Time Course of Depression and Outcome of Myocardial Infarction FREE

Depression is a common condition in patients with acute myocardial infarction (MI), affecting approximately 20% of patients during hospitalization and a similar proportion at some point during the first year after MI.^1- 2 Depressive symptoms, even in the absence of formal diagnosis of major depression, are strong independent predictors of morbidity and mortality after MI.^3- 4

Most studies examining the prognostic role of depressive symptoms in patients with MI have measured depressive symptoms at the time of the hospitalization for MI. Depressive symptoms at this stage may reflect transient “reactive” depression that may undergo spontaneous remission and may not influence outcome after MI. Alternatively, depressive symptoms after MI may reflect persistent depression. It is unknown whether patients with transient depression at the time of hospitalization for MI or those with persistent depression have an elevated risk of adverse outcomes compared with nondepressed patients. The prognostic role of new depression that develops after discharge in patients not depressed during the MI admission is also unknown. These questions have clinical implications because they may guide optimal screening and management of depression in patients after MI. Accordingly, the objective of this study was to evaluate the prognostic influence of different patterns of depressive symptoms (transient, new, or persistent) on 6-month outcomes after MI.

Participants for this study were prospectively recruited between January 2003 and June 2004 at 19 US centers as part of the Prospective Registry Evaluating outcomes after Myocardial Infarction: Events and Recovery (PREMIER) study.⁵ The PREMIER study examined processes of care and patients' health outcomes as a function of sociodemographic, clinical, and health status characteristics, with the purpose of improving MI care.⁵ Participants gave informed consent, and the research protocol received institutional review board approval at each participating center.

The methods of the PREMIER study have previously been described.⁵ Briefly, consecutive patients admitted with MI were eligible to participate in the PREMIER study if they were 18 years or older; had an elevated level of troponin or creatine phosphokinase of muscle band within 24 hours of hospital admission; and demonstrated supporting evidence of a MI (eg, prolonged ischemic signs/symptoms or electrocardiographic ST changes). Patients were excluded if they had been transferred to the recruiting institution from another facility longer than 24 hours after their original presentation, were incarcerated, refused participation, were unable to provide informed consent, did not speak English or Spanish, or lacked interview data for depression either at baseline or at 1 month.

A comprehensive medical chart abstraction was performed, and patients were interviewed at each site in person to collect information on sociodemographic, behavioral, psychosocial, and health status measures. Patients were interviewed by telephone again at 1 and 6 months after discharge by a national follow-up center.

Depressive symptoms were assessed in-hospital (baseline interview) and at 1 month using the 9-question Primary Care Evaluation of Mental Disorders Brief Patient Health Questionnaire (PHQ).^6- 7 For each of the 9 depressive symptoms, patients indicated whether symptoms had bothered them “not at all,” “several days,” “more than half of the days,” or “nearly every day” during the previous 2 weeks, yielding a score from 0 to 3. The severity index of PHQ ranges from 0 to 27. A PHQ score of 10 or higher corresponds to a level of at least moderate depression and represents the minimum number of symptoms required for the diagnosis of major depression.^8- 9 The PHQ is a questionnaire with a sensitivity of 88% and a specificity of 88% for major depression, with a standard cutoff score of 10 or higher.^6- 7

Health status, at baseline and follow-up, was assessed using the Seattle Angina Questionnaire (SAQ), a 19-item scale measuring 3 components of health status (symptom burden, functional status, and disease-specific quality of life [QOL]) over the preceding 4 weeks.^10- 11 All SAQ domains have been validated and predict death and subsequent cardiac events in patients with coronary heart disease.¹¹ Responses are scored from 0 to 100, with higher scores indicating better health (ie, less angina, less physical limitation, and better QOL).¹⁰ History of depression was defined as receiving medications or counseling for depression.

Outcome measures were (1) a combined end point of all-cause rehospitalization or mortality at 6 months and (2) health status (angina frequency, physical limitation, and QOL) at 6 months. A combined end point was used for rehospitalization and mortality owing to the low number of mortality events. However, in an exploratory fashion, these 2 end points were also examined separately. Vital status at 6 months after discharge was obtained from contacts with family members and the Social Security Death Master File. In the follow-up interview of survivors at 6 months, patients were queried whether they were readmitted to a hospital after the index MI admission.

The prevalence of at least moderate depression (PHQ score ≥10) was calculated at baseline and at 1 month after discharge. By using baseline and 1-month PHQ data, patients were classified into 4 categories of depressive symptoms: no depression (PHQ score <10 at baseline and at 1 month); new depression (PHQ score <10 at baseline and ≥10 at 1 month); transient depression (PHQ score ≥10 at baseline and <10 at 1 month); and persistent depression (PHQ score ≥10 both at baseline and at 1 month).

Bivariate analyses were conducted to compare demographic characteristics, medical history, clinical characteristics at admission, baseline health status, cardiac catheterization data, and in-hospital events among the 4 depressive symptom groups (ie, no, new, transient, and persistent depressive symptoms). We used 2-sample t tests or analyses of variance for continuous variables and χ² tests for categorical variables. We compared the unadjusted frequencies of outcomes (all-cause mortality, all-cause rehospitalization, and health status outcome) among the 4 categories of depressive symptoms.

The SAQ QOL was examined as a continuous variable, and SAQ angina frequency and physical limitation were examined as categorical variables. Because angina frequency and physical limitation scores were highly skewed, consistent with previous studies,¹² we divided the angina frequency symptom burden scores into any angina (0-99) or no angina (100) and the physical limitation scores into any physical limitation (0-99) or no physical limitation (100).¹⁰ As a sensitivity analysis, we further stratified angina frequency and physical limitation scores into 3 categories reflecting daily or weekly (0-60), monthly (61-99), or absent (100) angina and moderate to severe (0-49), mild (50-74), or minimal (75-100) physical limitation.

Multivariable proportional hazards regression models were used to evaluate the association between depressive symptom status and 6-month all-cause rehospitalization or mortality. Multivariable linear regression was used to model QOL as a continuous outcome variable. Multivariable logistic regression was used for analyses with angina and physical limitations at 6 months as dependent variables. Multivariable ordinal regression models were fitted for the categories of angina and physical limitation in the sensitivity analyses. In all multivariable analyses, patients with transient, new, or persistent depression were compared with patients without depressive symptoms, adjusting for baseline health status and other clinical variables that are known to influence outcome and/or differ significantly by depression status on a bivariate basis (P<.05). The covariates in the final model included patient demographics (age, race, and sex), medical history (congestive heart failure [CHF], diabetes, hypertension, chronic obstructive lung disease, smoking, and prior MI), severity of MI (ST-segment elevation MI [STEMI]), and left ventricular ejection fraction (LVEF) using 40% as a cutpoint.

Because the degree of depression may change with antidepressants, we also adjusted for antidepressant use at 1 month in a separate step. In additional analyses, we evaluated whether a history of depression acts as an effect modifier for the association between depressive symptoms and outcomes by testing the interaction between history of depression and depressive symptoms status in the regression models.

In each model examining an SAQ outcome (angina, physical limitation, and QOL), the corresponding baseline health status was also included as a covariable. Proportional hazards assumptions were tested using Schoenfeld residuals and were verified for all proportional hazards models. Logistic regression model fit and discrimination were assessed using the Hosmer and Lemeshow goodness-of-fit test and the model c statistic, respectively. All tests for statistical significance were 2-tailed with an α level of .05. All analyses were performed using SAS version 9.1 (SAS Institute Inc, Cary, NC).

In these analyses, patients who were too ill to be interviewed at 6 months (n = 55; 2.6%) were excluded because of missing outcome data. To evaluate potential bias resulting from this, a sensitivity analysis was performed by including these 55 patients and assigning them the “worst case” scores for each outcome. All 55, therefore, were modeled as having angina, having a lowest decile of distribution of SAQ QOL score, and being rehospitalized within 1 month of hospital discharge.

We also evaluated potential bias deriving from the exclusion of missing 6-month outcome data due to patients who were lost to follow-up or refused the 6-month interview. In the overall sample including these patients (but excluding patients who were deceased or too ill to be interview at the 6-month follow-up), we calculated a propensity score of having a missing 6-month interview using logistic regression. The variables used for propensity analysis were site of enrollment, sex, age, race, insurance, economic burden, living at home, ischemic symptoms on arrival, STEMI or non-STEMI, baseline PHQ score, renal failure, diabetes, CHF, shock, pulmonary embolism, coronary angiography, previous percutaneous coronary intervention, previous MI, history of cocaine and alcohol abuse, physical function, taking part in decision making, instructions received by patient on whom to call for an emergency if they had any questions, lipid evaluation, cardiac rehabilitation, and length of stay in-hospital. The propensity score measured the probability of having a missing 6-month interview. The reciprocal of this score was then used as a weight in the analyses, resulting in lower weight given to observations from patients with missing outcome data.

In the study period, 3953 patients met MI eligibility criteria and were approached for study and 2498 (63%) patients consented for enrollment in the PREMIER study. Patients consenting to enroll in PREMIER were younger and more likely to be male and white, but they were as likely to use antidepressants on admission and discharge compared with patients eligible for enrollment. Of these 2498 patients, 2096 (84%) were eligible for the present analysis because they had interview data for depression both at baseline and at 1 month.⁵ Of these patients, 55 were too ill to be interviewed at 6 months and were excluded from analysis. Of the remaining patients, 1873 (89%) had 6-month mortality or rehospitalization data and 1703 (81%) had 6-month health status data. Patients lost to follow-up at 6 months were more likely to be African American and to have a history of CHF, chronic obstructive lung disease, prior MI, and smoking and alcohol abuse. Patients lost to follow-up at 6 months were also more likely to have depressive symptoms at baseline compared with patients who were not lost to follow-up (29% vs 21%; P = .006).

The overall prevalence of moderate to severe depressive symptoms was 20.6% (n = 387) during the hospitalization for MI and 13.1% (n = 246) at 1-month after discharge. Considering the presence of depressive symptoms at both time points, 7.1% of patients (n = 134) had persistent depression, 6.0% (n = 112) had new depression, 13.5% (n = 253) had transient depression, and 73.5% (n = 1382) had no depression either during hospitalization or at 1 month after discharge. There appeared to be a dose-response effect of depression severity based on in-hospital and 1-month PHQ scores. As depression severity increased at baseline, the patients became more likely to be depressed at 1-month. Also, as depression severity decreased at baseline, the patients became less likely to be depressed at 1 month (Figure 1).

Patients with depressive symptoms at either assessment (persistent, new, or transient) were more likely than nondepressed patients to be younger, women, African American, unmarried, unemployed, and to have Medicaid as insurance (Table 1). They were also more likely to smoke and to have a history of hypertension, diabetes, chronic obstructive lung disease, CHF, and prior MI. In addition, patients with depressive symptoms had worse health status at baseline compared with nondepressed patients. However, there was no association between depressive symptom status and measures of disease severity such as LVEF, extent of coronary artery stenoses on coronary angiography, peak troponin, STEMI, and TIMI (Thrombolysis in Myocardial Infarction) risk scores^13- 14 for STEMI or non-STEMI. Finally, there was no significant association between depressive symptom status and hospital complications such as reinfarction, renal failure, or atrial fibrillation.

Patients with persistent, new, or transient depression were more likely to have a history of depression compared with nondepressed patients (41.4%, 18.8%, 19.5%, and 7.9%, respectively). At 1 month after discharge, patients with persistent, new, transient, or no depression were prescribed antidepressants at the rates of 25.4%, 9.8%, 14.2%, and 7.5%, respectively (P<.001) (Table 1). Antidepressants at discharge were predictive of 1 month depression scores after controlling for age, sex, race, CHF, diabetes, severity of MI, and baseline depression score. Use of these medications was associated with a 1-point lower PHQ depression score at 1 month (P = .003).

Patients in any of the categories of depressive symptoms (persistent, new, or transient) had higher rates of the combined end point of rehospitalization (n = 533) or mortality (n = 42) at 6 months compared with patients who were not depressed (Table 2). Findings from unadjusted Kaplan-Meier analyses showed significant differences in rehospitalization or mortality rates between the 4 categories of depressive symptoms (Figure 2). Patients in any of the categories of depressive symptoms also had more angina, more physical limitation, and worse QOL, often with a gradation toward worse health status going from nondepressed patients to patients with persistent depression.

After adjusting for baseline health status and other covariables, all categories of depressive symptoms remained significantly associated with each of the study outcomes (Figure 3). The adjusted hazards ratios for rehospitalization or mortality were 1.42, 1.71, and 1.34 for patients with persistent, new, or transient depression, respectively, compared with nondepressed patients (all P<.05). Corresponding adjusted odds ratios (ORs) were 2.64, 2.73, and 1.62 (all P<.01) for angina and 3.27, 2.25, and 1.69 (all P<.05) for physical limitation. All categories of depressive symptoms were also associated with progressively worse QOL, ranging on average from 6.9 to 15.8 points lower (β coefficient), going from transient to new to persistent depression, compared with that of nondepressed patients (Figure 3). In contrast, traditional prognostic indicators after MI were not associated with angina and QOL, including prior MI (OR for angina, 0.90; 95% confidence interval [CI], 0.64 to 1.24; and β coefficient for QOL, −1.63; 95% CI, −3.86 to 0.61), STEMI (OR for angina, 1.15; 95% CI, 0.88 to 1.49; and β coefficient for QOL, −1.06; 95% CI, −2.81 to 0.70), and reduced LVEF (OR for angina, 0.85; 95% CI, 0.64 to 1.24; and β coefficient for QOL, −1.66, 95% CI, −3.75 to 0.43). Of these traditional indicators, only LVEF was associated with physical limitation (OR, 1.9; 95% CI, 1.31 to 2.75). These results remained unchanged after adjusting for antidepressant use at 1 month. In addition, there was no significant interaction between history of depression and any of the categories of depressive symptoms.

In the secondary exploratory analyses, we evaluated whether the significant overall effect on the combined outcome end point of all-cause rehospitalization or all-cause mortality comes from rehospitalization or mortality. Although mortality rates were almost twice as high for persistent and new depression compared with nondepressed patients, the number of events in each category was extremely small, and overall there were no significant differences in mortality rates among the 4 depressive symptom categories (Table 2). The number of events was too small to do multivariable regression analysis for mortality separately from rehospitalization.

In secondary analyses, multivariable ordinal regression models confirmed that depressive symptoms were predictive of higher angina frequency and more physical limitation, with results consistent with those of the primary analyses. New, transient, or persistently depressed patients were significantly more likely to be in a “worse” angina or physical limitation category compared with those who were never depressed.

Similarly, in the sensitivity analysis in which the “worst case” scores were assigned to the 55 patients who were too ill to be interviewed, the models yielded comparable results to those of the primary analyses. The results of logistic regression analyses using propensity scores were also similar to those of the primary analyses, suggesting no significant bias in our results.

Depressive symptoms after MI may be transient during hospitalization and subside shortly after discharge, or they may persist for a longer time or develop only after discharge. These differences in time course of depressive symptoms may have clinical implications for patients' outcomes, yet such distinctions are seldom made. We found that 1 in 5 patients was depressed during the hospitalization for MI, of whom 65% had transient depression and 35% had persistent depression at 1 month. An additional 6% of patients had new depression at 1 month after discharge. Transient, new, and persistent depressive symptoms were all significantly associated with the combined outcome of rehospitalization or mortality as well as with patient health status, including more angina, more physical limitation, and worse QOL. The increase in risk persisted after adjustment for baseline patient characteristics, medical comorbidities such as CHF and prior MI, STEMI, and LVEF. Overall, the association of depressive symptoms with patient health status at 6 months was stronger and more consistent than traditional measures of disease severity such as prior MI, STEMI, and LVEF. These findings indicate that depressive symptoms, present either during the hospitalization for MI or at 1 month after discharge or at both times, are important, independent predictors of post-MI outcomes.

The results of our study expand on previous literature in several ways. First, it has been argued that transient depression after MI may not have prognostic importance.^1,15 In contrast, our data clearly demonstrate that even transient depression after MI is associated with worse outcomes. Second, only 1 previous study investigated the time course of depression and its association with outcome after MI.¹⁶ In that study, the level of depressive symptoms during admission for MI was more closely linked to survival at 5 years compared with the level of depressive symptoms at 1 year.¹⁶ However, 1-year assessment of depression may be too distant, as many patients at risk may have already had adverse outcomes. To our knowledge, no study has previously investigated in-hospital and 1-month depression status after MI and subsequent outcomes. Furthermore, our finding that new depression at 1 month after MI predicts adverse outcome is novel. Taken together, the results of our study suggest the importance of both in-hospital and early (ie, 1-month) evaluation for depressive symptoms in patients with MI.

A history of depression or antidepressant use did not change our study results. In some previous studies, patients with MI who had a history of depression were at significantly increased risk of mortality compared with those with first depression during admission for MI,¹⁶ while others showed contradictory results.¹⁷ Our results show that any depressed patients, irrespective of history of depression, are at higher risk of adverse outcomes after MI.

Biological mechanisms such as autonomic dysfunction, immunological dysregulation, proinflammation, and platelet aggregation as well as behavioral factors such as smoking and poorer self-care and compliance in depressed patients may be associated with poor outcomes after MI.^18- 19 In our study, even transient depression was associated with poorer outcomes. It is plausible that even if depressive symptoms decrease with time in some patients with MI, the biological or behavioral effects of depression take longer to subside. It is also possible that these patients remain in a trajectory of higher risk compared with patients who were never depressed because of the initial adverse effects of depression.

Our study has several possible limitations. We do not have data on whether eligible and enrolled patients differed by depression because we were only able to interview consenting patients. However, there may not be a substantial difference in depression status between enrolled and nonenrolled patients, since antidepressant use was similar in the 2 groups. We did not obtain a clinical diagnosis of major depression but focused on depressive symptoms. However, the PHQ has high sensitivity and specificity for major depression, and previous studies have shown that even mild depressive symptoms, without a clinical diagnosis of depression, are associated with adverse outcomes after MI.^3,20 In addition, PHQ administration is more feasible than a structured clinical interview for major depression in a large sample of patients and does not require a mental health specialist. Since this is an observational study, there is the possibility of unmeasured confounding and cause and effect cannot be proved. However, we adjusted for an extensive set of demographic, social, and clinical variables. Although persistent depression was usually a numerically stronger predictor of adverse outcomes and transient depression was a weaker predictor, we cannot resolve whether persistent and transient depression carry similar or different cardiac risks because our study was not powered to answer this question. Our study was designed to establish that patients with each of the 3 types of depression course have worse outcomes compared with those without depression. We could not adjudicate cause of death or rehospitalization and thus cannot comment on cardiovascular death or rehospitalization. In addition, since the number of deaths in our cohort was small, we cannot show a significant effect of depressive symptoms on mortality by itself as outcome and cannot determine where the significant overall effect on the combined end point comes from. Finally, patients lost to follow-up at 6 months were more likely to have depression compared with patients who were not lost to follow-up. However, there was no difference in study results when these patients were included in sensitivity analyses that used a broad range of variables to account for missing data, suggesting no significant bias in our results. Nonetheless, unmeasured variables between responders and nonresponders can confound the results.

The results of this study indicate that future efforts to improve post-MI outcomes should include screening for depression both in the hospital and at 1 month after discharge. Despite recent emphasis on depression screening and treatment,^21- 22 consistent with previous literature, depression was largely undertreated in our study.¹⁵ In observational studies, improvement in depressive symptoms is associated with lower cardiac mortality.¹⁶ Data from clinical trials, however, remain limited. The Sertraline Anti-Depressant Heart Attack Randomized Trial (SADHART) found a trend toward improved cardiovascular outcomes with treatment using sertraline.²³ However, SADHART was a small safety trial and was not powered for assessing differences in these outcomes.²³ The Enhancing Recovery in Coronary Heart Disease patients (ENRICHD) study did not demonstrate a benefit on mortality and reinfarction with cognitive behavior therapy in post-MI depressed patients.²⁴ However, in post hoc analyses, antidepressant use was associated with a significantly lower risk of death or MI.²⁵ In a secondary analysis among ENRICHD participants who were more likely to have persistent depression, there was no difference in late mortality between the intervention and usual care arm, suggesting that severity or natural history of depression (persistent vs transient) was not an explanatory factor in the null findings of this trial.²⁶ Thus, our data emphasize the need of future study on whether treatment of any depressive symptoms after MI, including transient depressive symptoms, can improve cardiovascular outcomes. In the meantime, depression remains an important illness and the results of our study suggest that when depression is found at either time point (in-hospital or at 1 month after discharge), considerations should be given to depression treatment.

In conclusion, depressive symptoms in the first month after MI, whether persistent, of new onset after discharge, or even transiently present after MI, are strongly and independently associated with a broad array of 6-month outcomes. Depressive symptoms, irrespective of their course, are more consistently associated with patient health status at 6 months compared with traditional measures of disease severity such as prior MI, STEMI, and LVEF. Our results suggest the importance of early screening and identification of depressive symptoms at the time of hospitalization for MI as well as 1 month after discharge to potentially improve post-MI outcomes.

Correspondence: Susmita Parashar, MD, MPH, MS, Division of General Medicine, Department of Medicine, Emory University School of Medicine, Faculty Office Bldg, 49 Jesse Hill Jr Dr, Atlanta, GA 30303 (smallik@emory.edu).

Accepted for Publication: June 23, 2006.

Author Contributions: Drs Parashar and Spertus had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Parashar, Rumsfeld, Spertus, Reid, Amin, Lichtman, and Vaccarino. Acquisition of data: Parashar, Rumsfeld, Spertus, Reid, Lichtman, Dawood, and Vaccarino. Analysis and interpretation of data: Parashar, Rumsfeld, Spertus, Reid, Wenger, Krumholz, Weintraub, Lichtman, and Vaccarino. Drafting of the manuscript: Parashar, Spertus, Reid, and Dawood. Critical revision of the manuscript for important intellectual content: Rumsfeld, Spertus, Reid, Wenger, Krumholz, Amin, Weintraub, Lichtman, and Vaccarino. Statistical analysis: Reid and Lichtman. Obtained funding: Spertus. Administrative, technical, and material support: Parashar, Spertus, Wenger, Amin, Dawood, and Vaccarino. Study supervision: Vaccarino.

PREMIER Registry Investigators:Mid America Heart Institute, Kansas City, Mo: John Spertus, MD, MPH, Carole Decker, RN, PhD, Phillip Jones, MS, Kimberly Reid, MS; Baptist Health System, Little Rock, Ark: Gary Collins, MD; Barnes Jewish Hospital/Washington University, Saint Louis, Mo: Richard Bach, MD; Beth Israel-Deaconess Medical Center/Harvard University, Boston, Mass: David Cohen, MD, MSc; Denver General Health System, Denver, Colo: Frederick Masoudi, MD, MSPH, Edward Havranek, MD; Denver VA Medical Center, Denver: John Rumsfeld, MD, PhD; Duke University, Durham, NC: Eric Peterson, MD, MPH; Emory University, Atlanta, Ga: Susmita Parashar, MD, MPH, MS, Viola Vaccarino, MD, PhD, William S. Weintraub, MD; Henry Ford Medical Center, Detroit, Mich: Sanjaya Khanal, MD, Jane Jie Cao, MD, MPH; Kaiser Permanente, Denver: David Magid, MD, MPH; MeritCare, Fargo, ND: Wallace Radtke, MD, Mohamed Rahman, MD; Sentara Health System (both Sentara and Sentara Lee Hospitals), Norfolk, Va: John E. Brush, Jr, MD; Stanford University/Palo Alto VA Medical Center, Palo Alto, Calif: Paul Heidenreich, MD; Swedish Medical Center, Seattle, Wash: Timothy Dewhurst, MD; Truman Medical Center and the University of Missouri–Kansas City: Annette Quick, MD; University of Alabama, Birmingham: John Canto, MD, Vijay Misra, MD; University of Colorado Health System, Denver: John Messenger, MD; and Yale University, New Haven, Conn: Harlan Krumholz, MD, SM.

Financial Disclosure: None reported.

Funding/Support: This study was supported by grant MO1-RR00039 from the Emory University General Clinical Research Center and a grant from CV Therapeutics. Dr Parashar is supported by grants K12RR17643 and K23 RR023171 from the National Center for Research Resources, a component of the National Institutes of Health, and by an American Heart Association Scientist Development Award (0630084N); Dr Vaccarino is supported by grants K24 HL077506, R01 HL68630, and R01 AG026255 from the National Institutes of Health; Dr Rumsfeld is supported by a VA Advanced Health Services Research Career Development Award (VA ARCD 98341-2); and Dr Lichtman is supported by a Centers for Disease Control and Prevention Career Development Award (K01-DP000085-01).

Role of the Sponsor: The grants and sponsors had no role in the design and conduct of the study; in the collection, analysis, and interpretation of the data; in the preparation of the data; or in the preparation, review, or approval of the manuscript.