Early respiratory acidosis is a new risk factor for pneumonia after lung resection

Abstract

Postoperative pneumonia (POP) is a life-threatening complication of lung resection (LR). Its risk factors, bacteriological profile and outcome are not well known. The aims of this study were to describe the outcome and causal bacteria and to identify risk factors for POP. We reviewed all cases admitted to intensive care after LR. Clinical parameters, operative and postoperative data were recorded. POP was suspected on the basis of fever, radiographic infiltrate, and either leucocytosis or purulent sputum. The diagnosis was confirmed by culture of a respiratory sample. Risk factors for POP were identified by univariate and multivariate analysis. We included 159 patients in this study. POP was diagnosed in 23 patients (14.4%) and was associated with a higher hospital mortality rate (30% versus 5%, P = 0.0007) and a longer hospital stay. Members of the Enterobacteriaceae and Pseudomonas species were the most frequently identified pathogens. Early respiratory acidosis (ERA; OR, 2.94; 95% CI, 1.1–8.1), blood transfusion (OR, 3.8; 95% CI, 1.1–13.1), bilobectomy (OR, 7.26; 95% CI, 1.2–43.1) and smoking history (OR, 1.84; 95% CI, 1.1–3) were identified as independent risk factors. ERA may be a risk factor for POP and could serve as a target for therapeutic interventions.

INTRODUCTION

Lung resection (LR) is the first-line treatment for early non-small cell lung cancers. Postoperative pneumonia (POP) is a frequent complication of LR [1]. It is reported in 3.3–25% of patients after LR and is the leading cause of postoperative mortality [1–3]. POP is associated with a 5.4 % increased chance of postoperative mortality [3, 4]. In previous studies, the diagnosis was not systematically confirmed by microbiological evaluations of bronchial secretions. Only a few studies have described the causal pathogens of POP [1, 5, 6]. Known risk factors for POP include being older, preoperative lung function, cardiovascular comorbidity, preoperative chemotherapy, type of surgery and smoking status [6–11]. However, these risk factors cannot be targeted by specific therapeutic interventions.

The objectives of this study were: (i) to describe the prevalence and clinical impact of bacteriologically confirmed POP after LR; (ii) to identify the causal bacteria; and (iii) to identify preoperative and postoperative risk factors for POP.

METHODS

Patients

We reviewed all consecutive patients’ charts undergoing thoracic surgery from January 2000 to October 2005 at a single centre who were subsequently admitted to the respiratory intensive care unit (ICU). We included only patients scheduled for LR in this study.

Operability was determined according to the existing guidelines [12]. Admission to the ICU after surgery was decided at a weekly multi-specialty meeting on the basis of associated comorbidities, preoperative lung function and type of resection. Antibiotic prophylaxis was administered in accordance with national guidelines (cefazolin or cefamandole, vancomycin in case of β-lactam allergy). Patients were extubated before ICU admission and were examined twice daily. Pain was managed with systemic opioids released by a patient-controlled administration system. Epidural analgesia or intercostal blocks was not routinely performed. Physiotherapy was started twice-daily from the day after surgery. All patients were maintained in a semi-recumbent position and were confined to a chair on the first day after surgery. They were allowed to eat normally on the first day after surgery. The chest tube was removed if there was an absence of bubbling and excreta was less than 150 ml/day. Blood-gas analysis was performed at least once daily during the first two days after surgery. Chest X-ray was examined daily.

Data collection

Demographic data were age, sex, body mass index, preoperative American Society of Anaesthesiologists (ASA) physical status, smoking status (packs per day and timing of cessation), preoperative treatments, haemoglobin, serum creatinine and lung function test results (forced expiratory volume in 1 s (FEV₁), forced vital capacity, blood-gas analysis). Comorbid disease was defined as diabetes mellitus, congestive heart failure and immunosuppression.

Surgical data collected were indication, type (pneumonectomy, lobectomy, bilobectomy or wedge resection) and duration of surgery, resection side, node picking, type of antibiotic prophylaxis, total number of red blood cell units (RBCUs) transfused.

Postoperative evaluation extended from admission to the ICU to discharge from the hospital. ERA was defined as a pH < 7.38 and PaCO₂ > 42 mmHg for the first arterial blood gas recorded during the first 2 days after surgery. If POP occurred during the first 2 days after surgery, we collected only blood gas values preceding the onset of POP. Microbiological data were recorded.

Definition of POP

POP was suspected on the basis of: (i) temperature > 38°C or <36°C, (ii) new infiltrate or change in radiographic infiltrate, and (iii) one of the following criteria: purulent sputum or leucocytosis > 10⁹/l or <2.5 × 10⁹/l. In cases of suspected POP, a microbiological sample was obtained by bronchoalveolar lavage (BAL), protected specimen brush (PSB) sampling, bronchoscopic tracheal aspiration or sputum culture (SC). SCs were carried out for patients with respiratory failure for whom bronchoscopy was judged harmful. Bronchoscopy (with PSB or BAL) was performed for other patients. POP was confirmed in cases of positive culture of a bronchial specimen exceeding the following cutoff values: 10⁴ CFU/ml for BAL, 10³ CFU/ml for PSB. Sputum and tracheal aspirate culture results were considered positive if culture yielded a single microorganism.

Statistical analysis

Data are summarized as means ± SD or medians (25th–75th percentiles) for continuous variables and as percentages for categorical variables. We used chi-squared and Fisher's exact tests to compare categorical variables between patients with and without POP. We used Student's t-tests or Mann–Whitney tests to compare continuous variables. Logistic regression analysis was performed to assess associations between the occurrence of POP and preoperative, surgical or early postoperative factors preceding POP. The results are expressed as odds ratios and associated 95% confidence intervals. Variables significantly associated with POP at the 0.15 level in univariate analysis were considered in a multivariate backward analysis. To prevent potential overfitting, we used Firth's penalized maximum likelihood estimation where parameter shrinkage is performed during the fitting of the model (Steyerberg EW. Clinical prediction models: a practical approach to development, validation, and updating. New York). Two-sided P-values < 0.05 were considered to be significant. All analyses were performed with SAS v.9.2 software (SAS Institute Inc., Cary, NC, USA).

RESULTS

Study population

During the study period, 292 consecutive patients were admitted to our ICU after thoracic surgery; 133 of these patients did not meet the inclusion criteria. The remaining 159 patients constituted the study population (Fig. 1). In 109 patients (79%), antibiotic prophylaxis was administered once during surgical procedure and consisted of cefazolin or cefamandole. Eighteen patients received an alternative antibiotic regimen because they were allergic to β-lactams.

Figure 1:

Flow diagram of the study population selection process. Asterisk denotes surgical chest or pericardial tube (n = 51), other surgical procedures (videothoracic surgery, mediastinal surgery, sternotomy; n = 49), thoracic traumatism (n = 13), phrenic pacemaker insertion (n = 5), interventional bronchoscopy (n = 15).

Postoperative pneumonia

POP occurred in 23 patients during the study period (14.4%) at a median of 3.5 days (1–28 days) after surgery.

The main characteristics of the patients with and without POP are presented in Tables 1 and 2. The 23 patients with POP were more likely to be heavy smokers, to have undergone bilobectomy, to have received blood transfusions during surgical procedures and to have ERA. A history of chronic heart failure was also more frequent among patients with POP. Conversely, lung function tests did not differ significantly between the two groups (Table 1).

Table 1:

Demographic characteristics and preoperative pulmonary function in patients with and without POP

	Missing data	All patients (n = 159)	Without postoperative pneumonia (n = 136)	With postoperative pneumonia (n = 23)	P-value
Demographic characteristics
Age (years ± SD)	0	59 ± 14	58 ± 14	63 ± 13	0.11
Male (n, %)	0	101 (63%)	83 (61%)	18 (78%)	0.12
Body mass index (kg/m2 ± SD)	1	25 ± 13	25 ± 13	22 ± 4	0.18
Diabetes mellitus (n, %)	1	21 (13%)	16 (10%)	5 (22%)	0.2
Congestive heart failure (n, %)	1	23 (14%)	19 (12%)	4 (17%)	0.04
Immunosuppression (n, %)	1	23 (14%)	19 (14%)	4 (17%)	0.67
Preoperative chemotherapy (n, %)	2	39 (31.5%)	23 (22.1%)	6 (30%)	0.3
Cumulative smoking (packs/year ± SD)	0	30 ± 24.3	27 ± 23	43 ± 27	<0.01
Smoking in the 60 days before surgery (n, %)	0	54 (34%)	43 (32%)	17 (74%)	0.07
Preoperative lung function
FEV1, ml ± SD (% predicted)	15	2331 ± 631 (82%)	2353 ± 675 (82%)	2216 ± 714 (77%)	0.26
FEV1/FVC (mean ± SD)	25	72 ± 11	72 ± 11	71 ± 12	0.84
PaO2 (mmHg ± SD)	33	84 ± 14	81 ± 15	82 ± 11	0.16
PaCO2 (mmHg ± SD)	33	38 ± 4	38 ± 4	39 ± 4	0.42

FEV₁: forced expiratory volume in 1 s; FVC: forced vital capacity; SD: standard deviation.

Table 2:

Surgical data for patients with and without postoperative pneumonia

	Missing data	All patients (n = 159)	Without postoperative pneumonia (n = 136)	With postoperative pneumonia (n = 23)	P
Surgical and perioperative data
Intervention
Pneumonectomy (n, %)	0	39 (24.5%)	32 (30.7%)	7 (30.4%)	0.03
Lobectomy (n, %)	0	77 (78.4%)	65 (62.5%)	12 (52.2%)
Bilobectomy (n, %)	0	7 (5.7%)	4 (3.8%)	3 (13%)
Wedge resection (n, %)	0	36 (22.6%)	35 (17.3%)	1 (4.4%)
Carcinological resection (n, %)	0	124 (78%)	104 (76.4%)	20 (87%)	0.26
Left side resection (n, %)	0	86 (54.1%)	71 (52.2%)	15 (65.2%)	0.25
Surgery duration (min; mean ± SD)	2	209 ± 101	203 ± 105	204 ± 70	0.94
Transfusion (n, %)	0	19 (12%)	13 (10%)	6 (26%)	0.03
Postoperative data
Chest tube duration (days, mean ± SD)	5	6.1 ± 4.5	5.8 ± 3.9	8.7 ± 7.6	0.33
Early respiratory acidosis (n, %)a	29	48 (30%)	36 (26%)	12 (52%)	0.065
pH (mean ± SD)	29	7.4 ± 0.1	7.4 ± 0.1	7.3 ± 0.1	<0.002
PaCO2 (mmHg, mean ± SD)	29	44 ± 8.7	43.5 ± 8.9	46.4 ± 7.8	0.17
HCO3− (mmHg, mean ± SD)	29	25.1 ± 2.6	25.2 ± 2.5	24.7 ± 3	0.34

^aEarly respiratory acidosis was defined as pH < 7.38 and PaCO₂ > 42 mmHg for at least the first postoperative arterial blood-gas sample recorded during the first 2 days after surgery. SD: standard deviation.

We isolated 29 different bacteria: a single species from each of 19 patients and several species from four. Gram-negative bacilli were the most frequent pathogens identified (21 of the 29 bacterial isolates). Species from the Enterobacteriaceae (n = 10), Pseudomonas aeruginosa (n = 7) and Haemophilus influenzae (n = 4) were isolated. Streptococcus species (n = 3) and Staphylococcus aureus (n = 2) were recovered in four episodes. No difference was found between bacteria in the time interval between surgery and POP.

Risk factors for POP

The results of multivariate analysis are given in Table 3. Preoperative smoking status was identified as a risk factor for the development of POP. Perioperative RBCU transfusion and bilobectomy were also associated with POP. Postoperative ERA was identified as independently associated with POP. Results were consistent when replacing ERA defined as a binary event by its two defining continuous variables: pH value (OR of 4.6 for every 0.1 point decrease; 95% CI, 1.9–10.6; P < 0.001) or PaCO₂ (OR of 1.9 for every 10 mmHg increase; 95% CI, 1.1–3.3; P = 0.027).

Table 3:

Multivariate analysis for POP risk factors (analysis carried out for 130 patients)

Variable	Odds ratio	95% CI	P-value
Cumulative smoking (+25 PY)a	1.84	1.12–3.04	0.017
RCB transfusion	3.79	1.1–13.1	0.035
Early respiratory acidosis	2.94	1.06–8.14	0.038
Bilobectomy resection	7.26	1.2–43.1	0.029

^aOdds ratio associated with an increase of 25 packs/year (PY); RCB: red cells blood.

Hospitalization outcome

The median duration of hospital stay after surgery was 11 days (3–45 days) for patients without POP and 23.5 days (8–155 days) for patients with POP (P < 0.001). Fourteen deaths (in-hospital postoperative mortality rate of 8.8%) occurred during the study period. Seven patients without POP (4.4%) and seven patients with POP (30.4%; P < 0.007) died in hospital. The leading cause of death was acute respiratory distress syndrome (ARDS; 10 cases). Six of the seven deaths in the POP group were caused by ARDS.

DISCUSSION

We found that POP was frequent and the leading cause of death after LR. ERA, cumulative smoking history, bilobectomy and red blood cell transfusion during surgery were all independently associated with POP.

The 14.4% prevalence of POP in this study is within the range of values (3.3–25%) reported in previous studies [1, 2]. Consistent with previous reports, POP is associated with a longer hospital stay [1, 6].

As previously reported, Gram-negative bacilli were the leading group of pathogens causing POP, especially Enterobacteriaceae, and most of them had innate antibiotic susceptibility (data not shown). The high prevalence of Enterobacteriaceae suggests that POP may be mediated by inhalation. Most of the causal bacteria identified were saprophytic bacteria of the oral cavity and upper airways. Our results confirm the high prevalence of P. aeruginosa which is not well understood. The value of active antibiotic prophylaxis against Pseudomonas remains unclear.

Our results are not consistent with older age, male sex, cardiopulmonary comorbidity, previous chemotherapy or radiotherapy or altered lung function tests being risk factors for POP. A large cohort study, including more than 4000 patients, recently showed that older age and male sex were risk factors for death after LR, but POP was not reported [3].

Lung function test results were not identified as a risk factor for POP. Previous studies resulted in various conclusions regarding the role of preoperative lung function tests as predictors of postoperative outcome. Some studies have shown that an altered preoperative FEV₁ is linked to higher postoperative morbidity and mortality rates, but this was not found to be the case in several other studies [1, 5].

Bilobectomy, current smoking and red blood cell transfusion during surgery were identified as independent risk factors for POP in this and previous studies [3, 5, 13, 14]. Current smokers have been found to have a 5.5 times higher risk of developing postoperative pulmonary complications than patients who have never smoked [13]. Abstinence from smoking decreases blood carboxyhaemoglobin concentration and bronchial inflammation, hyper-responsiveness and secretion. Intraoperative blood loss has been shown to be associated with infectious complications after LR [14]. Tissue hypoxia, induced by hypovolaemia and anaemia, may limit tissue repair. Transfusion-related immunomodulation may promote POP.

We identified ERA as an independent risk factor for POP in the multivariate analysis. ERA, measured during the first 2 days after surgery and before the diagnosis of POP, was twice as frequent in patients who went on to develop POP than in those who did not develop POP. Respiratory acidosis may be related to postoperative analgesia with systemic opioids, mostly delivered by patient-controlled intravenous pump systems. Respiratory acidosis can also be induced by prolonged pleural drainage because of induced pain. Chest tubes were left in place for a longer period in patients with POP than in patients without POP, but this difference was not significant. If the role of ERA as a triggering factor for POP is confirmed in other settings, an evaluation of early non-invasive ventilation (NIV) for the prevention of POP could be carried out. NIV seems to be well-tolerated and could decrease length of hospital stay [15].

Our study has several limitations. First, we observed a relatively small number of events (23 POPs). As the minimum number of events per predictor required for obtaining reliable predictions may be 10, we cannot exclude that our multivariate model is overfitted. However, we used a newer approach to statistical modelling that is less prone to overfitting. Second, there were some missing data due to the retrospective design of the study. Blood-gas analysis was not performed systematically during postoperative stay, so this information was lacking for 33 patients. In multivariate analysis, we analysed patients with complete predictor and outcome values only. It led to discarding of information for 19 patients (one had POP) and to a decrease in precision. Third, we limited our analysis to patients referred to the ICU by the surgical team, and this may account for the high mortality rate and the relatively high proportion of pneumonectomy cases in our series.

In conclusion, our data confirm that POP is a frequent and severe complication of LR and prolonged hospital stay. The results also confirm that current smoking, bilobectomy and preoperative blood transfusion are independent risk factors for POP and suggest that postoperative ERA is a risk factor for POP. These results may help to identify patients at high risk of POP, for whom early postoperative interventions, such as NIV or broader prophylactic antibiotic regimens may be evaluated to decrease POP incidence.

Conflict of interest: none declared.