigation and, before obtaining access to them, the objectives of our investigation were clearly reported in a request form.

Treatment and assessment
The IRL GREY trial design included three treatment phases, each lasting 12 weeks. During phase 1, eligible patients who gave their written informed consent were openly treated with venlafaxine XR (up to 300 mg/day). At the end of this phase, participants with incomplete response were randomised to receive either aripiprazole (2-15 mg/day; target dose: 10 mg/day) or placebo in addition to venlafaxine (phase 2), with the goal of achieving remission (two consecutive assessments with a ≤10 MADRS score). Finally, phase 3 patients who had remitted were maintained in the treatment arm to which they had been assigned during phase 2, in order to determine the stability of remission (Joel et al., 2014). Our analysis was based on data collected during phase 1.
Cognitive assessment was performed including the MMSE (Folstein et al., 1975) for global cognitive functioning, the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) (Randolph et al., 1998) to evaluate memory performance, and two tests [i.e. the Colour-Word Interference Test (CWIT) -Inhibition part, and Trail-Making Test (TMT)] included in the Delis-Kaplan Executive Function System (D-KEFS), which is a neuropsychological test battery evaluating executive functioning (Fine and Delis, 2011). Health assessment was completed by the Cumulative Illness Rating Scale for Geriatrics (CIRS-G) (Linn et al., 1968;Miller et al., 1992), which rates 13 organ systems from 0 (no problem) to 4 (organ failure) and the 36-item Short Form Health Survey (SF-36) (Ware and Sherbourne, 1992) as a measure of health-related quality of life whereas side effects due to antidepressant treatment were ascertained using the Udvalg for Klinske Undersogelser Side Effects Scale (Lingjaerde et al., 1987), Abnormal Involuntary Movement Scale (AIMS) (Lane et al., 1985) and Barnes Akathisia Scale (Barnes, 1989).

Statistical analyses
The association of current age and age of onset with clinical variables, such as baseline depression score, suicidal ideation, antidepressant response, and cognitive domains, was explored in the IRL GREY sample. After a preliminary power analysis conducted using G*Power 3 (Faul et al., 2007), age (young-old: 60-74 years vs. oldold: ≥75 years) and age of onset (EOD: <65 years vs. LOD: ≥65 years) groups were compared by means of univariate tests (Student's t or χ 2 tests as appropriate).
The statistical significance threshold was set at α = 0.025 due to the number of comparisons, but without a formal correction for multiple testing given the a priori hypothesis (Amrhein et al., 2019).
In subsequent steps, age and age of onset were included in multiple regression or logistic regression models to ascertain their interacting effects as predictors of clinical variables.
The results of the IRL GREY trial's primary analyses have been reported elsewhere (Joel et al., 2014;. Preliminary power analysis suggested that our sample was enough powered (0.80) to detect small-medium effect sizes (d = 0.36) in comparisons between 60-74 and ≥75 years groups, corresponding to differences of 2 points for MADRS, 1.6 points for suicidal ideation, 6 points for memory tasks, and 1.3 points for executive function tests. As regards age of onset groups, the minimum detectable effect size was d = 0.42 which corresponded to differences of 2.3 points for MADRS, 1.9 points for suicidal ideation, 7 points for memory tasks, and 1.5 points for executive function domains.

Discussion
This study aimed to investigate the role of current age and age at onset of first depressive episode as specifiers for late-life depressive disorder. These two parameters were found to be highly interrelated in the IRL GREY sample, therefore it was interesting to disentangle their specific effects.
Most characteristics reported in LOD did not appear to be typical manifestations of the disorder but, rather, consequences of older age. In particular, older individuals had less severe depression symptoms but more cognitive impairment than their younger counterparts. The negative correlation between age and depression severity was already known in previous studies. It had been suggested, for example, by an analysis of the National Comorbidity Survey-Replication in which researchers had observed that major depressive episodes were less prevalent and severe in individuals aged 75 or more compared to younger age groups (Kessler et al., 2010). More recently, the association between depressive symptomatology and age in LLD has been characterised by an inverted U-shape, with a peak between age 70 and 80, followed by a significant decrease in >80 years old patients (Barrenetxea et al., 2022). A possible explanation for such an age-related decrease in depressive symptoms is that only a minority of elderly patients could maintain the same depressive manifestations as adults whereas in most others affective symptoms would be replaced by somatic or cognitive manifestations (Gonda et al., 2009;Skoog, 2011) that are neither included among DSM criteria for major depressive episode nor assessed by common scales for adult depression such as the MADRS. It is worth noting that, although our ≥75 years old subjects were less severely depressed, their levels of suicidal ideation were similar to those detectable in younger individuals. This highlighted the presence of a complex network of suicidal risk factors operating in old age depressive disorders (Fernandez-Rodrigues et al., 2022). In particular, since depression and physical illnesses are both contributors to suicide risk in the elderly population (Obuobi-Donkor et al., 2021), the lower levels of depressive symptoms reported in older patients could be counterbalanced by their greater physical burden. The effectiveness of antidepressant medications for LLD is still unclear. If some researchers have found positive results, which led them to recommend antidepressant medications as first-line treatments for elderly depression (Pruckner and Holthoff-Detto, 2017), for others these drugs would only warrant questionable benefits and, therefore, they should be cautiously prescribed to older people (Mallery et al., 2019). The IRL GREY study had already shown a positive response to antidepressant treatments in patients aged >60 (Hall et al., 2015). In our reanalysis, a new finding was that such a response could be maintained in older age. Similarly, a recent meta-analysis has estimated an overall response rate of >50% in elderly depressed patients, reporting no significant age-related variations (Gutsmiedl et al., 2020). However, by controlling for baseline severity, we observed that depressive symptoms had a more clear-cut decrease during antidepressant treatment in ≥75 years old group. Thus, not only antidepressant response was present in older age, but it seemed to be heightened. This possibility has already been discussed commenting on the results of a large population-based study on completed suicide. In that cohort, individuals aged >80 treated with antidepressant medications were less likely to commit suicide than their counterparts without antidepressant therapy, but this difference was not seen in younger age groups (Erlangsen and Conwell, 2014). The impact of antidepressant medications in elderly samples is affected by biological and clinical variables. Biological research has emphasised the role of age-related changes in serotonin function in different brain regions (Goldberg et al., 2004). At the clinical level, it is important how antidepressants are titrated. As a general rule, slow titration should be preferred as it has been associated with better antidepressant response and greater tolerability (Olgiati et al., 2014;Olgiati and Serretti, 2015).
Cognitive impairment is one of the best characterised traits of LLD (Masse et al., 2022) that has been associated with response and attrition during antidepressant treatment, as well as with suicidal behaviour (Cristancho et al., 2018;Jordan et al., 2020;Lin et al., 2021;Pan et al., 2022). The degree of cognitive dysfunction was suggested to be higher in LOD in comparison with the EOD subtype (Ly et al., 2021). Before concluding that such differences are real; however, it is necessary to unravel possible confounding factors. For example, Cheng et al. (2020) have reported no associations between cognitive function and age of onset in a sample of older adults with remitted depressive disorders. To comment on their negative findings these researchers have argued that in previous studies the relationship between cognitive impairment and LOD could have been biased by active or residual depressive symptoms. We analysed the confounding effect of age. In our sample, there were several cognitive differences between LOD and EOD groups: although most of them were related to chronological age rather than to the age of onset, suggesting they could simply reflect the impact of brain ageing (Diniz et al., 2021), memory dysfunction has emerged as a true feature of LOD. This finding was consistent with prior studies such as Mackin et al. (2014), who reported a correlation between memory impairment and LOD in older adults with depression and executive dysfunction and, more recently, a study of healthy older individuals who later developed depression, among whom mild cognitive impairment preceded depression onset (Almdahl et al., 2022). Another characteristic that was associated with LOD in our sample was an increased risk of developing dyskinesias, notwithstanding the assumption of lower antidepressant doses. This could be more easily understood by considering that in older subjects who develop Parkinson's disease the onset of this disorder is frequently preceded by LOD (Gustafsson et al., 2015;Wang et al., 2018;Kazmi et al., 2021). Hence, there might be a greater instability of dopamine circuits in some patients who develop depression in late life (Taylor et al., 2022), which would also facilitate the onset of movement disorders.
The most evident weakness of our analysis was the small size of the LOD group. This caused a loss of statistical power and, as a consequence, we cannot exclude that a number of existing differences between LOD and EOD were overlooked. A further weakness was to have assessed depressive symptomatology by means of the MADRS. In fact, this scale has not been specifically developed for geriatric depression and does not cover somatic and other commonly reported depressive symptoms that are common in the elderly population. On the other hand, a strength of this study was the possibility of analysing a broad set of clinical variables thanks to the accurate phenotypic characterisation of this sample.
In conclusion, our results suggest that ageing process might exert a strong pathoplastic effect on depressive disorders which firstly occurred in young adulthood; further research is needed to elucidate the underlying biological mechanisms in order to better tailor LLD treatments according to patients' ages. On the other hand, specific changes in brain function may occur in old age, thereby facilitating the onset of depression, memory dysfunction, and movement disorders in previously healthy individuals.