Introduction
Interstitial pneumonia (IP) has an incidence of 5.6 per 100 000 person-years and an annual cumulative prevalence of approximately 18 per 100 000 people [1]. Opportunities for patients with IP to undergo surgery are common. Acute exacerbation (AE) may occur in patients with postoperative IP, and the mortality rate associated with AE-IP is extremely high [2–5]. However, no evidence-based prophylaxis exists, and the complete prevention of AE-IP is impossible. Therefore, the risk factors for AE-IP must be clearly understood and risk assessments should be conducted. To date, AE-IP risk factors have been reported to include male sex, preoperative steroid administration, history of AE-IP, low lung capacity, usual IP pattern on computed tomography (CT), high serum sialylated carbohydrate Krebs von den Lungen-6 (KL-6) antigens levels, elevated C-reactive protein (CRP) levels, lung surgery involving more than a segmental resection, prolonged duration of surgery, oxygen exposure, ventilation before surgery, and unilateral lung ventilation [3,5–18]. However, with the exception of one report [6], all studies have had small sample sizes with a limited number of events and only a few have appropriately controlled for confounding factors using adequate statistical analysis. Thus, no established risk factors have been identified. Furthermore, despite the existence of a large noteworthy study on lung resection [6], sufficient evidence has not been established for either lung or non-lung surgeries. Therefore, this study aimed to determine whether AE-IP is associated with worse mortality postoperatively and to identify the factors associated with the development of AE-IP.
Materials and Methods
The Clinical Research Ethics Committee of Kyushu University, Fukuoka, Japan, approved this observational retrospective study (research approval number: 2019-251) on August 5, 2020, and the need for individual written informed consent was waived due to its retrospective design. This study was conducted in compliance with the Declaration of Helsinki, 2013 and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. Overall, 492 patients with a preoperative diagnosis of IP who underwent noncardiac surgery at Kyushu University Hospital between December 2012 and April 2020 were included in this study. IP was either already diagnosed and followed up by a respiratory physician or incidentally diagnosed by a radiologist on preoperative CT. After excluding five patients aged < 18 years and five identical patients, 482 participants were enrolled in this study. None of the participants had preoperative AE-IP.
We extracted surgery and anesthesia data from the participants’ electronic anesthetic charts and preoperative, postoperative, and survival data were collected manually by reviewing each participant’s electronic medical records individually. Data on sex, age, height, weight, Brinkman index (calculated by multiplying the number of cigarettes by the number of years), preoperative supplemental oxygen, and preoperative ventilator use were collected for the study participants. Additionally, we collected data on KL-6, CRP, and lactate dehydrogenase (LDH) levels; percent vital capacity (%VC); and forced expiratory volume in 1 s (FEV1) on spirometry measured within 1 month preoperatively. The following surgical and anesthesia data were collected: type of surgery (lung surgery or other), anesthesia method (general or local anesthesia), presence of differential lung ventilation, surgery duration, anesthesia duration, urine volume, blood loss volume, infusion volume, in-out balance, volume of oxygen used, presence of intraoperative blood transfusion, and duration of fraction of inspired oxygen (FiO2) ≥ 60%. Continuous intraoperative data, such as the FiO2 were automatically stored in electronic anesthetic charts, and minute-by-minute data were used for the analysis. The threshold of 60% FiO2 was established based on a previous report [9]. Although no protocol for FiO2 during anesthesia has been established at our facility, FiO2 is generally adjusted at the discretion of the individual anesthesiologist so that SpO2 does not fall below 90% for both lung and non-lung surgeries when anesthetizing patients with IP.
The primary outcome of this study was the incidence of postoperative AE-IP. We defined AE-IP as the worsening of respiratory symptoms within 1 month postoperatively, with an AE-IP diagnosis confirmed by a radiologist on CT. Postoperative mortality information was collected from medical records, and the association between AE-IP and postoperative mortality was examined. AE-IP-related deaths were defined as deaths for which IP was listed as the cause of mortality in the medical record. The follow-up period for death was calculated from the date of surgery to the date of death confirmed in the medical record or to the last date recorded (e.g., the date of the last outpatient visit).
Statistical analysis
The baseline characteristics of the participants in the AE and non-AE groups were compared. Continuous variables are presented as mean ± SD or median (Q1, Q3) values and categorical variables as percentages. We used the Student’s t test for normally distributed continuous variables, Mann–Whitney U test for non-normally distributed continuous variables, and chi-square test for frequencies. We calculated the cumulative survival rates for all-cause and IP-related deaths between the AE and non-AE groups using the Kaplan–Meier method and compared them using a log-rank test. The Cox proportional hazards model was used to estimate hazard ratios (HRs) with 95% CIs for all-cause and IP-related deaths. The risk estimates in the multivariable analysis were adjusted for potential confounding factors at baseline, including age, sex, body mass index, preoperative supplemental oxygen, preoperative ventilator use, and type of surgery (lung surgery or other). Logistic regression was used to estimate odds ratios (ORs) with 95% CIs for AE-IP to identify factors associated with AE-IP. The backward method was used to select significant variables that could be candidate factors associated with AE-IP, and P < 0.20 was used as the selection criteria. In the variable selection model, we included potential confounding factors previously reported as AE-IP risk factors and those with P < 0.20 in the univariate analysis. A total of 270 participants, excluding 212 participants with missing data, were included in the multivariable analysis using the variable selection model and the subsequent case-control study using propensity score matching. A logistic regression analysis was performed for sensitivity analysis to examine the validity of the factors associated with AE-IP obtained from the variable selection, including the 212 participants who were excluded due to missing data. Owing to the insufficient number of events and overall number of participants, case-control studies with propensity score matching were subsequently conducted for each of the variables selected as candidate factors associated with AE-IP. For each factor selected through variable selection, propensity scores for AE-IP with other factors and potential confounding factors previously reported as AE-IP risk factors were generated, and 1:5 matching was performed using conditional logistic regression analysis to examine whether the factors of interest differed. Additionally, to visualize the relationship between the duration of FiO2 ≥ 60% and AE-IP, a generalized additive model was used to calculate the predicted AE-IP incidence and establish the spline curves. Spline curves showing these associations were drawn for both lung and non-lung surgeries. All statistical analyses were performed using SAS software package (version 9.4; SAS Institute). Two-sided P values < 0.05 were considered statistically significant for all analyses.
Results
Baseline characteristics of participants
Of the 482 participants included in the analysis, 19 developed AE-IP postoperatively. Participants who had AE-IP were more likely to be male. The mean values for weight, median values for KL-6 and CRP levels, and frequencies of FiO2 ≥ 60% were greater in the AE group than in the non-AE group (Table 1). Details regarding the surgical sites are provided in Supplementary Table 1.
Crude cumulative survival rates
The median (Q1, Q3) follow-up periods for the AE and non-AE groups were 212 (29, 550) and 729 (306, 1702) days, respectively. During the follow-up period, 10 and 119 patients died in the AE and non-AE groups, respectively. Additionally, five and 28 IP-related deaths occurred in the AE and non-AE groups, respectively (Table 2). A significant difference was found in the crude cumulative survival rates for all-cause (A) and IP-related (B) deaths between the AE and non-AE groups (Fig. 1).
Hazard ratios for all-cause and IP-related deaths in the AE and non-AE groups
The crude HR for all-cause death was significantly higher in the AE group (HR: 4.82, 95% CI [2.51−9.26]) than in the non-AE group. In the multivariable-adjusted analysis, this association remained significant even after adjusting for potential confounding factors (HR: 5.16, 95% CI [2.62−10.14]). Similar results were observed for IP-related deaths.
Odds ratios for AE-IP in the crude and variable selection analyses
The crude OR for AE-IP was higher in males (OR: 10.64, 95% CI [2.17−192.15]) than in females, and higher CRP levels significantly increased the risk of AE-IP (OR: 1.12, 95% CI [1.03−1.22]). Multivariable analysis using the variable selection method showed that the candidate factors associated with AE-IP were male sex, higher CRP levels, FiO2 ≥ 60%, and lung surgery; however, statistical significance was observed only for CRP levels (OR: 1.21, 95% CI [1.05−1.40], P = 0.006), FiO2 ≥ 60% (OR: 5.22, 95% CI [1.37−22.68], P = 0.018), and lung surgery (OR: 0.24, 95% CI [0.06−0.96], P = 0.047) (Table 3). The results of the sensitivity analysis, which included the 212 participants who were excluded due to missing values, showed that the ORs for AE-IP after multivariable adjustment were 14.54 (95% CI [1.32−159.70]) for male vs. female, 1.17 (95% CI [1.07−1.28]) for every 1 mg/dl elevation in CRP and 5.26 (95% CI [1.74−15.89]) for FiO2 ≥ 60% vs. FiO2 < 60% (Supplementary Table 2); these findings were almost identical to the main results of multivariable analysis in Table 3. A positive relationship also existed between the duration of FiO2 ≥ 60% and AE-IP incidence: the predicted AE-IP incidence increased linearly with the duration of FiO2 ≥ 60% (Fig. 2). This linear relationship was similar for both lung and non-lung surgeries (Supplementary Fig. 1).
Case-control analysis using propensity score matching
Multivariable analysis using the variable selection method was used to detect candidate factors associated with AE-IP. However, since many variables existed for the number of AE-IP cases, a case-control analysis was subsequently performed using propensity score matching. After matching with the propensity scores created using potential confounding factors previously reported as AE-IP risk factors and three additional factors other than CRP, CRP levels were significantly higher in the AE group than in the non-AE group (P = 0.044). Similarly, for FiO2 ≥ 60%, after matching other factors, frequency of FiO2 ≥ 60% was significantly higher in the AE group than in the non-AE group (P = 0.035). Sex and type of surgery did not differ significantly between the groups (Table 4).
Discussion
In this study, we found that postoperative AE-IP was associated with a higher risk of mortality, particularly IP-related deaths. Furthermore, variable selection analysis showed higher CRP levels, male sex, non-lung surgery, and FiO2 ≥ 60% as candidate factors associated with AE-IP, and a subsequent propensity score matching analysis revealed that higher CRP levels and FiO2 ≥ 60% were the most likely factors associated with AE-IP. We also showed that the risk of AE-IP increased linearly with the duration of FiO2 ≥ 60%.
The frequency of postoperative AE-IP in this study was 3.9%. Although various values for this incidence have been reported, 3.9% is generally consistent with previous reports [9,10,15,19,20]. Notably, overall (34.5%) and IP-related (20.8%) deaths were higher in patients with AE-IP than in those without AE-IP. These results suggest that identifying the risk factors for AE-IP and predicting and preventing AE-IP are clinically important. In this context, the clinical relevance of this study, which identified higher CRP levels and FiO2 ≥ 60% as factors associated with AE-IP based on data from 482 participants at a single institution, is highly significant. In particular, the finding that high intraoperative FiO2 is associated with AE-IP incidence is clinically significant for anesthesiologists. The results of this study could have profound clinical implications as they may influence future perioperative management of patients with IP.
Although many studies on risk factors for AE-IP exist, only a few have focused on FiO2 during anesthesia. The incidence of AE-IP on the non-operative side after lung surgery has been reported in a few studies, suggesting a relationship between high FiO2 levels and AE-IP. However, no studies have provided statistical analyses to support this relationship. The mechanism by which high FiO2 levels are associated with AE-IP may be explained by endogenous production of oxygen radicals by some cell types, including inflammatory cells [21]. Furthermore, oxygen radicals can inactivate various enzymes within alveolar epithelial and capillary endothelial cells to damage cellular membranes and intranuclear genes [22]. Production of various cytokines by activated neutrophils and alveolar macrophages is associated with the pathogenesis of lung injury [21,22]. The administration of higher oxygen concentrations to model mice with peplomycin-induced IP has been reported to result in more severe pulmonary fibrosis, demonstrating the pulmonary toxicity of high supplemental oxygen concentrations [23]. Furthermore, although atelectasis and small airway closure normally does not occur until 100 min of FiO2 at 30%, this can occur within 10 min at 100% FiO2 [24]. This mechanism may be related to an increase in the time-dependent risk of AE-IP due to the production of reactive oxygen species and increased atelectasis. However, the possibility of reverse causation in the relationship between FiO2 and AE-IP should also be considered, though further prospective studies are required to confirm this association.
A large multicenter study involving 1763 patients at 61 centers found that a history of AE-IP, male sex, preoperative steroid use, high KL6 levels, and low %VC were significantly associated with AE-IP [6]. Furthermore, a meta-analysis of 2655 patients across 12 studies identified male sex, high KL6 and LDH levels, elevated white blood cell counts, low oxygen pressure in arterial blood (PaO2) levels, and prolonged surgery as risk factors for AE-IP [19].
Several previous reports, despite their small sample sizes, have reported high CRP levels as a risk factor for postoperative AE-IP, which is consistent with the findings of this study [7,10,13,16,22]. High preoperative CRP levels have been reported to reflect a strong inflammatory state, and the induction of inflammatory cytokines is considered a possible trigger for AE-IP [10]. Notably, the two large studies mentioned earlier reported that male sex, high %VC, high KL6 and LDH levels, and prolonged surgery were risk factors for AE-IP [6,19]. However, we did not observe similar results in our study. The most likely explanation is that we had an insufficient number of participants and AE-IP cases; therefore, our study lacked the statistical power to identify all factors in every case. Previous studies have also reported similar observations that not all risk factors can be identified [4,25]. Therefore, as this exploratory analysis was designed only to generate hypotheses about relationships, it provides no conclusive evidence about these associations.
Strengths and limitations
Our results are valuable because this is the first study that uses appropriate statistical methods to show that intraoperative FiO2 ≥ 60% is associated with AE-IP and suggest a time-response relationship wherein the incidence of AE-IP increases with time of administration. Another strength of this study, compared with previous reports, is the large number of participants at a single institution. In multicenter studies, differences in perioperative management among centers may hinder the identification of risk factors, whereas in a single-center study, these extraneous factors do not need to be considered, and consistent results are more likely to be obtained.
However, a limitation of this study is that even though the number of participants was large for a single center, it was not sufficient to identify all the factors associated with AE-IP, as previously noted. Second, even though the risk estimates were adjusted for possible confounding factors, residual confounders may exist. Third, as this was an observational study, the FiO2 levels during surgery were arbitrarily set by the respective anesthesiologists based on the clinical situation. Additionally, as we had insufficient data regarding PaO2, it was impossible to determine whether AE-IP was attributable to FiO2 or PaO2 and whether it was the use of or need for FiO2 ≥ 60% that was a risk factor for AE-IP in this study. Fourth, insufficient information on the participants’ IP diagnosis prevented separate analyses of each type of IP, and the lack of data on AE-IP classification as unilateral or bilateral prevented its inclusion in the study. Furthermore, information on the history of AE-IP and preoperative steroid use was lacking. Finally, this study was conducted on a single race at one institution; therefore, the generalizability of our findings is limited.
In conclusion, this study demonstrates that, in addition to elevated preoperative CRP levels, anesthesia management with a higher FiO2 during surgery is associated with postoperative AE-IP.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement1
Print
