While depression is a heterogeneous entity, findings from this study suggest that the CVLT may be a useful tool for characterizing and differentiating between learning and memory processes in late-onset MDD and the more prevalent miD in older adults. As expected, those with miD performed comparably with controls and significantly better than those with MDD on the six CVLT indices most consistently used to differentiate among healthy individuals and those with cortical and subcortical pathologies.
Specifically, the MDD group scored significantly lower than both the miD and control groups in acquisition (total recall on Trials 1–5 of List A), rate of learning, and short-delay free recall. Compared with controls, the MDD group also showed significantly greater improvement in recall when provided with recognition prompts. This latter finding supports a retrieval deficit in MDD and bolsters the notion that subcortical pathology is likely to result in reduced performance on more effortful recall tasks while recognition is intact. Semantic clustering and intrusion errors on cued recall of List A were the only two of the six CVLT indices that did not differ significantly among the three groups. The lack of significant differences on semantic clustering is consistent with the results of two other studies (Elderkin-Thompson et al., 2003
) examining CVLT performance in late-life miD and MDD when the index episode was the first mood episode, as it was in the current study; in Elderkin-Thompson and colleagues (2007)
, semantic clustering did significantly differ between the groups when individuals with a history of depressive episodes were also included in the analyses. Intrusions on cued recall were not specifically examined in these studies (Elderkin-Thompson et al., 2003
). Since the semantic clustering ratio did not discriminate among the groups in this or the two prior studies by Elderkin-Thompson and colleagues (2003
), it is possible that the CVLT findings from this study may readily be generalized to other verbal list learning tests that do not measure semantic organization (e.g., Rey Auditory Verbal Learning Test). However, this warrants further study with similar samples, particularly in light of the findings from Elderkin-Thompson and colleagues (2007)
that semantic clustering appeared to mediate scores on some CVLT indices.
The findings also provide support for the use of verbal learning and memory classification groups to characterize the heterogeneity among elderly individuals with depressive disorders. Three distinct verbal learning and memory profiles emerged from cluster analysis of the three CVLT indices previously used to discriminate among “normal” abilities and subcortical and cortical pathologies (total recall on Trials 1–5 of List A, the difference between recognition discriminability and Trial 5 recall, and intrusion errors produced on the cued recall of List A). Within the miD group, roughly 74% were classified as normal, nearly 15% were categorized as subcortical, and approximately 11% were classified as cortical. The majority of MDD patients (~54%) were assigned to the subcortical verbal learning and memory profile, almost 31% were classified as normal, and about 15% were categorized as cortical. The proportion of individuals assigned to each cluster differed significantly, and the three clusters resembled theoretical representations of normal, cortical, and subcortical profiles, as hypothesized.
Consistent with this study's hypotheses, most individuals in the miD group demonstrated normal verbal learning and memory profiles. While MRI studies have documented that prefrontal lobe volumes in elderly individuals with miD fall between age-matched controls and those with MDD (Kumar et al., 1997
), such findings do not appear to translate into performance differences in verbal memory between those with miD and controls. Although Elderkin-Thompson and colleagues (2007)
found similarly deficient CVLT performance between miD and MDD groups, at least half of both of these groups had previous depressive episodes, and the study did not examine cued intrusions or the difference between recognition discriminability and free recall.
The pattern of performance exhibited by the MDD group is also consistent with the study hypotheses, as well as a subcortical dysfunction hypothesis of verbal learning and memory deficits for a subgroup of depressed patients (Massman et al., 1992
). Compared with Massman's group, this study's MDD sample showed a higher proportion of subcortical and cortical profiles. This discrepancy may be due to the compounding effects of age and depression on cognition of MDD patients in this study (e.g., King, Caine, Conwell, & Cox, 1991
). The finding that classification status for both patient groups was not significantly influenced by severity of depression was also congruent with findings from Massman and colleagues (1992)
and other investigators (Sweeney, Wetzler, Stokes, & Kocsis, 1989
It is notable that while there were significant differences in verbal learning and memory profile cluster assignments (“normal,” “subcortical,” or “cortical”) between
the groups, there was heterogeneity in profile assignments within
each group. While some variability may be due to moderating influences not examined in this study (e.g., use of estrogen replacement therapy [ERT] in women, non-psychotropic medications, psychotherapy, previous hospitalizations), it should also be considered that elderly individuals with miD are at a significant risk for subsequent development of MDD (Cuijpers et al., 2005
; Judd & Akiskal, 2002
; Lyness et al., 2009
). Furthermore, MDD may present as an early sign of a later developing dementia (e.g., Agbayewa, 1986
; Kral & Emery, 1989
; Nussbaum, Kaszniak, Allender, & Rapcsak, 1995
). Indeed, Lyness and colleagues (2009)
reported that untreated miD conferred a 7-fold risk for developing MDD at 1-year follow-up compared with non-depressed controls. Thus, within both patient groups, perhaps neurobiologically
, those who showed normal profiles were experiencing miD, while those with subcortical profiles had MDD and those with cortical profiles had an incipient dementing illness.
This study found no significant gender differences on the six selected CVLT indices as a group
. There were also no significant gender differences in cluster assignments to normal, subcortical, or cortical profiles. Reduced power due to the modest sample size in this study may have limited the potential of detecting an interaction with gender. Direct comparison with gender effects in the Massman and colleagues (1992)
study, on which this study's classification strategy was based, was not possible as nearly all depressed participants from the Massman study were male. However, when examining the six CVLT indices individually, women performed significantly better than men on TOT and SDFR. Other studies have reported similar sex differences on TOT and SDFR indices, both in healthy (Kramer et al., 1988
; Lamar et al., 2003
; Saykin et al., 1995
) and depressed samples (Otto et al., 1994
). While females have also been found to use semantic clustering strategies to a significantly greater extent than males, findings have either been restricted to children and adolescents (Kramer et al., 1997
) or represented a broad age range and were not specific to older adults (Kramer et al., 1988
). In the two previous studies of CVLT performance in late-life miD (Elderkin-Thompson et al., 2003
), the influence of gender has been equivocal as it was either not specifically examined (Elderkin-Thompson et al., 2007
) or no gender effects were found on several CVLT indices when analyzed as a group (effects were not assessed on individual indices) (Elderkin-Thompson et al., 2003
These results should also be interpreted in the context of limitations related to sample size and characteristics, and the potential influence of moderator variables. As this study focused on community-dwelling elderly patients who were predominantly Caucasian and had a variety of comorbid medical illnesses, the findings can only be generalized to similar populations. However, the study may provide clinical utility by the inclusion of outpatients with diverse medical problems as clinicians are often asked to treat such individuals (Lichtenberg, Ross, Millis, & Manning, 1995
Another limitation is that the effect of psychotropic medication (low-dose antidepressants and/or anxiolytics in therapeutic dosages) on one third of the MDD group is unknown. Improvement in cognitive functioning resulting from antidepressant medications has been documented (Van den Berg, Oldehinkel, Brilman, Bouhuys, & Ormel, 2000
); however, any such substantive effects in this sample seem unlikely because the pattern of cognitive findings fits the hypotheses. Although these medications, particularly the anxiolytics, may have contributed to reduced CVLT performance of MDD participants (Mintzer & Griffiths, 2000
; Settle, 1998
), most studies have found such effects to be small (Berg & Dellasega, 1996
). Notably, the scatterplots of CVLT scores did not show subsets of MDD participants who were outliers in either direction, although medication effects cannot be entirely ruled out. Participants in the miD group were drug-free, and most were drug naïve.
Third, an ongoing consideration in the assessment of memory and cognition in adults with depression is the issue of reduced effort, which itself might be reasonably regarded as a feature of depression (Rohling et al., 2002
). Our study did not include embedded or stand-alone cognitive effort tests. However, in an attempt to address this important issue, we applied a post hoc
detection method developed by Millis and colleagues for the CVLT with individuals with traumatic brain injury (Millis et al., 1995
; Millis & Volinsky, 2001
). Based on the results of this approach, reduced effort was determined to be an unlikely source of influence on our findings. Nonetheless, future studies, particularly those employing the CVLT-II, should take care to index effort in accord with emerging standards of practice (Heilbronner et al., 2009
Finally, not all factors that may have influenced CVLT performance were analyzed. For example, psychotherapeutic interventions in the miD and MDD were not assessed. It was also not known whether women in this study were receiving ERT, and estrogen levels were not obtained. Additionally, variables such as comorbid medical problems, the use of multiple non-psychotropic medications, and hospitalizations prior to 1 year before the study may have the potential to affect CVLT scores. Although Massman and colleagues (1992)
did not find any significant relationships between these variables and verbal learning and memory profile classification status, their depressive sample was younger than the elderly individuals who participated in this study.
To address these limitations, future investigations of cognition in late-onset miD should be replicated with a larger sample size, which is more ethnically and racially diverse, to confirm the findings and to consider the influence of aforementioned moderating variables including potential interactions with gender. It is also important that researchers operationally define miD explicitly. The current DSM-IV research criteria are tentative, and there are no consistent guidelines that researchers follow when investigating correlates of miD. As noted, the miD duration criterion in this study was increased from 2 weeks to 1 month to enhance the likelihood that patients were not merely experiencing a transient dysphoria. Additional efforts to examine the neurobiological (e.g., neuroimaging) and neuropsychological correlates of miD will likely reveal more specific characteristics that are common to this putative and heterogeneous disorder. Comparing cognitive profiles of miD to DyD, as well as longitudinal studies of cognition in miD, may be valuable for better understanding the functional impact of miD and the extent to which cognition may vary as a function of mood state.
In summary, the findings from this study suggest that although elderly individuals with miD may manifest similar mood symptoms to those with MDD (e.g., depressive mood, anhedonia), their verbal learning and memory profiles are different and have different functional implications. The miD group performed comparably with controls on the CVLT, both in terms of magnitude and patterns, and significantly better than the MDD group. Furthermore, the majority of participants in both the miD and control groups showed prototypical “normal” verbal learning and memory profiles, while the majority of those in the MDD group exhibited prototypical “subcortical” profiles. These findings highlight the value of assessing both mood symptoms and cognition in older adults with depression, particularly given the considerable public health impact of both miD and MDD, including evidence that untreated miD is a significant risk factor for developing MDD (Cuijpers et al., 2005
; Judd & Akiskal, 2002
; Lyness et al., 2009
) and that MDD may be an early sign of progression to dementia (e.g., Agbayewa, 1986
; Kral & Emery, 1989
; Nussbaum et al., 1995
). The results also raise important treatment considerations for clinicians. For example, therapeutic interventions for those with verbal learning and memory deficits would need to incorporate strategies to monitor medication adherence. Positive behavior supports that help individuals compensate for memory problems (e.g., prompts, cues, environmental modifications) would also be beneficial, as would education of family members about cognitive difficulties associated with the illness and compensatory adaptations.
While there is heterogeneity in learning and memory functions that is not sufficiently explained by diagnosis alone, clinicians can generally assume that mild or subsyndromal forms of depression in the elderly typically do not include memory problems, while memory problems do accompany MDD and require specific attention. However, 25% of the miD sample in this study exhibited either subcortical or cortical profiles, suggesting that a substantial minority of older adults with miD have memory problems at a rate exceeding expectations related to normal aging alone. Establishing the implications for risk of progression to MDD and dementia, as well as its functional impact and potential for compensation, require additional research attention with longitudinal investigations.