Adult samples

Seven adult sample studies (total n = 661) examined age as a variable related to cognitive change. One shorter study^{Reference Thomas, Puig and Twamley25} reported mixed significant results.

Thomas et al^{Reference Thomas, Puig and Twamley25} conducted a 12-week RCT for compensatory cognitive training (n = 77). Age was a significant moderator for category fluency, for which younger participants showed more of an improvement over time. However, results were not analysed separately for the two broad diagnostic groups of severe mood disorder (n = 46) and schizophrenia or schizoaffective disorder (n = 31), so the effect of age on cognitive change for patients with mood disorder alone cannot be commented on.

The remaining six studies (total n = 584) found no significant relationship between age and cognitive change.^{Reference Hoffman, Blumenthal, Babyak, Smith, Rogers and Doraiswamy16,Reference Jeon, Woo, Lee, Kim, Chung and Ha17,Reference Nadeau, Bowers, Jones, Wu, Triggs and Heilman19,Reference Miskowiak, Vinberg, Macoveanu, Ehrenreich, Køster and Inkster22,Reference Miskowiak, Rush, Gerds, Vinberg and Kessing23,Reference Lee, Kim, Kang, Hong and Kim26} Jeon et al (n = 164)^{Reference Jeon, Woo, Lee, Kim, Chung and Ha17} included age in an analysis, but did not report finding a significant relationship. It is assumed that age did not significantly affect cognitive change.

Older adult samples

Four shorter studies (total n = 960) examined age as a predictor of cognitive change in older adult samples.

One shorter study^{Reference Barch, D'Angelo, Pieper, Wilkins, Welsh-Bohmer and Taylor33} (n = 166) found mixed results depending on cognitive area. Barch et al^{Reference Barch, D'Angelo, Pieper, Wilkins, Welsh-Bohmer and Taylor33} measured cognitive functioning before and after a 12-week course of sertraline in adults aged 60 years and older with MDD. Older age was associated with less improvement for the areas of working memory (r = −0.25, P < 0.005), executive function (r = −0.17, P < 0.05) and psychomotor speed (r = −0.18, P < 0.05). No significant relationship for language function or episodic memory was found.

The three remaining studies (total n = 794) found no significant relationship between age and cognitive change.^{Reference Devanand, Pelton, Marston, Camacho, Roose and Stern28,Reference Doraiswamy, Krishnan, Oxman, Jenkyn, Coffey and Burt29,Reference Raskin, Wiltse, Siegal, Sheikh, Xu and Dinkel31}

Longer follow-up studies

Pelton et al (n = 35)^{Reference Pelton, Harper, Roose, Marder, D'Antonio and Devanand38} examined age as a predictor of cognitive change over 48 weeks in their older adult study, with non-significant results.

Age as a predictor of cognitive change

Only one adult study and one older adult study found significant results, despite 11 studies examining age. Thomas et al^{Reference Thomas, Puig and Twamley25} had a sample that was mixed between patients with mood disorders and patients with schizophrenia, so it is possible this influenced results. They also found the significant result for one domain only, despite running multiple cognitive tests. The combined sample size of the studies with significant results was also small compared with the combined negative studies. Thus, there is limited evidence from the studies identified that age is a predictor of cognitive change. A confounding methodological issue is, however, that none of the studies were conducted over an extended age range.

Adult samples

Gender in relation to cognitive change was investigated in five shorter studies (total n = 382). None of these studies found a significant relationship between gender and cognitive change.^{Reference Jeon, Woo, Lee, Kim, Chung and Ha17,Reference Nadeau, Bowers, Jones, Wu, Triggs and Heilman19,Reference Miskowiak, Vinberg, Macoveanu, Ehrenreich, Køster and Inkster22,Reference Miskowiak, Rush, Gerds, Vinberg and Kessing23,Reference Lee, Kim, Kang, Hong and Kim26}

Older adult samples

In the two studies examining it (total n = 755),^{Reference Doraiswamy, Krishnan, Oxman, Jenkyn, Coffey and Burt29,Reference Raskin, Wiltse, Siegal, Sheikh, Xu and Dinkel31} gender was not found to have a significant relationship with cognitive change.

Longer follow-up studies

Pelton et al (n = 35)^{Reference Pelton, Harper, Roose, Marder, D'Antonio and Devanand38} examined gender as a predictor of cognitive change over 48 weeks in older adults, with non-significant results.

Adult samples

One of six shorter studies (total n = 615) that examined education as a predictor of cognitive change showed significant results. In an RCT of bupropion and escitalopram for patients with MDD (n = 41), Soczynska et al^{Reference Soczynska, Ravindran, Styra, McIntyre, Cyriac and Manierka20} reported that for immediate verbal memory, higher level of education was correlated with greater improvement. No significant relationship between education and cognitive change was found in the remaining studies (total n = 574).^{Reference Porter, Robinson, Malhi and Gallgher6,Reference Hoffman, Blumenthal, Babyak, Smith, Rogers and Doraiswamy16,Reference Jeon, Woo, Lee, Kim, Chung and Ha17,Reference Miskowiak, Rush, Gerds, Vinberg and Kessing23,Reference Buschert, Prochazka, Bartl, Diemer, Malchow and Zwanzger27}

Older adult samples

The one older adult study (n = 39) that examined education found no significant relationship with cognitive change.^{Reference Devanand, Pelton, Marston, Camacho, Roose and Stern28}

Longer follow-up studies

Pelton et al (n = 35)^{Reference Pelton, Harper, Roose, Marder, D'Antonio and Devanand38} examined education as a predictor of cognitive change in their 48-week, older adult study, with non-significant results.

Older adult samples

Raskin et al (n = 311)^{Reference Raskin, Wiltse, Siegal, Sheikh, Xu and Dinkel31} reported no significant interaction between ethnicity and treatment for change in composite cognitive score.

Adult samples

Nadeau et al^{Reference Nadeau, Bowers, Jones, Wu, Triggs and Heilman19} investigated handedness in their shorter study and found no significant relationship with cognitive change in the areas of language, visuospatial function, executive function, verbal episodic memory and attention.

Adult samples

Depression severity as a predictor of cognitive change was investigated by four studies (total n = 238).^{Reference Nadeau, Bowers, Jones, Wu, Triggs and Heilman19,Reference Mikoteit, Hemmeter, Eckert, Brand, Bischof and Delini-Stula21,Reference Miskowiak, Rush, Gerds, Vinberg and Kessing23,Reference Lee, Kim, Kang, Hong and Kim26} Lee et al (n = 86)^{Reference Lee, Kim, Kang, Hong and Kim26} found a significant relationship between depression severity and the subjective Perceived Deficits Questionnaire-Korean version (PDQ-K), for total score as well as in the domains of attention/concentration and organisation/planning. For these areas, greater baseline depression severity was associated with greater subjective cognitive improvement. A significant relationship for prospective memory was not found.^{Reference Lee, Kim, Kang, Hong and Kim26}

The three remaining studies (n = 152) found no significant relationship between depression severity at baseline and cognitive change.^{Reference Nadeau, Bowers, Jones, Wu, Triggs and Heilman19,Reference Mikoteit, Hemmeter, Eckert, Brand, Bischof and Delini-Stula21,Reference Miskowiak, Rush, Gerds, Vinberg and Kessing23}

Older adult samples

Three shorter studies (total n = 921) examined baseline depression severity in association with cognitive change, with mixed results.

Barch et al^{Reference Barch, D'Angelo, Pieper, Wilkins, Welsh-Bohmer and Taylor33} found higher baseline Montgomery–Åsberg Depression Rating Scale score was correlated with less improvement in visual episodic memory (r = −0.16, P < 0.05); however, they had non-significant results for working memory, executive function, language function or psychomotor speed in their 12-week, open-label trial (n = 166). The remaining two studies found no significant relationship between depression severity at baseline and cognitive change.^{Reference Doraiswamy, Krishnan, Oxman, Jenkyn, Coffey and Burt29,Reference Raskin, Wiltse, Siegal, Sheikh, Xu and Dinkel31}

Longer follow-up studies

Four longer studies (total n = 948) examined baseline depression severity in association with cognitive change, with mixed results.

Han et al^{Reference Han, McCusker, Cole, Čapek and Abrahamowicz32} assessed 281 in-patients aged 65 years and older over 12 months. Participants were taking antidepressant medication, and were pooled from a RCT and an observational cohort study. Han et al analysed groups separately based on major, minor or no depression, according to DSM-IV criteria. Only the group with minor depression showed a significant improvement on the Mini-Mental State Examination (MMSE) during their trial.

Lohman et al^{Reference Lohman, Rebok, Spira, Parisi, Gross and Kueider36} examined the relationship between depressive symptoms and memory performance with training, using participants from a memory training group and a control group (n = 310) followed up over 5 years. Participants were community-dwelling adults aged over 65 years. A significant association between elevated depressive symptoms, measured with the 12-item version of the Centre for Epidemiological Studies-Depression Scale (CESD-12), and change in recall and recognition memory was reported, indicating those with elevated depressive symptoms experienced a faster decline in memory. Parisi et al^{Reference Parisi, Franchetti, Rebok, Spira, Carlson and Willis37} examined participants from the reasoning training arm of the same study (n = 322), at the 10-year follow-up. CESD-12 scores at baseline were not significantly associated with change in reasoning performance, which was the only domain assessed.

Pelton et al (n = 35)^{Reference Pelton, Harper, Roose, Marder, D'Antonio and Devanand38} found no significant correlation between baseline HRSD scores and changes on cognitive measures.

Depression severity as a predictor of cognitive change

There is conflicting evidence between age groups. Among the adult samples, the only study to find significant results for severity used a subjective measure of cognitive functioning as the outcome measure.^{Reference Lee, Kim, Kang, Hong and Kim26} The measure being subjective may have some issues, and the authors acknowledge the lack of objective tests as a limitation, although they do note that their measure (PDQ-K) has been reported to significantly correlate with objective measures in the literature.^{Reference Lee, Kim, Kang, Hong and Kim26,Reference McIntyre, Lophaven and Olsen40} It might be expected that greater depression severity is associated with greater cognitive impairment and therefore greater chance of improvement with treatment. This may in fact underlie the significant result in Lee et al^{Reference Lee, Kim, Kang, Hong and Kim26} However, it may also be that greater depression severity in the elderly is associated with greater neurobiological disturbance, which may be irreversible. For example, vascular changes may be associated with more severe depression,^{Reference Holt, Phillips, Jameson, Cooper, Dennison and Peveler41} which, as discussed below, appears to be associated with less cognitive improvement in response to treatment.^{Reference Barch, D'Angelo, Pieper, Wilkins, Welsh-Bohmer and Taylor33}

There was some evidence for baseline depression severity predicting cognitive change in longer studies. Significant results indicated that greater severity was associated with worsening cognitive change, i.e. less improvement, quicker decline or no significant improvement. As the studies were older adults, this may indicate progression of degenerative disease or vascular disease.^{Reference Alexopoulos42}

Adult samples

Three shorter studies (total n = 516) investigated diagnosis in relation to cognitive change, one of which found a significant result.

In 353 in-patients, prescribed antidepressant medications as appropriate by their psychiatrist, Lin et al^{Reference Lin, Xu, Lu, Ouyang, Dang and Lorenzo-Seva18} investigated the effect of depressive subtypes on cognitive change.⁴³ A significant relationship was found for psychomotor speed. Those with melancholic depression showed a greater increase in psychomotor speed (Trail Making Test A) over time than those with atypical or undifferentiated subtypes. There was no significant effect of subtype on the remaining five domains of cognitive function.

Miskowiak et al^{Reference Miskowiak, Vinberg, Macoveanu, Ehrenreich, Køster and Inkster22,Reference Miskowiak, Rush, Gerds, Vinberg and Kessing23} (n = 84) examined diagnosis of unipolar or bipolar, and found no effect of diagnosis on improvement in cognitive function.

Older adult samples

Two shorter studies (total n = 239) of older adult samples investigated age at depression onset as a predictor of cognitive change. One study (n = 166) of the three reported a significant relationship across their 12-week, open-label trial. Later age at onset was correlated with less improvement in executive functioning (r = −0.24, P < 0.005).^{Reference Barch, D'Angelo, Pieper, Wilkins, Welsh-Bohmer and Taylor33} No relationship was found between age at onset and cognitive change in the areas of working memory, language function, psychomotor speed or episodic memory.^{Reference Barch, D'Angelo, Pieper, Wilkins, Welsh-Bohmer and Taylor33} In the remaining study, Nebes et al^{Reference Nebes, Pollock, Houck, Butters, Mulsant and Zmuda30} (n = 73) found no significant relationship between age at depression onset and cognitive change.

Longer follow-up studies

Olaya et al (n = 291)^{Reference Olaya, Moneta, Miret, Ayuso-Mateos and Haro39} investigated age at onset in their 3-year, naturalistic follow-up study of older adults, with non-significant results.

Age at onset as a predictor of cognitive change

Late-onset depression with a first episode after around the age of 65 years is associated with greater cognitive impairment and brain changes (vascular changes and hippocampal size^{Reference Paranthaman, Burns, Cruickshank, Jackson, Scott and Baldwin44,Reference Lloyd, Ferrier, Barber, Gholkar, Young and O'Brien45} ), and might therefore predict less propensity for improvement in cognitive function during treatment.

The non-significant results from Nebes et al^{Reference Nebes, Pollock, Houck, Butters, Mulsant and Zmuda30} and Olaya et al^{Reference Olaya, Moneta, Miret, Ayuso-Mateos and Haro39} were surprising, as later age at onset of depression in elderly patients tends to be associated with greater cognitive dysfunction.^{Reference Bora, Harrison, Yücel and Pantelis46} Indeed, Nebes et al did find significant differences in performance between earlier- and later-onset groups, with later-onset patients performing worse on two of the cognitive tests.^{Reference Nebes, Pollock, Houck, Butters, Mulsant and Zmuda30} The use of self-report may have affected results of Olaya et al,^{Reference Olaya, Moneta, Miret, Ayuso-Mateos and Haro39} as well as their lack of consideration of dementia or mild cognitive impairment, which could have been screened for with a short questionnaire such as the MMSE or the Dementia Rating Scale. Further research could help clarify age at onset as a predictor of cognitive change.

Adult samples

Miskowiak et al^{Reference Miskowiak, Rush, Gerds, Vinberg and Kessing23} (n = 79) showed that longer duration of illness was associated with a significant increase in verbal memory (16%) following erythropoietin infusions when adjusted for in their logistic regression with baseline subjective cognitive difficulties, measured with the Cognitive and Physical Functioning Questionnaire, as the predictor. There was no significant increase when baseline memory dysfunction was the predictor, as measured with the Rey Auditory Verbal Learning Test. Number of mood episodes had no significant association with cognitive change in either analysis.

Lee et al^{Reference Lee, Kim, Kang, Hong and Kim26} (n = 86) examined three other illness-related factors in relation to cognitive change: sick days in the past week before study entry, functional impairment as measured by the Sheehan Disability Scale (SDS) and quality of life as measured by the EuroQol-5 dimensions questionnaire (EQ-5D). For their subjective measure of prospective memory, a higher score on EQ-5D and fewer sick days were associated with greater cognitive improvement. The SDS did not have any significant correlations with cognitive change.

Older adult samples

In their eight-week study (n = 311), Raskin et al^{Reference Raskin, Wiltse, Siegal, Sheikh, Xu and Dinkel31} investigated number of previous episodes of depression, and number of previous medications received for depression. For their composite score, no significant interaction of treatment and these factors was found.

Adult samples

Baseline objective cognitive performance as a predictor of cognitive change was investigated by eight papers for seven shorter studies (total n = 727); five (total n = 318) reported significant results.

Miskowiak et al^{Reference Miskowiak, Rush, Gerds, Vinberg and Kessing23} examined objective baseline verbal memory dysfunction related to cognitive change in an RCT of erythropoietin infusion for moderately depressed patients with unipolar depression and patients with bipolar disorder in partial remission (n = 79). For patients with baseline cognitive dysfunction, when compared with patients with no dysfunction, the odds of clinically relevant verbal memory improvement increased by a factor of 290.6. Ott et al^{Reference Ott, Vinberg, Kessing and Miskowiak24} performed a secondary analysis in the same study. When cognitive dysfunction was defined as a score of >1 s.d. below the norm on more than two of the eight cognitive tests, there was deemed to be a significant relationship, in which baseline cognitive dysfunction increased chances of achieving clinically relevant improvement. Patients with cognitive dysfunction were 9.7 and 9.9 times more likely to achieve this at treatment end (week 9) and at 14 weeks, respectively.

Mikoteit et al (n = 25)^{Reference Mikoteit, Hemmeter, Eckert, Brand, Bischof and Delini-Stula21} examined the relationship between baseline cognitive performance and cognitive change (alertness: reaction time r = –0.82, P < 0.001; working memory: correct responses r = –0.40, P < 0.1, reaction time r = –0.50, P < 0.01, ratio of correct responses/reaction time r = –0.20, P > 0.1; divided attention: correct responses r = 0.29, P > 0.1, reaction time r = –0.80, P < 0.001, ratio of correct responses/reaction time r = –0.50, P < 0.05). Those who had poorer performance at baseline showed greater cognitive improvement. Nadeau et al (n = 48)^{Reference Nadeau, Bowers, Jones, Wu, Triggs and Heilman19} reported mixed results. Stronger executive functioning at baseline was associated with greater gains in executive functioning. For language, visuospatial function, verbal memory and attention, there was no significant relationship between the baseline score and change within any domain. Barch et al^{Reference Barch, D'Angelo, Pieper, Wilkins, Welsh-Bohmer and Taylor33} similarly had mixed results. They found that poorer performance at baseline, as measured by the MMSE, predicted less improvement for working memory; however, they found no significant relationship for executive function, processing speed, episodic memory or language function.^{Reference Barch, D'Angelo, Pieper, Wilkins, Welsh-Bohmer and Taylor33}

The three remaining studies found no significant relationship between baseline cognitive performance and cognitive change with treatment (total n = 409).^{Reference Hoffman, Blumenthal, Babyak, Smith, Rogers and Doraiswamy16,Reference Jeon, Woo, Lee, Kim, Chung and Ha17,Reference Buschert, Prochazka, Bartl, Diemer, Malchow and Zwanzger27}

Older adult samples

Two shorter studies investigated baseline cognitive functioning (n = 517) and found no evidence to suggest baseline cognitive functioning affected cognitive change over their 12-week antidepressant trials.^{Reference Doraiswamy, Krishnan, Oxman, Jenkyn, Coffey and Burt29,Reference Nebes, Pollock, Houck, Butters, Mulsant and Zmuda30}

Adult samples

Two papers investigated subjective cognitive function at baseline in relation to cognitive change in one sample of patients with current unipolar depression or bipolar disorder in partial remission (n = 79).^{Reference Miskowiak, Rush, Gerds, Vinberg and Kessing23,Reference Ott, Vinberg, Kessing and Miskowiak24} Miskowiak et al^{Reference Miskowiak, Rush, Gerds, Vinberg and Kessing23} examined subjective cognitive difficulties as measured by the Massachusetts General Hospital Cognitive and Physical Functioning Questionnaire (MGH-CPFQ). Consistent with their own, and Ott et al's objective verbal memory findings above, a significant relationship was reported between subjective baseline dysfunction and objective cognitive outcomes. Odds of the patient having clinically relevant memory improvement increased if they had subjective cognitive complaints at baseline. However, subjective cognitive complaints measured with the MGH-CPFQ, showed no significant relationship between baseline subjective cognitive complaints and change in subjective cognitive function.^{Reference Ott, Vinberg, Kessing and Miskowiak24}

Baseline cognitive performance as a predictor of cognitive change

Three studies showed that reduced cognitive performance at baseline predicted greater improvement. Baseline objective cognitive performance therefore appears to be one of most consistent predictors of cognitive change. The obvious reason for this is that patients with impaired cognitive performance at baseline are more likely to change. This is likely not only for obvious clinical reasons but also statistically it is more likely improvement can be shown if there are greater deficits. One study conversely found that lower baseline performance predicted less improvement,^{Reference Barch, D'Angelo, Pieper, Wilkins, Welsh-Bohmer and Taylor33} although this was a significant result for only one of five domains investigated, making it more likely that this is a chance finding. Additionally, just one study found better baseline cognitive performance was associated with greater cognitive gains, a result they attributed to the phenomenon of those with superior cognitive function being better able to build on this given recovery from depression.^{Reference Nadeau, Bowers, Jones, Wu, Triggs and Heilman19} Again, despite examining five domains, only executive functioning showed a significant result.

Older adult samples

Barch et al^{Reference Barch, D'Angelo, Pieper, Wilkins, Welsh-Bohmer and Taylor33} found that more severe hyperintensities correlated with less improvement in psychomotor speed (r = −0.16, P < 0.05) in a 12-week trial (n = 166). No significant relationship was found for working memory, executive function, language function or episodic memory. Higher vascular risk score correlated with less improvement for working memory (r = −0.24, P < 0.005) and executive function (r = 0.17, P < 0.05), No significant relationship was found for language function, psychomotor speed or episodic memory.

One shorter study (n = 17) investigated grey matter volumes in relation to cognitive change.^{Reference Marano, Workman, Lyman, Munro, Kraut and Smith34} Marano et al^{Reference Marano, Workman, Lyman, Munro, Kraut and Smith34} conducted a 12-week trial of citalopram in out-patients with late-life depression, aged over 55 years old. Larger grey matter volumes were associated with greater cognitive improvement, depending on the test and brain area. For verbal learning and memory and letter fluency, greater improvement was associated with larger grey matter volumes in primarily frontal areas. There was also an association between greater improvement in verbal learning and memory and smaller grey matter volumes in the bilateral superior frontal gyrus (BA 8), and an association between greater improvement in letter fluency and smaller grey matter volumes in the bilateral praecuneus (BA 7).

In a 12-week, placebo-controlled trial of donzepil in 18 out-patients with depression and cognitive impairment, Pelton et al reported that worse olfactory performance (University of Pennsylvania Smell Identification Test) at baseline significantly correlated with better episodic verbal memory after the trial.^{Reference Pelton, Soleimani, Roose, Tabert and Devanand35}

In one shorter study by Raskin et al^{Reference Raskin, Wiltse, Siegal, Sheikh, Xu and Dinkel31} (n = 311), several comorbid conditions were investigated as predictors of cognitive change. These were anxiety, insomnia, arthritis, diabetes, vascular disease and mild dementia. No significant interaction of treatment with any of these comorbid disorders was found by their composite cognitive score.

Biological factors as predictors of cognitive change

There is evidence in older adult samples that a higher vascular risk score, more severe white matter hyperintensities^{Reference Barch, D'Angelo, Pieper, Wilkins, Welsh-Bohmer and Taylor33} and lower grey matter volumes^{Reference McIntyre, Lophaven and Olsen40} are associated with less cognitive improvement. Both may be indicators of greater risk for cognitive decline, but both studies were short term.