| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the Unit of Thoracic Surgery, "Umberto I°" Regional Hospital, Ancona, Italy.
Correspondence to: Alessandro Brunelli, MD, Via S. Margherita 23, 60129 Ancona, Italy; e-mail: alexit_2000{at}yahoo.com
Abstract
Purpose: To assess whether the presence and duration of air leaks after lobectomy are associated with an increased incidence of cardiopulmonary complications.
Methods: Propensity score analysis was used on 726 patients undergoing pulmonary lobectomy from 1995 through 2004 to form three well-matched pairs of patients: patients with prolonged air leak (PAL) [> 7 days] and without air leak; patients with short air leak (SAL) [
7 days] and without air leak; and patients with SAL and PAL. These matched groups were then compared to assess postoperative hospital stay and early outcome.
Results: Patients with SAL had a longer postoperative hospital stay compared to patients without air leak (8.6 days vs 7.8 days, respectively; p < 0.0001) but had similar morbidity and mortality. Patients with PAL had a longer postoperative hospital stay compared to patients without air leak (16.2 days vs 8.3 days, respectively; p < 0.0001) and with SAL (16.9 days vs 9 days, respectively; p < 0.0001), but similar cardiopulmonary complications were noted between the groups. Patients with PAL had a higher rate of empyema compared to patients without air leak and with SAL (8.2% vs 0%, p = 0.01 and 10.4% vs.1.1%, p = 0.01, respectively).
Conclusions: The presence of air leak was not associated with an increased incidence of cardiopulmonary morbidity but was associated with an increased risk of empyema. Future prospective studies are needed to confirm safety of fast track in patients with air leak.
Key Words: lobectomy • lung cancer • morbidity • prolonged air leak • propensity matching
Air leak after lung resection is still a frequent occurrence that can prolong hospital stay123 and increase hospital costs.1 For this reason, a fast-track policy in patients with air leak (ie, early discharge with a Heimlich valve unless otherwise indicated) seems a logical approach. However, some authors124 have reported an increased incidence of postoperative morbidity in those patients with air leak, questioning the safety and cost-effectiveness of this policy. The objective of this study was to verify whether the occurrence and duration of air leaks after pulmonary lobectomy are associated with an increased incidence of cardiopulmonary morbidity in propensity case-matched groups of patients.
Materials and Methods
Seven hundred twenty-six patients (141 women and 585 men) underwent pulmonary lobectomy or bilobectomy for non-small cell lung cancer at our institution from January 1995 through December 2004 and were taken into consideration for the present study. Since the study was conceived to assess morbidity in association with air leak after lobectomy, patients who died in the postoperative period (n = 21) were included in the analysis irrespective of the duration of their air leak.
This is a retrospective analysis performed using an electronic, single-center, audited database in which data are entered at the time of patient discharge. The local institutional review board approved this work, and informed consent was obtained from all patients to enter their data in the database for clinical and investigational purposes.
Patients were excluded from operation if they had a predicted postoperative FEV1 of < 30%5 in association with a poor cardiopulmonary reserve (maximum oxygen consumption < 10 mL/kg/min or height reached at stair-climbing test < 12 m).6 The same surgical team performed all the procedures through a muscle sparing lateral thoracotomy whenever possible. The following operations were performed in order of frequency: right upper lobectomy (n = 203), left upper lobectomy (n = 196), left lower lobectomy (n = 124), right lower lobectomy (n = 90), right lower bilobectomy (n = 45), middle lobectomy (n = 37), and right upper bilobectomy (n = 31). Mechanical staplers were used to develop incomplete fissures in 80% of cases, and to close the bronchus in all patients. Twenty percent of patients had completely developed or filmy fissures that did not require the use of staplers. After completion of the lobectomy, a mediastinal lymphadenectomy was performed in all patients. Moreover, after re-inflation of the lung, air leaks were pinpointed by squirting sterile water over the lung and sutured prior to chest closure. Buttressed staple lines or chemical sealants were occasionally used in this series. A total of 210 upper lobectomies or bilobectomies in the present data set were performed using a pleural tent procedure.7 Two chest tubes were positioned before the closure of the thoracotomy, one anteriorly into the apex and one in a posterior-inferior position. In most of the patients, the chest tubes were placed on suction (– 20 cm H2O) and then converted to water seal when minimal or no air leak was evident, or in case of a persistent air leak as an attempt to stop it. From June 2001, some of the patients were enrolled in prospective trials89 of different chest tube managements (water seal, suction, or alternate suction). These different air leak preventive measures (pleural tent and chest tube modalities) were accounted for in the propensity score models. Chest tubes were removed when the quantity of the effusion was < 200 to 300 mL in 24 h and when no evidence of air leak was present (after a 24-h clamping trial). This chest tube management was applied to all patients in this series, and it is the standard policy at our institution. During the postoperative period, chest physiotherapy and incentive spirometry were administered to all patients, in addition to bronchodilators if needed. In patients with air leak, oral antibiotics were continued until the air leak stopped.
For the purpose of this study, an air leak that persisted 7 days was termed short air leak (SAL), and an air leak lasting > 7 days was termed prolonged air leak (PAL). Sporadic latent air leaks occurred (all evident by the second postoperative day); however, for the purpose of this study, in these patients the air leak duration was calculated from the first postoperative day. A standardized quantitative analysis of the air leak was not possible, as during the period of the study not all the drainage apparatuses utilized had a leak meter.
The presence of an air leak was checked twice daily during the morning and evening rounds. Chest radiographs were not routinely performed and were obtained only when clinically indicated (reduced breath sounds at auscultation, increased sputum production, fever and leukocytosis, reduced oxygen saturation, a suspicion of chest tube malfunctioning).
Statistical Analysis
The sample sizes of the matched groups were adequate to obtain a statistical power of at least 0.8, with a significance level of 0.05 to detect an odds ratio of 2.85 for morbidity in air leak patients, in accordance with previous findings.1 In the analysis of the causal effect of treatment or risk factor, the key assumption is that the receipt of the treatment is independent of the potential outcomes with and without treatment if certain observable covariates are held constant. When this fundamental assumption is violated in empirical situation, researchers should take measures to remove bias in estimating the treatment effect. Propensity score matching is such an attempt to correct the violation of the assumption by matching subjects on conditional probability of receiving treatment or being exposed to a given risk factor, and through directly modeling the process of assignment of treatment conditions to study subjects. In sum, it models the heterogeneity of causal effects of treatment.10
We performed three propensity score matching procedures to match the following groups of patients1011: (1) patients with SAL (lasting 7 days) and patients without air leak; (2) patients with SAL and patients with PAL (lasting > 7 days); and (3) patients with PAL and patients without air leak. Before matching patients, a parsimonious explanatory model was developed by logistic regression analysis and bootstrap bagging.12 Bootstrap analysis was used to assess reliability of the variables included in the model. In the bootstrap procedure, 1,000 samples of the same number of observation as the original data set were selected with replacement from the original set of patients. In each of these samples, logistic regression analysis was performed and the stability of the final model can be assessed by identifying the variables that enter most frequently in the repeated bootstrap models and comparing those variables with the variables in the final model. If the final model variables occur in a majority (> 50%) of the bootstrap models, the original regression model can be judged to be stable.11 The probabilities of SAL or PAL developing after lobectomy (propensity score) were estimated by logistic regression analysis incorporating the variables identified in the parsimonious models plus additional baseline and operative variables. Therefore, the variables used in the propensity models were the following: age; gender; body mass index; FEV1; carbon monoxide diffusion capacity (DLCO); FEV1/FVC ratio; diabetes; cardiac comorbidity; side and site (upper vs lower) of resection; smoking history (pack-years); neoadjuvant chemotherapy; percentage of functioning lung parenchyma removed during operation (percentage of functional loss); presence of pleural adhesions; presence of pleural tent; chest tube modalities (water seal, suction, alternate suction); length of stapled parenchyma (millimeters); length of operation (minutes); and period of operation (early, before January 2000; late, from January 2000). An explanation of variables is given in the Appendix.
The risk of "overmatching" may be inherent to all matching approaches. In fact, the variables involved in the matching may be involved in or closely connected with the mechanism whereby the independent variable affects the outcome. To check for this phenomenon, we assessed the individual discrimination of the propensity models for prediction of air leak and cardiopulmonary complications. The c-index values of the propensity models for predicting the presence of air leak were all > 0.75. When the same models were applied to predict cardiopulmonary complications, their discriminations were poor (all < 0.65). Furthermore, only few variables were common significant risk factors for air leak and complications (age and DLCO). The propensity score models had the following discrimination and calibration parameters, as assessed by the c-index and the Hosmer-Lemeshow goodness of fit statistics: model A, c-index 0.76, Hosmer-Lemeshow 6.7 (p = 0.6); model B, c-index 0.801, Hosmer-Lemeshow 5.1 (p = 0.7); model C, c-index 0.799, Hosmer-Lemeshow 7.7 (p = 0.5).
The method of generating a parsimonious model and then augmenting it with other factors to develop the propensity model was described elsewhere.11 As stated by its developers,13 the propensity model is not parsimonious. In fact, the goal is to balance patient characteristics by incorporating "everything" recorded that may relate to either systematic bias or simply bad luck that has otherwise unbalanced the comparison groups of interest, ignoring usual concerns about model overdetermination.11 All variables were at least 95% complete and sporadic missing values were imputed by taking the most frequent response category or averaging nonmissing values for continuous variables. The matching technique included a nearest neighbor 1-to-1 matching with no replacement and with caliper. A caliper size of one fourth of SD of propensity score was used for each matching process.14 The procedures yielded the following well-matched pairs: 212 pairs with SAL and without air leak; 85 pairs with PAL and without air leak; 96 pairs with PAL and with SAL. The respective pairs of propensity score matched patients were then compared by using the McNemar test for matched pairs for categorical variables and the Wilcoxon matched-pairs signed-rank test for continuous variables. The Bonferroni correction for multiple comparisons was used. Outcome variables were the following: length of postoperative hospital stay, in-hospital or 30-day cardiopulmonary complications and mortality. An explanation of variables is given in the Appendix. All tests were two tailed with a significance level of 0.05 and were performed using statistical software (Stata 8.2; StataCorp; College Station, TX).
Results
Four hundred seventeen patients had an air leak present on the first postoperative day (57.4% of patients). One hundred twenty-eight (17.6% of patients) had PAL (> 7 days). Cardiopulmonary complications developed in 143 patients (19.7%), and the mortality rate was 2.9% (n = 21). The median air leak duration in patients with SAL was 2 days (range, 1 to 7 days). The median air leak duration in the patients with PAL was 14 days (range, 8 to 62 days).
All patients with air leak were treated conservatively with the exception of one patient who required reoperation for suture of a parenchymal injury caused by a fractured rib. Propensity score analyses yielded well-matched pairs. The Appendix provides results of comparison of baseline and operative characteristics between groups.
Table 1 shows the results of the comparison between propensity-matched patients with SAL and without air leak. Although the postoperative hospital stay was longer in the patients with SAL (p < 0.0001), no differences were noted in terms of morbidity and mortality rates between the groups.
|
|
|
The objective of this study was to assess whether the occurrence and the duration of air leaks after pulmonary lobectomy were associated with an increased incidence of cardiopulmonary complications, which may question the safety and appropriateness of an early discharge of these patients. Propensity score case-matched analysis was used to adjust for confounding background characteristics (common risk factors of air leak and cardiopulmonary complications). This form of analysis is considered the most rigorous method available for the so-called "apples-to-apples" investigation of causal effects on outcome in a nonrandomized setting.11
We found that both SAL (lasting 7 days) and PAL (> 7 days) prolonged the postoperative hospital stay, confirming previous reports.123 However, SAL and PAL were not associated with an increased incidence of cardiopulmonary complications or mortality compared to patients without air leak. Likewise, patients with PAL had a longer hospital stay compared to patients with SAL but did not have an increased rate of morbidity. However, patients with PAL had an increased rate of empyema (8 to 10%). This may warrant an audit analysis on the type and schedule of antibiotic prophylaxis in these patients.
In this series, patients with a PAL stayed in the hospital approximately 8 days longer than patients without air leak. This finding underlines the need of a fast-tracking policy to shorten hospital stay and reduce hospital costs.15 The lack of association between air leak and cardiopulmonary morbidity are at variance with the results reported in previous studies14 in which, however, no balancing scores were used to match patients.
In a recent elegant study analyzing propensity-matched groups of patients with and without air leak, Okereke and Coll2 concluded that any air leak, irrespective of its duration, should be regarded as a surgical complication, for it increases the risk of cardiopulmonary morbidity. Differences in the type of surgical population, chest tube management, use of pleural tent in our upper lobectomy patients, and in the definition and recording of complications (always an important matter when dealing with morbidity among different centers) may explain the discrepancy between the work of Okereke and Coll2 and our work.
Our study has potential limitations. First, it is a retrospective analysis from a single center. This is not a randomized trial, and even though the propensity score analysis constitutes the most rigorous method for investigating the causal effects in this setting, it cannot account for unknown variables affecting outcome that are not correlated strongly with measured variables. Moreover, "overmatching" is always a risk in clinical studies using matching approaches. In fact, variables involved in the matching may be involved in or closely connected with the outcome or strongly correlated with known or unknown factors affecting the outcome. Therefore, the results generated by this analysis should be cautiously interpreted under this perspective.
Second, a quantitative study of the magnitude of air leak and its possible association with duration and complications was not possible.15 Finally, during the period of the study some intraoperative (pleural tenting in upper lobectomies) and postoperative measures (water seal or alternate suction chest tube modality) were introduced that could have influenced the rate of occurrence of PAL. However, these factors were accounted for in the construction of propensity score models, which yielded well-matched pairs of patients.
In conclusion, at variance with previous reports,124 we found that the occurrence of an air leak after lobectomy, irrespective of its duration, was not associated with an increased incidence of cardiopulmonary morbidity. Therefore, based on these results, we think that a fast-tracking policy (ie, early discharge with a Heimlich valve) for these patients may be appropriate and safe in order to reduce hospital costs. However, particularly in light of the increased rate of empyema observed in these patients, this assumption needs to be further verified by future independent prospective follow-up studies.
Appendix
Baseline and Operative Variables
For the purpose of this study, the following spirometric variables were considered: FEV1, DLCO, and FEV1/FVC ratio. Pulmonary function tests were performed according to American Thoracic Society criteria. DLCO was measured using the single-breath method. Results of spirometry were collected after bronchodilator administration and were expressed as percentage of predicted for age, sex, and height (with the exception of FEV1/FVC ratio).
The number of functioning segments removed during operation (percentage of functional loss) was estimated by means of CT scan and bronchoscopy findings. In patients with a calculated predicted postoperative FEV1 < 50%, a quantitative perfusion lung scan was used, according to Markos et al.5
A concomitant cardiac disease was defined as follows: previous cardiac surgery, previous myocardial infarction, history of coronary artery disease, current treatment for hypertension, arrhythmia, or cardiac failure. We computed the number of pack-years of smoking as the total number of years smoked, times the average number of cigarettes smoked per day, divided by 20.
Other variables included in the propensity score analysis as baseline characteristics were the following: age, gender, diabetes, and neoadjuvant chemotherapy. Operative variables included the followings: side and site of resection (upper or lower), presence of pleural adhesions (only dense pleural adhesions occupying > 30% of a lobe or more than one lobe were taken into consideration for the analysis), length of stapled parenchyma (millimeter), length of operation (minute), period of operation (early, before January 2000; late, from January 2000), presence of pleural tent, chest tube modality (water seal, – 20 cm H2O suction, – 10 cm H2O alternate suction). Results of the comparison of baseline and operative characteristics between the propensity case-matched groups analyzed are shown in Tables 456 .
|
|
|
Pulmonary Complications: Respiratory failure requiring mechanical ventilation for > 48 h; pneumonia (as defined by chest radiographic findings, increased WBC count, and fever); atelectasis requiring bronchoscopy; pulmonary edema; pulmonary embolism.
Cardiovascular Complications: Myocardial infarction (as defined by cardiac enzymes increase and suggestive ECG findings); hemodynamically unstable arrhythmia requiring medical treatment; cardiac failure (defined by suggestive symptoms, physical examination and chest radiographic findings). Any patient with one or more of the above-mentioned complications was regarded as having a cardiopulmonary complication.
Footnotes
Abbreviations: DLCO = carbon monoxide diffusion capacity; PAL = prolonged air leak; SAL = short air leak
All authors have no conflicts of interest to disclose.
Received for publication December 13, 2005. Accepted for publication March 28, 2006.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
