Postoperative but not intraoperative transfusions are associated with respiratory failure after pneumonectomy

Abstract

OBJECTIVES

Open in new tabDownload slide

Transfusion of blood products has been associated with increased risk of post-pneumonectomy respiratory failure. It is unclear whether intraoperative or postoperative transfusions confer a higher risk of respiratory failure. Our objective was to assess the role of transfusions in developing post-pneumonectomy respiratory failure.

METHODS

We performed a retrospective cohort study using prospectively collected data on consecutive pneumonectomies between 2005 and 2015. Patient records were reviewed for intraoperative/postoperative exposures. Univariable and multivariable analyses were performed.

RESULTS

Of the 251 pneumonectomies performed during the study period, 24 (9.6%) patients suffered respiratory failure. Ninety-day mortality was 5.6% (n = 14) and was more likely in patients with respiratory failure (7/24 vs 7/227, P < 0.001). Intraoperative and postoperative transfusions occurred in 42.2% (n = 106) and 44.6% (n = 112) of patients, respectively and were predominantly red blood cells. On univariable analysis, both intraoperative (P = 0.03) and postoperative transfusion (P = 0.004) were associated with a higher risk of respiratory failure. The multivariable model significantly predicted respiratory failure with an area under curve (AUC) = 0.88 (P = 0.001). On multivariable analysis, the only independent predictors of respiratory failure were postoperative transfusions [adjusted odds ratio (aOR) 6.54, 95% confidence interval (CI) 1.74–24.59; P = 0.005] and lower preoperative forced expiratory volume (adjusted OR 0.96, 95% CI 0.93–0.99; P = 0.03). Estimated blood loss was not significantly different (P = 0.91) between those with (median 800 ml, interquartile range 300–2000 ml) and without respiratory failure (median 800 ml, interquartile range 300–2000 ml).

CONCLUSIONS

Respiratory failure occurred in 9.6% of patients post-pneumonectomy and confers a higher risk of 90-day mortality. Postoperative (but not intraoperative) transfusion was the strongest independent predictor associated with respiratory failure. Intraoperative transfusion may be in reaction to active/unpredictable blood loss and may not be easily modifiable. However, postoperative transfusion may be modifiable and potentially avoidable. Transfusion thresholds should be assessed in light of potential cost-benefit trade-offs.

INTRODUCTION

Pneumonectomy is associated with a high risk of respiratory complications. Despite advances in perioperative care, this procedure still carries with it a high rate of respiratory complications including a risk of respiratory failure of up to 18% [1, 2]. Postoperative respiratory failure is the greatest cause of perioperative death in adult thoracic surgery patients [3]. Furthermore, the overall all-cause mortality for those with severe respiratory failure is 45% as reported by Bime et al. [4] Thus, respiratory failure is an important complication with significant mortality implications.

Blood transfusions have been shown to increase the risk of respiratory complications after pneumonectomy [5, 6]. Restrictive transfusion practices have been increasingly used to reduce the risk of morbidity and mortality in various populations [7, 8]. Moreover, they appear to have demonstrated protective effects [7]. Restrictive transfusion practices gained more widespread acceptance after the publication of the TRICC trial in 1999, demonstrating that restrictive transfusion practices were as safe as liberal transfusion and could possibly reduce various complications and in-hospital mortality [9]. A previous study by Kidane et al. [10] in different settings demonstrated that perioperative transfusions appear to be an independent risk factor for ARDS and respiratory failure after pneumonectomy. Furthermore, packed red blood cells (pRBCs) transfusion showed a significant dose–response relationship with ARDS and respiratory failure [10]. In that study, it was unclear whether intraoperative and postoperative transfusions exert similar effects on the rate of postoperative respiratory failure.

Our primary objective was to determine if perioperative transfusions are associated with increased rates of postoperative respiratory failure in adults undergoing elective pneumonectomy in the era of restrictive transfusion practices. Our secondary objective was to assess whether intraoperative and postoperative transfusions exert similar effects on the rate of postoperative respiratory failure.

MATERIALS AND METHODS

After approval from the University Health Network Research Ethics Board, a retrospective cohort study was conducted using prospectively collected data. Data were supplemented with a review of anaesthetic records. The study population consisted of consecutive adult patients who underwent elective pneumonectomies between 2005 and 2015 at a single Canadian quaternary care centre with 7 thoracic surgeons. This was a retrospective cohort study.

Transfusion details

The clinical decisions of the surgeon and anaesthesiologist were based on the risk of blood loss, haemodynamics and minimization of intraoperative fluid overload to determine intraoperative transfusion. A procedure for pRBC transfusion outside the operating room has been established at our centre. In cases of asymptomatic anaemia, transfusion is used to keep haemoglobin over 70 g/l. Any haemoglobin level <100 g/l caused by symptomatic anaemia will prompt transfusion.

Outcomes

The outcome assessed was respiratory failure defined as a need for new mechanical or non-invasive positive pressure ventilation for ≥24 h.

Pneumonectomy practices

The 7 thoracic surgeons performing the pneumonectomies were consistent in conducting the preoperative work-up, operations, and postoperative care of the patients. Included in the preoperative work-up were quantitative ventilation-perfusion scans, pulmonary function tests, echocardiogram, and stress test if necessary. Patients received one dose of preoperative antibiotics. At the discretion of the anaesthesiologist most patients received an epidural. Posterolateral thoracotomy was used for all patients, and the bronchial stump was stapled in most cases, occasionally over-sewn, and then reinforced using pericardial fat pad tissue, the posterior pericardium or the intercostal muscle flap. One chest tube was used, which was routinely removed within the first 24 h postoperatively, with the exception of extrapleural pneumonectomy (EPP) for mesothelioma. Postoperative antibiotics were employed for 48–96 h. Goal-directed fluid management was not employed, while there was a restrictive approach to intraoperative fluids and transfusions. Total daily fluid intake (both enteral and parenteral) postoperatively was limited to ≤1500 ml. Preoperative radiation in our cohort was given for mesothelioma patients undergoing EPP, as part of an investigational protocol of hemithoracic intensity-modulated radiation therapy followed by EPP within 1 week of completing radiation [11]. The chemotherapy regimens used were predominantly platinum-based doublet therapies (mostly, cisplatin and vinorelbine or etoposide). No agents with pulmonary toxicity were used in this cohort of patients [12].

Statistical analysis

In our univariable analysis, Fisher’s exact test was employed for categorical data, independent t-test for continuous normally distributed data, and Mann–Whitney U-test was employed for continuous non-normally distributed data. Univariable analysis of risk factors was used to assess factors associated with respiratory failure. These factors included demographics (age, gender, body mass index, smoking status), comorbidities, fluid balance, preoperative chemotherapy or radiation, estimated blood loss (EBL). In our multivariable logistic regression, we included variables with P-value <0.2 on univariable analysis. Backward stepwise regression modelling was performed using likelihood ratios. Model fit was assessed using the likelihood ratio test and the Hosmer–Lemeshow test. Due to the concern about the differences in patients undergoing EPP and also undergoing pneumonectomy for non-cancer indications, a sensitivity analysis was performed excluding these 2 subgroups.

RESULTS

Table 1 describes the characteristics of the study population. Of the 251 pneumonectomies performed during the study period, the majority (n = 243, 96.8%) were for cancer. There were 8 pneumonectomies performed for the following non-cancer indications: 6 were performed for refractory/complicated non-tuberculous mycobacterial disease and 2 were performed as definitive management of recurrent haemoptysis due to bronchiectasis. There were 121 EPPs performed; our centre is a high-volume centre specializing in mesothelioma, thus 105 of the EPPs were for mesothelioma, 6 EPPs were for locally advanced thymoma, 5 EPPs were for non-small-cell lung cancer, 4 EPPs were for local control of chemo-resistant metastatic disease (the majority of which were sarcomas), and 1 EPP was for a locally invasive solitary fibrous tumour. Twenty-four (9.6%) patients suffered respiratory failure, 18 of whom required invasive mechanical ventilation. Figures 1 and 2 show chest radiographs of patients who developed respiratory failure after pneumonectomy and EPP, respectively. The incidence of in-hospital mortality was 4.8% (n = 12/251). Ninety-day mortality was 5.6% (n = 14) and was more likely in patients with respiratory failure (7/24 vs 7/227, P < 0.001). Intraoperative and postoperative transfusions occurred in 42.2% (n = 106) and 44.6% (n = 112) of patients, respectively and were predominantly with red blood cells rather than fresh frozen plasma or platelets. Although there was some overlap of patients who received intraoperative and postoperative transfusions, these were not the same patients; only 68 of the 106 people who received intraoperative transfusions went on to have postoperative transfusions. However, receipt of intraoperative transfusion was significantly associated with receipt of postoperative transfusion (P = 0.004). The EPP population was significantly more likely to get postoperative transfusions (P < 0.001) but not intraoperative transfusions (P = 0.12).

On univariable analysis, both intraoperative (P = 0.031) and postoperative transfusion (P = 0.004) were associated with a higher risk for respiratory failure (Table 2). The average amount of postoperative transfusion was 1.1 units (2.4 SD). There was a trend towards a significant correlation between the amount of transfusion and respiratory failure (P = 0.059); thus, this suggests a potential dose–response phenomenon. Nineteen percent (n = 21/112) of those with postoperative transfusions developed respiratory failure (P = 0.004) while 15% (n = 16/106) of those with intraoperative transfusions developed respiratory failure (P = 0.03). This appeared to have an effect on mortality such that 83% (n = 10/12) of those who died had postoperative transfusion (P = 0.01) whereas 58% (n = 7/12) had intraoperative transfusion (P = 1.0). The multivariable model significantly predicted respiratory failure with an AUC = 0.73 (P = 0.001) (Fig. 3). On multivariable analysis, the only independent predictors of respiratory failure were postoperative transfusions [adjusted odds ratio (aOR) 6.54, 95% confidence interval (CI) 1.74–24.59; P = 0.005] and lower preoperative forced expiratory volume (adjusted OR 0.96, 95% CI 0.93–0.99; P = 0.03). EBL was not significantly different (P = 0.91) between those with (median 800 ml, interquartile range 300–2000 ml) and without respiratory failure (median 800 ml, interquartile range 300–2000 ml) (Table 2). Patients who had higher EBL were likelier to receive both intraoperative (P < 0.001) and postoperative (P < 0.001) transfusions.

A sensitivity analysis was performed excluding those patients undergoing EPP and also those undergoing pneumonectomy for non-cancer indications. This sensitivity analysis replicated the findings of our primary analysis in that the only independent predictor associated with respiratory failure was the occurrence of postoperative transfusion. There were 7 patients who required reoperation for postoperative bleeding; of these 7 patients, 2 went on to develop respiratory failure. We performed a sensitivity analysis in which we ran our multivariable analysis excluding these 2 patients; this sensitivity analysis also replicated the findings of our primary analysis in that the only independent predictor associated with respiratory failure was the occurrence of postoperative transfusion.

DISCUSSION

Respiratory failure occurred in 9.6% of patients after pneumonectomy and conferred a higher risk of 90-day mortality. Postoperative (but not intraoperative) transfusion was the strongest independent predictor associated with respiratory failure. Due to the observational nature of this study, this is an association; causation cannot be determined. Preoperative forced expiratory volume was the only other factor independently associated with the occurrence of respiratory failure, with lower levels being associated with a higher risk of respiratory failure. EBL was not significantly different between those with and without respiratory failure, thus our findings were not likely due to confounding relationships between increased intraoperative bleeding complications, reactionary transfusions and respiratory failure. The incidence of post-pneumonectomy respiratory failure in this study was lower than that reported in Kidane et al.’s previous study in a different setting but was within the range reported in the literature [1, 2, 5, 6, 13, 14].

Postoperative transfusion appeared to be significantly associated with an increased incidence of post-pneumonectomy respiratory failure, but intraoperative transfusion did not. This may be due to more liberal, potentially modifiable post-pneumonectomy transfusion policies. Post-pneumonectomy patients are often placed under fluid restriction to prevent post-pneumonectomy oedema. When these fluid-restricted patients experience volume contraction, one strategy to restore volume without risking overload may be to transfuse patients with blood products as opposed to crystalloid volume replacement. However, this attempt to reduce the hydrostatic consequences of crystalloid infusion may inadvertently be increasing the inflammatory consequences of using blood products. Blood transfusions lead to an inflammatory state in recipients at an estimated frequency of 1 in 5000 [15]. It has been shown that RBC transfusions are associated with an immune response in patients already in an inflamed state [16], and in pre-injured endothelial cells exposed to blood products in vitro [15]. This may be due to the minute but non-trivial presence of leucocytes, cytokines and growth factors even in leucoreduced blood [17]. This postoperative blood transfusion may contribute to the risk of post-pneumonectomy respiratory failure. The post-pneumonectomy hydrostatic risk must be balanced with the inflammatory risk of blood transfusion.

Our findings support those of other studies regarding risk factors for respiratory failure after pneumonectomy. Several studies have identified the impact of perioperative blood transfusion on the incidence of post-pneumonectomy respiratory failure [2, 6]. Marret et al. [2] found that, in addition to perioperative blood transfusion, an ASA class ≥3 was a risk factor of respiratory complications. The other predictors of post-pneumonectomy respiratory failure found in the literature are various predictive comorbidities including coronary artery disease, advanced pathologic states, preoperative septic complication, pre-existing comorbidity as defined by the Charlson index of comorbidity ≥3 [18, 19]. Our findings of the association between lower preoperative forced expiratory volume and respiratory failure are also consistent with the literature [20]. The effect of transfusion on respiratory failure development may be so large that it dampens or nullifies the effect other prognostic variables usually exert; for example, our study did not identify a higher risk of respiratory failure associated with right pneumonectomy as has been identified in other studies [21]. The previous study by Kidane et al. found a post-pneumonectomy respiratory failure rate that was about 10% higher than that identified by this current study (19.23% vs 9.56%). There are several potential reasons for this, including different selection criteria between centres in pneumonectomy eligibility as well as potential differences in the patient populations. The previous study examined transfusions in a perioperative manner, while this study differentiated preoperative and postoperative transfusion. Despite these distinctions, both studies found that perioperative transfusion was significantly associated with post-pneumonectomy respiratory failure with a large effect size. The novel finding in this study is that it identified a key role for postoperative transfusion as opposed to and independent of intraoperative transfusion.

Limitations

This study has limitations. It is a retrospective study and is subject to selection and reporting biases. While it is possible that intraoperative complications including bleeding may predispose a patient to both respiratory failure and postoperative blood transfusions, we found that EBL was not significantly different between those with and without respiratory failure (P = 0.91). Thus our findings are not likely due to confounding relationships between intraoperative bleeding complications, reactionary transfusions, and respiratory failure. It is not known what proportion of these transfusions would have been deemed truly necessary. The postoperative setting is less likely to result in massive exsanguination requiring liberal transfusions and, in general, allows for clear criteria for transfusion compared to the intraoperative setting. Kidane et al.’s previous study compared the first years to the final years of the study period to adjust for the effect of changing secular trends in improved medical/surgical care. This also had the effect of adjusting for the changes in transfusion practices over time. Our current study does the same. Prospectively collected tidal volumes and peak airway pressures would have allowed us to adjust for the effects of intraoperative volutrauma/barotrauma on the development of respiratory failure; however, this data was inconsistently collected (i.e. varying intervals) in our anaesthetic records and thus limits our conclusions and inferences about the true nature of these variables. That being said, the anaesthesiologists at our centre tend to share a consistent philosophy of minimizing airway pressures and tidal volumes during pneumonectomies. Future studies should ensure the collection of this data in real time to minimize confounding effects.

CONCLUSION

In conclusion, our findings suggest that postoperative pRBC transfusion is an independent risk factor for post-pneumonectomy respiratory failure, while intraoperative transfusion is not. Intraoperative transfusion may be in reaction to active/unpredictable blood loss and may not be easily modifiable. However, postoperative transfusion may be modifiable and potentially avoidable. Although the differing clinical effect of intraoperative versus postoperative transfusion on post-pneumonectomy respiratory failure is not completely understood, our results suggest that postoperative transfusion may be an independent driver of respiratory failure in this population and can potentially be a target for quality improvement initiatives. Regardless, perioperative pRBC transfusion remains an independent risk factor even in the era of leucoreduction and restrictive transfusion practices and may be modifiable. Of note, almost half of the patients received a transfusion. Future studies should prospectively investigate the trade-offs between the deleterious effects of un-transfused anaemia and the increased risk associated with transfusion, as well as the trade-off between hydrostatic and inflammatory risk after pneumonectomy. For now, our findings support a continued push to minimize pRBC use during the perioperative care of pneumonectomy patients, especially in the postoperative setting.

Conflict of interest: none declared.

Author contributions

Biniam Kidane: Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Project administration; Writing—original draft; Writing—review & editing. Nithin Jacob: Data curation; Funding acquisition; Resources; Writing—review & editing; data collection. Allan Bruinooge: Methodology; Writing—original draft; Writing—review & editing. Yu Cindy Shen: Data curation; Methodology; Validation; Writing—original draft; Writing—review & editing; data collection. Shaf Keshavjee: Funding acquisition; Investigation; Resources; Supervision; Writing—review & editing. Marc E. dePerrot: Data curation; Funding acquisition; Resources; Supervision; Writing –review & editing. Andrew F. Pierre: Data curation; Funding acquisition; Resources; Supervision; Writing—review & editing. Kazuhiro Yasufuku: Data curation; Funding acquisition; Resources; Supervision; Writing—review & editing. Marcelo Cypel: Data curation; Funding acquisition; Resources; Supervision; Writing—review & editing. Thomas K. Waddell: Data curation; Funding acquisition; Resources; Supervision; Writing—review & editing. Gail E. Darling: Data curation; Funding acquisition; Resources; Supervision; Writing—review & editing.

Presented at the 25th Annual Meeting of the European Conference on General Thoracic Surgery, Innsbruck, Austria, 28–31 May 2017.

REFERENCES

Abbreviations

 
• CI

  Confidence interval

 
• EBL

  Estimated blood loss

 
• EPP

  Extrapleural pneumonectomy

 
• OR

  Odds ratio