Incidence, aetiology and outcomes of major postoperative haemorrhage after pulmonary lobectomy

Abstract

Open in new tabDownload slide

OBJECTIVES

Post-lobectomy bleeding is uncommon and rarely studied. In this study, we aimed to determine the incidence of post-lobectomy haemorrhage and compare the outcomes of reoperation and non-operative management.

METHODS

We conducted a single-institution review of lobectomy cases from 2009 to 2018. The patients were divided into two groups based on the treatment for postoperative bleeding: reoperation or transfusion of packed red blood cells with observation. Transfusion correcting intraoperative blood loss was excluded. One or more criteria defined postoperative bleeding: (i) drop in haematocrit ≥10 or (ii) frank, sustained chest tube bleeding with or without associated hypotension. Covariates included demographics, comorbidities and operative characteristics. Outcomes were operative mortality, complications, length of hospital stay and readmission within 30 days.

RESULTS

Following 1960 lobectomies (92% malignant disease, 8% non-malignant), haemorrhage occurred in 42 cases (2.1%), leading to reoperation in 27 (1.4%), and non-operative management in 15 (0.8%). The median time to reoperation was 17 h. No source of bleeding was identified in 44% of re-explorations. Patients with postoperative haemorrhage were more often male (64.3% vs 41.2%; P < 0.01) and more likely to have preoperative anaemia (45.2% vs 26.5%; P = 0.01), prior median sternotomy (14.3% vs 6.0%; P = 0.04), an infectious indication (7.1% vs 1.8%; P = 0.01) and operative adhesiolysis (45.2% vs 25.8%; P = 0.01). Compared with non-operative management, reoperation was associated with fewer units of packed red blood cells transfusion (0.4 vs 1.9; P < 0.001), while complication rates were similar and 30-day mortality was absent in either group.

CONCLUSIONS

Haemorrhage after lobectomy is associated with multiple risk factors. Reoperation may avoid transfusion. A prospective study should optimize timing and selection of operative and non-operative management.

INTRODUCTION

Pulmonary lobectomy is one of the most common thoracic operations performed worldwide, representing thousands of procedures per year for both malignant and non-malignant indications [1, 2]. Although rare, postoperative haemorrhage is a major complication and the most common indication for early reoperation [3]. The study of haemorrhage after lobectomy is important because questions remain regarding the need for and timing of packed red blood cell (pRBC) transfusion and operative re-exploration. Postoperative bleeding deserves separate study from intraoperative bleeding as these 2 scenarios represent unique clinical events. Unlike intraoperative bleeding, which mandates haemostasis prior to closure with continued management thereafter, major postoperative bleeding not apparent at the time of closure and may be treated with (i) reoperation, (iii) transfusion of pRBC or (iii) a combination thereof. The study of postoperative bleeding is challenging as the indication for transfusion and reoperation in a retrospective study faces the task of translating individualized surgical judgement into variables of clinical care.

Risk factors for early mortality, prolonged hospitalization and development of respiratory failure have been explored in multiple studies [2, 4–6]. However, risk factors specific to postoperative haemorrhage are difficult to study due to the lack of coding granularity within national databases. Adhesions, prior ipsilateral thoracic interventions, operative approach and anticoagulation during or prior to lobectomy have been inconsistently associated with increased transfusion and postoperative bleeding [3, 7–9]. Determining the incidence and risk factors for postoperative haemorrhage in a single institution, with access to complete chart review, may help to optimize the operative approach and direct evidence-based interventions [10, 11]. In turn, the addition of recognized risk factors to models of operative quality may improve risk stratification [12].

Towards better understanding of the aetiology, risk factors and management of major post-lobectomy haemorrhage, we undertook this study with the a priori hypothesis that reoperation would reduce transfusion requirement and length of hospital stay. We focused on patients receiving pRBC transfusions due to postoperative bleeding and those undergoing re-exploration for haemorrhage. We excluded patients with minor postoperative bleeding who did not require additional intervention.

METHODS

Data source

We conducted a single-institution retrospective review of consecutive patients (≥18 years) who underwent pulmonary lobectomy between January 2009 and July 2018. Patients undergoing chest wall resection, bronchial sleeve and bilobectomy, were included. Patients undergoing completion pneumonectomy were excluded. We included the institutional sample of the Society of Thoracic Surgeons General Thoracic Database (Version 2.07–2.30), a prospective patient registry for lobectomy with patient demographics, pulmonary function data, and intraoperative and postoperative outcomes. The institutional sample included specific variables regarding intraoperative and postoperative pRBC transfusion. In addition, the Partners Research Patient Data Registry (RPDR) was used, a centralized health system repository of clinical and administrative data for inpatient and outpatient encounters containing mortality data through linkage with the Social Security Death Index. These data were linked to the institutional electronic health record, providing admission notes, procedure/imaging reports (operative, radiology and endoscopy) and discharge summaries. The Partners institutional review board approved the study’s conduct.

Preoperative, operative and postoperative treatment

Open and video-assisted thoracoscopic surgery (VATS) was performed by 11 surgeons in an era of changing operative approaches: the proportion of VATS procedures increased from 41% in 2009 to 67% in 2017. The surgical approach selected by the individual surgeon tended to favour thoracotomy for larger tumours, adjacent organ invasion, and after treatment with neoadjuvant therapy. In general, preoperative aspirin for specific indications other than undefined prophylaxis was continued throughout the study period, while vitamin K antagonists, direct thrombin inhibitors and other antiplatelet agents were held before and in the immediate postoperative period as per institutional guidelines, as well as those of the American College of Chest Physicians and the American College of Surgeons [13–15]. Electrocautery was used regularly, while argon beam, ultrasound coagulators or other energy devices were applied infrequently. Throughout the study period, the pulmonary artery and vein were either suture ligated or divided with staples. Typically, patients were extubated in the operating room before transport to the postoperative anaesthesia care unit. Direct intensive care unit (ICU) admission was uncommon (∼15%) and occurred when postoperative anaesthesia care unit beds were unavailable, for complex index procedures, or for patients with severe comorbidities. Patients routinely received prophylactic subcutaneous heparin or enoxaparin on postoperative day 1 and were continued on these medications until hospital discharge. The choice between heparin and enoxaparin prophylaxis, and whether the first postoperative dose was delayed, was made based on the use of epidural analgesia, the judgement of the surgeon, and in consideration of chronic kidney disease.

Patient characteristics

Demographics including age, sex and race/ethnicity were obtained. Patient body mass index, the American Society of Anesthesia (ASA) score, tobacco use and most recent pulmonary function tests were recorded at the time of surgery. Preoperative comorbidities included coronary artery disease (CAD), cerebrovascular disease, peripheral arterial disease, anaemia (haemoglobin <13.5 g/dl for males, <12.0 g/dl for females), dialysis dependence and thromboembolic disease. Use of preoperative chemoradiotherapy was recorded, as well as any prior median sternotomy or ipsilateral thoracotomy/thoracoscopy. Given the focus on postoperative haemorrhage, data on preoperative anticoagulant use were collected from the electronic health record.

The indication and approach (open, VATS or robotic assisted) were recorded for each procedure and re-exploration. Review of operative reports was used to determine the presence of intraoperative adhesiolysis, tumour chest wall involvement and performance of nodal dissection. Database and chart review were used to determine the presence of intraoperative transfusion, postoperative transfusion, duration of operation and procedure end-time. We chose to examine procedure end-time in order to measure the interval before reoperation. We categorized procedure end-time into morning (before 12 PM), afternoon (12 –5 PM) and evening (after 5 PM) to study its association with re-exploration. A board-certified radiologist specialized in thoracic imaging (F.J.F.) reviewed the chest radiographs of patients either undergoing reoperation and/or transfusion for postoperative bleeding to determine the presence of haemothorax manifesting as one or more of the following: enlarging ipsilateral opacity at the hilum, opacification of the chest wall or an effusion >50% of the pleural space.

Major haemorrhage after lobectomy

Patients with major postoperative haemorrhage were identified by ≥1 unit pRBC transfusion within 30 days of the index operation or reoperation for the documented indication of bleeding. Patients who did not require reoperation or transfusion of pRBC were not considered to have major postoperative haemorrhage. The decision to transfuse was made at the discretion of the attending surgeon, but in general was reserved for patients with haemoglobin <7.0, haemoglobin <9.0 in patients with significant CAD, and symptomatic patients (tachycardia and hypotension) with a dropping haemoglobin in concordance with ASA guidelines for blood management [16]. Because there are many indications for postoperative pRBC transfusion, among patients who were treated only with transfusion, major postoperative haemorrhage was further defined by (i) ≥10 point drop in haematocrit between postoperative values or (ii) clinical evidence of frank blood from a pleural catheter. Frank blood from a pleural catheter was determined through review of clinical notes documented during postoperative hospitalization, in which this finding was cited as the indication for transfusion. The chest tube output volume could not be reliably separated from the interval before or after intervention and was therefore stated only as the total amount. Patients who received postoperative transfusion solely to correct intraoperative blood loss during the index operation or due to chronic anaemia were not considered to have postoperative haemorrhage. In further analysis, patients with postoperative haemorrhage were subcategorized based on treatment which included reoperation or pRBC transfusion without reoperation.

Outcomes: length of stay, mortality and complications

Outcomes were compared between patients with and without postoperative haemorrhage, including: hospital length of stay (LOS), ICU LOS, units of pRBC transfused, discharge disposition, 30-day hospital readmission and 30-day mortality. Postoperative 30-day complications as recorded in the Society of Thoracic Surgery (STS) database were also compared between the 2 groups [17].

Statistical analysis

Categorical variables were compared using Fisher’s exact test and continuous variables using the Wilcoxon rank-sum tests, under the assumption each patient and their operation were independent events. Given the rarity of post-lobectomy haemorrhage, we elected not to perform a multivariable analysis as the selection of variables would not have been rigorous and subject to bias. Computations were carried out using Stata software, version 15.1 (StataCorp, College Station, Texas). Two-tailed P-values <0.05 were considered to be statistically significant.

RESULTS

Among 1960 consecutive lobectomies, 27 (1.4%) underwent reoperation for postoperative haemorrhage. In addition, 80 patients (4.1%) required ≥1 pRBC transfusion within 30 days of the index operation. Of these, 15 received transfusions in the setting of ongoing postoperative haemorrhage, while 65 were transfused due to intraoperative blood loss or chronic anaemia associated with prolonged hospitalization and illness. Therefore, 2.1% of patients (27 + 15/1960) met the criteria of major post-lobectomy haemorrhage. Among surgeons with more than 50 lobectomies during the study period, the rate of postoperative bleeding ranged from 0% to 2.38% (P = 0.34) and reoperation from 0% to 1.47% (P = 0.26). The CONSORT diagram in Fig. 1 illustrates the composition of this group.

Patient and operative characteristics

The median patient age was 67 and most were female (58.3%) and white (88.2%) (Table 1). Patients with postoperative haemorrhage were more often male (64.3% vs 41.2%; P < 0.01), had pre-existing anaemia (45.2% vs 26.5%; P = 0.01), a prior median sternotomy (14.3% vs 6.0%; P = 0.04) and a lower predicted diffusion capacity for carbon monoxide (69.6% vs 78.1%; P = 0.02). Patient factors such as preoperative use of anticoagulants, steroids, other comorbidities, and ASA score were not significantly different. However, preoperative anaemia was more common after neoadjuvant therapy (55% vs 24%; P < 0.001).

The most common indication for lobectomy was non-small-cell lung cancer (NSCLC) in 84.6% (Table 2), but postoperative haemorrhage occurred more often in patients with metastasis, infectious disease and benign neoplastic disease. The rate of post-lobectomy haemorrhage specific for NSCLC was 1.7%. Patients with postoperative haemorrhage more often underwent operative adhesiolysis (45.2% vs 25.8%; P = 0.01) and were admitted to the ICU (28.6% vs 14.8%; P = 0.03). There were no cases of postoperative haemorrhage among the 47 patients who underwent robotic assisted lobectomy, while the rate of postoperative haemorrhage among patients who underwent VATS lobectomy was 1.7%.

Reoperation versus non-operative management of postoperative haemorrhage

Among the 42 patients with postoperative haemorrhage, 27 underwent reoperation and 15 were managed non-operatively (Table 3). Both treatment cohorts had similar rates of hypotension (systolic blood pressure <90 mmHg) and vasopressor requirement, but direct admission to the ICU after the index procedure was less common before reoperation (7% vs 67%; P < 0.001). Over 75% of patients in both groups had radiological evidence of haemothorax. Among these patients, 24 had opacification of ≥½ of the affected haemothorax, 28 had opacification of the chest wall, and 1 had evidence of mediastinal shift. Patients selected for reoperation had shorter index procedures (160 min vs 249 min; P < 0.001), a higher proportion of index operations concluded in the morning, and a higher proportion of VATS (80% vs 48%; P = 0.03) than those treated with transfusion alone. The median time to reoperation was 17 h: 11 (41%) returned to the operating room in <12 h, another 11 (41%) returned between 12 and 48 h, and a final 5 (19%) >48 h after lobectomy. The most common reason for a late (after 12 h) conversion to operative management was an expanding haemothorax (46.7%), followed by new or worsening hypotension (26.7%), inappropriate response to pRBC transfusion (20.0%) and unremitting pleural catheter output (6.7%). The median total chest tube output during hospitalization was lower in patients treated with reoperation, but this did not reach statistical significance (Fig. 2). In both groups, other blood products were used sparingly; only 3 patients received platelet transfusion and 4 fresh frozen plasma.

At re-exploration, the operative approach was the same as that during the index operation (13 VATS, 13 open thoracotomy) except in 1 case of intraoperative conversion. The most common site of bleeding was either diffuse or unidentified (44%), followed in frequency by chest wall (23%), bronchial artery or lymph node (19%) and pulmonary artery or lung parenchyma (15%) (Fig. 3). The transfusion requirement in the reoperation cohort was significantly lower than in the cohort receiving non-operative treatment (0.4 units pRBC vs 1.9 units pRBC; P < 0.01), and transfusion was not required in more than half of reoperations (52%).

Hospital course and complications

Hospital course, 30-day mortality, 30-day complications and 30-day readmission are summarized in Table 4. There was no operative mortality in patients with postoperative haemorrhage and no difference in ICU LOS between those with and without haemorrhage. Patients who developed postoperative haemorrhage had a longer LOS (6 days vs 4 days, P = 0.01) as well as higher rates of wound infection (2.4% vs 0.3%, P = 0.04) and thromboembolic disease (4.8% vs 2.2%, P = 0.03) than those who did not develop postoperative haemorrhage; however, all of these rates and intervals were low. In both groups, the majority of patients were discharged home (>88%), while the 30-day readmission rate among patients with post-lobectomy haemorrhage was markedly higher than those in the absence of post-lobectomy haemorrhage (16.7% vs 2.8%, P = 0.01).

In subgroup analysis, comparing patients with post-lobectomy haemorrhage that were treated operatively and those that were managed non-operatively, there was no significant difference in hospital course or 30-day readmission. There was a borderline reduction in the rate of thromboembolic disease among patients who were treated with reoperation (0% vs 13.3%; P = 0.12).

DISCUSSION

In this retrospective study of close to 2000 consecutive lobectomies, we found a 2.1% incidence of major postoperative haemorrhage and a 1.7% rate of postoperative haemorrhage in patients who underwent lobectomy for NSCLC. In a univariable analysis, male gender, preoperative anaemia, diffusion capacity and operative circumstances such as adhesions and non-malignant disease were associated with increased postoperative bleeding. Among patients with postoperative haemorrhage, reoperation was associated with a significant reduction in units of pRBC transfused compared with non-operative management. This information may guide surgeons when selecting these different treatment modalities in the rare event of post-lobectomy haemorrhage.

The rates of haemorrhage after lobectomy have recently been assessed in both single-institution and multi-institution studies. The incidence of postoperative haemorrhage in these studies is similar to that found in our review, ranging from 0.9% to 2.4% [3, 4]. However, in many larger studies, post-lobectomy haemorrhage is not directly addressed or intraoperative transfusion and postoperative transfusion are combined as a single variable [18–21]. Our work adds meaningful information to the understanding of post-lobectomy haemorrhage and is unique in its assessment and differentiation between postoperative haemorrhage treated with reoperation or transfusion alone.

As with other recent studies, we did not observe an increased risk of postoperative haemorrhage in patients on antiplatelet medications [8, 9]. While these medications were held when possible, a large number of patients continued on aspirin without significantly worse perioperative outcomes. Dual antiplatelet (e.g. aspirin and clopidogrel) use was rarely encountered, and our study is underpowered to adequately assess its safety. However, in a recent retrospective study by Atay et al., the rate of transfusion in patients being treated with clopidogrel was 16% [22].

A key finding was the decreased rate of transfusion in patients treated with reoperation. This is important because rare but serious events may occur due to blood transfusion, such as transmission of blood-borne viruses, exposure to bacterial contaminants, development of haemolytic reactions and transfusion-mediated immunomodulation [23]. Furthermore, blood transfusion has been independently associated with worse perioperative and oncological outcomes in a variety of cancers [24–26]. In addition, blood transfusion is a risk factor for subsequent development of hepatic cancer as well as non-Hodgkin lymphoma [27].

Patients who underwent non-operative management of post-lobectomy haemorrhage were more likely to have complex index procedures, and preoperative anaemia; concern for the risk of a second operation may have contributed to the clinical decision-making, creating a partly obscured selection criterion. While preoperative anaemia was associated with increased risk of postoperative haemorrhage, we defined postoperative bleeding as a ≥10 point drop in haematocrit between postoperative values in order to distinguish transfusion for intraoperative blood loss and chronic anaemia. The greater instability and clinical complexity of patients who underwent non-operative management (as demonstrated by a higher rate of postoperative ICU use) may have influenced this treatment decision. While re-exploration after cardiac surgery often identifies a clear source of bleeding (in 64–88% as reported by Karthik and Frojd and their co-authors [28, 29]), this study found diffuse bleeding without a clear source in 44% of reoperations. Whether every postoperative haemorrhage after lobectomy would have resolved without reoperation cannot be determined; however, reoperation was associated with decreased need for pRBC transfusion.

The rating for programme performance after lobectomy considers reoperation a major complication: each event threatens the patient with the loss of a performance star [12]. By contrast, transfusion for postoperative haemorrhage is not associated with the same penalty risk. How these penalties may affect treatment is unknown. We do not suspect this circumstance directly enters the surgeon’s mind; however, previous publications have shown that they may subtly influence decision-making [30]. Specific factors that could adjust risk for postoperative haemorrhage such as prior median sternotomy and adhesiolysis are not included as covariates in the STS model.

Limitations

The strengths of this study include its granularity. We were able to review operative reports from all procedures and achieve a level of detail absent from national and multi-institutional databases. However, we recognize certain limitations. We focused on patients with major postoperative bleeding requiring transfusion or reoperation. Patients with minor bleeding and no intervention were excluded from detailed analysis as they represent less of a clinical dilemma. In addition, a single-institution study may not be representative of other institutions or regions. The decision between observation and reoperation was likely influenced by unmeasured variables including surgeon’s preference and time of diagnosis, and thus optimal timing and threshold for surgical intervention remain undefined. A cost analysis comparing operative versus non-operative management of postoperative haemorrhage was not part of our study. Given the high-cost associated with operative time as well as hospitalization this concern is important and should be addressed in future studies. Furthermore, absent data on the detailed amount of pleural catheter output by hour reduces the usefulness of this important clinical bleeding signal in our report. Other modalities to test for bleeding such as drain haematocrit or thrombocyte aggregation were not used in this study and remain a topic for further investigation. Lastly, the retrospective nature of this study allows us to demonstrate associations. There may be confounding and unmeasured variables that contribute to these findings which can be better controlled in prospective studies and through increased statistical power. Unfortunately, given ethical concerns as well as required time and cost, a prospective study is unlikely.

This study has refined the definition of postoperative haemorrhage and described its incidence and risk factors in greater detail than previously reported. Early reoperation may spare patients additional transfusion of pRBC, while non-operative monitoring may be applied to patients with higher operative risk. As in all clinic decisions, unique patient factors dictate the time of management that is most efficacious. Clinical deterioration with continued expansion of haemothorax, rising pressor requirement, and inadequate response to blood transfusion seem to mandate re-exploration. Lastly, unique risk factors for postoperative haemorrhage should be considered in future refinements of current quality composite scores for lobectomy.

Conflict of interest: none declared.

Author contributions

Brooks V Udelsman: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Validation; Writing – original draft; Writing – review & editing. Monica Soni: Conceptualization; Data curation; Writing – review & editing). Maria Lucia Madariaga: Conceptualization; Writing – review & editing. Florian J Fintelmann: Conceptualization; Data curation; Formal analysis; Investigation; Writing – review & editing. Till D Best: Formal analysis; Investigation; Writing – Review & Editing. Selena Shi-Yao Li: Formal analysis; Investigation; Writing – review & editing. Davic C Chang: Investigation; Methodology; Writing – review & editing. Henning A Gaissert: Conceptualization; Formal analysis; Investigation; Project administration; Writing – original draft; Writing – review & editing.

Presented at the 27th European Conference on General Thoracic Surgery, Dublin, Ireland, 9–12 June 2019.

REFERENCES

ABBREVIATIONS

ABBREVIATIONS
 
• ASA

  American Society of Anesthesia

 
• CAD

  Coronary artery disease

 
• ICU

  Intensive care unit

 
• IQR

  Interquartile range

 
• LOS

  Length of stay

 
• NSCLC

  Non-small-cell lung cancer

 
• pRBC

  Packed red blood cells

 
• RPDR

  Partners Research Patient Data Registry

 
• STS

  Society of Thoracic Surgery

 
• VATS

  Video-assisted thoracoscopic surgery