Validation of a novel risk score to predict mortality after surgery for acute type A dissection

Abstract

 

Open in new tabDownload slide

OBJECTIVES

The aim of this study was to externally validate a lab-based risk score (lactate, creatinine, aspartate aminotransferase, alanine aminotransferase or bilirubin) by Ghoreishi et al. to predict perioperative mortality in patients undergoing surgical repair for acute type A aortic dissection.

METHODS

The risk score to predict operative mortality was applied to a large and homogenous validation cohort that consisted of 632 patients undergoing surgery for acute type A aortic dissection in 2 centres. Multivariable regression analysis was performed to determine the impact on survival. Receiver operating characteristics with deduced area under the curve were used to assess the ability to predict perioperative mortality.

RESULTS

A total of 632 patients (54% male, mean age 62 ± 14 years) were assigned to 3 different risk groups according to the calculated mortality score [low risk <7 (31.2%), moderate risk 7–20 (36.1%) and high >20 (32.7%)]. Perioperative mortality was 8% in the low-risk group, 10% in the moderate-risk group and 24% in the high-risk group (P < 0.0001). Receiver operating characteristic analysis of this new score revealed an area under the curve of 0.69 with adequate calibration. In addition, multivariable analysis revealed an independet assocation with perioperative mortality (odds ratio 1.509; 95% confidence interval 1.042–2.185). While overall survival differed between the risk groups (P < 0.0001), the score does not serve as an independent predictor of long-term mortality when adjusted for relevant covariates.

CONCLUSIONS

The external validation process confirmed that a newly proposed risk score offers clinicians a helpful and reliable tool to improve the preoperative risk assessment of acute type A aortic dissection patients based on easily accessible and broadly available laboratory parameters.

INTRODUCTION

Acute type A aortic dissection (ATAAD) is a life-threatening condition, where rapid diagnosis and therapy is mandatory to prevent adverse outcome. Due to the aggressive natural course of the disease ATAAD and the unfavourable outcome of conservative treatment, rapid correct initial diagnosis is of great importance and immediate surgical repair is recommended [1, 2]. The presentation of aortic dissections is diverse and often misleading, depending on the extent of the dissection and the associated complications related to organ malperfusion. Unfortunately, initial misdiagnosis is common and is associated with the delay of surgical repair [2].

Various risk factors for operative or in-hospital mortality in patients undergoing surgical repair have been identified. Besides advanced age, a critical preoperative state due to organ ischaemia, pericardial tamponade, left ventricular systolic dysfunction and the need for preoperative resuscitation has a negative impact on operative outcome [3–6].

Due to the acuity of the underlying disease, time for thorough examination is limited. Risk stratification at admission is necessary to evaluate the potential outcome and to determine and schedule the appropriate treatment strategies. Several attempts have been made to better predict perioperative mortality. Scorecards and web applications—based on clinical variables as well as imaging data—for the prediction of early mortality after surgical repair have been created in recent years [7, 8]. The Penn Classification has emerged as a helpful tool to assess patients according to the degree of malperfusion [9]. Inspired by the Penn Classification but still in search of improved predictive power Ghoreishi et al. [10] proposed a new risk stratification, which adds additional information for preoperative risk assessement in ATAAD. They created an algorithm to predict operative mortality of ATAAD based on laboratory results including lactic acid, creatinine and liver-associated parameters at admission. The study was limited by a relatively small sample size and did not include a validation cohort. The aim of this study was to externally validate the proposed risk stratification model using a large and homogenous cohort of patients undergoing surgical repair for ATAAD.

PATIENTS AND METHODS

Patient population

We retrospectively included patients from 2 high-volume centres that received surgical treatment for ATAAD between 2000 and 2019. Clinical data were retrospectively evaluated. Local ethics committees approved aortic databases in both centres (Berlin: EA2/096/20, Innsbruck: UN5106).

Exclusion criteria for score calculation were lack of laboratory values, preoperative chronic kidney disease or preoperative dialysis (n = 51) and death prior to cardiopulmonary bypass (unsuccessfull mechanical as well as medical re-animation in patients with aortic rupture during transportation to the operation room or induction of anaesthesia; n = 17).

According to Ghoreishi, the risk score was calculated according to the proposed formula: 5.5 × (lactate) + 8 × (creatinine) ± 8 (liver malperfusion). Normal values for the used parameters were defined as following: blood lactate <1 mmol/l, creatinine for female patients 0.51–0.95 mg/dl and male patients 0.67–1.17 mg/dl, aspartate aminotransferase and alanine aminotransferase 10–50 U/l and bilirubin <1.28 mg/dl.

The definition of liver malperfusion was based on any increase of aspartate aminotransferase, alanine aminotransferase or bilirubin. Based on the resulting points, patients were stratified into 3 risk score groups. Both the formula to calculate the score and the according stratification into 3 groups were adopted from the original publication by Ghoreishi et al. [10]

Low risk was defined as a score of <7, moderate risk as a score of 7–20 and high risk as a score of >20.

Surgical repair

Once diagnosis was confirmed, patients were transferred to the operation theatre immediately, while computed tomography scans were reviewed and treatment strategy was set. Operative strategy has been previously described [11]. In short, arterial cannulation was predominantly performed via the axillary (n = 374; 59%) and femoral artery (n = 220; 35%). Direct cannulation of the aorta (n = 42; 7%) as well as carotid cannulation (n = 12; 2%) or cannulation of the innominate trunk (n = 10; 2%) was chosen less frequently.

The majority of surgical repairs were performed in hypothermic circulatory arrest and selective antegrade cerebral perfusion (n = 568; 90%) with a mean circulatory arrest time of 46 ± 29 min. Antegrade cerebral perfusion with cold blood (20–25°C) at a flow rate of 10 ml/kg/min body weight has been implemented as standard technique in both centres within the last 10 years (n = 287; 45%). Straight deep hypothermic circulatory arrest without additional cerebral perfusion (n = 30; 5%) as well as isolated retrograde cerebral perfusion (n = 251; 40%) via an angled cannula, which was inserted in the superior vena cava and snared during hypothermic circulatory arrest, was utilized mainly in the early study period (2000–2009).

A primarily tear-oriented approach towards surgical repair was followed in both surgical centres. Depending on the extent of intimal defects or a pre-existing dilatation, root (n = 212; 33%) and/or total arch replacement (n = 166; 26%) were performed. Postoperative treatment was standardized for every patient at a dedicated, experienced intensive care unit. Perioperative mortality was defined as death within first 30 days after surgery.

Statistical analysis

Statistical analysis was performed using SPSS 24 (SPSS Inc, Chicago, IL, USA) and R Version 4.0.0 [R Development Core Team (2019). R: A Language and Environment for Statistical Computing]. Categorical variables are presented as frequencies with corresponding percentages. Continuous variables were expressed as mean ± standard deviation. Differences between groups were tested by means of χ² test or one-way analysis of variance, as appropriate.

Risk factor identification for perioperative and long-term mortality was performed using univariable and multivariable logistic and Cox regression analyses. Variables with a P-value of <0.2 in univariable testing were considered for the multivariable model. Results of regression analyses were displayed either as odds ratio (OR) or as hazard ratio (HR) with corresponding 95% confidence interval (CI). Survival analysis was depicted by means of Kaplan–Meier method and tested by log-rank test. Receiver operating characteristic with area under the curve (AUC) was applied to evaluate discrimination. Hosmer–Lemeshow test was used to assess calibration. A P-value of <0.05 was considered as statistically significant.

RESULTS

Demographics and risk groups

In this retrospective validation analysis, 632 patients were included. Patients’ mean age was 62 ± 14 years, and 54% were male.

According to the proposed risk score calculation, 197 (31%) patients were in the low-risk group, 228 (36%) patients were in the moderate-risk group and 207 (33%) patients were in the high-risk group (Fig. 1). The mean values of the relevant laboratory parameters for score calculation differed significantly between the 3 risk groups.

Besides elevated laboratory values, patients in the high-risk group presented in an extremely critical clinical status with the highest rate of preoperative malperfusion syndrome (55% vs 39% in the moderate-risk group and 22% in the low-risk group, P < 0.001) and pericardial tamponade (38% vs 11% in the moderate-risk group and 7% in the low-risk group, P < 0.001) (Table 1).

Surgical repair

While the majority of patients received hemiarch aortic replacement, 166 patients (26%) underwent total arch repair due to aortic arch dilatation or entry tear located in the arch. There were no significant differences in aortic cross-clamp or circulatory arrest time but a trend towards longer bypass time in the high-risk group. The axillary artery was predominantly chosen as an arterial cannulation site in all risk groups. The femoral artery was cannulated more often in the high-risk group (42% vs 32% in moderate or 31% in low risk, P = 0.028). In 26 patients, double arterial cannulation—femoral cannulation in addition to direct aortic cannulation (n = 11) or axillary cannulation (n = 15)—was performed to address abdominal and/or femoral malperfusion. This strategy was primarily applied in moderate- and high-risk patients. Further details are depicted in Table 2.

Perioperative mortality and risk score

Perioperative mortality of the total population was 13.8% (n = 87). The mean mortality score of the total cohort was 17.7 ± 17. Perioperative mortality differed significantly between the 3 risk groups (P < 0.001) with the highest death rate of 23.7% in the high-risk group (Fig. 1).

Multivariable logistic regression analysis identified age (OR 1.021, 95% CI 1.001–1.042; P = 0.046), previous cardiac surgery (OR 3.843, 95% CI 1.437–10.275; P = 0.007), malperfusion syndrome (OR 2.036, 95% CI 1.144–3.625; P = 0.016), preoperative tamponade (OR 1.997, 95% CI 1.085–3.574; P = 0.026) and the proposed risk groups (1.509, 95% CI 1.042–2.185; P = 0.029) as independent predictors for perioperative mortality (Table 3).

The receiver operating characteristic analysis for the prediction of perioperative mortality was calculated for the total cohort and revealed an AUC of 0.69 with adequate calibration. C-statistics varied between the 3 proposed risk groups (Fig. 2).

Long-term survival

In accordance with previous studies, the COX regression analysis identified age (HR 1.035, 95% CI 1.021–1.050; P < 0.001), previous cardiac surgery (HR 2.116, 95% CI 1.130–3.960; P = 0.019), malperfusion syndrome at presentation (HR 1.541, 95% CI 1.056–2.375; P = 0.025) and preoperative tamponade (HR 1.579, 95% CI 1.050–2.375; P = 0.028) as independent predictors for cumulative mortality during follow-up. The mortaliy score did not show a statistically significant association with cumulative mortality (HR 1.274, 95% CI 0.999–1.624; P = 0.050) (Table 4).

In Kaplan–Meier analysis, there was a stepwise increase in long-term mortality according to the 3 different risk groups, mainly driven by the high early mortality in the high-risk group (Fig. 3).

DISCUSSION

The proposed risk score is a valuable and easy accessible tool to predict perioperative mortality in patients undergoing surgical repair for ATAAD. The main findings of present study can be summarized as follows:

• The new risk score proposed by Ghoreishi et al. is a predictor of perioperative mortality in patients with ATAAD, independent of established risk factors.

• Receiver operating characteristics confirmed its ability to predict perioperative mortality following ATAAD, regardless of risk groups.

• There is no statistically significant association between the proposed score and long-term mortality.

Scorecards and risk scores

Adequate preoperative evaluation of patients suffering from ATAAD remains a challenge. Most risk stratification tools are driven by parameters reflecting end-organ malperfusion or reflecting a critical preoperative state. The implementation and validation of the Penn classification for the preoperative risk assessment is one example [9, 12]. Leontyev et al. [7] extended the risk assessment by incorporating the presence of critical preoperative state—defined as inotropic support, cardiopulmonary resuscitation or mechanical ventilation—in their scorecard in addition to age, malperfusion and coronary artery disease. By omitting age and replacing malperfusion syndrome by the Penn classification, the proposed scorecard underwent external validation and adjustment, improving the ability to predict in-hospital mortality in ATAAD [13].

In addition to this Leipzig–Halifax scorecard, further risk classifications and some web applications to predict mortality in ATAAD and to identify high-risk patients have been developed [8, 14]. The most recent web-based application by Czerny et al. [8] was created based on parameters of 2537 patients enrolled in the German Registry of Acute Type A Dissection (GERAADA). Luehr et al. [15] externally revalidated the score’s accurateness early after implementation. To calculate the proposed GERAADA score, detailed information on the clinical status, aortic valve function, imaging data (malperfusion, extension of dissection), and knowledge regarding location of the primary tear is required. This score is of good predictive ability (AUC 0.72), but arguably its application might be challenging in the preoperative setting. Detailed echocardiography might be lacking or the entry tear location may remain indefinite. As a consequence, some of the needed score parameters might not be easy to retrieve. In contrast to all previously discussed risk tools, the recently proposed score based on readily available laboratory values is of similar robustness (AUC 0.69).

Impact of preoperative laboratory parameters

The score developed by Ghoreishi et al. [10] was exclusively based on serum lactate, creatinine and liver enzymes (aspartate aminotransferase, alanine aminotransferase) or bilirubin.

Isolated elevated serum lactate as predictor for impaired outcome in ATAAD has been discussed ambiguously in literature [16, 17]. Serum lactate could be interpreted as a marker indicating a critical state, reflecting the presence of malperfusion, cardiac tamponade or other life-threatening clinical conditions.

Abnormalities in creatinine or liver parameters could be side effects of cardiac tamponade or low cardiac output, but more importantly interpreted as a direct evidence for abdominal malperfusion. Hypotension affects the liver as one target organ and leads to hepatocellular dysfunction, being reflected in a sudden rise of liver enzymes within 4.5 h [18]. A rise in absolute serum creatinine in acute kidney injury is directly related to baseline kidney function, Waikar and Bonventre [19] could show a 50% increase of serum creatinine in normal baseline function after 4 h after onset of severe acute kidney injury. Insights on the impact of number of organs involved in preoperative malperfusion in type A dissection patients were given in a GERAADA registry analysis [20]. Presence of malperfusion of 2 organ systems was associated with a 30.2% death rate or a 2.4-fold elevated risk of death [5, 20].

Similar results could be obtained when applying the risk score proposed by Ghoreishi et al. [10]. Elevated creatinine levels in combination with liver injury accounted for high scores (>20) in the model and were associated with 37% operative mortality rate. The implemented laboratory parameters in this score seem to act as reliable markers for abdominal malperfusion, resulting in liver and/or kidney injury, critical state or tamponade.

Process of external validation

The original presentation of this novel risk score by Ghoreishi et al. contained certain limitations and differed from our study in some important aspects.

The number of patients retrospectively analysed for score calculation was limited to 134, in contrast to our large two-centre cohort including 632 patients. In contrast to the previous report, we were able to create a statistically robust multivariable risk model for perioperative mortality enhancing the importance of this new risk score. In addition, established risk factors like age, previous cardiac surgery, malperfusion syndrome and preoperative tamponade were confirmed as independent risk factors for perioperative mortality. Corresponding to the GERAADA risk score, previous cardiac surgery seems to increase operative complexity with impact on postoperative survival [8]. One might also argue that patients with a history of a cardiac surgical procedure represent a subgroup with advanced age and progressive cardiovascular disease, especially coronary artery disease [21, 22].

The numerical distribution of patients into the 3 different risk groups was rather homogenous in our cohort, while half of all patients in the original manuscript were assigned to the moderate-risk group [10]. Due to the larger sample size, we were able to investigate the score regarding its prognostic value in each risk group and detected similar discriminative ability throughout all risk groups.

Clinical implications

Due to the high morbidity and mortality in ATAAD, the identification of patients at very high risk is essential for a risk stratification tool. As the lab results needed for this score calculation are normally available within 1 h in a 24/7 fashion at most centres, risk assessment could be performed preoperatively or during induction of general anaesthesia. The validated score could help to counsel the patients and their families being confronted with this acute diagnosis. In addition, high scores should act as red flags for detailed investigations. This could be useful to identify patients who might benefit from additional treatment strategies to address malperfusion preoperatively. Aortic fenestration and endovascular repair prior to surgery have been reported as alternative interventional approaches [23]. From the surgical perspective, the frozen elephant trunk technique or newer technologies (e.g. Ascyrus Medical Dissection Stent) as well as double arterial cannulation may be considered in high-risk patients as treatment option to prevent ongoing distal organ malperfusion [24, 25].

Limitations

Major limitations in this validation study originate from its retrospective nature. Due to the long enrolment period of this study, surgical strategies, especially perfusion techniques, have changed over time. As data from 2 tertiary centres with standardized perioperative care were merged, local differences (centre volume, level of surgical expertise) might have affected the outcome.

Score calculation was based on preoperative variables only. Time from onset on symptoms to blood draw was not documented in our databases and therefore conclusions of different variations in time of different blood parameters cannot be provided. Patients with known kidney disease requiring dialysis were excluded from this study. Many patients had their index admission at the hospital and pre-exisiting data on medical history or organ function might have been missed.

Conflict of interest: Volkmar Falk: Medtronic GmbH, Biotronik SE & Co., Abbott GmbH & Co. KG, Boston Scientific, Edwards Lifesciences, Berlin Heart, Novartis Pharma GmbH, JOTEC GmbH and Zurich Heart. All other authors declared no conflict of interest.

Author contributions

Markus Kofler: Conceptualization; Formal analysis; Methodology; Visualization; Writing—original draft; Writing—review & editing. Roland Heck: Data curation; Methodology; Project administration. Fabian Seeber: Data curation; Formal analysis. Matteo Montagner: Conceptualization; Data curation. Simone Gasser: Data curation. Lukas Stastny: Data curation. Stephan D. Kurz: Supervision; Validation. Michael Grimm: Supervision. Volkmar Falk: Supervision; Validation. Jörg Kempfert: Conceptualization; Supervision; Validation. Julia Dumfarth: Conceptualization; Data curation; Formal analysis; Methodology; Project administration; Supervision; Validation; Writing—original draft; Writing—review& editing.

Reviewer information

European Journal of Cardio-Thoracic Surgery thanks Daniel-Sebastian Dohle, Maximilian Luehr, Sven Peterss and the other, anonymous reviewer(s) for their contribution to the peer review process of this article.

REFERENCES

ABBREVIATIONS

 
• ATAAD

  Acute type A aortic dissection

 
• AUC

  Area under the curve

 
• CI

  Confidence interval

 
• GERAADA

  German Registry of Acute Type A Dissection

 
• HR

  Hazard ratio

 
• OR

  Odds ratio

Author notes