Depressive Symptoms are Associated with More Negative Global Metacognitive Biases in Combat Veterans, and Biases Covary with Symptom Changes over Time

Metacognitive awareness, insight into one ’ s abilities, is thought to be disrupted in depression and posttraumatic stress disorder (PTSD), with more negative metacognitive biases and reduced awareness, respectively. However, studies have rarely examined global aspects of metacognitive awareness in depression and PTSD, instead using task-speci ﬁ c measures. In 467 trauma-exposed post-9/11 veterans, we administered assessments of PTSD and depression, self-report and objective measures of cognitive functioning (neuropsychological battery of memory, attention


Introduction
Metacognition refers to the self-reflective activity of monitoring, planning, interpreting, and regulating the contents of information processing [1]. One important aspect of metacognition is awareness of one's traits, abilities, or performance. This awareness has two components: metacognitive sensitivity (i.e., accuracy) refers to a person's capacity to correctly evaluate themself, whereas metacognitive bias measures the difference between subjective and objective traits, abilities, or performance (i.e., tendency to under-vs. over-estimate, see [2]). Metacognitive awareness is posited to be impaired in depression, where pervasive negative selfevaluations are frequent [3,4], and in PTSD, where increased autonomic arousal and deficits in cognitive control may impair awareness [5,6]. However, studies examining metacognition in depression and PTSD have primarily measured distress-related metacognitive beliefs using self-report questionnaires (e.g., I think my worrying is dangerous; for a review, see [7]), and far fewer have examined more general metacognitive awareness. Further, studies examining metacognitive awareness in PTSD and depression have typically measured task-specific trial-by-trial confidence (e.g., How confident are you that you got that memory trial correct? [8]; for exceptions, see [9,10]). Recently, researchers studying metacognition in clinical populations have recommended approaches to metacognitive awareness that are "global" rather than task-specific, by measuring domaingeneral self-beliefs and abilities [11]. This more global approach was applied to measure metacognition in the current study.

Metacognitive Awareness in Depression.
Though depression is often associated with negative self-evaluations, evidence for metacognitive awareness deficits has been mixed. Most studies in this area have compared participants' performance on objective tasks to their subjective judgments of performance, both trial-by-trial and after a block of trials [12], allowing researchers to quantify metacognitive sensitivity and bias. Studies have observed a negative metacognitive bias across both subclinical and clinical depression with some inconsistencies, whereas studies have not generally observed metacognitive sensitivity deficits (for a review, see [12]). For example, in two large online experiments, an anxious-depressive dimension was associated with systematically negative biases (i.e., underconfidence) in both a perceptual decision task, where participants had to discern which of the two boxes had more dots, and a general knowledge task, where individuals had to discern a correct fact between two choices, despite a lack of impairment in objective performance [13]. In two studies using the same perceptual decision-making task, depressive symptoms were associated with a more negative bias and an enhanced metacognitive sensitivity [14], while there were no differences reported in metacognitive bias or sensitivity between those with and without depression in another study [15]. Other studies comparing depressed and control groups have observed negative metacognitive biases in tasks including object recognition, general knowledge, facial emotion recognition, line-length perceptual judgment [16], adjective recognition [17], the Stroop task [18], and verbal memory performance [19]. A prominent model explaining dysfunctional metacognitive awareness in depression implicates patterns of negative thinking, such as attending to negatively valenced material [20] or the increased salience of negatively valenced memories [21], that become activated during negative moods [22,23]. These negative biases in perception and memory are resilient to feedback and inform more negative global self-performance estimates [4,24,25]).
Despite the evidence for depression-related negative metacognitive biases in trial-by-trial tasks, several weak or nonsignificant depression/metacognitive bias associations have been reported, including working memory (digit span backwards, rho = −0:10) and executive functioning (Wisconsin Card Sorting Task (WCST), rho = −0:04, [26]). One explanation for these inconsistent findings in trial-bytrial metacognitive biases is due to the diversity of cognitive tasks. While indexing metacognitive awareness in an isolated domain, specific cognitive tasks (e.g., WCST, [26]) may not generalize to global self-beliefs about cognitive abilities. Further, global self-beliefs such as self-esteem have been more directly related to subjective and functional experiences in depression [3,11]. One recent study found significant associations between negative global metacognitive biases and depressive symptoms in adolescents and emerging adults by comparing confidence ratings on two perceptual tasks with the Rosenberg Self-Esteem Scale [10]. These findings are consistent with recent results from Hoven et al. [9] in a general population showing greater negative global metacognitive bias in those with more anxious-depressive symptoms. Overall, metacognitive sensitivity has not shown to be impaired in depression, while studies examining trial-by-trial metacognitive biases have been mixed, though generally observe depression-related impairments. However, few studies have examined more global aspects of metacognitive biases.

Metacognitive Awareness in PTSD.
Compared to depression, studies examining metacognitive awareness in PTSD are relatively scarce, with most PTSD metacognition studies focusing on participants' beliefs about their trauma (e.g., hyperresponsibility and a belief about being irreparably damaged by trauma; [27]). With regard to metacognitive awareness, even after controlling for depression, hyperarousal symptoms have been associated with metacognitive deficits in PTSD on the Metacognition Assessment Scale. This scale measures a clinician's rating of a patient's awareness of illness, cognitive insight, and memory accuracy [28]. One explanation for this association is that PTSD-related increases in autonomic arousal impair processes necessary for successful self-reflection (e.g., inhibitory control; [5,29]). Additionally, Sacher et al. [8] measured trial-by-trial confidence in episodic memory performance and found that individuals with PTSD had worse awareness of their memory deficits and more negative bias, though they did not control for depression. Studies of metacognitive therapy as developed by Wells and Sembi [30] seek to correct dysfunctional beliefs about an individual's trauma and have been shown to be effective in decreasing PTSD symptoms ( [31]; see also [32]). Metacognitive therapy offers an alternative to direct trauma confrontations in prolonged exposure therapy or cognitive behavior therapy (CBT), such as by practicing detached mindfulness through the abandonment of threat monitoring; in contrast, CBT discusses cognitive content, rather than cognitive processes, and corrects the metacognitive evaluations of this content [33].
1.3. The Current Study. Our goal was to extend the previous literature in several important ways. First, instead of taking a task-specific approach, the current study measured global metacognitive awareness by contrasting an objective cognitive battery and five physical health factors with self-2 Depression and Anxiety reported cognition and health, respectively. This is one of the first studies to examine how global metacognitive awareness, which may be particularly clinically and functionally meaningful, is associated with depression (see [10]), the first to do so in PTSD, and the first to measure metacognitive awareness using a domain-general battery of objective cognitive tasks. Second, by repeating our assessment battery approximately two years later in a subset of participants, the current study is the first to measure how longitudinal changes in metacognitive awareness are related to changes in depressive and PTSD symptoms. This can begin to address the critical question of whether metacognitive awareness deficits are a stable trait or rather covary with depressive and PTSD symptoms. Finally, to develop a better mechanistic understanding of the relative importance of PTSD and depressive symptoms to metacognitive awareness deficits, mediation analyses of the observed effects were performed.
To address these questions, a large group of 467 traumaexposed post-9/11 veterans with a range of depressive and PTSD symptoms were recruited and tested. Depression and PTSD are common sequelae of trauma exposure that are frequently comorbid in veterans (see [34,35]). Relevant to metacognitive awareness, while 72% of post-9/11 veterans report moderate to very severe cognitive impairment [36], evidence suggests that objective cognitive deficits are milder and less frequently observed (e.g., 35% of post-9/11 veterans have shown to have DSM-5-defined mild cognitive dysfunction; [37]), suggesting a potential negative metacognitive bias. In the current study, to better characterize metacognitive awareness in both sensitivity and bias, objective cognition on a validated battery of neuropsychological tasks was contrasted with subjective cognitive functioning within the cognition subscale of the World Health Organization Disability Assessment Schedule-II (WHODAS II). In addition, metacognitive health awareness was measured through contrasting general subjective health with validated objective indices of cardiometabolic health (e.g., blood pressure, cholesterol, and waist circumference; see Table S1). Finally, in 267 veterans who were reassessed approximately two years later, associations between changes in metacognitive awareness and changes in PTSD and depressive symptoms over time were examined.

Transparency and Openness
Data and syntax can be accessed pending approval from the VA Boston Health Care System IRB. Please contact Dr. Joseph DeGutis at degutis@wjh.harvard.edu. The current study is part of an ongoing longitudinal study [38]. We report how we determined our sample size, all data exclusions, manipulations, and measures in the study. The VA Boston Healthcare System IRB approved this study (#2354, Translational Research Center for Traumatic Brain Injury and Stress Disorders: Human Characterization Core B), written consent was obtained from all participants, and research was conducted in accordance with the Declaration of Helsinki.

Methods
Participants were drawn from a pool of 813 post-9/11 combat-deployed veterans recruited into the Translational Research Center for TBI and Stress Disorders (TRACTS; for a more in-depth description of the sample and measures, see [38]) who participated in data collection at time 1 and time 2 visits (e.g.,~2 years later) between the years 2010 and 2019. The TRACTS exclusionary criteria excluded participants who had a history of neurological/physical impairments (n = 5), moderate to severe traumatic brain injury (n = 38), or psychiatric disorders (n = 9), including bipolar disorder and/or suicidal/homicidal ideation requiring crisis intervention at either assessment. Additional participants were removed due to evidence of reduced effort on the Medical Symptom Validity Test ( [39], n = 67) at either time point. A total of 46 participants who did not complete the WHODAS II were also excluded, as well individuals not reporting any significant interference in daily life on the WHODAS II to remove ceiling effects (n = 177, similar to our recent paper [40]). Notably, all key analyses replicated when including individuals that reported no significant interference in daily life (see Supplementary Materials (available here)). Participants may have been excluded for more than one reason. This left a final sample of 467 participants at time 1 and 267 at time 2.
3.1. Clinical Measures. PTSD diagnosis and severity were assessed using the clinician-administered PTSD scale for the DSM-IV (CAPS-IV, [41]). The CAPS-IV is a clinicianadministered interview corresponding to the DSM-IV symptoms for PTSD with excellent reliability (test-retest reliability of 0.89, [42]). The current CAPS score was used as a summation of all PTSD symptoms, with total scores ranging from 0 to 136. The depression subscale from the Depression, Anxiety and Stress Scale (DASS) was used to measure continuous depressive symptoms, with total scores ranging from 0 to 42 [43]. Diagnostic measures from the Structured Clinical Interview for DSM-IV were used to measure a current depression diagnosis, as well as anxiety disorder, and substance use disorder (SCID-I/NP; [44]). Please note that while the DSM-IV measures all mood disorder, in the current sample >95% of those with a mood disorder had major depressive disorder. CAPS-IV and SCID-I/NP were kept for continuity across time points. Pain was assessed using the McGill Pain Questionnaire (MPQ; [45]), sleep was assessed using the Pittsburgh Sleep Quality Index (PSQI; [46]), and traumatic brain injury was assessed using the Boston Assessment of Traumatic Brain Injury-Lifetime [47].

Subjective Cognitive Functioning. The World Health
Organization Disability Assessment Schedule II (WHODAS II, [48]) is a validated self-reported functional outcome measure that has been highly recommended to assess functioning in trauma-exposed post-9/11 veterans [49]. The WHODAS II is a 42-item measure with six subscales: understanding and communicating, getting around, self-care, getting along with people, life activities, and participation in society. We focused our analyses on the understanding 3 Depression and Anxiety and communicating subscale, which ranged from 0 to 100, with higher scores indicating greater disability in the cognitive domain. The subscale asked 6 questions such as "in the past 30 days, how much difficulty did you have in analyzing and finding solutions to problems in day to day life?" (see Table S1). Self-reported cognition was reverse scored in analyses, such that higher scores represent better self-reported functioning for ease of interpretation and to be consistent with objective cognition and metacognitive bias measures. After participants report their functional difficulties across the six subscales, participants are separately asked to indicate how much their functioning interferes with their daily life on a scale of 0 (none) to 4 (extreme). The final sample included participants with overall interference scores greater than 0 to avoid issues with skewness/ceiling effects [40], though analyses of interest were replicated with all veterans in the Supplementary Materials.

Objective Cognitive Functioning.
In order to measure objective cognitive functioning, a global cognitive composite was created as an average of the standardized age-adjusted z -scores for three cognitive domains: executive functions, memory, and attention (see [37]). The executive function composite included the Delis-Kaplan Executive Function System (D-KEFS) Trail Making Test Number/Letter Switching subtest, i.e., Trails B, as a measure of working memory/ switching [50], the D-KEFS letter fluency as a mixed executive function measure of working memory/switching (FAS category switching and letter fluency subtests, [50]), the D-KEFS Color-Word Test (i.e., the Stroop test) as a measure of inhibitory control, the CANTAB Intra-Extra Dimensional Set Shift Task (number of stages completed) as a measure of task-switching [51], and the Auditory Consonant Trigrams as a measure of working memory (ACT, [52]). The ACT was not administered to all participants at time 2; therefore, within-group changes were not examined with this measure. Verbal learning and memory were measured using the California Verbal Learning Test-Second Edition [53]. The verbal memory composite score measures encoding, recall, and recognition and consists of the mean age-adjusted z -scores of total learning, short-delay free recall, long-delay free recall, and long-delay recognition hits (see [37]). The attention composite included the Test of Variables Attention [54], the Digit Span Forward trials (WAIS-IV; [55]), and the Trail Number, Sequencing subtest [50], i.e., Trails A.

Calculating Global Metacognitive Sensitivity and Bias.
Global metacognitive sensitivity, i.e., accuracy, was calculated using nonparametric Spearman's correlations between self-reported and objective cognition measures at the group level. In calculating the global metacognitive bias score, self-reported cognition (reverse scored so that higher values equate to higher subjective ability) in the WHODAS II was subtracted from the global objective cognition score as averaged across executive function, memory, and attention tasks within an individual. Bias, i.e., calibration or confidence, is usually calculated as the difference between self-reported cognitive abilities and objective task performance [12].
Scores below 0 refer to a negative metacognitive bias relative to the sample, such that an individual reports more selfreported cognitive deficits than are objectively measured. Self-reported cognition was z-scored within the total veteran sample in lieu of appropriate normative data, while objective cognition was calculated based on normative data [37]. The objective cognition scores were then z-scored within the sample so that the mean and SD of both objective and selfreported measures were 0 and 1, respectively. Because of this and dissimilarities in self-reported and objective cognition distributions, the difference score represents a relative value. For example, an individual with balanced metacognitive bias (i.e., self-reported-objective cognition = 0) should not be interpreted as having no metacognitive bias, but rather a lack of bias within the relative distribution of deployed post-9/11 veterans in the sample.
3.5. Self-Reported, Objective, and Metacognitive Health. Metacognitive awareness of health was measured by comparing self-reported and objective measures of physical health. Metacognitive health bias and sensitivity scores were calculated identically to their cognition counterparts. For subjective health, participants rated a single item, "How would you rate your general health?", on a Likert scale ranging from 1, "excellent," to 4, "poor." Objective health was assessed using comprehensive physical and physiological measures of five cardiometabolic syndrome risk factors: obesity (waist circumference), triglycerides, high density lipoprotein, blood pressure, and glucose levels (for more information, see [38,56]). The dependent variable was the total cardiometabolic risk factor score for each participant, ranging from 0 to 5, representing the total number of elevated/abnormal risk factors. The sum of the risk factors was used because it represents a unified construct (cardiometabolic health) and has been used in previous studies (e.g., [57]). For consistency with cognition measures, both self-reported and objective health were reverse scored such that higher scores represent better health.

Longitudinal Analyses.
Because not all participants returned for time 2, longitudinal analyses were performed in a reduced sample (n = 267). Global metacognitive sensitivity was computed as before and compared between time points. In calculating a longitudinal difference score in global metacognitive bias and clinical measures of interest, time 2 scores were subtracted from time 1. Changes in global metacognitive bias and self-reported and objective cognition were then associated using Spearman's correlations with changes in clinical measures to see if they tracked with symptom changes and to further explore specificity. Finally, in order to more completely characterize changes in metacognitive bias scores, a reduced sample was analyzed after grouping by changes in diagnoses of a current PTSD or depression disorder between time points: improved depression, worsened depression, improved PTSD, and worsened PTSD. For example, the improved depression group had depression at time 1 and no longer met criteria for depression at time 2, while the worsened group had depression at time 1 and subsequently developed depression at time 2 (full 4 Depression and Anxiety analyses available in the Supplementary Materials). A 2 × 2 repeated measures ANOVA was run, with time 1/time 2 metacognitive bias and group (e.g., no depression diagnosis to depression diagnosis vs. depression diagnosis to no depression diagnosis) as factors.
3.7. Sample Size Justification. The current study is part of an ongoing longitudinal study that has consistently found significant associations between PTSD/depressive symptoms and self-reported and objective cognitive functioning [37] as well as inhibitory control and PTSD [29,40]. Previous studies have found significant associations between objective performance on cognitive tasks and functional behaviors (n = 489, [9]; n = 57, [10]). Relevant to the current investigation, a recent study found significant associations between greater depressive symptoms and more negative metacognitive bias by comparing two perceptual tasks with global self-esteem estimates (a similar measure to our subjective cognition score), finding a small effect size of β = 0:117 [9]. Considering that Hoven et al. [9] used two brief tasks and we used a considerably more reliable 3-hour objective cognitive battery, we estimate that we would have at least a 25% larger effect size (β = 0:146). Using a α = 0:05 and 1 − β = 0:80, our study should require 362 participants to adequately detect these associations. Therefore, given our sample size (n = 467), we estimate that we will have enough power to detect an effect size approximately the same as the previous literature (β ≥ 0:146, where a small effect is 0.2).

Demographics and Clinical
Characteristics. As can be seen in Table 1

Associations between Changes in Depression/PTSD and Changes in Metacognitive Awareness.
To understand if metacognitive awareness is a stable trait or rather fluctuates with depressive and PTSD symptoms, associations between changes in metacognitive measures and changes in PTSD and depressive symptoms were examined over an approximately 2-year period. Changes in metacognitive bias from time 1 to time 2 were associated with changes in both depressive (ρ = −0:25, p < 0:001) and PTSD (ρ = −0:33, p < 0:001) symptoms, such that decreases in PTSD and depressive symptoms were associated with more positive changes in metacognitive bias. In a joint regression, the changes in PTSD symptoms and changes in depressive symptoms explained unique variance in changes in metacognitive bias (R 2 = 0:12; β = −0:244, p < 0:001; β = −0:182, p = 0:004, respectively). Changes in PTSD significantly mediated the effect of changes in depressive symptoms in predicting changes in metacognitive bias, such that worse PTSD symptoms partially explained the relationship between depressive symptoms and more negative metacognitive bias (β = −0:11, 95% CI: -0.20, -0.04) with a direct effect of PTSD remaining significant (β = −0:16, p = 0:005). In order to more completely characterize changes in metacognition, veterans were separated into groups depending on their changes in PTSD or depression diagnoses. In general, findings in metacognitive bias and health were numerically consistent with continuous analyses, where biases became more negative when veterans developed PTSD or depression and inversely became more positive as diagnoses remitted, though these analyses lacked the power to reach significance (see Supplementary  Materials and Tables S2-4).
Changes in metacognitive bias in health showed a similar trend as cognition in their relationship with both changes in depressive and PTSD symptoms but failed to reach significance (see Table 3). Global metacognitive sensitivity in cognition and health was not significantly related to changes in depression or PTSD (see Supplementary Materials (available here)). Associations between baseline clinical and demographic variables with changes in metacognitive bias measures were generally nonsignificant (see Table S7).

Discussion
The current study is one of the few to examine associations between global metacognitive awareness and depressive/ PTSD symptoms and the first to examine how longitudinal changes in metacognitive awareness relate to changes in depressive/PTSD symptoms. More severe depressive and PTSD symptoms were significantly associated with more negative metacognitive biases in cognition, i.e., underconfidence (ρ = −0:32, -0.23, respectively), but were not associated with differences in metacognitive sensitivity. A similar pattern of more negative metacognitive bias with greater depressive/PTSD symptoms was also observed in the domain of health, indicating that metacognitive biases in depression and PTSD may be quite general. Notably, depressive symptoms partially mediated the relationship between PTSD symptoms and negative metacognitive bias in cognition but not vice versa, suggesting that depressive symptoms are driving this relationship. Longitudinal changes in metacognition were also examined, where changes in metacognitive bias in cognition were significantly associated with changes in both depressive and PTSD symptoms. This result is more consistent with metacognitive bias being closely linked with PTSD and depressive symptoms rather than being a stable precursor or risk factor for the development of depression/PTSD. Together, these findings advance models of metacognition in depression and PTSD as well as have important clinical and treatment implications.
The current results suggest that depressive and PTSD symptoms are consistently associated with a negative global metacognitive bias in awareness. In contrast to mixed results linking negative metacognitive bias to depressive symptoms in single, idiosyncratic cognitive tasks (multiple studies reporting null or positive biases, e.g., [12]), the current 7 Depression and Anxiety results indicate a robust negative bias on a more general objective cognitive measure composed of standardized neuropsychological tests. Additionally, this negative bias was present when participants made judgments about their general health. The current study's consistent and robust PTSD/depression associations with global measures of metacognitive bias contrasts mixed results with local trial-by-trial tasks and supports the potential benefit of using global measures of metacognition and their transdiagnostic relation to functioning, as some researchers have recently advocated for [9,10]. In these studies, global measures were informed by local trial-by-trial confidence ratings [58,59]. The current study extends these studies by contrasting a battery of objective tasks with self-reported cognitive functioning during daily activities (i.e., not specifically asking about task performance), which may capture additional daily life cognitive awareness information. These results are consistent with a recent theoretical review by Seow and colleagues [11] calling for more global, transdiagnostic approaches to metacognitive awareness, as opposed to measuring metacognitive awareness exclusively in isolated abilities, which make up the vast majority of the clinical metacognitive awareness literature. The consistency of the current study's results across both cognition and health provides preliminary evidence that there is a general global metacognitive bias associated with greater depressive and PTSD symptoms.
The results also showed that metacognitive awareness deficits were particularly associated with greater depressive symptoms. In a joint model, greater depressive symptoms, but not PTSD symptoms, explained unique variance in predicting a more negative metacognitive bias in cognition. Depressive symptoms also partially mediated the association of PTSD symptoms with metacognitive bias in cognition, while PTSD symptoms did not mediate depressive symptoms. Consistent with this, comorbid PTSD and depression were associated with significantly more negative biases in   Note. * p < 0:05 and * * p < 0:01.

8
Depression and Anxiety cognition than those with PTSD only, while bias scores in those with PTSD only and those with neither PTSD or depression were not significantly different. These analyses were repeated using self-reported and objective health scores and similarly observed a significant association between more negative metacognitive biases in general health and greater depressive symptoms, though not PTSD symptoms. These findings are consistent with previous research implicating negative metacognitive bias in depression (e.g., [10,13]). Past studies have found weak or nonsignificant associations using task-specific measures (e.g., [26]) and have only recently begun to compare global and taskspecific measures of metacognitive awareness (e.g., perceptual knowledge; [9,10]). The current findings extend these studies by providing evidence for the domain-general nature of this depressive negative metacognitive bias through associating a cognitive battery spanning memory, attention, and executive functions as well as objective health measures with self-reported functioning. Further, the current results suggest that, in a trauma-exposed population, this relationship is driven by depressive rather than PTSD symptoms. Future studies would be useful to examine whether these depression-related negative biases extend to other functional domains (e.g., social skills or emotion regulation). One reason why metacognitive bias could be more related to depressive symptoms than PTSD symptoms may be the particular tendency in depression to attend to negatively valenced material or have more readily available access to negatively valenced memories (e.g., [20,24]). For example, one study examining emotional memory biases in an implicit learning task found that only depressive symptoms explained unique variance in a bias to remember negatively valenced adjectives in a joint model with anxiety, ADHD, and autism symptoms [21]. Related to this point, the selfreported cognition measure asked participants to think back about their cognitive abilities over the last 30 days. This long-term memory component is much greater than in trial-by-trial metacognitive awareness studies and could be another reason we found more robust effects between metacognitive bias and depression than the previous studies. Other behaviors cardinal to depression such as rumination [60] and belief inflexibility [61] may further exacerbate these memory biases as depression worsens, forming rigid, global distortions in an individual's abilities and plausibly leading to an overestimation of dysfunction in self-reported functioning.
Notably, metacognitive sensitivity was not reduced in those with depression or PTSD, suggesting that deficits are specific to biases. Metacognitive sensitivity in the current sample (e.g., diagnosis of depression and PTSD: ρ = 0:34) was similar in strength to studies of metacognitive awareness using local cognitive tasks (e.g., word and number recall r = 0:25 [62]; perceived knowledge r = 0:27 [63]) and comparable to metacognitive awareness measured using broad measurements of daily functioning (e.g., academic ability r = 0:21 -0:39; vocational skills r = 0:19 -0:36, [64]). Most abilities engender little feedback for one's self, let alone one's relative abilities compared to others. It could be that the better-than-average effect, the tendency for a person to perceive themselves as superior compared to their peers [65], may restrict the range of self-reported ability, potentially reducing metacognitive sensitivity. In contrast, individuals with depression are less susceptible to this effect [66] and may use a larger range of self-reported ability (e.g., comorbid depression and PTSD, SD = 19:00, and neither disorder, SD = 16:03). This expanded range could explain why we saw numerically greater metacognitive awareness sensitivity in those with PTSD and depression (ρ = 0:34) than those with neither disorder (ρ = 0:21). Rather than representing a scaling issue, another possibility is that numerically greater sensitivity reflects the "depressive realism" hypothesis, which argues that depressed individuals are better able to make certain judgments than nondepressed individuals [67]. The depressive realism hypothesis posits that depressed individuals would have increased metacognitive sensitivity along with relatively negative metacognitive biases, which the current study offers some support for, and is consistent with a recent meta-analysis finding a weak depressive realism effect [66].
The current study supports a model of metacognitive bias as being closely linked with depressive/PTSD symptom changes rather than being a more stable trait. Longitudinal analyses over a two-year period revealed significant associations between changes in depressive and PTSD symptoms with changes in metacognitive bias (ρ = −0:25, -0.33, respectively). This was driven by changes in self-reported cognition while objective cognition remained relatively stable. To our knowledge, this is the first study relating longitudinal changes in global metacognitive awareness with changes in PTSD or depressive symptoms. One potential explanation is that increases in symptoms of depression and PTSD, such as depressed mood and feelings of worthlessness, could lead to a more negative metacognitive bias across cognition and health. Alternatively, changes in metacognitive bias could lead to changes in depressive symptoms. A recent study found that metacognitive beliefs on the Metacognitions Questionnaire, a self-reported measure of metacognitive beliefs, predicted anxiety symptoms two months later [68]. Positive beliefs around rumination (i.e., that rumination is a useful coping strategy) also predicted greater depressive symptoms two months later [69]. However, it remains unclear if more global metacognitive biases precede changes in depressive symptoms.
Metacognitive bias changes co-occurring with or supporting changes in depression and PTSD symptoms have important clinical implications. For example, one core component of cognitive behavioral therapy for depression involves challenging and correcting perceived incompetence [70]. Indeed, improvements in executive functions were associated with greater treatment response from CBT in adults with depression or anxiety [71]. However, therapies specifically targeting metacognitive biases or sensitivity are rare. One approach, called metacognitive training, has been used in depression to successfully improve metacognitive sensitivity and bias (e.g., in false memories, [72]) as well as depressive symptoms generally [73]. In a nonclinical sample, another study training metacognitive memory ability by providing feedback to participants' predictions after each 9 Depression and Anxiety trial found both improved bias and sensitivity on that task ( [74]; see also [75,76]). A recent study found domaingeneral improvements on metacognitive bias, where feedback after completion of a perceptual discrimination task was associated with metacognitive awareness improvements on that task and another untrained recognition memory task [77]. Generally, cognitive feedback training has shown that it is possible to improve metacognitive bias and sensitivity, though additional research is needed to examine if these improvements consistently generalize to reduced depressive and PTSD symptoms and/or improved functioning in daily life.

Limitations
Though the current findings provide important insights into global metacognitive bias and accuracy, they have several limitations. Participants were combat-exposed post-9/11 veterans, which were predominantly white, male, and middle aged, and therefore may not generalize to other populations. Additionally, the self-reported cognition measure showed floor effects that necessitated exclusion of individuals without any daily life functional interference. However, the results were nearly identical when these participants were included (see Supplementary Materials (available here)). Further, normative data for American adults on the WHODAS II were not available, while the objective cognition measures were age-adjusted and z-transformed based on normative samples. As a result, the zero point of the metacognitive bias score was relative to the sample rather than to the general population. Another potential limitation is that objective cognition in a laboratory setting could be slightly overestimated in those with greater depressive/PTSD symptoms, where dynamics and stressors of daily functioning could further impair cognition [78,79]. That being said, the metacognitive bias in health results were very similar to cognition, and it is unlikely that objective health measures differ between the lab and real world. Still, it could be interesting for future studies to investigate objective and self-reported abilities in more real-world settings using ecological momentary assessments [80]. Finally, while consistent relationships between metacognitive bias and depression over PTSD were found, the sample had considerable comorbidity between PTSD and depression. A sample with more individuals with depression alone would be useful to dissociate the effects of PTSD from depression.

Summary
The current study implicates depressive symptoms above and beyond PTSD symptoms in negative global metacognitive biases and suggests that changes in metacognitive bias fluctuate with depressive and PTSD symptoms two years later. While metacognitive sensitivity was unrelated to the severity of depressive and PTSD symptoms, robust associations between depressive symptoms and negative global metacognitive biases across both cognition and health were observed, attesting to the generality of this association. The current results provide evidence for the utility of more global measures of metacognitive awareness in characterizing psy-chiatric symptoms and their relation to symptom changes and outline future clinical targets to reduce metacognitive bias and potentially enhance outcomes in those suffering from depression and PTSD.

Data Availability
The clinical data used to support the findings of this study are restricted by the VA Boston Health Care System IRB in order to protect veterans' patient privacy. Data are available by contacting Dr. Joseph DeGutis at degutis@wjh.harvard.edu for researchers who meet the criteria for access to confidential data.

Conflicts of Interest
The authors declare that there were no conflicts of interest with respect to the authorship or the publication of this article. and Stress Disorders, a VA Rehabilitation Research & Development TBI National Research Center (B9254-C). M.E. was supported by a VA Clinical Science Research & Development Merit Review (I01CX001653), and W.M. was supported by a National Institute of Health NCCIH grant (R21 AT009430-01).

Supplementary Materials
For a detailed listing of self-reported and objective cognition measures, see Table S1. For the results of metacognitive sensitivity in health and PTSD, see Figures S1-S2. For the results of metacognitive bias in health by depressive symptoms, see Figure S3. For the changes in metacognitive sensitivity and bias across time points, see Tables S2-4. For correlations between self-reported, objective, and metacognitive bias measures, see Table S5. For the metacognitive biases' relationship with clinical and demographic variables, see Table S6. For the correlations between baseline clinical and demographic variables with changes in self-reported, objective, and metacognitive measures, see Table S7. (Supplementary Materials)