Chapter  123:  Post-Myocardial Infarction Depression

Literature Search

The EPC team performed a comprehensive search that included electronic and hand searching. In March 2004, we searched the following electronic databases: MEDLINE^®, the Cochrane CENTRAL^® Register of Controlled Trials (Issue 1, 2003), the Cochrane Database of Methodology Reviews (CDMR^®), the Cumulative Index of Nursing and Allied Health Literature (CINAHL^®), the Psychological Abstracts (PsycINFO^®), and EMBASE^®.

Hand searching for possibly relevant articles was performed by three techniques. First, the EPC team identified 16 journals that we thought were most likely to contain relevant studies and scanned the table of contents of each of these journals for relevant citations from October 2003 to April 2004. Second, we reviewed references cited in recent review articles for inclusion. Third we examined the reference lists of eligible articles for additional articles that might be relevant

Two members of the EPC team independently reviewed the abstracts identified by the search to exclude those that did not meet the eligibility criteria. Primary studies were eligible if they addressed one of the key questions, included original human data, were not case reports, and were written in the English language. Individual key questions had additional exclusion criteria. When two reviewers agreed that an abstract was not eligible, it was excluded from further review.

To focus the evidence report on the studies that would be most valuable in addressing the key questions, we used the following additional eligibility criteria:

• For key question 4 we excluded studies that did not include a concurrent comparison group.

• For key question 5, we excluded studies that did not use a validated reference standard.

Review Process

Paired reviewers assessed the quality of each eligible article. Differences between the paired reviewers were resolved by face-to-face discussion. The reviewers assigned points for the quality of the studies based on information about the representativeness of the patients included in the study, the potential for bias and confounding, the description of the intervention or evaluation, the adequacy of followup, and the appropriateness of the statistical methods. The score for each category of study quality was the percentage of the total points available in each category for that study, and could range from zero to 100 percent.

One reviewer in each pair was the primary reviewer who abstracted data from the article. The second reviewer confirmed the accuracy of the first reviewer's work.

Key Question 1

In patients diagnosed with and hospitalized for acute MI what is the prevalence of depression during initial hospitalization for MI?

• Twenty-five articles met criteria for inclusion in this review.¹–⁵,⁹–²⁷,⁴⁸

• Articles were published between 1986 and 2004.

• Eight studies used a structured clinical interview,¹–³,⁵,¹⁰,¹⁶,¹⁷,⁴⁹ and 17 used a validated questionnaire. ⁴,⁵,⁹,¹¹–¹⁵,¹⁸–²⁵,⁵⁰

• Major depression was reported in about one of every five patients hospitalized for an MI. The reported prevalence of potentially significant symptoms of depression varied widely (range 10 to 47 percent).

• In general, the reported prevalence of potentially significant symptoms of depression was higher when it was based on a Beck Depression Inventory (BDI)⁵,⁹,¹⁸–²³,⁵⁰ than when based on a Hospital Anxiety and Depression Scale (HADS)¹²–¹⁵; this may be because the BDI includes somatic symptoms that may overlap with MI symptoms, whereas the HADS does not.

Key Question 1a. What is the prevalence of depression during initial hospitalization for an acute MI, with and without a history of previous depression as reported by study investigators?

There was insufficient data to address this question.

Key Question 2

What percentage of patients with post-MI depression continues to have depression (or depressive symptoms) one or more months after initial hospital discharge?

• We found 22 articles that met criteria for inclusion in this review.²,¹²,¹⁴,¹⁸–²⁰,²²,²³,²⁵,²⁶,³⁸,⁵¹–⁶¹

• Nine of the 22 used a standardized clinical interview to diagnose depression.¹⁰,²⁶,⁵¹–⁵⁷

• Only three studies reported the prevalence of depression in patients during the MI hospitalization and then specifically re-assessed and reported the prevalence in these same patients at followup.²,¹⁸,²³

• Based on these three studies, most patients with depression during the initial MI hospitalization continue to have depression 1 to 4 months later.

Key Question 3

What is the association of post-MI depression with outcomes independent of other predictors of post-MI outcomes?

• Sixteen studies addressed the relationship of post-MI depression and mortality.⁵–¹⁰,²⁰,²⁸–³⁶

• Mortality has been assessed as early as 4 months⁵ and as late as 10 years after MI.⁷

• The evidence indicates that post-MI depression is associated with a significantly increased risk of death.

• A single study indicated that post-MI depression is associated with increased cardiac re-admission in the first year after MI.³⁹

• Six studies reported on cardiac events in relationship to post-MI depression.²⁵,²⁶,³¹,³⁷–³⁹ The three studies reporting a positive relationship between post-MI depression and cardiac events³¹,³⁷,³⁹ were generally larger than the three studies finding no relationship²⁵,²⁶,³⁸ suggesting that the latter may have had insufficient power to detect differences if they, in fact, were present.

• Depression during the initial hospitalization was related to poor quality of life in the first year after MI.¹³,³⁰,⁵⁹,⁶²

Key Question 3a. What is the association of post-MI depression with surrogate markers of cardiac risk independent of other predictors of post-MI outcomes?

• Three studies examined the association of post-MI depression with heart rate variability, platelet activity, and inflammatory markers (one study for each surrogate marker).⁵⁷,⁶³,⁶⁴

• All three studies reported surrogate markers of increased risk in patients with post-MI depression, even after adjustment for covariates.

Key Question 4

Do post-MI patients with depression have better outcomes with depression treatment compared to those without depression treatment?

• Twelve studies, 11 of which were randomized controlled trials, addressed this question.¹⁰,⁴⁰–⁴⁷,⁶⁵–⁶⁷ The studies were published between 1991 and 2003.

• In post-MI patients with depression, psychosocial intervention improves depression but not other outcomes.¹⁰,⁴⁴

• In post-MI patients with depression, selective serotonin reuptake inhibitors improve depression and some surrogate markers of cardiac risk, but no studies of sufficient power address the question of whether this treatment improves survival.⁴⁵,⁴⁶,⁶⁵,⁶⁶

Key Question 4a. Do outcomes differ with or without improvement in depression for post-MI patients with depression that do receive depression treatment?

There was insufficient data to address this question.

Key question 4b. Do outcomes differ with or without improvement in depression for post-MI patients with depression that do not receive depression treatment?

There was insufficient data to address this question.

Key Question 5

What are the performance characteristics (e.g., sensitivity, specificity, reliability, and predictive value) of instruments or methods that are used to screen for depression (or depressive symptoms) following an acute MI?

• We found six studies published between 1968 and 1988 meeting criteria to address this issue.¹⁴,¹⁸,⁵⁶,⁶⁸–⁷⁰

• Of the six studies, four were from Europe,¹⁴,⁵⁶,⁶⁸,⁷⁰ one from Canada¹⁸ and one from the United States.⁶⁹

• All included studies reported exclusively on post-MI populations.

• None of the instruments reported had been normalized specifically for post-MI patients.

• The BDI tended to be more sensitive to lower levels of depressive symptoms but less sensitive to more severe depression compared to the HADS and the Symptom Checklist-90 Depression scale.

Key question 5a. What are the performance characteristics of instruments or methods that are used to screen for depression (or depressive symptoms) following an acute MI during hospitalization?

There was insufficient data to address this question.

Key question 5b. What are the performance characteristics of instruments or methods that are used to screen for depression (or depressive symptoms) following an acute MI, within three months after hospitalization?

There was insufficient data to address this question.

Key Question 6

Does the use of cardiac treatment for patients with acute MI differ for those with and without depression?

• Nine studies published between 1982 and 2004 met criteria for review on this question. ²³,²⁶,³⁷,⁵⁰,⁷¹–⁷⁵

• Four studies compared prescribed discharge medications and were inconsistent in their findings: United States and United Kingdom⁵⁰,⁷¹ studies suggested decreased prescriptions of beta-blockers and aspirin, while European²⁶ and Canadian²³ studies found no difference.

• Three studies compared adherence to prescribed medications and lifestyle modifications and consistently found decreased adherence among depressed patients.⁷¹–⁷³

• Two studies compared use of cardiac procedures and reached divergent conclusions about the use of procedures in post-MI patients.²³,³⁷

• Two studies assessed completion of cardiac rehabilitation but had insufficient numbers to reach conclusions about the influence of depression on completion of rehabilitation.⁷⁴,⁷⁶

Key Question 1

Major depression is reported in about one of every five patients hospitalized for MI. This proportion is fairly consistent among the eight studies that used a structured clinical interview to establish this diagnosis. The reported prevalence of potentially significant symptoms of depression varies more widely (range 10 to 47 percent). This wide range of reported prevalence rates appears to be due almost exclusively to differences in measurement instruments used, and even to differences in threshold criteria applied from study to study when the same instrument was used. In general, the reported prevalence of potentially significant symptoms of depression is higher when this diagnosis is based on a BDI score of 10 or higher than when it is based on a HADS score of either 8 or higher or 11 or higher. This difference may be attributed to the BDI's inclusion of somatic symptoms that may overlap with MI symptoms, whereas the HADS does not include somatic symptoms and is designed for use in hospitalized patients.

Additional studies also are needed to define the most clinically-relevant measure of depression during the initial MI hospitalization. Studies are needed to determine the clinical or demographic factors that are associated with post-MI depression.

Key Question 2

Although 22 studies reported the prevalence of depression in patients 1 month or longer after initial hospital discharge, only three reported the prevalence of depression in patients during the MI hospitalization and then specifically reassessed and reported the prevalence of depression in these same patients at followup. These studies suggest that most patients (60 to 70 percent) with depression during the initial MI hospitalization continue to have depression (or depressive symptoms) 1 to 4 months later.

Additional studies are needed that assess depression (or depressive symptoms) in groups of patients during the initial hospitalization and at various time points after MI. Studies of patients who are reassessed for depression at multiple time points post-MI are also needed.

Key Question 3

Sixteen studies evaluated the relationship between depression, measured shortly after an acute MI, and subsequent mortality. Studies have assessed this relationship as early as 4 months post-MI and as late as 10 years post-MI. Despite the facts that various measures of depression have been used, that different subgroups of depressed patients have been evaluated, and that different post-MI survival times have been assessed, the weight of the evidence is strikingly consistent. Overall, the evidence supports the notion that post-MI depression is associated with a significantly increased risk for subsequent death, whether by cardiac or other causes. Depression appears to be associated with about a 3-fold increased risk of cardiac mortality per se based on at least three studies that addressed cardiac mortality in a total of almost 2,000 patients.³⁹,⁷⁷,⁷⁸ Depression during the initial hospitalization is associated with poor quality of life in the first year after an MI.

During the first year after MI, depression during the initial MI hospitalization has been found to be inversely related to physical quality of life, social quality of life of women, sexual activity and satisfaction among men, return to work of employed men, and to physical, psychological, and social health and function. Limitations of the above mentioned studies included the variety of diagnostic instruments used to assess depression; the lack of agreement on what aspects of quality of life are of greatest import or how to measure included studies; the degree to which potential confounders were adequately considered; and the absence of data in early post-MI time points.

Additional studies are needed to determine the major cause(s) of mortality among depressed post-MI patients. Additional studies also are needed to determine whether patients with depression are at higher risk for malignant arrhythmias than comparable post-MI patients without depression.

Key question 3a. A small amount of evidence suggests that post-MI patients with depression have alterations in autonomic function as reflected by decreased heart rate variability, increased platelet activity, and increased levels of soluble adhesion molecule 4. These studies suggest that the risk associated with post-MI depression could be transmitted by multiple biological pathways.

Additional studies are needed to elucidate the mechanism(s) responsible for increased mortality in patients with post-MI depression. Particular emphasis should be placed on surrogate markers which have been previously associated with increased risk without regard to depression, including markers for sudden death including heart rate variability, T-wave alternans, etc, and inflammatory markers including C-reactive protein, interleukins, adhesion molecules and others. Studies are needed that evaluate the hemostatic and platelet function of patients with post-MI depression. Future studies also should address whether responses to commonly used antiplatelet agents differ among post-MI patients with versus without depression.

Key Question 4

No studies of sufficient power have yet been performed that directly address the question as to whether treatment with antidepressants improves survival in depressed patients after an MI. Some evidence suggests that selective serotonin reuptake inhibitor antidepressants have beneficial effects on surrogate markers of post-MI risk (e.g., heart rate variability, aortic time velocity integral). There is evidence that both psychosocial intervention and selective serotonin reuptake inhibitor antidepressants improve depression in post-MI patients. However, the possibility of increases in rare adverse events cannot be excluded.

Studies are needed to determine whether patients with depression who are treated for depression, especially with highly effective drugs, differ in outcomes from patients who are not treated. Future studies should also determine whether treatment for depression per se or resolution of depression is associated with different outcomes.

Key Question 5

There are insufficient data to allow an adequate assessment of the performance characteristics of instruments or methods used to screen for depression during the initial MI hospitalization. The very low positive predictive values of these screening instruments (generally in the 25 to 50 percent range) may be acceptable clinically if followed by a more thorough assessment of those who screen positive; however, the low positive predictive values are particularly problematic if used to detect relationships to outcome variables in the research setting. When compared with the HADS and Symptom Checklist-90 Depression scale, the BDI tends to diagnose less significant symptoms of depression at higher rates. It may be less effective in accurately diagnosing major depression.

Additional studies are needed to determine the performance characteristics of instruments or methods used to screen for depression (or depressive symptoms) during the initial MI hospitalization. Studies are needed in post-MI patients that examine the ability for depression screening instruments or methods to distinguish symptoms of depression from symptoms attributable to the MI, to poor physical health, or to the hospitalization itself.

Key Question 6

It remains unclear whether there are significant differences in cardiac medications prescribed to post-MI patients based on the presence or absence of depression. Three studies evaluated adherence to prescribed medications and secondary prevention measures in post-MI patients and consistently found lower adherence in those with depression than those without depression. Two good-quality studies, using different methods, came to diverse conclusions about whether the frequency with which cardiac procedures are used varies between post-MI patients with depression and those without depression.

Additional large studies are needed to examine whether the use of diagnostic and therapeutic procedures differs between depressed and non-depressed post-MI patients. Future studies should also address whether potential differences in procedures are due to differences in provider recommendation or to differences in patient acceptance. Further studies are needed to determine whether the treatment prescribed to post-MI patients with depression differs from those without depression. Future studies should address whether the non-pharmacologic interventions (including diet, exercise and cardiac rehabilitation) recommended to post-MI patients differ between those patients with and without post-MI depression. Future studies should examine the adherence behavior of post-MI patients and evaluate measures that could improve adherence to recommended treatment.

Introduction

Most studies that examined the prevalence of depression after MI examined patients at the time of the initial hospitalization and at various times after discharge. Because question 2 refers to the percentage of patients who continue to have depression after hospital discharge, these two questions are interrelated, and therefore questions 1 and 2 are combined for the purposes of this report.

Important technical and methodological issues should be clarified before examining this review. First, depression is defined and addressed differently from study to study. For the purposes of this review, we have defined depression as “symptoms meeting established clinical threshold criteria for depression as measured by validated questionnaires or standardized psychiatric interviews.” It must be noted, however, that the manner in which “clinical threshold criteria” is interpreted varies considerably across studies. For example, one study used a score of 8 or higher on the depression subscale of the Hospital Anxiety and Depression Scale (HADS) to report a prevalence of “clinically significant levels of depression”,⁶³ whereas another used a cutoff of 11 to denote “clinical levels of depression”.⁶⁴ Three other studies referred to a score in the 8 to 10 range as “possibly clinically significant” or “borderline” depression and a score of 11 or higher as “probably clinically significant” depression.⁶⁵–⁶⁷ This review found similar variability in threshold criteria and designation when other screening instruments were used. Wherever possible, this has been specifically noted given that prevalence data may become confusing when “depression” is characterized according to different threshold criteria. In addition, although the definition of depression for the purposes of this review includes measurements by validated questionnaires and standardized interviews, the former typically measure symptoms of depression whereas the latter typically use a version of the Diagnostic and Statistical Manual (DSM) to establish a diagnosis of major or minor depression or dysthymia. Again, wherever possible, this is noted in the report. Finally, it should be noted that the screening instruments typically used for depression are referred to in this report as “validated questionnaires or rating scales.” Some validation evidence is available for all of these, but the quality varies and no systematic validation has been carried out in patient populations with recent MI. Indeed, question 5 of this report addresses the performance characteristics of instruments or methods used to screen for depression after an acute MI.

Results of Literature Search

After the final article review, 51 articles were eligible pertaining to question 1. Of these, 22 were studies reporting data that were also reported in at least one other study. For example, some investigators reported the prevalence of in-hospital depression and outcome at 6 months. In another study by the same investigators, the prevalence of in-hospital depression was again reported but with a 12-month outcome. In such instances, only one article was considered eligible for question 1. After these 22 articles were excluded, 5 others were excluded because the prevalence of depression was determined by using a non-standard screening instrument for which no validation was provided. This left 25 articles for question 1.

After article review, 30 articles were eligible pertaining to question 2. One of these was excluded as duplicative, and six others were excluded because the prevalence of depression was determined by using a non-standard screening instrument for which no validation was provided. One study was excluded because the follow-up data were collected 2 weeks after discharge rather than the minimum 1 month after discharge that was required for this question. This left 22 articles for question 2. Of these 22 articles, seven reported prevalence data both at hospitalization and follow-up. However, only three of these separately reported prevalence data at follow-up for those depressed during hospitalization. The other four articles and the 15 that reported only follow-up data are reported separately because they did not directly address question 2.

Characteristics of Studies

Table 1

Summary of Studies on the Prevalence of Depression (more...)

Table 1

Summary of Studies on the Prevalence of Depression During Hospitalization for Acute Myocardial Infarction (Question 1)

Study Author, Year	Study Design	Location	No. of Subjects	Mean Age	Male (%)	White (%)	HTNa(%)	DMb(%)	Smoking (%)	Lipidc(%)	Method & Time of Assessment of Depression	Prevalence (%)
Taylor, 1986	RCTd	USA	173	52	100	NRe	NR	NR	NR	NR	HAMDf; 3 wkg post-MIh	13
Bennett, 1988	Pro cohorti	Europe	37	62	73	NR	NR	NR	NR	NR	HADSj; In hospitalk	13
Davis, 1988	Pro cohort	Canada	52	51	90	NR	NR	NR	NR	NR	SCIDl, and BDIm;In hospitalk	6, 10
Carney, 1990	Case control	USA	70	53	76	NR	43	16	48	NR	DISn; In hospital	23
Silverstone 1990	Pro cohort	Europe	100	NR	NR	NR	NR	NR	NR	NR	Montgomery-Asberg; In hospital	19
Gilutz, 1991	Pro cohort	Europe;	Europe: 98;	NRo	NR	NR	NR	NR	Europe: 32;	NR	Holland Sgroi Anxiety Depression Scale;	31, 35
		Middle East	Middle East: 87						Middle East: 35		In hospital, and 10–15 dp post-MI
Schleifer, 1991	Pro cohort	USA	335	63	64	NR	NR	NR	NR	NR	Nurse Interview; 8–10 d post-MIk	30, 17
Forrester, 1992	Cross-sectional	USA	129	59	74	62	NR	NR	NR	NR	PSEq; In hospital within 10 d post-MI	19
Legault, 1992	Pro cohort	Canada	52	55	78	NR	NR	NR	NR	NR	BDI; In hospitalk	18
Kaufmann, 1999	Pro cohort	USA	331	NRr	66	NR	50	27	28	38	DIS; In hospital 3–15 d post-MI	27
O'Rourke, 1999	Pro cohort	Europe	70	58	74	NR	NR	NR	NR	NR	HADS; 3 – 5 d post-MI	17
Mayou, 2000	Pro cohort	Europe	344	63	73	NR	NR	34	58	NR	HADS;In hospital, and within 3 d post-MI	18, 8
Bush, 2001	Pro cohort	USA	267	65	58	82	43	33	27	59	SCID, and BDI; In hospital within 2–5 d post-MI	17, 20
Brink, 2002	Pro cohort	Europe	114	68	68	NR	35	15	31	NR	HADS; In hospital, and within 1 wk post-MI	11, 8
Lane, 2002	Pro cohort	Europe	288	63	75	93	39	13	43	72	BDI; In hospital within 15 d post-MIk	31
Lesperance, 2002	Pro cohort	Canada	896	59	68	NR	35	16	47	NR	BDI; In hospitalk	32
Luutonen, 2002	Pro cohort	Europe	85	61	77	NR	NR	NR	NR	NR	BDI; In hospitalk	21
Watkins, 2002	Cross-sectional	USA	204	59	58	65	NR	Non- depressed 31, Depressed 49	Non- depressed 53, Depressed 81	NR	DIS; In hospital 3–9 d post-MI	18
Barefoot, 2003	Pro cohort	USA	196	61	63	67	NR	NR	NR	NR	HAMD, and BDI; within 2 wk post-MIk	28, 37
Berkman, 2003	RCT	USA	9279s	63	61	26	60t	33t	64t	57t	SCID; 2–4 wk post-MI	27
Lauzon, 2003	Pro cohort	Canada	550	60	80	96	35	16	40	38	BDI; Within 2 – 3 d of hospitalizationk	35
Martin, 2003	Pro cohort	Europe	335	67	67	NR	NR	NR	NR	NR	HADS; In hospital, and within 1 wk post-MIk	15, 6
Rafanelli, 2003	Pro cohort	Europe	61	61	85	NR	NR	NR	NR	NR	SCID; Within 1 mou post-MI	Minor 10, Major 2
Strik,2003	Pro cohort	Europe	318	52	1	NR	28	9	54	20	SCL-90; 1 mo post-MI	47
Steeds,2004	Pro cohort	Europe	131	NR	NR	NR	NR	NR	NR	NR	BDI-II; In hospital	47

a

Hypertension

b

Diabetes mellitus

c

Hyperlipidemia

d

Randomized controlled trial

e

Not reported

f

Hamilton Rating Scale for Depression

g

Week

h

Myocardial infarction

i

Prospective cohort study

j

Hospital Anxiety and Depression Scale

k

Also assessed at a later point. See Question 2

l

Structured Clinical Interview for Diagnostic & Statistical Manual of Mental Disorders - IV

m

Beck Depression Inventory

n

Diagnostic Interview Schedule

o

In the European population 21% were less than 45 years old, 54% were 46 – 55 years old and 25% were 56 – 60 years old. In the Middle Eastern population 30% were less than 45 years old, 46% were 46 – 55 years old and 24% were 56 – 60 years old

p

Day

q

Psychological Stress Evaluator

r

45% were less than 65 years old and 56% were greater than 65 years old

s

Medically eligible and screened for depression

t

Of those with depression or low social support

u

Month

Table 2

Summary of Studies on the Prevalence of Depression (more...)

Table 2

Summary of Studies on the Prevalence of Depression After Hospitalization for Acute Myocardial Infarction (Question 2)

Study Author, Year	Study Design	Location	No. of Subjects	Mean Age	Male (%)	White (%)	HTNa(%)	DMb(%)	Smoking (%)	Lipidc(%)	Method & Time of Assessment of Depression	Prevalence
Trelawny, 1987	NRd	Europe	32	NR	100	NR	NR	NR	71	NR	Goldberg's CISe; In hospital,
											10 df post d/cg,	At 10 d: 20,
											2 moh post d/c,	At 2 mo: 26,
											6 mo post d/c	At 6 mo: 26
Davis, 1988	Retro cohorti	Canada	52	51	90	NR	NR	NR	NR	NR	BDIj;
											6 – 8 wkk post-MIl	At 6 – 8 wk 10
Follick, 1988	RCTm	USA	238	55	72	NR	NR	NR	53	NR	SCL-90n;
											Baseline,
											1 mo,
											3 mo,
											9 mo	At 9 mo: 10
Schleifer, 1991	Retro cohort	USA	335	64	64	NR	NR	NR	NR	NR	RDCo (Nurse Interview);	At 3 – 4 mo:
											8–10 d,	Majorp 14,
											3–4 mo	Minorq 19
Legault, 1992	Pro cohortr	Canada	52	55	78	NR	NR	NR	NR	NR	BDI;
											In hospital,
											3 mo post-MI,	At 3 mo: 7,
											12 mo post-MI	At 12 mo: 9
Garcia, 1994	Cross-sectional	Europe	97	50	100	NR	NR	NR	NR	NR	RDC;	Major 11,
											In hospital,	Minor 27
											1 mo post-MI
Travella, 1994	Pro cohort	USA	70s	58	74	NR	NR	NR	NR	NR	HAMDt,	At 3 mo:
											PSE;	Major 15,
											3 mo post-MI,	Dysthymia 4 ;
												At 6 mo: Major 21,
											6 mo post-MI,	Dysthymia 3 ;
												At 9 mo: Major 28,
											9 mo post-MI,	Dysthymia 3;
												At 12 mo: Major 16,
											12 mo post-MI	Dysthymia 13
Clarke, 1996	Pro cohort	Canada	52	NR	100	NR	NR	NR	NR	NR	ZDSu;
											3 mo post-MI	At 3 mo: 24
Lesperance, 1996	Pro cohort	Canada	222	60	78	NR	NR	NR	NR	NR	DISv;
											6 mo post-MI,	At 6 mo: 20,
											12 mo post-MI	At 12 mo: 9
Bennett, 1998	Pro cohort	Europe	37	62	73	NR	NR	NR	NR	NR	HADSw;
											In hospital,
											3 mo post-MI	At 3 mo: 3
Lehto, 2000	Retro cohort	Europe	101	62	69	NR	28	15	12	63	DEPSx;
											20.5 mo (median) post-MI	20.5 mo: 15.8 (median)
Strik, 2001	Pro cohort	Europe	206	60	76	NR	NR	NR	NR	NR	SCIDz,	At 1 mo
											HAMD,	Major 11
											SCL-90,	Minor 8
											BDI,
											HADS;
											1 mo post-MI
Lane, 2002	Pro cohort	Europe	288q	63	75	93	NR	NR	NR	NR	BDI;
											In hospital,
											4 mo post-MI,	At 4 mo: 38
											12 mo post-MI	At 12 mo: 37
Luutonen, 2002	Pro cohort	Europe	85	61	77	NR	NR	NR	NR	NR	BDI;
											6 mos,	At 6 mos: 30
											18 mos	At 18 mos 34
Shiotani, 2002	Pro cohort	Asia	1042	64	80	NR	48	32	66	37	ZDS;	Within first 3 mo: 42
											within 3 mo post-MI
Strik, 2002	Retro cohort	Europe	140	58	76	NR	NR	NR	NR	NR	BDI, HADS, SCL-90 if any one positive assessment with SCID;	Major 11
											3 mo post-MI	Minor 2
Aben, 2003	Pro cohort	Europe	200	60	77	NR	NR	NR	NR	NR	BDI, HADS, SCL-90 if any one positive assessment with SCID and HAMD;
											1 mo after MI,	At 1 mo: 14
											3 mo after MI,	At 3 mo: 19
											6 mo after MI,	At 6 mo: 21
											9 mo after MI,	At 9 mo: 27
											12 mo after MI	At < 1 yraa: 28
Barefoot, 2003	NR	USA	196	61	63	67	NR	NR	NR	NR	HAMD, BDI;	At 2 wks by HAM- D: 17;
											In hospital,	by BDI: 27
											2 wk after first assessment hospitalization
Lauzon, 2003	Pro cohort	Canada	550	60	80	96	35	16	40	38	BDI;
											In hospital,
											1 mo post-MI,	At 1 mo: 39
											6 mo post-MI,	At 6 mo: 39
											12 mo post-MI	At 1 yr: 30
Martin, 2003	Cross-sectional	Europe	335	67	67	NR	NR	NR	NR	NR	HADS;
											6 wk post-MI,	At 6 wk: 5o
											6 mo post-MI	At 6 mo: 5.4o
												6 wk: 13p
												6 mo: 10p
Strik, 2003	Pro cohort	Europe	318	52	1	NR	28	9	54	20	SCL-90;
											1 mo post-MI	At 1 mo: 47
Lesperance, 2004	Pro cohort	Canada	481	60	81	NR	66	NR	15	NR	SCID;
											Approx. 2 mo post-MI	Approx 2 mo: 7

a

Hypertension

b

Diabetes mellitus

c

Hyperlipidemia

d

Sample of suspected myocardial infarction patients 28 of 32 confirmed

e

Clinical Interview Schedule

f

Day

g

Discharge

h

Month

i

Retrospective cohort study

j

Beck Depression Inventory

k

Week

l

Myocardial Infarction

m

Randomized controlled trial

n

Symptom Check List 90

o

Research Diagnostic Criteria

p

Major depression

q

Minor depression

r

Prospective cohort study

s

With variable numbers at different time points

t

Hamilton Rating Scale for Depression

u

Zung Depression Scale

v

Diagnostic Interview Schedule

w

Hospital Anxiety and Depression Scale

x

Depression scale

z

Structured Clinical Interview for Diagnostic & Statistical Manual of Mental Disorders - IV

aa

Year

Of the 25 articles for question 1, eight used a standardized clinical interview for the diagnosis of depression (e.g., Structured Clinical Interview or the Diagnostic Interview Schedule).⁵⁸, ⁶⁸–⁷⁴ One of these⁷² interviewed patients within a month of their MI, but the exact timing of the assessment relative to the hospitalization is unclear, and so data from this study are not included in summary tabulations. Seventeen used one or more validated questionnaires or rating scales such as the Beck Depression Inventory (BDI),²⁵ ,⁷⁴–⁸¹ Hamilton Rating Scale for Depression (HAMD),⁷⁸,⁸² HADS,⁶³–⁶⁷ Montgomery and Asberg Depression Rating Scale,⁸³ or Symptom Check List-90 (SCL-90) Depression Subscale.⁸⁴

Of the 22 articles for question 2, nine used a standardized clinical interview for the diagnosis of depression.¹⁰,⁶⁹,⁸⁵–⁹¹ Seventeen used one or more validated questionnaires or rating scales such as the BDI,¹⁰,⁷⁵,⁷⁶,⁷⁸,⁸⁰,⁸¹,⁸⁸,⁸⁹,⁹² HAMD,¹⁰,⁷⁸,⁹³ Zung Depression Rating Scale (ZDS),⁹⁴,⁹⁵ HADS,¹⁰,⁶⁴,⁶⁶,⁸⁸,⁸⁹ or SCL-90 Depression Subscale.¹⁰,⁸⁴,⁸⁸,⁸⁹,⁹⁶

Quality of Studies

Results of Studies

Table 1.1 (Appendix G) describes the patient populations and prevalence of depression during the initial MI hospitalization. Table 1.2 (Appendix G) describes the patient populations and prevalence of depression 1 month or longer after hospital discharge.

Of the 25 articles for question 1, seven included fewer than 100 subjects,⁶³,⁶⁴,⁶⁸,⁷³,⁷⁵,⁷⁶,⁸⁰ and 17 studied more than 100 individuals.²⁵,⁵⁸,⁶⁵–⁶⁷,⁶⁹–^{72 77}–⁷⁹,⁸¹–⁸⁴,⁹⁷ The range of subjects studied was 37 to 2,481 patients. Thirteen studies were from the United States and/or Canada,²⁵,⁵⁸,⁶⁸–⁷²,⁷⁴–⁷⁶,⁷⁸,⁸⁰,⁸² and 12 studies were from Europe.⁶³–⁶⁷,⁷³,⁷⁷,⁷⁹,⁸⁰,⁸³,⁸⁴,⁹⁷ The mean age of study subjects ranged from 52 to 67 years. Sex was reported in all but two studies; men alone were studied in three reports and the remainder reported depression prevalence from a population mix of men and women. Only three of 25 studies reported race;⁵⁸,⁷⁰,⁷⁸ in these three, non-whites made up 33 to 38 percent of the sample. Nine studies reported the prevalence of individual cardiac risk factors²⁵,⁵⁸,⁶⁷,⁶⁸,⁷¹,⁷⁴,⁷⁷,⁸⁰,⁸⁴; whereas one study reported the aggregate number of cardiac risk factors in the population.⁷⁶ Killip class and/or LVEF was reported in seven studies.²⁵,⁶⁹–⁷¹,⁷³,⁷⁴,⁸¹ All studies reported depression prevalence data from MI survivors alone or from a study sample in which prevalence data from MI survivors were reported separately.

Although in question 1a the prevalence of depression in patients with a history of depression (as reported by study investigators) was to be addressed, only two studies reported history of depression,⁷¹,⁷⁸ and none provided separate prevalence data for this segment of the sample. Therefore, there is insufficient evidence to answer this question.

The prevalence of depression ranged from 17 to 27 percent in the seven studies that used a Structured Clinical Interview for DSM IV (SCID) to establish the diagnosis. The largest of these studies, the ENRICHD study,⁵⁸ which examined 2,481 patients, reported a prevalence of 20 percent with major depression.

In the 17 studies that used a validated questionnaire or rating scale, 10 to 47 percent of patients had symptoms of depression considered above a clinical threshold criterion. Of the nine studies that used the original version of the BDI, five used a cutoff score of 10 or higher.²⁵,⁷⁴,⁷⁷,⁷⁸,⁸⁰,⁸¹ The prevalence of at least mild to moderate symptoms of depression using a cutoff of 10 or higher on the BDI was 21 to 37 percent. In the four studies that used a HADS cutoff score of 8 or higher,⁶³,⁶⁵–⁶⁷ the prevalence of possible clinical depression was 11 to 18 percent. In the four studies that used a HADS cutoff score of 11 or higher,⁶⁴–⁶⁷ the prevalence of probable clinical depression was 6 to 13 percent.

In summary, the prevalence of major depression during the initial MI hospitalization is fairly consistent across studies. Based on these studies, about one of every five patients has major depression during the initial MI hospitalization. By contrast, the prevalence of different levels of depressive symptoms varies more widely across studies assessed by questionnaire or rating scale. Variation in screening instruments used, as well as variation in cutoff criteria within instruments, may account for the difference. It is impossible given current research on the assessment of depression in MI patients to compare “possible clinical depression” based on a HADS depression subscale with “at least mild-to-moderate symptoms of depression” based on a BDI. In general, the prevalence of symptoms of depression that meet standard threshold criteria on the BDI (10 or higher) is greater than that using the HADS (either 8 or higher or 11 or higher) in patients during the initial MI hospitalization. For example, the two studies in this report that included the most patients used the BDI for 870 patients, reporting a prevalence of 32 percent²⁵ and the HADS for 688 patients, reporting prevalences of 18 percent and 10 percent depending on the cutoff.⁶⁴ This is consistent with the prevalence ranges for all studies using these two instruments as noted above. This discrepancy is not surprising in that the BDI includes somatic symptoms that may overlap with MI symptoms in hospitalized patients, whereas the HADS is designed for use in hospitalized patients and does not include somatic symptoms.

Evidence Grade Table 1. Grading of the Quality of Evidence (more...)

Evidence Grade Table 1. Grading of the Quality of Evidence on the Prevalence of Depression After Myocardial Infarction (Questions 1 and 2)

	During Initial Hospitalization	One or More Months after Hospitalization
Quantity of Evidence:
Number of studies	23	23
Total number of patients studied	14,017	5,216
Quality and Consistency of Evidence:
Were study designs appropriate for determining prevalence? (yes = high quality; no = low quality)	No	No
Did the studies have serious (-1) or very serious (-2) limitations in quality? (Enter 0 if none)	0	0
Did the studies have important inconsistency? (-1)	-1	-1
Was there some (-1) or major (-2) uncertainty about the directness or extent to which the people, measures and outcomes are similar to those of interest?	0	-1
Were data imprecise or sparse? (-1)	0	0
Did the studies have high probability of reporting bias? (-1)	0	0
Did the studies show strong evidence of association between depression and myocardial infarction? (“strong” if significant relative risk or odds ratio > 2 based on consistent evidence from 2 or more studies with no plausible confounders (+1); “very strong” if significant relative risk or odds ratio > 5 based on direct evidence with no major threats to validity (+2))	Not Applicable	Not Applicable
Did the studies have unmeasured plausible confounders that most likely reduced the magnitude of the observed association? (+1)	Not Applicable	Not Applicable
What was median (and range) of estimated prevalence?	~26%	~21%*
	Range 2– 47%	Range 2.7– 47%
Overall quality grade (high, medium, low, very low)	Medium	Medium

*

NOTE - Numbers based on data at 1–3 months (N=17)

Evidence grading scheme as described in: Grade Working Group. Grading quality of evidence and strength of recommendations. BMJ 2004; 328: 1490–7.

As summarized in Evidence Grade Table 1, we concluded that the overall body of evidence on question 1 merited a medium quantity of evidence and reasonable quality of evidence despite the presence of some inconsistencies in the evidence.

Introduction

A number of studies have reported that depression after an MI is associated with increased mortality and poorer outcomes on a variety of psychosocial and health-related variables. However, not all of these studies accounted for the potential role of confounding factors that might have influenced these observations, including cardiovascular disease severity, socioeconomic factors, psychological variables, and psychiatric conditions apart from depression. To better understand the role of depression per se in predicting negative outcomes after an acute MI, we reviewed the English language literature for articles that evaluated the relationship between depression in days to weeks following an acute MI and a variety of health and psychosocial outcome measures.

We also reviewed the literature for articles that examined the relation between post-MI depression and surrogate markers of cardiac risk that might be associated with adverse outcomes—e.g., increased ventricular arrhythmias, reduced HRV, higher levels of inflammatory markers, and abnormal platelet function.

Results of Literature Search

Table 3

Summary of Studies on Relation of Depression to Survival (more...)

Table 3

Summary of Studies on Relation of Depression to Survival after Myocardial Infarction (Question 3)

Study Author, Year	No. Enrolled	Depression Instrument	Outcome Reported			Stata	Dep vs Non Depb	Multivariate Comparison
			Cardiac Mortality	Total Mortality	Other
Ahern, 1990	265	BDIc profile of mood states	NAd	NA	TMe or cardiac arrest at baseline	RRf	NRg	1.38 (CIh 0.99 – 1.93)
Ladwig, 1991	560	NR	Low: 0.9; Medium: 2.4; High: 7.5	NA	Sustained VTi admission	ORj	Mantel-Haenszel p<0.001	2.8 low/med depression, 4.9 low/high depression p=0.07
Frasure-Smith, 1993	222	MIMH DISk	NA	At 6 mosl	NA	HRm	5.74	4.29 (CI 3.14 – 5.44)
Frasure-Smith, 1999	904	BDI	NA	NA	Arrhythmia; MI recurrence; Revascularization Any hard event	OR	CVn death 3.23 (CI 1.65 – 6.33);	3.66 (CI 1.68 – 7.99)
							Arrhythmia 3.11 (CI 1.32 – 7.37);
							MI recurrence 1.62 (CI 0.93 – 2.8);
							Revascularization 0.82 (CI 0.53 – 1.26);
							Any hard events 1.97 (CI 1.25 – 3.13)
Irvine, 1999	703	BDI	Sudden cardiac death	NA	NA	HR	AMIo 0.52 (CI 0.15 – 1.76)	NR
Denollet, 2000	319	ZDSp	Cardiac events	NA	QOLq poor perceived health at 5 yrsr, Depressive effect	OR	NR	1.31 (CI 0.53 – 3.24)
Lane, 2000	288	BDI	Dep 9%, NonDep 0.7%	Dep 10.1%, NonDep 8%	Dartmouth chart	OR	NR	NR
Welin, 2000	275	ZDS	17%	24%	Non-fatal recurrent MI; Stroke; Cancer	HR	TM: 2.45 (CI 1.49 – 4.02),	TM 1.75 (CI 1.02 – 2.99),
							CMs3.54 (CI 1.85 – 6.79)	CM 3.16 (CI 1.38 – 7.25)
Bush, 2001	285	SCIDt BDI	NA	Dep: 13%, NonDep: 3.8%	NA	RR	3.8 p=0.008	3.5 p=0.001
Druss, 2001	88241	NR	NA	NA	NA	HR	Mental 1.19 (CI 1.04 – 1.36),	NR
							Affective 1.11 (CI 1.03 – 1.18)
Lane, 2001	288	BDI	10%	10%	QOL	OR	CM 1.15 (CI 0.49 – 2.69)	NR
Lane, 2002	288	BDI	Dep: 10%	NA	NA	OR	TM 1.04 (CI 0.5 – 2.16),	NR
							CM 0.84 (CI 0.37 – 1.90)
Lesperance, 2002	896	BDI	< 5 yrs: 7.2%;	< 5 yr: 10.7%;	NA	HR	BDI 5 – 9: 1.94 (CI 1.16 – 3.25);	TM
			5 – 9 yr: 13.7%;	5 – 9 yr: 16.2%;			BDI 10 – 18: 2.8 (CI 1.68 – 4.66);	BDI 10 – 18: 2.35 (CI 1.53 – 3.61)
			10–18 yr: 18.5%;	10–18 yr: 23.2%;			BDI > 19: 1.32 (CI 2.4 – 7.75)	BDI ≥ 19 (CI 2.16 – 5.92) TM
			> 19 yr: 26.6%	> 19 yr: 32.9%				BDI 10 – 18: 3.17 (CI 1.79 – 5.6)
								BDI ≥ 19: 3.13 (CI 1.56 – 6.27)
Carney, 2003	NR	Depression interview: structured HAMDu, BDI	NA	Non-fatal AMI	NA	HR	TM 2.4 (CI 1.2 – 4.7)	TM 2.4 (CI 1.2 – 4.7)
							Non-fatal AMI 1.2 (CI 0.7 – 2.0)	Non-fatal AMI 1.2 (CI 0.7 – 2.0)
Frasure-Smith, 2003	NR	NR	At 5 yr	NA	NA	HR	NR	1.44 (CI 91.17-1.78)

a

Comparison statistic

b

Depressed versus non-depressed

c

Beck Depression Inventory

d

Not applicable

e

Total mortality

f

Relative risk

g

Not reported

h

Confidence Interval

i

Ventricular tachycardia

j

Odds Ratio

k

National Institute of Mental Health Diagnostic Interview Schedule

l

Month

m

Hazard ratio

n

Cardiovascular

o

Acute myocardial infarction

p

Zung Depression Scale

q

Quality of life

r

Year

s

Cardiac Mortality

t

Structured Clinical Interview for DSM-IV

u

Hamilton Rating Scale for Depression

Table 4

Summary of Studies on the Relation of Depression to (more...)

Table 4

Summary of Studies on the Relation of Depression to Cardiac Events after Myocardial Infarction (Question 3)

Study Author, Year	No Enrolled	Depression Instrument	Outcome			Stata	Dep vs. Non-Depb	Multivariate Comparison
			Cardiac Events	Cardiac Mortality	Other
Irvine, 1999	703	Psychological questionnaire	NRc	Sudden cardiac death 34,	Vascular death 1,	RRd	NRe	Amiodarone group:
				Other cardiac death 16	Non-cardiac death 12			BDIf somatic score 1.10 (CI 0.93,1.29);
								BDI cognitive-affective score 0.73 (CI 0.52,1.01);
								Placebo Group:
								BDI somatic score 1.00 (CI 0.88,1.13);
								BDI cognitive-affective score 1.09 (CI 0.99,1.89)
Frasure-Smith, 2000	222	Structured baseline interview	Recurrent cardiac eventsg	NAh	NA	ORi	Any cardiac event: 3.32 (CI 1.69,6.53);	Recurrent cardiac events 1.99 (CI 0.92,4.31)
							ACSj: 2.75 (CI 1.32,5.72);
							Arrhythmic event: 3.65 (CI 0.99,10.47)
Druss, 2000	NR	ICD-9CM	NA	TM at 1 mok	Likelihood of PTCAl or CABGm during index hospitalization	OR	TMn:	TM 0.63 ( p=0.2) in patients with affective disorders;
							Affective disorders 7.3%;	Use of PTCA and CABG in patients with affective disorders:
							No mental disorders 10.8%	PTCA: 0.51
								CABG : 0.63
Shiotani, 2002	1086	ZDSo	Annual cardiac event rate: Dep 31%, Non-dep 24%;	CMp: Dep 4, Non-dep 1	Readmission: Dep: 34, Non-dep: 26	OR	Cardiac events: 1.46 (CIq 1.11 – 1.92)	Cardiac events: 1.41 (CI 1.04 – 1.92)
			Cardiac events: Dep 138, Non-dep 145;
			MI: Dep 19, Non-dep 15;
			Arrhythmias: Dep 2, Non-dep 1;
			RePTCA: Dep 94, Non-dep 110;
			Heart Failure: Dep 7, Non-dep 5;
			Angina: Dep 6, Non-dep 5
Strik, 2003	318	SCL-90r	Non-fatal MI at 3–4 yrss	Fatal MI at 3 – 4 yrs	Consumption at 3–4yrst	HRu ORv	Consumptionx: 1.61 (CI 1.00 – 2.57)w	MI:CI 2.32 (CI 1.04 – 5.18);
								Consumption: 1.55 (CI 0.96 – 2.52)
Strik, 2004	422	Baseline: SCID; Follow-up: 3 psychiatric self rating scales BDI, SCL-90, HADSaa	Major cardiac events 1 mo – 3 yrsy;	NR	Consumption 1 mo – 3 yrs	HRab	Cardiac events: 1.1(0.36,3.42);	Cardiac Events: 0.88 (CI 0.26,2.93);
			Increased consumption			ORac	Consumption: 1.66 (CI 0.90 – 3.07)	Consumption: 1.98 (CI 1.0 – 3.93)

a

Comparison statistic

b

Depressed versus non depressed

c

Not reported

d

Risk ratio

e

Depression reported not to be significant predictor of sudden cardiac death in univariate analysis

f

Beck Depression Inventory

g

Survived and nonsurvived reinfarctions, admission for unstable angina, arrhythmic deaths, survived cardiac arrests

h

Not applicable

i

Odds ratio

j

Acute coronary syndrome

k

Month

l

Percutaneous transluminal coronary angioplasty

m

Coronary artery bypass graft

n

Total mortality

o

Zung Depression Scale

p

Cardiac mortality

q

Confidence interval

r

Symptom Checklist - 90

s

Years

t

Cardiac rehospitalization and/or frequent visits

u

Cardiac events

v

Health care consumption

w

Confidence intervals not reported as single variables

x

Heath care consumption

y

Death or recurrent myocardial infarction

z

> 6 visits at cardiac outpatient clinic during follow-up

aa

Hospital Anxiety and Depression Scale

ab

Depression as predictor of major cardiac event

ac

Depression as predictor of health care

Table 5

Summary of Studies on the Relation of Depression to (more...)

Table 5

Summary of Studies on the Relation of Depression to Quality of Life after Myocardial Infarction (Question 3)

Study Author, Year	No. Enrolled	Depression Instrument	Outcome		Statb	Dep vs. Non Depc	Multivariate Comparison
			QOLa	Other
Travella, 1994	129	Present state examination - modified,HAMDd	Johns Hopkins functioning inventory, Social functioning exam	NAe	Multivariate Rank	NRf	Social functioning exam with depression at
							baseline f=1.85 p=0.17;
							3 mosg f=7.73 p=0.1;
							6 mos f=4.38 p=0.45;
							9 mos f=13.45 p=0.001;
							12 mos f=7.94 p=09
Drory, 1998	276	BDIh	Frequency of sexual activity after MIi at 3 – 6 mos,	Depression	Pearson Correlation	Frequency of sexual activity 0.16 (p<0.01);	Frequency of sexual activity: 0.14 (p<0.01);
			Satisfaction with sexual activity after MI			Satisfaction with sexual activity 0.14 (p<0.01)	Satisfaction with sexual activity: 0.15 (p<0.01))
Irvine, 1999	703	BDI symptom checklist	Social perception and help, daily living	Sudden cardiac death	Coxj	ORk	NR
Ladwig, 1999	552	D-S	NA	Perception of angina pectoris	OR	NR	2.98 (CIl 1.50 – 5.90)
O'Rourke, 1999	70	HADSm	Illness perception questionaire	NA	Regression ANOVA	NR	NR
Soejima, 1999	134	Depression Index	RTW m	NA	Logistic regression OR	NR	RTW Extroversion: 3.72 (CI 1.33 – 10.4)
							Depressive sx o in hospital 0.15 (CI 0.02 – 0.87)
Bogg, 2000	220	HADS (II) Global Mood Scale	QOL after MI measure	NA	Regression	NR	Physical QOL male R2 =46, Physical QOL female R2 =27, Baseline anxiety R2 =54
Denollet, 2000	322	ZDS p	Global Mood Scale, Health complaint scale	NA	Regression OR	NR	Failure to quit smoking 2.3 (CI 1.2 – 4.5);
							Depressive s/x 3.3 (CI 1.9 – 5.8);
							Type D personality 2.2 (CI 1.2 – 3.8);
							EFq < 50% 2.0 (CI 1.0 – 3.9);
							Hyperlipidemia 2.0 (CI 1.1 – 3.4)
Lane, 2000	288	BDI	Dartmouth COOPr Charts, BDI	NA	Correlation score	Gender r=0.31, Partner status r=0.24,	BDI R2 0.11 p=0.0001, Partner status 0.05 p=0.002, Peel Index 0.05 p=0.001, State anxiety 0.02
						Living alone r=0.20,
						Employment status r=0.18,
						Frequency of exercise r= -0.21,
						Duration of exercise r= -0.17
						BDI r=0.37,
						State anxiety r=0.28,
						Treat anxiety r=0.32,
						Peel Index r=0.27,
						LOSs r=0.15
Mayou, 2000	344	HADS	SF-36 Scoret	NA	HR	At baseline:	NR
						51.2 (distressed),
						67.5 (non-distressed)
						p < 0.05;
						At 3 mos:
						38.7 (distressed),
						62.3 (non-distressed)
						p < 0.05;
						At 12 mos:
						47.2 (distressed),
						64.0 (non-distressed)
						p < 0.05
Lane, 2001	288	BDI	NR	NA	Regression Correlation	Gender r=0.2,	BDI 0.11,
						Partner status r=0.22,	Living alone 0.07,
						Living alone r=0.3,	Peel Index 0.07,
						Employment status r=0.18,	State anxiety 0.03
						Frequency of exercise r= -0.18,
						BDI r=0.32,
						State anxiety r=0.28,
						Treat anxiety r=0.24,
						Peel Index r=0.29,
						Killip class r=0.15,
						LOS r=0.25
Brink, 2002	134	HADS	Physical Component SF-36,	NA	Zero order correlation	NR	Physical Component: 0.47**;
			Mental Component SF-36				Mental Component: 0.66**
Drory, 2002	NR	BDI	Perceived health status	NA	Hierarchical regression	NR	Psychological well being
Drory, 2003	NR	BDI	Perceived health status	NA	Regression	NR	NR

a

Quality of life

b

Comparison statistic

c

Depressed versus non-depressed

d

Hamilton Rating Scale for Depression

e

Not applicable

f

Not reported

g

Month

h

Beck Depression Inventory

i

Myocardial infarction

j

Cox regression

k

Odds ratio

l

Hospital Anxiety and Depression Scale

m

Confidence interval

n

Return to work

o

Symptoms

p

Zung Depression Scale

q

Ejection fraction

r

Charts for Primary Care Practices

s

Length of stay

t

Medical Outcomes Study Short Form 36 Health Survey

Table 6

Summary of Studies on the Relation of Depression to (more...)

Table 6

Summary of Studies on the Relation of Depression to Biomarkers after Myocardial Infarction (Question 3a)

Study Author, Year	No. Enrolled	Depression Instrument	Outcome	Stata	Depressed vs Non-Depressed	Multivariate Comparison
			Biomarkers
Carney, 2001	Depb 380 NonDepc 424	Screening: ENRICHDd modified DSM-IVe DISHf present depressive episode BDIg severity of depression	In univariate analysis all 4 indices of 24 hour HRVh were significantly lower in patients with depression.	Linear regression	NRi	NR
			In multivariate analysis all 3 indices except 24 hour HRV were significantly lower in patients with depression.
Kuijpers, 2002	NR	DSM-IV	PF4j was significantly higher in Dep post-MI patients compared to NonDep post-MI patients, p=0.021.	Mann-Whitney U	PF4 mean rank 15.75 IU/ml vs. 9.25 IU/ml	NR
			There was a trend toward a significantly increased B-TGk level with p=0.08, inspite of use of aspirin		B-TG mean rank 15.04 IU/ml vs. 9.96 IU/ml
Lesperance, 2004	965	SCIDl	Dep patients had significantly higher sICAM 1 levels even after adjustment for confounders.These results were only slightly attenuated by adjustment for antidepressant treatment.	Linear regression	NR	sICAM
			No significant association between depression and IL6n.			0.095+/-0.044 (with Dep)
			Uncertain about the relationship of CRPo on depression as patients were on statins.			0.086+/-0.045 (with antidepressant added to the model)

a

Comparison statistic

b

Depressed

c

Non-depressed

d

Enhancing Recovery in Coronary Heart Disease Trial

e

Diagnostic & Statistical Manual of Mental Disorders

f

Depression Interview and Structured Hamilton

g

Beck Depression Inventory

h

Heart rate variability

i

Not reported

j

Platelet factor 4

k

β-thromboglobulin

l

Structured Clinical Interview for Diagnostic & Statistical Manual of Mental Disorders-IV

m

Soluble intracellular adhesion molecule 1

n

Interleukin 6

o

C-reactive protein

Characteristics of Studies

Of the 12 articles that addressed QOL, the number of subjects ranged from 62¹¹⁰ to 347.⁶⁵ One study was from the United States,⁹³ seven from Europe,²³, ⁶, ⁴⁰, ⁶³, ⁶⁵, ⁶⁷, ¹¹¹ three from the Middle East,¹⁰⁷,¹⁰⁹,¹¹⁰ and one from Asia.¹⁰⁸ The patient's sex was reported in all studies; men alone were studied in three reports, women alone in one report, and the remaining studies reported on QOL outcome in a gender-mixed population.

Quality of Studies on Depression and Survival

Quality of Studies on Depression and Cardiac Events

Quality of Studies on Depression and Quality of Life

Quality of Studies on Depression and Surrogate Markers of Cardiac Risk

Results of Studies on Depression and Survival

Four studies assessed mortality rates in post-MI depressed patients 1 year after the index event. One of these studies¹⁰² involved patients that were included in two other studies covered in this evidence report.²⁵,¹¹⁵ In addition to assessing the potential relation of depression to mortality, this study evaluated the potential influence of the person's sex on mortality in these patients. The study's major conclusion was that depression, which was a predictor of death at 3 months, remained a significant risk factor for death 1 year after MI. Gender did not appear to influence this risk, which was largely independent of other post-MI risks for death. A second study¹¹⁴ assessed depression on a continuous scale using the BDI and found that increased depression scores predicted mortality, after statistical adjustment for known cardiac risk factors. Another study,²³ which appears at odds with this conclusion, found that depression, measured using the BDI, did not predict either cardiac or all-cause mortality after MI.

Four studies evaluated patients approximately 1½–3 years after an MI.¹²,²⁴,⁵⁸,⁷⁷ Two studies involved a subgroup of patients in the ENRICHD trial who have been described elsewhere in this evidence report. The ENRICHD study⁵⁸ evaluated the influence of treatment with cognitive behavioral therapy (and, when indicated, an SSRI) on post-MI patients with depression or low perceived social support. Depression was assessed using the DIS. At 29 months of follow-up, there were no differences in survival in the treated versus the untreated patients, nor were there differences in survival between patients in the depressed and low perceived social support groups. A substudy of ENRICHD⁷⁴ compared depressed and nondepressed patients and found that depression was an independent risk factor for death after an acute MI, but the effect did not appear until nearly a year after the index event.

Two other studies reported on the 3- to 4-year survival of post-MI patients who had been evaluated for depression at the time of their index event. One study²⁴ evaluated the risk of sudden cardiac death 4 years after an MI and found that increased depressive symptoms (BDI score greater than or equal to 10) during the initial MI hospitalization were associated with a more than two-fold increase in mortality 4 years later. In contrast to these findings, another study⁷⁷ found that a BDI score of at least 10 after an MI did not predict survival 3 years after. Although the patients in these two trials were similar in age (63 years) and gender (75–82 percent) male the patient populations of the two studies differed significantly. The study of Irvine et al²⁴ was performed in post-MI patients with frequent PVCs. Depression was assessed 2 to 4 weeks after the index MI. In this study the overall two-year mortality was 9.4 percent (63/671) with a cardiac mortality of 7.5 percent 50/671 and among those with cardiac deaths 68 percent (34/50) were sudden. In contrast, in the study of Lane et al⁷⁷ depression was assessed during the index hospitalization. The overall mortality was 13 (38/288) with a cardiac mortality of 11 percent (33/288). However, of the cardiac deaths only 9 percent were sudden (3/33) and accordingly many more deaths were due to recurrent ischemic events 60 percent (20/33) or heart failure 30 percent (10/33). Authors of both of these studies²⁴,⁷⁷ hypothesized that somatic symptoms of depression and/or disease severity may be more predictive of death than is depression per se.

Three studies,⁶,²⁵,¹⁰⁵ compared mortality rates in depressed patients versus non depressed patients 5 years after an acute MI. In one study,⁶ patients were considered depressed when they scored in the upper tertile on a measure of depression. Symptoms of depression were not associated with an increased mortality risk, 5 years later. A second study,²⁵ conducted in a cohort of patients that had been previously evaluated 1 year post-MI,¹⁰² found that post-MI depression was more closely linked to mortality 5 years after MI than 1 year after MI. Another study by the same group of researchers¹⁰⁵ assessed the relation of depression, anxiety, and general health in post-MI patients to mortality 5 years after the index event. They concluded that an unidentified, but unique, aspect of depression predicts long-term cardiac mortality post-MI.

A single study¹⁰² reported on mortality in depressed post-MI patients 10 years after the index event. As with most of the studies that assessed the relationship between depression and mortality at earlier time points, this study found that death following MI was positively and independently associated with depressed mood, as measured by the ZDS.

Evidence Grade Table 2. Grading of the Quality of Evidence (more...)

Evidence Grade Table 2. Grading of the Quality of Evidence on the Independent Association of Measures of Depression with Clinical Outcomes in Patients with Acute Myocardial Infarction (Question 3)

	Survival	Cardiac Events	Quality of Life
Quantity of Evidence:
Number of studies	14	6	14
Total number of patients studied	92958	116,697	3, 381
Quality and Consistency of Evidence:
Were study designs appropriate for determining the association between depression and outcomes? (yes = high quality; no = low quality)	Yes	Yes	Yes
Did the studies have serious (-1) or very serious (-2) limitations in quality? (Enter 0 if none)	0	0	-1
Did the studies have important inconsistency? (-1)	-1	0	0
Was there some (-1) or major (-2) uncertainty about the directness or extent to which the people, measures and outcomes are similar to those of interest?	0	-1	-1
Were data imprecise or sparse? (-1)	0	0	-1
Did the studies have high probability of reporting bias? (-1)	0	0	0
Did the studies show strong evidence of association between depression and outcomes? (“strong” if significant relative risk or odds ratio > 2 based on consistent evidence from 2 or more studies with no plausible confounders (+1); “very strong” if significant relative risk or odds ratio > 5 based on direct evidence with no major threats to validity (+2))	+1	0	0
Did the studies have evidence of a dose-response gradient? (+1)	+1	0	0
Did the studies have unmeasured plausible confounders that most likely reduced the magnitude of the observed association? (+1)	0	0	0
What was median (and range) of estimated association?	Not Applicable	Not Applicable	Not Applicable
Overall quality grade (high, medium, low, very low)	High	Medium	Very low
Importance of outcome (critical, important, or not important)	Critical	Important	Important

Evidence grading scheme as described in: Grade Working Group. Grading quality of evidence and strength of recommendations. BMJ 2004; 328: 1490–7.

Results of Studies on Depression and Cardiac Events

Shiotani et al.⁹⁵ evaluated 1,042 consecutive patients for symptoms of depression with the ZDS within 3 months of acute MI. Patients with depression over age 65 had significantly higher cardiac event rates during the first year after MI than did patients over age 65 without depression. There was a trend toward an association of depression and cardiac events in younger patients, but this did not reach statistical significance (p = 0.11).

Druss et al.⁵² evaluated a group of over 100,000 Medicare patients who sustained an MI, approximately 5 percent of whom had mental illness. Of the group of patients with mental illness, some had an affective disorder (major depression, dysthymia, or bipolar disorder). The group of patients with a mental illness, as well as the subgroup of patients with an affective disorder, had a lower likelihood of receiving cardiac revascularization procedures (percutaneous transluminal coronary angioplasty, PTCA or coronary artery bypass graft surgery, CABG) during the initial MI hospitalization than did the group of patients without mental illness. In a study such as this one, cardiac revascularization procedures may be treated either as an MI treatment or as an outcome variable (i.e., a cardiac event), and therefore this study was included in this question.

Frasure-Smith et al.¹⁰⁶ studied 848 MI survivors and reported cardiac readmissions and procedures during the first year of follow-up. These authors found that depression (BDI score of 10 or higher) during the initial MI hospitalization was associated with a significantly increased number of cardiac readmissions and days of hospitalization. There was no relationship of depression during the initial MI hospitalization and cardiac procedures (cardiac catheterization, angioplasty, or coronary bypass surgery) in the first year post-MI.

Strik et al.⁸⁴ studied 315 men who completed measures of emotional distress, including depression, within the first month after the initial MI hospitalization. These authors then measured major adverse cardiac events, defined as cardiac death or recurrent MI, over an average follow-up period of 3.4 years. Although depression alone was a significant predictor of cardiac events (including cardiac death), this relation disappeared when a measure of anxiety was included in the multivariate model.

In a subsequent study with similar methodology, Strik et al.⁸⁹ found similar results with a sample of 206 men and women followed for 3 years after their first MI.

Results of Studies on Depression and Quality of Life

Studies that addressed QOL measures reported the relation of depression during the initial MI hospitalization to physical-behavioral function,⁶,²³,⁴⁰,⁶⁵,⁶⁷,⁹³,¹¹¹ psychological well-being,⁶,⁶³,⁶⁵,⁶⁷,¹⁰⁹–¹¹¹ social and work role performance,⁶,²³,⁶⁵,⁹³,¹⁰⁷,¹⁰⁸,¹¹¹ and personal perception of health⁶,²³,⁶⁵ (See Table and Evidence Table 3-4 in Appendix G). Several studies reported on multiple aspects of health-related QOL. Some studies reported the relation of depression at the time of the initial MI hospitalization to QOL measures within the first 12 months post-MI,⁶³,⁶⁷,¹⁰⁷,¹⁰⁸,¹¹¹ whereas others reported on the relation to QOL at least 12 months post-MI.⁶,¹¹⁰ Two articles reported on the same sample at 4-month follow-up⁴⁰ and at 12-month follow-up.²³ Three studies reported on both near- and long- term outcome.⁶⁵,⁹³,¹⁰⁹

One study reported that depression as assessed by the HADS during the initial MI hospitalization was prospectively associated with a reduced score on the mental component of the short-form health survey (SF-36 Mental Component Score) but not with the SF-36 Physical Component Score at 5 months post-MI.⁶⁷ Another study found that in-hospital depression for females and 1-month post-MI depression for males were prospectively related to the physical domain of QOL at 6 months as assessed by the Quality of Life After MI measure.¹¹¹ This study also reported that social aspects of QOL at 6 months were predicted by in-hospital and 1-month depression for women but not for men and that 6-month psychological QOL was not related to depression at either earlier time for either sex.¹¹¹ In-hospital depression measured with the BDI was prospectively related to sexual activity and sexual satisfaction of men 3 to 6 months post-MI in a multivariate analysis controlling for demographic (e.g., age, education) and medical (e.g., Killip class, diabetes mellitus) variables.¹⁰⁷ One study among Japanese men documented that in-hospital symptoms of depression after an MI were prospectively related to dichotomously measured return to work in multivariate analysis, but not to time until work resumption or returning to work at a reduced capacity.¹⁰⁸ Thus, during the 1-year rehabilitation period, post-MI depression has been found to be related to physical QOL, social QOL of women, sexual activity and satisfaction among men, return to work of employed men, and health care utilization.

Another study found that QOL 4 months after MI measured with Dartmouth Primary Care Cooperative Information Project (COOP) charts (combining physical, social, role functioning, general health, pain, change in health, social support, and perceived health) was related to in-hospital depression (BDI), having a partner, MI severity and state anxiety, even after controlling for sociodemographic, activity, and illness variables.⁴⁰ In a follow-up of this sample,²³ depression, living alone, MI severity (Peel Index score), and state anxiety were found to be independent predictors of 12-month QOL, even when controlling for sociodemographic, activity, and illness variables.

The three repeated-measures studies in this section had remarkable consistency across short-term and long-term follow-ups. Individuals with high in-hospital distress (i.e., above established cutoffs for either HADS-depression or HADS-anxiety scales) scored significantly lower on all eight subscales of the SF-36 (four physical and four psychosocial) at 3-month follow-up.⁶⁵ The mean HADS-depression score for the distressed group was significantly higher than for the nondistressed group. Consistent with the 3-month follow-up data, individuals with high in-hospital distress scored significantly lower on all eight subscales of the SF-36 at the 12-month follow-up.⁶⁵ DSM-III depressive disorder diagnosed in-hospital with use of the Present State Exam was associated with poorer physical functioning and greater depression severity only at discharge, and with less social connectedness only at the 6-month follow-up, whereas social functioning was significantly worse at the 3-, 6-, 9- and 12-month follow-ups.⁹³ The third repeated-measures study found that in-hospital depression measured with the BDI was significantly related to psychological distress among male survivors of MI as measured by using the Mental Health Inventory at the 3- to 6-month post-MI follow-up, with controlling for sociodemographic and medical variables.¹⁰⁹ Furthermore, psychological distress at 3 to 6 months moderated the effect of in-hospital depression on psychological distress at the 5-year follow-up among male survivors of MI, also with controlling for sociodemographic and medical variables.¹⁰⁹ Thus, at both early and late follow-up times, in-hospital depression has been shown to be related to physical, psychological, and social health and function.

The 5-year QOL outcomes of patients with acute MI were examined in two studies. In-hospital depression measured with the BDI was significantly related to psychological well-being of women at the 5-year post-MI follow-up, as measured using the Mental Health Inventory and controlling for sociodemographic and medical variables.¹¹⁰ Similarly, among female survivors of MI, in-hospital depression and self-reported concomitant medical problems were related to psychological distress 5 years later.¹¹⁰ One 5-year outcome study used cutoffs on the Health Complaint Scale's somatic complaints and disability subscales to define poor QOL. Multivariate analysis found a significant relation between symptoms of depression (ZDS) and poor QOL.⁶ Thus, even 5 years after MI, psychological distress and physical health and functioning were related to in-hospital depression independent of other important QOL determinants.

Results of Studies on Depression and Surrogate Markers of Cardiac Risk

Three studies reported data on surrogate markers of cardiac risk. (See Table Evidence Table 3-1, 3-2, 3-3, and 3-4 in Appendix G.) Carney et al. evaluated HRV in 673 patients with acute MI between October 1997 and January 2000 who were screened for participation in the ENRICHD trial.¹¹² The depressed subjects were eligible if they had a DSM-IV diagnosis of major or minor depression and a BDI score of 10 or higher within 28 days of admission for acute MI. The control group consisted of patients who were otherwise eligible for ENRICHD but did not meet the depression or social isolation criteria. Major exclusions included atrial fibrillation or flutter, the presence of an implanted electronic cardiac pacemaker, severe medical or severe psychiatric comorbidity, cognitive impairment, or substance abuse. Spectral HRV analysis was performed from ambulatory electrocardiographic monitors for the following bands: ultra low frequency (ULF, 1.15 × 10-5 Hz), very low frequency (VLF, 0.033 to 0.04 Hz), low frequency (LF, 0.04 to 0.15 Hz), and high frequency (HF, 0.15–0.40 Hz).

All log-transformed indices of HRV were significantly lower in the 307 patients with depression than in the 365 without depression. The depressed patients were significantly younger (mean of 57.1 years versus 60.9 years), more likely to be female (49.5 percent versus 32 percent), more likely to have diabetes mellitus (34.5 percent versus 22.1 percent), and more likely to smoke cigarettes (40.7 percent versus 23.5 percent). After adjustment for the baseline demographic differences, depression remained significantly associated with lower indices of all parameters of HRV except HF power. This study is consistent with greater autonomic dysregulation among post-MI patients with depression as compared with post-MI patients without depression.

Evidence Grade Table 3. Grading of the Quality of Evidence (more...)

Evidence Grade Table 3. Grading of the Quality of Evidence on the Independent Association of Measures of Depression with Surrogate Measures of Disease Severity in Patients with Acute Myocardial Infarction (Question 3a)

	Heart Rate Variability	Beta-Thromboglobulin	Platelet Factor 4	Soluble Intercellular Adhesion Molecule-1	Interleukin-6	C - Reactive Protein
Quantity of Evidence:
Number of studies	1	1	1	1	1	1
Total number of patients studied	804	24	24	481	481	481
Quality and Consistency of Evidence:
Were study designs appropriate for determining the association between depression and the surrogate measures? (yes = high quality; no = low quality)	Yes	Yes	Yes	Yes	Yes	Yes
Did the studies have serious (-1) or very serious (-2) limitations in quality? (Enter 0 if none)	0	-1	-1	-1	-1	-1
Did the studies have important inconsistency? (-1)	0	0	0	0	0	0
Was there some (-1) or major(-2) uncertainty about the directness or extent to which the people, measures and outcomes are similar to those of interest?	0	0	0	0	0	0
Were data imprecise or sparse? (-1)	0	-1	-1	0	0	0
Did the studies have high probability of reporting bias? (-1)	0	0	0	0	0	0
Did the studies show strong evidence of association between depression and the surrogate measures? (“strong” if significant relative risk or odds ratio > 2 based on consistent evidence from 2 or more studies with no plausible confounders (+1); “very strong” if significant relative risk or odds ratio > 5 based on direct evidence with no major threats to validity (+2))	0	0	0	0	0	0
Did the studies have evidence of a dose-response gradient? (+1)	0	0	0	0	0	0
Did the studies have unmeasured plausible confounders that most likely reduced the magnitude of the observed association? (+1)	0	0	0	0	0	0
What was median (and range) of estimated association?	Not Applicable	Not Applicable	Not Applicable	Not Applicable	Not Applicable	Not Applicable
Overall quality grade (high, medium, low, very low)	High	Low	Low	High	High	High
Importance of outcome (critical, important, or not important)	Important	Important	Important	Important	Important	Important

Evidence grading scheme as described in: Grade Working Group. Grading quality of evidence and strength of recommendations. BMJ 2004; 328: 1490–7.

Introduction

The preceding sections of this evidence review have focused on the risks associated with having depression after an MI. Most of this evidence is based on observational studies and statistical approaches to determine whether any association between depression and these outcomes is independent of other factors. Studies of treatment for depression can provide important information on causal inferences for depression and poor outcomes. Treatment studies also address the issue of greatest concern to patients and clinicians: Can treatment reduce any of the increased risks that may be associated with having depression? Question 4 addresses whether post-MI patients with depression have better outcomes with depression treatment than do those without such treatment. An important aspect of this question relates to determining whether it is the resolution of the depression, or the treatment itself, that reduces the risk. For example, if SSRIs improve outcomes, is it because the SSRIs have beneficial platelet effects, or is it because the SSRIs increase the rate of depression resolution? These issues are the motivation for question 4.

Results of Literature Search

Our comprehensive literature search identified 12 articles addressing this question. Eleven studies were randomized clinical trials, and one used a less rigorous method for developing comparison groups. A common reason for eliminating studies was that patients entered into the studies did not all have clinical depression, and the results for the depression subgroup were not presented separately. Such studies do not provide direct evidence regarding question 4. One meta-analysis¹¹⁶ addressed the question of whether the addition of psychosocial interventions improves the outcome of a standard rehabilitation regimen for patients with CAD. The authors used a broad definition of psychosocial interventions and identified 23 randomized clinical trials. They concluded that the addition of psychosocial treatments to standardized cardiac rehabilitation regimens reduces mortality and morbidity, psychological distress, and some biological risk factors.

Characteristics and Results of Studies

The largest studies with the highest quality data that most directly address the question are the SADHART or SADHART⁵⁹ and the ENRICHD study.⁵⁸ SADHART was conducted in 40 outpatient cardiology centers in seven countries. Participants were randomized to 24 weeks of double-blind treatment with either sertraline or placebo. Participants were recruited in the hospital by either chart review or physician referral. Inclusion criteria included being hospitalized for either an MI or unstable angina in the past 30 days. Diagnosis of MI was made according to standard criteria. Unstable angina was defined as having (1) typical ischemic symptoms lasting longer than 10 minutes, (2) being hospitalized, and (3) having changes on an electrocardiogram (ECG) within the previous 12 hours. Alternatively, one could meet criteria for unstable angina by being hospitalized with unstable angina and having known CAD. Depression was identified with the DIS completed by a trained interviewer and with the self-rated BDI. The DIS was completed within 30 days of the hospitalization. For the DIS the 2-week duration criteria and impairment criteria were eliminated for the diagnosis of major depression because of the complicating cardiac event. Exclusions included (1) uncontrolled hypertension, (2) cardiac surgery anticipated in the next 6 months, (3) MI or cardiac surgery in the previous 3 months, (4) congestive heart failure, (5) bradycardia, and (6) MI or unstable angina of a nonatherosclerotic etiology. Psychiatric exclusions included alcohol or substance abuse, psychotic illness, cognitive impairment, current use of an antidepressant, or initiation of psychotherapy in the past 3 months.

After meeting the initial eligibility criteria for the study, all participants received placebo for 14 days. During this time they completed their cardiovascular tests and a psychiatrist confirmed that their depressive symptoms were present for 2 weeks. Because many of the participants had mild depression, they were stratified by severity of depression, and it was hypothesized that those with more severe depression would have a greater benefit from treatment with sertraline. Participants were started on sertraline 50 mg daily and could be systematically increased to 200 mg daily. The only other psychotropic medications allowed were chloral hydrate and zopiclone. The primary goal for the study was to evaluate the cardiac safety of sertraline. Therefore, the primary outcome was cardiac function, specifically LVEF. Depression outcomes included the BDI and 17-item HAMD (assessed up to 16 weeks) and Clinical Global Severity (assessed up to 24 weeks). Cardiac events were adjudicated by a committee. Revascularization was not considered an endpoint for this study.

Of the 556 individuals who were initially eligible, 369 individuals remained eligible after the 2-week run-in phase and were randomized. Because of the run-in phase, the mean time to start the intervention was 24 days after MI. There were no significant baseline differences between the groups. Most patients were aged between 50 and 70 years and had least two cardiac risk factors. While most had mild to moderate levels of depression, 25 percent met criteria for more severe depression. Most participants completed treatment (mean of 149 days) and achieved a low dose of sertraline at the end (mean of 68 mg daily). Those on sertraline did report more nausea and diarrhea, common side effects of sertraline.

For depression outcomes, the results generally indicated that the sertraline group had better outcomes. The study groups did not differ on the HAMD ratings but these ratings stopped at 16 weeks. Clinical Global Inventory outcomes were assessed through 24 weeks and did differ between the study groups. For those with a history of recurrent depression, outcomes for those randomized to sertraline were better than for those receiving placebo for both the Clinical Global Severity and HAMD ratings.

About 20 percent of the participants in the study had not experienced a recent MI. Results were not presented separately for those with an MI. No Kaplan-Meier data were presented to determine when the sertraline began to affect cardiac outcomes. It also was not clear how drop outs were handled in the analysis.

A second paper from the SADHART study was also included in our review. This study focused on QOL outcomes.¹¹⁷ For this analysis more than 90 percent of all randomized patients provided data. Overall, both the sertraline and placebo groups showed improvement in QOL. According to results from the Medical Outcomes Study Short Form 36 (SF-36), the study groups did not differ in Physical Component or Mental Component change scores. However, for the recurrent depression group, there was a significantly positive change on the Mental Component score for those on sertraline compared with those receiving placebo, and there was a trend toward more improvement on the Physical Component score as well. Similar results were found using the Quality of Life Enjoyment and Satisfaction scale (Q-LES-Q).

A second major study addressed the efficacy of treatment for depression in individuals who have depression after an MI—the ENRICHD study.¹¹⁸ This multicenter trial was designed to measure the effect of a psychosocial and cognitive behavior therapy intervention. A total of 2,481 patients were recruited from eight clinical centers for the study. Eligibility requirements included meeting criteria for either depression (39 percent) or social isolation (26 percent), or both (34 percent). The depression measure was based on the Depression Interview and Structured Hamilton (DISH) instrument. The major modification with this instrument is that patients were eligible if they had depressive symptoms for 1 week, provided they had previously had an episode of major depression. DISH and BDI scores indicated most patients had moderate to minor levels of depression. The participants all had to be enrolled within 28 days of their MI. Initially the participants could not be taking antidepressants, but halfway through the study patients who remained depressed even after taking antidepressants for at least 14 days were included in the study. The criteria used to establish an MI were standard.

Patients were assigned randomly either to the intervention or routine care. Patients in both groups received written materials about cardiac risk factors based on the American Heart Association Active Partnership Program, and those in usual care had physician notification of their depression. The intervention included cognitive behavioral therapy provided first with three individual sessions and then within a group setting. Therapy could last up to 6 months. Intervention patients were referred to study psychiatrists for initiation of pharmacotherapy if they had scores higher than 24 on the HAMD or if they had less than 50 percent reduction in BDI scores after 5 weeks. Pharmacotherapy was usually sertraline and could be continued for up to 12 months.

Between 1996 and 1999, 2,481 patients were randomized. Vital status was ascertained for 93 percent of patients at 6 months and all patients were followed for at least 18 months. Ninety-two percent of those randomized to the intervention received the intervention as assigned. Median time to enrollment was 6 days post-MI, and the intervention started a median of 17 days after MI. Patients attended a median of 14 sessions. At baseline the intervention group was more likely to be on an antidepressant (9.1 percent versus 4.8 percent), and these differences were still apparent at the end of data collection (21 percent versus 14.6 percent).

Depression scores improved for both groups, but the intervention produced significant but modest differences in depression at 6 months as measured by the BDI and HAMD. At 30 months there were no differences between the intervention and usual care groups in terms of all-cause mortality, cardiovascular mortality, or nonfatal cardiac events. There was a statistically significant interaction between gender, intervention, and risk of death or recurrent nonfatal MI. Men in the intervention group had better outcomes than men in the usual care group, and women in the intervention group had worse outcomes than women in the usual care group.

Although not part of the randomized component of the study, the investigators presented an analysis of whether those on antidepressants had better cardiac outcomes. In both crude analyses and analyses adjusting for baseline variables including age, BDI score, Killip class, ejection fraction, serum creatinine, previous MI and prior diagnosis of congestive heart failure, stroke or transient ischemic attack, pulmonary disease or diabetes, the hazard rate for death or nonfatal MI was 0.63 (95 percent CI, 0.46 to 0.87) for those taking antidepressants compared with those not taking antidepressants. Antidepressant use was modeled as a time-dependent variable.

Overall, this study provided a high-quality counseling intervention that was initiated soon after an MI. It had a modest effect on depression but did not have any effect on cardiac outcomes early or up to 42 months of follow-up. Despite including individuals with only low social support, it did not seem to complicate conclusions related to depression care and outcomes. Any conclusions concerning the effects of antidepressants on cardiac outcomes are less certain. In light of the study protocol, which was to refer those only among the experimental group with high or persistent depression scores to be evaluated for antidepressants, it seems likely that those on antidepressants would have higher depression scores. This non-randomized component of the study could potentially produce selection bias, as antidepressants are related to both the exposure (psychosocial intervention) and the outcomes; thus it makes the experimental group look better than the control group, even if no true association exists. In their analysis the investigators adjusted for baseline BDI scores but not for subsequent BDI scores. It would appear that the analysis presented in the manuscript might underestimate the effect of antidepressants because it did not fully adjust for depression severity. It also would be reassuring to report if the effect of antidepressant therapy was detectable for both those in the usual care group and those in the intervention group. The analysis regarding antidepressants excluded all those who had entered the trial only for low social support. However, it was not clear who the comparison group was in this antidepressant analysis. Also, the study did not adjust for willingness to accept a recommendation to start an antidepressant, which may be a marker for overall adherence to multiple care recommendations.

Other psychosocial intervention studies. With numerous studies demonstrating that depression is a risk for poor outcomes after an MI, several investigations were conducted to determine whether psychosocial interventions might reduce the elevated risk. The minimal risks associated with psychosocial interventions make such interventions attractive for patients who usually require multiple medical interventions for their cardiac care.

Frasure-Smith and colleagues⁴⁸ completed a randomized clinical trial involving 1,376 post-MI patients in which they compared a supportive and educational home nursing intervention and usual care. Most of the participants did not have significant problems with anxiety and depression. The mean BDI score was only 8.3 at baseline with 33 percent above the usual cutoff for clinical depression. The intervention was designed to address psychological distress, not specifically depression. The investigators met their target for intervention frequency in 87 to 94 percent of the participants. Contact for most patients receiving the intervention was spread out over 6 to 7 months. Despite meeting their intervention target goals, the investigators found that mean BDI scores at 1 year had no greater reduction for those in the intervention group than for those in the usual care group. It is unlikely there would be large intervention effects for the subgroup with high levels of depression at baseline if there were no differences in mean change in BDI scores, but data on the subgroup of participants having a score above 10 on the baseline BDI would have been informative. The authors did not provide information on the percentage of patients receiving antidepressant medication or formal psychotherapy. In terms of cardiac death, noncardiac death, nonfatal reinfarctions, or hospital readmission, there were no statistical differences by intervention status. Women in the intervention group tended to do worse than those in the usual care group. For men, there were no significant differences between the intervention and usual care groups. The authors speculated that the intervention nurses did not receive specific training in psychiatric-disorder screening and psychotherapeutic techniques, and this deficiency might account for the lack of effect of the intervention. In addition, the low need for depression treatment at baseline effectively limited power to detect differences in outcome.

Taylor and colleagues¹¹⁹ tested the effect of a home-based, case-managed, multifactorial risk reduction program with 585 men and women aged 70 years or younger who were hospitalized for an acute MI in one of five San Francisco Bay Area hospitals. The intervention began in the hospital and included screening for psychological distress, monitoring with follow-up phone calls, and referring for mental health treatment when necessary. Patients were assessed at 6 months and 12 months. Of note is the unusual manner in which depressive symptoms were assessed. Depression was assessed with a single question about mood. With depression assessed by a single question, it is nearly impossible to detect any change. During follow-up, depression scores dropped significantly for both groups. There were no differences between the intervention and usual care groups. The intervention was not effective even for those with moderate-to-severe depression scores at baseline.

Johnston et al.⁴⁹ randomized by ward and recruited 100 patients from a Scottish hospital within 72 hours of admission for acute MI. Despite randomization at a group level, the study groups were balanced on important baseline characteristics. Randomization was to usual care, inpatient cardiac rehabilitation, or extended rehabilitation. The extended rehabilitation group was followed for up to 6 weeks after discharge and received a mean total of 9.5 sessions with 8.4 hours of contact. Patients and their spouses picked topics and received “nonjudgmental” counseling with a focus on helping patients achieve their objectives. Possible topics included explanation of heart attacks, risk factor modification, or emotional effects after a heart attack. Depression scores, which were measured at baseline, discharge, 2 months, 6 months, and 12 months after discharge with use of the HADS, indicated significantly lower levels of depression for both intervention groups. Unlike most studies the depression scores worsened for the usual care group after leaving the hospital. The authors suggested that their intervention had better results than others because the approach was one-to-one and not group, nurses worked from a treatment manual, and partners were included.

The study by Dracup et al.⁴⁷ compared 41 patients who participated in a structured outpatient cardiac rehabilitation program to 100 patients who did not complete cardiac rehabilitation. There was no randomization. All patients were living with a partner. Patient characteristics at baseline were similar for the two groups. Depression was measured with the Multiple Affect Adjective Checklist. Between baseline and 6 months, the cardiac rehabilitation patients had a reduction in their depression score, whereas the group without cardiac rehabilitation had no reduction. These differences were statistically significant, but the statistical adjustment for covariates was not clear.

Brown and colleagues⁴⁵ recruited patients from the cardiac rehabilitation centers of five major medical centers. All participants were required to have (1) MI or CABG within the past 4 and 4 to 24 months, (2) prognosis no worse than 3.3 (New York Heart Association Criteria), which indicates a moderately compromised cardiac status, (3) stable cardiac disease and no contraindications to physical activity, (4) onset of depression/anxiety associated with MI/CABG as assessed by the Schedule of Affective Disorders and Schizophrenia, (5) scores greater than 13 on the BDI or above 70 on the SCL-90 - Revised, and (6) spouses, friends, or relatives willing to participate in the treatment intervention. As compared with other study samples, this sample included individuals having higher levels of depression. Participants were randomized to either active treatment or attention control. Active treatment focused on principles of behavior therapy with relaxation training and some cognitive restructuring self-instruction. Treatment, which included 12 weekly sessions, was provided by three different therapists. Clinician ratings were completed by raters blinded to the intervention status of the participants. Both groups improved on depression ratings (self-report and clinician ratings), but at 12 and 15 months the behavior therapy group continued to improve whereas the attention control group began to relapse. No information on cardiovascular outcomes was collected.

Other antidepressant intervention studies. Most of the pharmacological studies have used SSRI antidepressant medications. A 1998 study compared nortriptyline (tricyclic antidepressant) with paroxetine (SSRI).¹²⁰ As part of this study, 81 individuals younger than 65 years who had ischemic heart disease and moderately severe major depression (16 or higher on the 17-item HAMD) were randomized to either paroxetine or nortriptyline. Each patient completed a 2-week placebo lead-in phase to establish the stability of the depression diagnosis and to complete their cardiovascular tests. About two thirds of the patients had a history of MI, but the MI was not recent. The primary aim of the study was to determine the effect of the two types of antidepressant medication on cardiovascular outcomes. Patients received the study medication in a double-blind fashion for 6 weeks. Both medications were equally effective in reducing depression (60 percent achieved a 50 percent reduction in depressive symptoms). However, more participants withdrew from the nortriptyline than the paroxetine group (10 versus 2). Paroxetine did not have a significant effect on heart rate, blood pressure, conduction intervals, or ventricular arrhythmias. Nortriptyline was associated with an 11 percent increase in heart rate, a significant decline in standing systolic blood pressure, and a decrease in HRV. There was no effect on cardiac conduction. For equally effective treatments, the SSRI had less cardiovascular effects than did the tricyclic antidepressant. This study was funded by Smith-Kline Beecham, the manufacturer of paroxetine.

Another small study looked at the safety and efficacy of fluoxetine in the treatment of patients with major depression after a MI.⁵⁰ For this study, depression was confirmed by a clinician-administered HAMD between month 3 and month 12 after an MI, and patients younger than 75 years were eligible. In this double-blind study, 54 patients were randomized to either fluoxetine or placebo for 9 weeks. Despite the small sample size there was nearly a statistically significant greater reduction in depression for the fluoxetine group than for the placebo group. For the majority of cardiovascular measures there were no differences between those randomized to fluoxetine or placebo. The investigators measured cardiac output using the aortic time velocity integral and found it increased in the placebo group and decreased in the fluoxetine group. The QRS interval was lower in those on fluoxetine than in the placebo group (p=.03). This study was funded by Eli Lilly, manufacturer of fluoxetine.

In 2000 McFarlane and colleagues¹²¹ reported the results of a double-blind, randomized, placebo-controlled study of sertraline in 38 post-MI patients. The total sample was small with 12 randomized to sertraline, 15 to placebo, and 11 non-depressed age-matched controls. Depression was diagnosed with a validated self-report instrument called the Inventory to Diagnose Depression (IDD). The IDD was administered before hospital discharge and again within 2 weeks of the acute infarct. Ejection fractions averaged over 50 percent in each group. Depression scores tended to decrease faster in the sertraline group than the placebo group. Standard deviation of normal R-R interval (SDNN), a measure of HRV, increased by 5 percent in the sertraline group and decreased 9 percent in the placebo group. Lower levels of SDNN have been reported to be a marker for increased cardiac mortality. Differences by study group were not evident in several other markers of HRV. This study was not supported by a pharmaceutical company but by the Heart and Stroke Foundation of Ontario, Canada.

Evidence Grade Table 4. Grading of the Quality of Evidence (more...)

Evidence Grade Table 4. Grading of the Quality of Evidence on the Efficacy of Depression Treatment for Myocardial Infarction Patients with Depression (Question 4)

	Medications	Psychosocial Interventions
Quantity of Evidence:
Number of studies	5	7
Total number of patients studied	Active group: 490	Active group: 2356
	Control group: 408	Control group: 2361
Quality and Consistency of Evidence:
Were study designs randomized trials (high quality), non-randomized controlled trials (medium quality), or observational studies (low quality)?	5 randomized controlled trials	6 randomized controlled trials, 1 prospective cohort
Did the studies have serious (-1) or very serious (-2) limitations in quality? (Enter 0 if none)	0	-1
Did the studies have important inconsistency? (-1)	No	No
Was there some (-1) or major (-2) uncertainty about the directness or extent to which the people, interventions and outcomes are similar to those of interest?	No	No
Were data imprecise or sparse? (-1)	-1	-1
Did the studies have high probability of reporting bias? (-1)	No	No
Did the studies show strong evidence of association between treatment and outcomes? (“strong” if significant relative risk or odds ratio > 2 based on consistent evidence from 2 or more studies with no plausible confounders (+1); “very strong” if significant relative risk or odds ratio > 5 based on direct evidence with no major threats to validity (+2))	No	No
Did the studies have evidence of a dose-response gradient? (+1)	No	No
Did the studies have unmeasured plausible confounders that most likely reduced the magnitude of the observed association? (+1)	No	No
What was median (and range) of estimated treatment effect?	Depression: unable to assess due to the different methods of assessing depression used, the different follow-up times and the different antidepressants evaluated.	Depression: difficult to assess due to the different methods of assessing depression used and the different follow-up times.
	Total mortality: unable to assess, as only one study reported results for total mortality (Relative Risk for death was 0.39 with 95% confidence intervals between 0.08 and 1.39 comparing sertraline to placebo).	Total Mortality (range) (follow-up from 12–18 months)
		Intervention group:4–13.6%
		Control group: 3–13.8%
Overall quality grade (high, medium, low, very low)	Medium	Low
Importance of outcome (critical, important, or not important)	Important	Important

Evidence grading scheme as described in: Grade Working Group. Grading quality of evidence and strength of recommendations. BMJ 2004; 328: 1490–7.

Quality of Studies

Introduction

Clinicians and investigators need accurate and reliable methods to screen for depression in post-MI patients. The screening assessment methods should have adequate validity, reliability, and diagnostic utility within post-MI populations. Population-specific evaluation of the performance characteristics of assessment tools is particularly important with MI survivors because somatic symptoms of depression are easily mistaken for, and often overlap with, the physical sequelae of MI. Moreover, the timing of the depression assessment may influence its reliability and accuracy given that the hospitalization itself (question 5a) may adversely affect sleep habits, appetite, and other aspects of the usual depression assessment. Adequate validity of an assessment instrument is achieved if the instrument can successfully measure depression and differentiate it from similar, but distinct constructs. Reliability refers to the degree of consistency of a measure. Diagnostic utility refers to the extent to which a measure correctly identifies individuals who meet or do not meet certain diagnostic criteria as determined by a “gold standard,” which for depression would typically be a structured interview process. Validity, reliability, and diagnostic utility are multidimensional concepts that gain strength with convergence of evidence. They cannot be established globally for a particular instrument, but rather defined in terms of specific groups or populations and the purpose for which they are used (i.e., screening versus research diagnosis).

As reviewed in the section of the report on questions 1 and 2, numerous measurement instruments have been used to assess depression in post-MI patient populations. Perhaps the two most frequently used instruments are the BDI and the HADS. Evidence was presented in the section of the report on questions 1 and 2 indicating that the BDI tends to produce higher prevalence estimates across studies than the HADS. This finding raises questions about the performance characteristics of these and other measures used with post-MI patients. This section will review evidence related to the validity, reliability, and diagnostic utility of assessment techniques that have been used to evaluate depression in survivors of MI.

Results of Literature Search

After article review, seven articles were eligible for review on question 5. One study was excluded at this point because it did not report specific cutoff criteria used for diagnosing depression. Of the six remaining studies, two also were deemed eligible for question 5a.¹¹⁹,¹²² In one of these two studies, the majority of participants (78 percent) were interviewed during hospitalization and the balance post-discharge, but within 28 days of the index MI.¹²² Three studies were eligible for question 5b as well as for question 5.⁷⁵,⁹⁰,¹²³ One study reported data relevant to both questions 5a and 5b.⁶⁶

Characteristics of Studies

A summary of key aspects of the six eligible studies is presented in Appendix G Tables 5. Included studies were published between 1988 and 2003. All of the studies were composed entirely of post-MI patients. Two studies reported data on diagnostic utility,⁷⁵,⁹⁰ two on factor/construct validity,⁶⁶,¹¹⁹ one on convergent validity,¹²² one on discriminant validity,¹²³ and two on internal consistency reliability.⁶⁶,¹¹⁹ Four studies were conducted in Europe,⁶⁶,⁹⁰,¹¹⁹,¹²³ one in Canada,⁷⁵ and one in the United States.¹²² Only one - which included over 2,000 participants, but provided limited data for this question - was a multicenter study.¹²² The other studies included between 52 and 335 participants. The mean age of participants ranged from 51 to 67 years. The ENRICHD study,¹²² which was designed to maximize diversity of participants, enrolled 44 percent women and 34 percent non-white participants. The other studies enrolled between 73 and 90 percent men. Only one other study provided data on race; it enrolled only white participants.⁹⁰ The ENRICHD study¹²² was the only eligible study to report cardiac risk factors, and none of the studies reported MI characteristics such as Killip class or LVEF.

Quality of Studies

Results of Studies

Internal consistency reliability for the HADS and the HADS Depression subscale (HADS-D) as measured by Cronbach's alpha was found to be 0.82 and 0.72, respectively, in a sample of 194 male and female patients.¹¹⁹ Another study by the same authors, which included 335 male and female patients, reported internal consistency reliability for the HADS as 0.87, 0.88, and 0.90 at 1 week in-hospital, 6 weeks, and 6 months post-MI and for the HADS-D as 0.76, 0.80, and 0.81 for the same follow-up times.⁶⁶ Nunnally has suggested a widely referenced guideline of 0.70 as a lower bound for acceptability of internal consistency reliability. These two studies also reported data from an exploratory factor analysis¹¹⁹ and confirmatory factor analyses across time⁶⁶ that supported the construct validity of the HADS-D as measuring depression distinct from anxiety.

The ENRICHD investigators¹²² reported evidence on the convergent validity of the self-report BDI and the HAMD, the items of which were embedded in an SCID. A Pearson correlation coefficient between the BDI and the HAMD was 0.64, close to the low end of the range of similar correlations that have been reported in other clinical settings (0.61 to 0.87).¹²⁴ The ENRICHD sample, however, included only individuals with symptoms of depression. This stipulation would be expected to restrict the range of scores on the BDI and HAMD and to lower the correlation coefficient compared to what might be found in a screening sample of both depressed and nondepressed individuals.

Three additional studies reported psychometric data within 3 months after hospitalization for the BDI, the HADS, the HAMD, the Symptom Checklist-90 Depression subscale (SCL-90-Dep), and the ZDS. The SCL-90 is a brief, multidimensional self-report inventory designed to screen for a wide range of symptoms of psychopathology that includes an index of depression.¹²⁵ The ZDS is a 20-item self-rating scale to screen for symptoms of depression.¹²⁶

Figure 3. Sensitivity and Specificity of Instruments (more...)

   Figure 3. Sensitivity and Specificity of Instruments used to Screen for Major and Minor Depression Following an Acute Myocardial Infarction

Figure 4. Sensitivity and Specificity of Instruments (more...)

   Figure 4. Sensitivity and Specificity of Instruments used to Screen for Major Depression Following an Acute Myocardial Infarction

Figures 3

4

One study¹²³ reported evidence on the convergent validity of the SCL-90-Dep and the ZDS for 143 patients with a Pearson correlation coefficient ranging from 0.70 to 0.77 across four time periods (1, 3, 6, and 12 months). Evidence related to discriminant validity was also reported for both the SCL-90-Dep and the ZDS as compared with a measure of “vital exhaustion” developed in the Netherlands. That these correlations ranged from 0.75 to 0.83 for the two measures across the four time periods suggests these measures do not discriminate well between depression and post-MI symptoms of fatigue. The construct validity of vital exhaustion, however, is unclear and the measurement instrument used included numerous symptoms of depression. Additionally, in this study patients with a diagnosis of clinical depression, as determined by the HAMD, at baseline were offered antidepressants and withdrawn from the study. The lack of patients with significant symptoms of depression in the study population may have contributed to the lack of differentiation reported between symptoms of depression and vital exhaustion.

Evidence Grade Table 5. Grading of the Quality of Evidence (more...)

Evidence Grade Table 5. Grading of the Quality of Evidence on the Performance Characteristics of Methods Used to Screen for Depression After Myocardial Infarction (Question 5)

	BDI	HADS	SCL-90 Dep.	Zung	HAM-D
Quantity of Evidence:
Number of studies	3	3	2	1	2
Total number of patients studied	2,739	735	349	143	2,687
Quality and Consistency of Evidence:
Were study designs appropriate for determining the performance characteristics of the screening methods? (yes = high quality; no = low quality)	Yes	Yes	Yes	No	Yes
Did the studies have serious (-1) or very serious (-2) limitations in quality? (Enter 0 if none)	-1	0	-1	-2	-1
Did the studies have important inconsistency? (-1)	0	0	0	0	0
Was there some (-1) or major (-2) uncertainty about the directness or extent to which the people, measures and outcomes are similar to those of interest?	0	0	0	0	0
Were data imprecise or sparse? (-1)	-1	0	-1	-1	-1
Did the studies have high probability of reporting bias? (-1)	0	0	0	0	0
Did the studies have unmeasured plausible confounders that most likely reduced the magnitude of the estimated validity and reliability? (+1)	+1	0	0	0	+1
What was median (and range) of estimated sensitivity, specificity and reliability?	Sensitivity: MDD* .82 (1 study)	Sensitivity: MDD .90 (1 study)	Sensitivity: MDD .96 (1 study)	Sensitivity: No data	Sensitivity: MDD .86 (1 study)
	Specificity: MDD .79 (1 study)	Specificity: MDD .84 (1 study)	Specificity: MDD .74 (1 study)	Specificity: No data	Specificity: MDD .92 (1 study)
	Reliability: No data	Reliability: .88 (.82, .90)	Reliability: No Data	Reliability: No data	Reliability: No Data
Overall quality grade (high, medium, low, very low)	Low	Medium	Low	Very Low	Low

*

Major depressive disorder

Evidence grading scheme as described in: Grade Working Group. Grading quality of evidence and strength of recommendations. BMJ 2004; 328: 1490–7.

Summary of Studies

The measures reviewed here seem to be measuring depression based on adequate sensitivities and specificities across studies as compared with the SCID-IV⁹⁰ and evidence from confirmatory and exploratory factor analyses.⁶⁶,¹¹⁹ Evidence also exists for convergent validity¹²²,¹²³ and adequate internal consistency¹²⁷ of the measures.

Given the limited number of studies and analyses performed, it is difficult to draw conclusions about performance characteristics at each time period. This is particularly true during the initial MI hospitalization. Overall, however, evidence is relatively scant regarding the adequacy of the assessment of depression in post-MI populations. None of the measures reviewed here has been normalized explicitly on post-MI populations. In this context, two interrelated concerns predominate. The first relates to construct validity and how well these measures accurately distinguish between symptoms of depression and somatic symptoms related to poor physical health. The study samples covered in this review tended to be small for these purposes. Furthermore, the psychometric methods that were used are relatively weak ones and provide limited information. Existing psychometric methods that are able to establish the equivalence of key measurement characteristics across post-MI and nonhospitalized depressed subjects would be useful in refining assessment practices for depression in MI survivors.

A second concern relates to the intended use of the instruments reviewed and the very low PPVs reported.⁹⁰ If these instruments are to be used for general screening that will be followed by a more thorough assessment of those who screen positive, then low PPVs and high NPVs are acceptable, and cutoff points to achieve this objective have been reported.⁹⁰ A research setting, however, in which relatively few of those diagnosed with major depression are actually depressed (as compared with diagnosis by SCID-IV) is problematic and reduces the power to detect relationships between depression and outcome variables. Strik and colleagues⁹⁰ reported utility statistics for prominent measures of depression that were based on cutoff scores generated by an ROC paradigm intended to maximize combined sensitivity and specificity. This resulted in cutoff points well below those used in normal settings where cutoff points are set pragmatically according to specific clinical or research objectives. A stronger analysis might have included data for commonly used cutoff points. Despite this flaw, the study pointed out a potential problem in research on the relationship between depression and post-MI outcomes. The slightly higher cutoffs normally employed would be expected to lessen but not completely alleviate this problem.

Furthermore, Strik's study provides data relevant to the discrepancy in prevalence rates generated by studies using the BDI as compared with studies using the HADS. For combined major and minor depression, the BDI tends to be highly sensitive relative to the other measures, but with lower specificity. When only major depression is considered, the BDI is the least sensitive measure and is in the middle to lower range of specificities. This suggests that compared with the HADS, SCL-90-Dep, and HAMD, the BDI tends to diagnose patients with less significant depressive symptoms at much higher rates than do the other measures, but it may be less effective in accurately diagnosing major depression. Given the expected distribution of symptoms of depression, the BDI would be expected to identify more patients as having significant symptoms of depression, but it would do a poorer job of consistently identifying those who are most depressed.

In summary, the quality of evidence varies for performance characteristics of instruments that measure symptoms of depression in post-MI populations, both in terms of numbers of patients studied and in terms of types of evidence that have been reported.

Introduction

Patients who are depressed after MI may receive different treatment than patients without depressive symptoms. Several factors could explain this difference in practice. Patients with depression may have a diminished capacity to make important decisions about their health. As a result, depressed patients may be less willing to undergo cardiovascular procedures such as catheterization, percutaneous coronary intervention (PCI), or CABG. Depression may also result in decreased adherence with prescribed medications, refusal to participate in cardiac rehabilitation, and reluctance to make lifestyle changes such as smoking cessation. In addition, physicians and other health care providers may not refer patients with depression for cardiovascular procedures or other therapies like cardiac rehabilitation even when medically appropriate. This practice, in turn, might reflect a bias against depressed patients on the part of health care providers. Alternatively, the depressed patient's perception and communication of cardiovascular symptoms may differ from those of a patient without depression in a manner that makes it less likely for certain treatments to be considered necessary. To evaluate this question we reviewed the English language literature that addressed the relation of depressed mood to use of cardiac treatment after an acute MI.

Results of Literature Search

Characteristics of Studies

Table 9

Summary of Characteristics of Studies on the Relation (more...)

Table 9

Summary of Characteristics of Studies on the Relation of Depression to Use of Treatment after Hospitalization for Myocardial Infarction (Question 6)

Study Author, Year	Study Design	Location	Study Group	No. of Subjects	Mean Age	Male (%)	HTNa (%)	DMb (%)	Smoking (%)	Lipidc (%)
Bennet, 2003	Pro cohortd	Europe	NAe	37	62	73	NRf	NR	NR	NR
Blumenthal, 1982	Pro cohort	USA	NA	35	54	94	NR	NR	NR	NR
Druss, 2000	Retro cohortg	USA	Mental disorders	5365	76	52	42	22	23	NR
			No mental disorders	108288	76	46	40	26	15	NR
Ziegelstein, 2000	Pro cohort	USA	BDIh ≥10	35	46	46	71	31	31	69
			BDI <10	169	55	59	62	30	27	61
			Major Depression/ Dysthymia	31	42	52	71	36	29	71
			No Major Depression/ Dysthymia	173	54	58	63	31	27	60
Romanelli, 2002	Pro cohort	USA	Depressed	35	75	57	71	49	NR	57
			Not depressed	118	73	55	70	33	NR	60
Bennet, 2003	Pro cohort	Europe	NA	37	62	73	NR	NR	NR	NR
Lauzon, 2003	Pro cohort	Canada	Depressed	191	60	75	34	16	40	35
			Not depressed	359	60	81	36	16	40	39
Whitmarsh, 2003	Pro cohort	Europe	NA	93	64	76	NR	NR	61	NR
Steeds, 2004	Pro cohort	Europe	BDIf ≥12	62	NR	NR	NR	NR	NR	NR
			BDI < 12	69	NR	NR	NR	NR	NR	NR
Strik, 2004	Pro cohort	Europe	Depressed	63	NR	67	NR	NR	10	35
			Not depressed	143	NR	80	NR	NR	13	29

a

Hypertension

b

Diabetes mellitus

c

Hyperlipidemia

d

Prospective cohort study

e

Not applicable

f

Not reported

g

Retrospective cohort study

h

Beck Depression Inventory

Quality of Studies

Results of Studies

Evidence Grade Table 6. Grading of the Quality of Evidence (more...)

Evidence Grade Table 6. Grading of the Quality of Evidence on Whether Use of Recommended Treatment for Patients with Acute Myocardial Infarction Differs For Those With and Without Depression (Question 6)

	Cardiac Procedures	Lifestyle Interventions	Medications	Cardiac Rehabilitation
Quantity of Evidence:
Number of studies	3	1	4	2
Total number of patients studied	109359	108	1244	128
Quality and Consistency of Evidence:
Were study designs appropriate for determining the association between depression and use of treatment? (yes = high quality; no = low quality)	Yes	Yes	Yes	Yes
Did the studies have serious (-1) or very serious (-2) limitations in quality? (Enter 0 if none)	0	0	0	0
Did the studies have important inconsistency? (-1)	-1	0	-1	0
Was there some (-1) or major (-2) uncertainty about the directness or extent to which the people, measures and outcomes are similar to those of interest?	-1	-1	-1	-1
Were data imprecise or sparse? (-1)	0	-1	0	-1
Did the studies have high probability of reporting bias? (-1)	0	0	0	0
Did the studies show strong evidence of association between depression and use of treatment? (“strong” if significant relative risk or odds ratio > 2 based on consistent evidence from 2 or more studies with no plausible confounders (+1); “very strong” if significant relative risk or odds ratio > 5 based on direct evidence with no major threats to validity (+2))	0	0	0	0
Did the studies have evidence of a dose-response gradient? (+1)	0	0	0	0
Did the studies have unmeasured plausible confounders that most likely reduced the magnitude of the observed association? (+1)	0	0	0	0
What was median (and range) of estimated association?	Not Applicable	Not Applicable	Not Applicable	Not Applicable
Overall quality grade (high, medium, low, very low)	Low	Very low	Low	Low

Evidence grading scheme as described in: Grade Working Group. Grading quality of evidence and strength of recommendations. BMJ 2004; 328: 1490–7.