Major Adverse Cardiovascular Events After Colorectal Cancer Surgery, Oncological Outcomes, and Long-term Mortality: A Nationwide Retrospective Propensity Score-Matched Cohort Study

Abstract

Objective:

To investigate the occurrence of major adverse cardiovascular events (MACE) following colorectal cancer (CRC) surgery and its association with long-term mortality and oncological outcomes.

Background:

Cardiovascular complications after noncardiac surgery are a leading cause of perioperative mortality. However, limited knowledge exists on how these complications impact on long-term mortality.

Methods:

This retrospective cohort study used data from 4 nationwide Danish health registries and included all patients undergoing elective surgery with curative intent for CRC between 2001 and 2019. Patients experiencing MACE, defined as acute myocardial infarction, stroke, new-onset heart failure, or nonfatal cardiac arrest, within 30 days of surgery were matched with those who did not using 1:1 propensity score matching (PSM). The outcomes were all-cause mortality within 1, 3, or 5 years of surgery, as well as 5-year cancer recurrence and disease-free survival.

Results:

Out of 39,747 patients, 900 (2.3%) had MACE. PSM resulted in 809 pairs of matched patients. Within 1 year of surgery, 110 (13.6%) patients with MACE and 2063 (5.4%) without MACE died (PSM-adjusted hazard ratio [HR] = 1.36; 95% confidence interval [CI] = 1.02–1.83). Within 3 years, 248 (30.6%) patients with MACE and 6268 (16.5%) without MACE died (PSM-adjusted HR = 1.32; 95% CI = 1.07–1.62). Within 5 years, 333 (41.1%) patients with MACE and 9232 (24.3%) without MACE died (PSM-adjusted HR = 1.25; 95% CI = 1.04–1.50). For recurrence and disease-free survival, no statistically significant differences were observed.

Conclusions:

MACE within 30 days of CRC surgery is associated with higher overall long-term mortality. Investigating causality and preventive measures is urgent in this group.

INTRODUCTION

Globally, colorectal cancer (CRC) is the third most frequently diagnosed cancer.¹ Surgery alone or in combination with radio- and/or chemotherapy is currently the only curative treatment. Despite the advances in the use of laparoscopic surgery, postoperative mortality and morbidity, as well as postoperative complications, still occur in up to 45% of the cases.² Surgery induces a stress response that causes systemic inflammatory reactions, which increases the risk of complications, including cardiovascular events.^3–5 Major cardiovascular complications after noncardiac surgery are the leading cause of perioperative mortality, as they account for one-third of deaths.⁶ Cardiovascular complications are an increasing challenge due to the growing number of elderly comorbid patients undergoing various surgical procedures due to increased life expectancy and improved surgical techniques such as minimally invasive surgery.^7–9 Myocardial injury and myocardial infarction are the most common cardiovascular complications after noncardiac surgery.^10–12 Studies examining myocardial injury after noncardiac surgery (MINS) have shown both increased short- and long-term mortality^12–14 and cardiac and noncardiac complications^10,14–16 in unselected groups of patients undergoing various surgical, mostly benign, procedures. Similarly, an association has been observed between MINS and short-term survival after major colorectal surgery.¹⁷ However, studies examining CRC surgery have not been able to find similar associations between MINS and long-term mortality or oncological outcomes, only with postoperative complications.^18,19 In this study, we aimed to investigate if the occurrence of major adverse cardiovascular events (MACE) after CRC surgery was associated with all-cause long-term mortality and oncological outcomes in a Danish nationwide cohort.

METHODS

We performed a propensity score-matched nationwide observational register-based retrospective cohort study on prospectively collected data from Danish health registries. The cohort included Danish residents undergoing a nonlocal resection of a CRC between 2001 and 2019. The study was approved by Data and Development support in Region Zealand (file no. REG-047-2020 and REG-102-2020) and reported according to the Strengthening the Reporting of Observational Studies in Epidemiology statement.²⁰ According to Danish Legislation, it was not a requirement to gain ethical approval before initiation.²¹

Data Sources

We utilized data from 4 Danish health registries to conduct this study. Data from the Danish Colorectal Cancer Group (DCCG) database²² were enriched with data from the Danish National Patient Registry (DNPR),²³ the Register of Laboratory Results for Research (RLRR),²⁴ and the Register of Pharmaceutical Sales (RPS).²⁵ In the DCCG database, patients who underwent surgery for CRC were identified.²² The DCCG database has existed since 2001 and collected data prospectively on all adult patients diagnosed with a primary colorectal malignant lesion. Registration in the database is considered mandatory, and data granularity is very high. The database has more than 95% completeness.²⁶ DNPR contains trajectory data on all patients admitted to a Danish hospital or visiting an outpatient clinic in the secondary healthcare sector. DNPR provides data on diseases, examinations, certain in-hospital medical treatments, and surgical procedures.²³ RLRR holds nationwide biomarker data on the Danish population,²⁴ and RPS has information on dispensed prescriptions in the Danish population.²⁵ DNPR was used to identify the exposure variables and the outcome variable. DCCG, DNPR, RLRR, and RPS were used to identify additional confounding variables. The data in DNPR, RLRR, and RPS is not as granular as in DCCG, but data from these databases provides an overview of the patient’s contact with the secondary health care sector and thus a better understanding of the patient's phenotype and trajectory, including comorbidities and long-term postoperative events.²³ All 4 databases were transformed into a single Observational Medical Outcomes Partnership Common Data Model (OMOP-CDM) using the analytical tools created by the Observational Health Data Sciences and Informatics community.^27,28

Patient Population

The total population in the OMOP-CDM was 76.849; however, in this study, we only included patients above the age of 18 undergoing elective curative-intended nonlocal resections for colorectal malignancy. Furthermore, patients had to survive the initial 30 days after surgery and have at least one day at risk.

Exposure and Outcomes

Our exposure was the occurrence of MACE zero to 30 days after the curative-intended elective colorectal surgery. MACE was defined as a composite clinical outcome of myocardial infarction, new-onset heart failure, stroke, or nonfatal cardiac arrest. MACE was defined as the first event if a patient had more than 1 event in the period of interest. Thus, the target cohort consisted of patients who experienced MACE in the first postoperative month, whereas the comparator cohort consisted of patients who did not experience MACE in the first postoperative month. The primary outcomes were all-cause mortality within 1, 3, and 5 years of surgery. The secondary outcomes were 5-year recurrence and disease-free survival (DFS), in which the time at risk was defined as starting at 180 days after the surgery and ending at either recurrence, death, or end of the follow-up period. Recurrence and DFS were determined using the validated Lash algorithm.²⁹ For the secondary outcomes, the cohorts also excluded patients with metastatic disease. No sample size calculations were done.

Statistical Analyses

We used a propensity score matching (PSM) method to generate similar cohorts with or without the exposure. We used a 1:1 ratio, matching the nearest neighbors with a prespecified maximum caliper width of 0.2 of the standard deviation of the logit of the propensity score.³⁰ Propensity scores were computed using all patient characteristics from all time before surgery to up to 30 days after surgery. The actual PSM was data-driven and executed on 76,263 covariates. These included all demographics, conditions (eg, diagnoses such as diabetes or hypertension), measurements (eg, blood samples such as hemoglobin or creatinine), and procedures (eg, scans, x-rays, surgeries, etc.). The unique characteristics were fitted using large-scale regularized regression using the Cyclops R package. Regularization was used when fitting the propensity model, due to the large number of covariates. We investigated whether the propensity score model has been adequately specified by visual inspection of preference and distribution plots (ie, distribution of all covariates should be similar in the matched groups) and by calculating the standardized difference in means (SMD). We used a priori-specified SMD threshold of 0.1 for all covariates as previously recommended.^31,32

A Cox regression model was fitted to investigate mortality at all three-time points, DFS, and recurrence. Time at risk was defined as time from postoperative day 31 to death or end of follow-up at postoperatively years 1, 3, and 5. Mortality was visualized using Kaplan–Meier plots. Results were presented as hazard ratios (HR) with 95% confidence intervals (CIs). The proportional hazards assumption was evaluated to ascertain the constancy of HRs over time with log-log survival plots and examination of Schoenfeld residuals. All reported P values are two-sided, and the P value statistical significance was set at below 0.05. Cohorts were defined using the open-source tool ATLAS provided by Observational Health Data Sciences and Informatics and analyses were run in R-studio (version 4.2.0) using the PopulationLevelEstimation package.

RESULTS

A total of 39,747 patients were included in this study, of which 900 (2.3%) patients experienced MACE within 30 days of CRC surgery. We excluded 962 (2.4%) patients, as they did not survive before entering the cohort on postoperative day 30 or had less than 1 day at risk. Of patients with MACE, 89 (9.9%) did not survive the initial 30 days after surgery, whereas 847 (2.2%) patients without MACE died within the first 30 days—a difference of 7.7% point between the 2 groups. In total 38,785 patients were available in the cohort for PSM. After 1:1 matching, 809 pairs of patients were included (Fig. 1).

FIGURE 1.

Flowchart of patient inclusion before and after propensity score matching.

Before PSM, the patients experiencing MACE were older and more comorbid. After the matching, the distribution of propensity scores was similar in the groups and the matched cohort had balanced characteristics for both patients with and without MACE. Table 1 presents the distribution of 8 clinically relevant variables before and after matching. Figure 2 shows the density plots for the propensity scores before and after matching and Figure 3 illustrates the SMD before and after matching. Figure 3 indicates that 1.22% of covariates (928 out of 76,263) exhibited SMD values below −0.1 or above 0.1 after matching.

TABLE 1.

Cohort Overview

	Before Propensity Score Matching					After Propensity Score Matching
Characteristics	MACE		No MACE		SMD	MACE		No MACE		SMD
N	900		38.847			809		809
Age, years
≤49	<6	0%	1.482	4%	−0.26	<6	0%	9	1%	−0.11
50–74	398	44%	22.928	59%	−0.30	373	46%	355	44%	−0.04
≥75	500	56%	14.437	37%	0.38	434	54%	445	55%	0.03
Sex
Female	329	37%	17.932	46%	−0.20	294	36%	292	36%	0.01
Male	571	63%	20.915	54%	0.20	515	64%	517	64%	−0.01
BMI
≤18.5	21	2%	999	3%	−0.02	17	2%	17	2%	0.00
>18.5–≤25.0	318	35%	14.975	39%	−0.07	289	36%	279	34%	0.03
>25.0–≤30.0	291	32%	12.874	33%	−0.02	270	33%	282	35%	−0.03
>30.0–≤35.0	104	12%	4.445	11%	0.00	97	12%	93	11%	0.02
>35.0	37	4%	1.540	4%	0.01	31	4%	40	5%	−0.05
Missing	129	14%	4.014	10%	N/A	105	13%	98	12%	N/A
Charlson comorbidity index
0	298	33%	24.732	64%	−0.64	265	33%	241	30%	0.06
1	253	28%	6.670	17%	0.26	238	29%	229	28%	0.02
2	157	17%	4.024	10%	0.21	144	18%	148	18%	−0.01
3	188	21%	3.376	9%	0.35	158	20%	189	23%	−0.09
Missing	<6	0%	45	0%	N/A	<6	0%	<6	0%	N/A
WHO Performance status
0	152	17%	11.050	28%	−0.28	146	18%	114	14%	0.01
1	122	14%	3.778	10%	0.12	111	14%	134	17%	−0.08
2	67	7%	1.000	3%	0.22	62	8%	38	5%	0.12
3	9	1%	218	1%	0.05	7	1%	11	1%	0.05
4	<6	0%	37	0%	N/A	<6	0%	<6	0%	N/A
Missing	550	61%	22.764	59%	N/A	483	60%	511	63%	N/A
ASA
1	46	5%	8.900	23%	−0.53	42	5%	32	4%	0.06
2	388	43%	21.810	56%	−0.26	356	44%	353	44%	0.01
3	413	46%	7.343	19%	0.60	366	45%	403	50%	−0.09
4	31	3%	311	1%	0.18	25	3%	11	1%	0.12
5	N/A	%	9	0%	N/A	N/A	%	N/A	%	N/A
Missing	22	2%	474	1%	N/A	20	2%	10	1%	N/A
Pathologic UICC stage
1	171	19%	8.093	21%	−0.05	158	20%	138	17%	0.06
2	320	36%	13.772	35%	0.00	279	34%	304	38%	−0.06
3	286	32%	11.585	30%	0.04	259	32%	258	32%	0.00
4	123	14%	5.397	14%	0.00	113	14%	109	13%	0.01
Tumor location
Rectum	273	30%	12.556	32%	−0.04	251	31%	244	30%	0.02
Sigmoid colon	249	28%	10.888	28%	0.01	226	28%	227	28%	0.00
Descending colon	50	6%	2.329	6%	0.02	42	5%	46	6%	−0.02
Transverse colon	51	6%	2.110	5%	0.01	48	6%	51	6%	−0.02
Ascending colon	171	19%	6.086	16%	0.09	146	18%	133	16%	0.04
Cecum	107	12%	4.988	13%	−0.03	96	12%	106	13%	−0.04

Baseline characteristics of the population undergoing surgery for colorectal cancer. Data are expressed as No. (%).

ASA indicates American Society of Anesthesiologist’s physical status classification system; UICC: Union for International Cancer Control.

FIGURE 2.

Distribution of patient propensity score before (left) and after propensity score matching (right).

FIGURE 3.

Standard mean difference before and after propensity score matching. The standard mean difference value of each covariate before and after matching is represented by a blue dot.

Within the first year of surgery, 2173 (5.6%) patients died, 110 (13.6%) in the MACE group and 2063 (5.4%) in the non-MACE group. Within 3 years of surgery, 6516 (16.5%) patients died, 248 (30.6%) in the MACE group and 6268 (16.5%) in the non-MACE group, and within 5 years of surgery, 9565 (24.6%) patients died, 333 (41.1%) in the MACE group and 9232 (24.3%) in the non-MACE group. After PSM, we observed an adjusted HR of 1.36 (95% CI = 1.02–1.83; P = 0.04) for 1-year mortality for the MACE group compared with the non-MACE group. Similarly, we observed an adjusted HR of 1.32 (95% CI = 1.07–1.62; P = 0.01) for 3-year mortality and an adjusted HR of 1.25 (95% CI = 1.04–1.50; P = 0.02) for 5-year mortality. Results are summarized in Table 2 and the 5-year survival curve is presented in Figure 4.

TABLE 2.

Hazard Ratios for All-cause Mortality in Patients With and Without MACE Within 30 Days of Colorectal Cancer Surgery

	Hazard Ratio (95% CI)	PSM-Adjusted Hazard Ratio (95% CI)
1-year mortality	2.61 (2.14–3.14)	1.36 (1.02–1.83)
3-year mortality	2.06 (1.81–2.33)	1.32 (1.07–1.62)
5-year mortality	1.94 (1.74–2.16)	1.25 (1.04–1.50)

FIGURE 4.

A 5-year mortality in patients with colorectal cancer following surgery. Shown are the Kaplan–Meier curves for patients with MACE within 30 days of surgery (red) versus no MACE within 30 days of surgery (blue).

Furthermore, we investigated recurrence and DFS in patients with nonmetastatic CRC. After 1:1 matching, we had 697 pairs of patients. For 5-year recurrence, the adjusted HR was 0.96 (95% CI = 0.71–1.31; P = 0.81) and for 5-year DFS, the adjusted HR was 1.10 (95% CI = 0.90–1.35; P = 0.36) in the MACE group compared with the non-MACE group.

DISCUSSION

In this cohort study, we included 39,747 patients undergoing curative-intended elective surgery for CRC. After PSM, we included 809 matched pairs. Patients experiencing MACE during the first postoperative month had an increased risk of mortality 1, 3, and 5 years after surgery compared with patients who did not have immediate postoperative MACE. The risk was increased in both unadjusted and adjusted analyses. The increased risk of long-term mortality in a mixed population of nonmetastatic and metastatic disease led us to question whether MACE also affected recurrence and DFS. When excluding patients with metastatic disease, we did not show a statistically significant change in the hazard of recurrence and DFS.

The pathophysiological mechanism between cancer surgery and cardiovascular complications is complex and not entirely understood and ranges from perioperative myocardial oxygen-supply-demand mismatch to age, endothelial dysfunction, and cardiovascular comorbidities.^6,33,34 Furthermore, inflammation might play a role in the pathophysiological mechanisms. Cancer surgery—in general—is complex with considerable tissue or organ debulking and resection³⁵ and surgery itself initiates a systemic stress response that causes an increase in cortisol, inflammation, and hypercoagulability.³⁶ Commonly used cardiac risk scoring systems do not consider these factors.³⁵ Findings from a study investigating systemic inflammation and incident cancer in patients with stable cardiovascular disease,³⁷ highlight a potential link between inflammation, cardiovascular risk, and cancer.³⁸ A meta-analysis of 2 studies showed that higher preoperative CRP levels were associated with a higher risk of MACE after surgery (OR = 2.11; 95% CI = 1.51–2.94).³⁹ Even though the surgical stress response and the inflammatory response to CRC surgery can be modulated by surgical modalities,⁴⁰ it has also been shown that the hypercoagulable state associated with cancer may persist for up to a month after tumor resection.⁴¹

In this study, all patients were matched on all available variables within the first 30 days after surgery and all time before that. Thus, a condition for entering the study cohort was that patients had to survive until day 30. This study only investigates the long-term association with mortality. Therefore, the increased risk of mortality is only applicable if a patient both experiences a MACE and survives the initial 30 days after surgery. However, the 7.7% point difference between the two groups also indicates a higher short-term mortality for patients with MACE.

The fact that having MACE and surviving the first postoperative month affects mortality even 5 years after surgery warrants strategies to prevent the occurrence of MACE and mitigate the increased risk of mortality. This requires improved stratification of patients before surgery. A recent study showed an increased hazard of heart failure in patients with no previous cardiovascular disease operated for nonmetastatic CRC and identified body mass index, American Society of Anesthesiologist score, and treatment with antihypertensive and lipid-lowering drugs as potential risk factors for incident cardiovascular events.⁴² Another study found that cardiopulmonary exercise testing is a useful perioperative risk assessment tool for patients undergoing surgery for esophageal and gastric cancer.⁴³ Likewise, another study found that wireless single-lead ECG monitoring was able to identify more cases of new-onset postoperative atrial fibrillation in patients having undergone major gastrointestinal surgery than the clinical staff.⁴⁴ We have previously shown that myocardial injury, which is diagnosed solely on increased troponin levels, was associated with increased 90 days morbidity after CRC.¹⁸ In a prospective cohort study, preoperative use of beta-blockers was not associated with decreased cardiac complications after noncardiac surgery,⁴⁵ however, another study found that in patients with ischemic heart disease with heart failure or recent myocardial infarction undergoing noncardiac surgery, use of beta-blockers was associated with lower risk of 30-day MACE.⁴⁶ Thus, there is a need for further investigation to improve preoperative risk stratification, improve perioperative monitoring, and improve interventions to prevent the occurrence of MACE.

Our findings also highlight the need for improving the long-term care of patients, who do experience MACE, to potentially mitigate their increased risk of long-term mortality, which should be seen as a sequela to MACE. It has previously been shown that there is an increased risk of cardiovascular deaths in patients with CRC and hypothesized that it may be due to the use of cardiotoxic chemotherapeutic agents.⁴⁷ Thus, trying to mitigate long-term mortality risk due to cardiovascular events or disease in patients with cancer would require a multidisciplinary approach by surgeons, oncologists, cardiologists, and general practitioners. Such an approach has also previously been recommended in the literature with regard to the cardiovascular adverse effects of oncological treatments.⁴⁸

The strengths of this study are that it is based on 4 large nationwide Danish health registries that are validated and have high completeness, which enables analyzing a large cohort of unselected patients undergoing the same type of surgery. The combination of 4 registries ensured a detailed and universal assessment of disease history, prescription drug use, and laboratory values. This allowed us to adjust for a high variety of possible confounders using PSM, which is a robust method that allows for an intuitive interpretation of the analysis.⁴⁹ Additionally, the use of prospectively collected data reduces the risk of recall bias. There are also limitations to this study that need to be addressed. The retrospective observational design does not allow for a final causal conclusion to be made. Based on the literature, we set an a priori SMD threshold of 0.1 for the covariates in our cohort, but 1.22% of the covariates had a higher SMD—maximum being 0.19—which indicates a potential systematic difference between the groups and is a limitation to this study. Likewise, there is a risk of residual confounding by unmeasured covariates. Furthermore, the incidence of MACE was low resulting in the matched cohorts being relatively small, which affects the statistical power of the findings—especially when investigating oncological outcomes.

In conclusion, we found increased long-term mortality in patients suffering and surviving MACE within the first 30 postoperative days after elective curatively intended CRC surgery, but no statistically significant difference in recurrence and DFS. Even though patients survive their cancer and postoperative MACE, they still have an increased mortality even 5 years after surgery. Our finding supports the need to optimize patient monitoring for complications after CRC surgery (eg, postoperative troponin measurements), identify patients at risk of MACE and mitigate this risk, and improve treatments and care for patients with MACE.