Assessment of the aggregate risk score to predict mortality after surgical biopsy for interstitial lung disease^† Free

Abstract

OBJECTIVES

An aggregate risk score (range 0–6 points) for predicting mortality after surgical biopsy for interstitial lung disease (ILD) was recently developed from four independent variables: intensive care unit treatment (2 points), age >67 years (1.5 points), immunosuppression (1.5 points), open biopsy (1 point). In the development cohort, patients were grouped in four classes of aggregate score (A, B, C, D) showing incremental risk of death within 90 days from biopsy. We tested this mortality risk model in an independent cohort.

METHODS

The aggregate risk score and the corresponding class of 90-day mortality risk was retrospectively determined in 151 consecutive patients undergoing biopsy for uncertain ILD at the Center for Thoracic Surgery, University of Insubria (Varese, Italy) in 1997–2012. We evaluated, by Spearman's ρ test, the correlation between aggregate risk score and mortality rate in the development cohort and in our cohort. Fisher's exact test was used for comparison of overall mortality rate between the two cohorts.

RESULTS

The mortality rate correlation with risk score differed in our cohort (ρ = 0.127; P = 0.06) compared with the development cohort (ρ = 0.352; P < 0.0001). In our dataset mortality polarized: it was minimal in Classes A and B (2% and 0%, respectively), 33% in Classes C and D. This skewed mortality distribution was possibly contributed by significantly lower overall mortality rate in our cohort than in the development cohort (2.6% vs 10.6%; P = 0.0017). Despite the difference in mortality distribution, in our dataset, we confirmed that ILD patients with aggregate score >2 (Classes C and D) were at exceedingly high risk of postoperative mortality.

CONCLUSIONS

The aggregate score is a simple and useful risk score for ILD. Our dataset confirms that lung biopsy is reasonably safe in Class A and B patients while, in Class C and D patients, it is indicated only if histology would substantially change management and prognosis.

INTRODUCTION

Interstitial lung disease (ILD) refers to a heterogeneous group of lung disorders characterized by varying degrees of inflammation and fibrosis that present with similar clinical and radiographic features, but have different treatment options and prognosis [1, 2]. ILD is typically diagnosed based on patient medical history, high-resolution computed tomography (HRCT) of the chest and broncho-alveolar lavage results; in selected cases, trans-bronchial/trans-thoracic needle biopsy of the lung is necessary [3]. In spite of increased quality of chest HRCT, bronchoscopy and non-surgical biopsy procedures, at least one-third of patients with suspected ILD require surgical lung biopsy (SLB) to achieve accurate and definitive diagnosis [4]. For diagnosis of idiopathic pulmonary fibrosis and other types of interstitial pneumonias, the American Thoracic Society/European Respiratory Society recommendations indicate SLB as the golden standard, because accuracy of lung biopsy is superior to that offered by radiographic assessment [2, 5]. Notably, a change in patient management after lung biopsy has been reported in 27–73% of ILD cases [6, 7]. However, the role of SLB is controversial because of possible postoperative morbidity and mortality. The most frequent complication of SLB in ILD patient is prolonged air leakage, occurring in 6–12% of cases. Other common complications are pneumonia, pneumothorax, haemothorax, pleural effusion, empyema and the need for prolonged ventilator dependence [5, 8, 9]. The postoperative mortality rate after SLB performed by video-assisted thoracoscopic surgery (VATS) has been reported to range from 0 to10.6% for patients not risk stratified [6, 8–12]. Consequently, many clinicians are reluctant to refer patients to surgical biopsy, as they are uncertain whether the benefits outweigh the risks of the procedure. A simple and reliable method is needed, to identify ILD patients at high risk and to predict postoperative morbidity and mortality. Recently, Fibla et al. developed an aggregate risk score for predicting mortality after surgical biopsy for ILD [12]. This score that could be used to evaluate risks/benefits in patient candidate to SLB for suspected ILD, is based on four variables that were significant predictors of 90-day mortality: age >67 years (score 1.5); preoperative intensive care unit (ICU) admission (score 2); immunosuppressive treatment (score 1.5); open surgery (score 1). In the score development cohort, the total score of each subject with ILD was computed and patients were grouped in four classes [Class A (Score 0); Class B (Score 1–2); Class C (Score 2.5–3); Class D (Score >3)], showing an incremental risk of death at 90 days, from 2% to 86%. Here we aimed to test this mortality risk model in an independent cohort of ILD patients consecutively observed at our centre.

MATERIALS AND METHODS

We retrospectively examined 151 patients undergoing SLB for uncertain ILD at the Center for Thoracic Surgery, University of Insubria, Ospedale di Circolo, Varese, Italy, between 1997 and 2012. Of each patient, we evaluated the following data: gender, age, smoking history, modality of diagnosis of suspected ILD, percentage of predicted forced expiratory volume in 1 s (FEV₁), history of preoperative ICU admission, immunosuppressive or steroid treatment, oxygen therapy, mode of surgical biopsy (open or VATS), number of lung biopsies, duration of surgery, final histological diagnosis, 90-day postoperative complications and 90-day mortality. Table 1 summarizes baseline characteristics of the 151 patients [mean age, 54 years ± 13 standard deviation (SD); 60.3% male]. Half of the cohort had no history of active smoking; the 74 ever smokers had median smoking history of 21 pack-years (range 3–75 pack-years). In 21 patients (13.9%), the diagnosis of ILD was incidental, following a chest X-ray (CXR) or CT exam done for reasons unrelated to lung disease. The vast majority of patients (86%) had respiratory symptoms (dyspnoea, cough, chest pain, haemoptysis) and/or general symptoms (fever, weight loss, arthritis, itching). Preoperatively, pulmonary function tests were performed in 137 cases (90.7%). In these patients, mean FEV₁% was 89 ± 24% of predicted and in 19.7% of them, it was <70% of predicted. At the time of lung biopsy, 22 patients (14.6%) were on immunosuppressive therapy (corticosteroids or methotrexate, or both in one patient). Preoperatively supplemental oxygen was necessary in 21 patients (13.9%). Only 6 patients (4%) were preoperatively admitted to ICU, to treat respiratory failure.

In all patients, preoperative imaging assessment included CXR and HRCT exam, to evaluate ILD distribution in the lung parenchyma and to guide surgical biopsy. Preoperative bronchoscopy was carried out in 121 patients (80.1%).

For all patients, we retrospectively calculated the ILD aggregate risk score, by sum of individual scores, assigning to each patient the corresponding class of 90-day mortality risk, as described by Fibla et al. [12]. In this independent cohort, we evaluated the patients' distribution by class of risk [12], the postoperative mortality rate in each class and the correlation between ILD score and postoperative mortality rate. Logistic regression model was used to investigate the relationship between risk of death and ILD score, adjusted for age.

These results in our cohort were then compared with those reported by Fibla et al. in the ILD score development cohort [12], to assess reproducibility and value of the aggregate risk scoring system.

Fisher's exact test was used to compare mortality rates. Correlation of risk class with mortality was examined by Spearman's ρ test. P values <0.05 were considered statistically significant. Statistical analysis was performed with MedCalc 13.2.2 (MedCalc software, Acacialaan 22, Ostend, Belgium).

The study was approved by our Institutional Ethics Committee.

RESULTS

Lung biopsies were performed with staplers by the same surgical team throughout the study, by VATS in 141 cases (93%). Ten patients (7%) required open thoracotomy due to diffuse pleural adhesions (n = 4), or because single-lung ventilation was not tolerated (n = 6). Two or more lung biopsy samples were obtained in 67% of cases; one-third of patients had only one sample taken. Mean operative time was 56 ± 21 min for VATS, 100 ± 63 min for thoracotomy procedures. No major intraoperative complications were reported.

Mean postoperative stay was 7 days (range 1–105 days). Postoperative complication occurred in 12.6% of patients, as detailed in Table 2. Mortality rate at 90 days was 2.6% (4 cases); these severely compromised patients died within 15 days after SLB (3 for acute respiratory failure; 1 for myocardial infarction).

Specific definitive histological diagnosis was obtained in 144 cases (95.4%); in the other 7 cases (4.6%), the pathology report was non-specific, descriptive. Histological results are summarized in Table 3. Therapy was modified according to SLB results in 80 patients (53%): in 61/110 cases, (55%) of Class A; in 16/35 cases (46%) of Class B; in 1/3 cases (33%) of Class C; in 2/3 cases (67%) of Class D.

The data necessary to compute the ILD aggregate risk score [12] were available for all patients in our cohort, and we obtained for each class (A, B, C, D) the corresponding mortality rate (Table 4). The distribution of classes in the Varese cohort revealed a marked prevalence of patients in Class A (73%) and was significantly different from that of the development cohort of Fibla et al. (Yates’ P < 0.001; Table 4). Analysis of the relationship between risk of death and ILD score, adjusted for age, showed odds ratio 4.715 (95% confidence interval CI 1.92–11.55; P = 0.001) by 1-level increment of ILD score. Mortality rate correlation with class risk score in our cohort (ρ = 0.127; P = 0.06) differed from that in the development cohort (ρ = 0.352; P < 0.0001). Overall mortality rate was significantly lower in our cohort than in the development cohort (2.6% vs 10.6%; P = 0.0017). Moreover, in our dataset mortality polarized: it was minimal in Classes A and B (2% and 0%, respectively), and 33% in Classes C and D.

DISCUSSION

When a specific diagnosis of ILD cannot be confidently obtained from clinical and radiological information, SLB may become essential, especially to differentiate between usual interstitial pneumonia (UIP) and other forms of ILD. As indicated by current British Thoracic Society guidelines, the decision to pursue an SLB in ILD has two caveats: (i) the recognition of limitations in the histological assessment in ILD and (ii) the risk associated with surgery [3].

Therefore, before embarking on SLB for suspected ILD, the risk/benefit ratio of the procedure should be carefully assessed. The specificity of diagnosis obtained by SLB varies widely in the published series, ranging from 37 to 100%, depending on the prevalence of difficult cases and on local expertise [3, 6, 11–13]. In our series, a specific diagnosis was obtained in 95.4% of SLBs. Clearly, a multidisciplinary team approach with expertise in ILD histopathology is a key point for diagnostic yield [3]. Risk of complications and mortality inherent in the SLB procedure has been emphasized in several reports and correlates with the severity of patient illness [8, 12–14]. Interestingly, perioperative mortality of SLB for ILD was nil in a recent multicentric study that excluded compromised patients and those with mechanical ventilation or with PaO₂ <60 mmHg [15].

Moreover, it should be considered that morbidity and mortality are generally higher for open compared with video thoracoscopic surgery (4.8–50% vs 11–19% and 0–21.4% vs 0–10.6%, respectively) [8–10, 12–15].

Factors identified as predictors of morbidity and mortality after surgical biopsy for ILD include: preoperative ventilator dependence, age, immunocompromised status, UIP diagnosis, <35% of predicted diffusing capacity of the lung for carbon monoxide, acute exacerbation of ILD at the time of biopsy and an open thoracotomy procedure [8, 14, 16]. Fibla et al. selected the most relevant among these risk factors, and developed a simple aggregate risk score for predicting mortality after SLB. They proposed a risk score based on four independent variables (preoperative ICU stay, age >67 years, immunosuppressive treatment, open surgery), to stratify SLB candidates in four classes of risk [12]. According to these authors, SLB can be recommended in Class A patients, as their 90-day mortality risk is 2%, while it is not advisable in Class D patients (90-day mortality risk: 86%). SLB could be proposed to Class B and Class C patients (90-day mortality risk, respectively: 12 and 40%) only after exhaustive evaluation of benefit/risk ratio. To our knowledge, the value of the aggregate score of Fibla et al. was not previously tested in an independent cohort. Therefore, we applied this score, which is straightforward to calculate, in the context of our 151 consecutive patients undergoing SLB for ILD. In our series, mortality distribution was skewed (2% and 0% in Class A and B, respectively; 33% in Class C and D), possibly contributed by significantly lower mortality rate in our cohort compared with the development cohort (2.6% vs 10.6%). Likely due to inclusion of many high-risk patients, the 10.6% postoperative mortality in the series of Fibla et al. [12] is high also compared with the 2–6.7% postoperative mortality generally reported for ILD cohorts that were not risk-stratified [9–11]. Although the mortality distribution in our dataset was different from that in the Fibla cohort, we found that SLB in ILD patients with aggregate score <2 (Classes A and B) has a minimal mortality risk, and confirmed that ILD patients with aggregate score >2 (Classes C and D) are at high risk (33%) of postoperative mortality.

There are limitations to the present study. This is a single-centre study, with a retrospective design, although the data were prospectively collected since 1997 under regular revision by a data manager. ILD patients in Varese were younger compared with the Mayo Clinic group of Fibla [12] (mean age 54 ± 13 vs 60.9 ± 14 years; P < 0.0001), had a different histological distribution and a lower proportion of poor prognosis diseases (UIP/idiopathic pulmonary fibrosis: 22.5% vs 39.2%; P = 0.0005). Moreover, in our series, the proportion of ILD Class A patients was significantly larger, and that of Class D was lower, compared with Fibla's cohort, because we had a conservative indication to SLB for high-risk patients. This selection bias is likely responsible for the lower mortality in Class D in our cohort (33%) compared with the development cohort (86%) [12]. Points of strength are the consecutive series of SLB for ILD, no missing data and a uniform mode of surgical management throughout the study.

In conclusion, we found that the aggregate score is a simple and useful risk score for ILD that may facilitate patient counselling and decision-making for SLB candidates. Our retrospectively evaluated dataset confirms that lung biopsy is a reasonably safe procedure in Class A and Class B patients, while in Class C and Class D patients SLB is indicated only if histology would substantially change ILD management and prognosis.

Conflict of interest: none declared.

REFERENCES

APPENDIX. CONFERENCE DISCUSSION

Dr T. Hakala(Joensuu, Finland): 90-day mortality was 2.6%. Was the cause of death related to the lung biopsy or the lung disease itself?

Dr Imperatori: As cause of death we recorded one due to myocardial infarction and three due to respiratory failure. The three respiratory failures were related to the interstitial lung disease. Two of these patients were in the ICU before surgery.

In our mind, it was a reactivation, or a significant activation, of the inflammation, probably due to the underlying disease and also to the surgery; it was probably a combination of the two causes. But, notably, in the other 66% of cases in classes C and D, the biopsy was able to give us some important information enabling us to change the management, the treatment, and the prognosis in at least 50% of the patients.

Dr D. Van Raemdonck(Leuven, Belgium): I have a question regarding your methods. If I remember well from your introduction, in the aggregate risk score from Fibla, one point was given for open biopsy.

Dr Imperatori: Yes.

Dr Van Raemdonck: And from your abstract I concluded that all patients in your series had VATS biopsy. So how did you compare that score versus your patients?

Dr Imperatori: We performed 93% by VATS and 7% of cases by mini-thoracotomy; in four cases we had to convert to mini-thoracotomy due to adhesions, while six cases were directly approached by open surgery because single-lung ventilation was not tolerated. So we recorded these as open surgery, open biopsy in 7% of our cases. It means 10 patients of the 151.

Author notes