Prognostic Value of Troponin and Creatine Kinase Muscle and Brain Isoenzyme Measurement after Noncardiac Surgery: A Systematic Review and Meta-analysis

• Increased troponin or creatinine kinase, muscle and brain isoenzyme (CK-MB) after surgery may independently predict a patient's intermediate or long-term risk of death or a major cardiovascular event

• In this systematic review, increased troponin and, to a lesser extent, CK-MB, after surgery independently predicted mortality, particularly within the first year

• These findings might be applied clinically to identify and manage patients with high risk for postoperative cardiac mortality

NONCARDIAC surgical interventions offer the ability to cure disease and improve patients' quality of life. The number of patients undergoing noncardiac surgery is growing. Worldwide estimates suggest 200 million adults annually undergo major noncardiac surgery.1,2Despite the procedural benefits, several million of these patients suffer a major cardiovascular complication (i.e ., cardiovascular death, nonfatal myocardial infarction, or nonfatal cardiac arrest) during the perioperative period (≤ 30 days after surgery),3and many more patients die or suffer a major cardiovascular event in the subsequent 1–2 yr after surgery.4,5 

Recent studies suggest that a troponin or creatine kinase muscle and brain isoenzyme (CK-MB) measurement after surgery may independently predict a patient's intermediate- (≤ 12 months) or long-term (more than 12 months) risk of death or a major cardiovascular event.6,7Based on these findings, some investigators have advocated monitoring perioperative troponin measurements in patients undergoing noncardiac surgery to identify patients at risk.7,8A recommendation to monitor one of these measurements after surgery requires an accurate understanding of the prognostic value of a perioperative cardiac enzyme or biomarker through a systematic, unbiased, comprehensive assessment of the evidence. We therefore conducted a systematic review and meta-analysis to address the following question: In patients undergoing noncardiac surgery, is a troponin or CK-MB measurement after surgery an independent predictor of death or a major cardiovascular event in the years after surgery?

We included all noncardiac surgery studies that fulfilled the following criteria: Patients had at least one troponin or CK-MB measurement after surgery; the study had one or more patients who suffered a major cardiovascular event or died more than 30 days after surgery; and the study assessed the prognostic value of postoperative troponin or CK-MB measurement through multivariable analysis or provided us with the data to undertake the multivariable analysis. There were no language or publication restrictions. We excluded studies that did not evaluate adults or evaluated patients having cardiac surgery.

Strategies to identify studies included the following: searching six bibliographic databases ( appendix 1reports databases and search terms); searching our own files; consulting with experts; reviewing reference lists from articles fulfilling our eligibility criteria; searching PubMed using the “related articles” feature for studies fulfilling our eligibility criteria; and searching Web of Science for cited references of key publications.

Teams of two screeners independently screened the title and abstract of each citation identified in our search. These screeners selected any citation that they suspected had any possibility of fulfilling our eligibility criteria to undergo full review. If either of the two screeners identified a citation as potentially relevant, we obtained the full text article for detailed review.

Teams of two reviewers independently determined the eligibility of all studies that underwent full text evaluation. Disagreements were resolved through discussion between the two reviewers; when this did not resolve differences, a third reviewer made a final decision on the study's eligibility.

We abstracted the following data from all eligible studies: study period, type of surgery, number of patients, length of follow-up, measurement assessed (i.e ., troponin I or T or CK-MB), assay manufacturer, measurement threshold, frequency of measurements, proportion of patients with increased cardiac biomarker or enzyme, and number of deaths or major cardiovascular events. We abstracted the following study quality criteria: completeness and method of patient follow-up (e.g ., direct patient contact), blinding of outcome adjudicators to cardiac biomarker or enzyme results, and factors adjusted for in multivariable analysis.

Two individuals independently abstracted data from all studies that fulfilled our eligibility criteria, and we resolved disagreements using the same consensus process discussed above. We contacted the authors of all eligible studies to obtain missing data or confirm the accuracy of the abstracted data.

A κ statistic was calculated to quantify chance-corrected interobserver agreement for study eligibility decisions. We determined raw agreement for abstracted variables.

Ten of the studies included in our meta-analyses did not directly provide an adjusted odds ratio (OR).6,7,9–16Two of these studies did not report a multivariable analysis, but the authors provided us with the individual patient data and we conducted our own multivariable analyses to obtain adjusted ORs.9,10In these multivariable logistic regression analyses, we included coronary artery disease as the adjustment variable.

Eight of these 10 studies performed time-to-event analyses and reported hazard ratios (HRs) from Cox proportional hazards models.6,7,11–16One of these studies also performed a logistic regression analysis, but did not report the numerical results.16The authors of this study provided us with the ORs from this analysis. In the remaining seven studies that reported a HR, we computed the relative risk (RR) at 1 yr, using the following formula:

where HR is the hazard ratio at 1 yr and P⁰is the proportion of patients without an abnormal troponin level who had died within 1 yr of surgery. When contacted, the authors provided us with crude mortality rates at 1 yr. We used this to estimate P⁰. Once we obtained the RR, we were able to calculate the OR at 1 yr, based on the following formula by Zhang and Yu17:

One study did report a multivariable OR for troponin, but did not report one for CK-MB.18Upon contacting the authors, they provided us with the CK-MB multivariable OR.

We pooled ORs using the DerSimonian and Laird random-effects model.19An I²value was calculated to assess heterogeneity across study results. An I²value more than 25% was considered to represent substantial heterogeneity.20Our a priori  hypotheses to explain substantial heterogeneity across study results were: stronger prediction in intermediate- versus  long-term follow-up, stronger prediction in troponin T versus  I assay, stronger prediction in early troponin measurement versus  late troponin measurement, stronger association in vascular versus  other types of surgery, stronger association in retrospective versus  prospective studies, stronger association in studies without versus  with blinding of outcome adjudicators, and stronger association without versus  with 100% completeness of follow-up. We tested for interactions by performing a z-test on the natural logarithms of the ORs across subgroups in an attempt to explain heterogeneity. The test of interaction was statistically significant for two of our a priori  hypotheses, and we therefore performed a metaregression analysis where we included both a priori  hypotheses in one model to simultaneously test these two possible sources of heterogeneity. Analyses were performed using S-PLUS 8.0 (TIBCO Software, Inc., Palo Alto, CA) and Stata 8.2 (StataCorp LP, College Station, TX).

Our search strategy identified 1,808 citations. After an initial screening of titles and abstracts, 1,714 studies were eliminated. Of 94 studies selected for full text evaluation, 15 studies fulfilled our eligibility criteria and are included in this systematic review (fig. 1).6–16,18,21–23 

Interobserver agreement for study eligibility was excellent (κ, 0.86). Raw agreement across individual abstracted variables ranged from 75–100%.

Table 1presents the characteristics of the 15 included studies that enrolled 4,040 patients (minimum and maximum sample size 88 and 722 patients, respectively). The majority of studies were prospective cohort studies or clinical trials. The duration of follow-up varied from 3 to 48 months. Twelve studies included patients undergoing vascular surgery,6–13,15,16,18,21seven studies included patients undergoing orthopedic surgery,7,10,14,18,21–23four studies included patients undergoing general or abdominal surgery,7,18,21,23and three studies included patients undergoing gynecologic and urologic surgery.7,21,23 

Table 2presents information related to the troponin measurements, and table 3presents information related to the CK-MB measurements. All studies evaluated the prognostic properties of troponin measurement after surgery (14 evaluated mortality,6–16,18,22,23and five evaluated major cardiovascular events12,14,16,21,22), whereas only four studies evaluated the prognostic properties of CK-MB measurement.6,10,11,18Ten studies evaluated troponin I exclusively,8,9,12–16,18,22,23three evaluated troponin T exclusively,7,11,21and two studies evaluated troponin I and T.6,10There was wide variation across studies in the threshold for an increased troponin and the timing and frequency of troponin measurements. An increased troponin measurement was observed in 8.4–52.9% of patients across studies, and an increased CK-MB was observed in 7.6–23.7% of patients across studies.

Table 4reports study quality. Overall completeness of follow-up was high, and 12 studies achieved complete follow-up.6,7,9–16,18,22Eight studies blinded outcome adjudicators to the cardiac biomarker and enzyme values.6–9,12,15,16,21There was substantial variation across studies regarding which variables were adjusted for in the multivariable analyses.

Figure 2reports the meta-analysis of the 14 studies for which we were able to obtain an adjusted OR for a postoperative increased troponin measurement to predict all-cause mortality. The 14 studies enrolled a total of 3,318 patients, among whom 459 died during follow-up. Ten of the 14 studies had statistically significant adjusted ORs, suggesting an association between increased troponin and mortality. The remaining four studies all showed nonsignificant ORs greater than 1.0. The meta-analysis of the 14 studies demonstrated that an increased troponin measurement after surgery was an independent predictor of mortality (OR 3.4, 95% CI, 2.2–5.2), but there was substantial heterogeneity in these results (I²= 56%).

Table 5reports the adjusted ORs for the studies that had at least one patient who suffered a major cardiovascular complication more than 30 days after surgery and reported the results of an adjusted analysis determining the prognostic capabilities of a postoperative increased troponin measurement to predict a major cardiovascular event. The cardiac composite outcome varied across studies, but all five studies demonstrated that an increased troponin measurement after surgery was an independent predictor of a major cardiovascular event. Four of the studies included myocardial infarction in their composite outcome,12,14,21,22and three of these studies12,14,22used the European Society of Cardiology/American College of Cardiology definition of myocardial infarction.24 

Figure 3presents the meta-analysis of the four studies for which we were able to obtain an adjusted OR for a postoperative increased CK-MB measurement to predict all-cause mortality. The four studies enrolled a total of 1,165 patients, among whom 202 died during follow-up. The pooled meta-analysis of the four studies demonstrated that an increased CK-MB measurement after surgery was an independent predictor of mortality (OR 2.5, 95% CI, 1.5–4.0), and there was minimal heterogeneity observed (I²= 4%).

Two of our a priori  hypotheses (type of surgery and length of follow-up) to explain heterogeneity demonstrated a statistically significant interaction (P = 0.001 and P < 0.001, respectively). We then undertook a metaregression analysis that included both of these subgroups, and length of follow-up stayed significant (P = 0.01), but type of surgery did not (P = 0.72). Figure 4reports the adjusted OR for an increased postoperative troponin measurement to predict all-cause mortality, based on duration of follow-up. The 10 studies with a duration of follow-up ≤12 months demonstrated a pooled OR of 6.7 (95% CI, 4.1–10.9; I²= 0%), whereas the four studies with a duration of follow-up more than 12 months demonstrated a pooled OR of 1.8 (95% CI, 1.4–2.3; I²= 0%).

Our meta-analysis indicates that an increased postoperative troponin measurement is an independent predictor of mortality, particularly during the first year after surgery and therefore may help physicians to risk stratify their patients after noncardiac surgery. Although fewer studies evaluated the prognostic capabilities of an increased postoperative CK-MB measurement, this also appears to provide independent prognostic information.

Strengths of our systematic review include: a comprehensive search; conducting eligibility decisions and data abstraction in duplicate with very good agreement; obtaining data from several authors to allow inclusion of their study in our meta-analysis; and we followed reporting standards for systematic reviews.25 

Our systematic review has several limitations. All but two studies6,10were at substantial risk of developing unreliable models, as they included an excess of variables in their models relative to the number of events in each study. Simulation studies demonstrate that logistic models require 12–15 events per predictor to produce stable estimates.26,27Most studies were small and had few events; as a result, many of the established intermediate- and long-term predictors of death were not included in the models. The studies used various types, manufacturers, and generations of troponin assays and employed various thresholds to label a value as increased. We evaluated all-cause mortality as opposed to cardiovascular mortality. If an increased troponin value after surgery does not predict noncardiovascular mortality, our results would represent an underestimation of the association between an increased troponin measurement and cardiovascular mortality. It is, however, frequently complicated to accurately discern the cause of death after surgery, and we were limited to what the eligible studies evaluated (i.e ., all-cause mortality).

We are unaware of any other systematic reviews that have evaluated the prognostic relevance of a troponin or CK-MB measurement after noncardiac surgery. A prior study that only followed patients to 30 days after surgery suggested an increased troponin I was an independent predictor of mortality within 30 days.28Our systematic review is an extension of this research, for we demonstrate an association between an increased postoperative troponin and mortality during intermediate- and long-term follow-up.

Our meta-analysis demonstrated that increased postoperative troponin is an independent predictor of mortality, but there was substantial heterogeneity across study results. Factors supporting the apparent subgroup effect suggesting a stronger association for troponin with death in the first year versus  thereafter include the fact that it was one of a relatively small number of a priori  hypothesis; we specified the direction of the effect a priori , the difference in ORs was relatively large and consistent, was biologically plausible, and the interaction was statistically significant in both univariable analysis and metaregression. It was, however, based on between- rather than within-study findings. Our findings suggest increased postoperative troponin is a strong predictor of mortality—the point estimate of effect suggests it increases the risk more than 6-fold—during the first year after surgery.

Given that the majority of patients who will develop a troponin elevation and myocardial infarction after surgery will not experience ischemic symptoms,29–31this supports the monitoring of troponin after surgery to enhance risk stratification. Detection of increased troponin after surgery may also provide an opportunity to change a patient's risk of death. At present, there are no randomized, controlled trials evaluating interventions among patients who have suffered an increased postoperative troponin measurement. Therefore, it remains unproven that detecting an increased troponin after surgery can improve patient outcomes, but there is a rationale to suspect that it can enhance patient outcomes.

Most patients with an increased postoperative troponin measurement have suffered a perioperative myocardial infarction.30Between 10% and 20% of patients suffering a perioperative myocardial infarction will die before hospital discharge,30,32,33and it is logical to assume that early detection of a myocardial infarction will afford physicians the greatest opportunity to prevent death, as is the case in the nonoperative setting. Strategies for managing a patient with an increased postoperative troponin that are more likely to benefit than harm patients include: (1) more frequent monitoring of vital signs to allow early detection and reversal of cardiovascular instability; (2) management in a telemetry-monitored unit or cardiac care unit to allow early detection and treatment of serious arrythmias34,35; (3) vigilant screening and correction of potential contributing factors (e.g ., hypoxia, anemia); and (4) optimal intravascular volume management to minimize the risk of heart failure.

We believe that it is also reasonable to suspect that even patients who would survive to hospital discharge, despite having suffered an undetected perioperative troponin elevation, can benefit from detection of their increased troponin. Given that the majority of these patients likely have some degree of underlying coronary artery stenosis,36–38it seems prudent to consider offering these patients management with known beneficial secondary prophylactic cardiac interventions (e.g ., aspirin, angiotensin I–converting enzyme inhibitor, β-blocker, and statin). Studies suggest that a minority of patients who will develop a positive troponin after surgery are taking aspirin,7,8,11,14,21,22an angiotensin I–converting enzyme inhibitor,7,11,14,23a β-blocker,7,8,14,21,22or a statin,11,16,22,23highlighting that there is substantial potential for improved medical management of these patients.

In considering whether it is more appropriate to monitor CK-MB or troponin measurements after surgery, several points are relevant. First, surgical trauma can result in the release of CK-MB from skeletal muscle and a false-positive CK-MB value for myocardial infarction.39–41Second, a substantial proportion of perioperative myocardial infarctions occur in the first 2 days after surgery when serum CK values are high secondary to surgical trauma. The high CK values can result in a low, and thus false-negative, ratio of CK-MB to total CK.41,42Third, troponin is more sensitive and specific than CK-MB for myocardial infarction in patients undergoing noncardiac surgery.41Considering the results of our meta-analyses, these points suggest that physicians should monitor troponin values after surgery, as opposed to CK-MB values.

Although our research demonstrates that an increased troponin value after surgery is a strong independent predictor of mortality at 1 yr, many important questions remain.43Although most studies have used, and guidelines have recommended, the 99th percentile for a healthy population to represent an increased troponin level, there is a need for large studies to determine what level of troponin should be considered abnormal after surgery and whether there is one or multiple troponin T thresholds that substantially influence risk. Further, there is a need for research to evaluate the new troponin high-sensitive assays that will replace the current fourth-generation troponin assays. We are currently conducting a prospective cohort study to address these questions and have enrolled a representative sample of over 20,000 adults undergoing noncardiac surgery in countries throughout the world, in the Vascular events In noncardiac Surgery patIents cOhort evaluatioN (VISION) Study.∥∥∥∥

The results of this systematic review suggest that an increased troponin measurement after surgery is an independent predictor of mortality, particularly within the first year. Further, our data suggest that the burden of mortality is higher within the first year after surgery for patients with a postoperative troponin rise. Physicians may find these data helpful in risk stratifying those patients who should undergo more intensive monitoring and management after noncardiac surgery.