DOI: https://doi.org/10.21203/rs.3.rs-1865700/v1
Background
The ILD-GAP scoring system has been widely used to predict the prognosis of patients with interstitial lung disease (ILD). The ability of the ILD-GAP scoring system combined with the Charlson Comorbidity Index score (CCIS) (ILD-GAPC) to predict ILD prognosis was investigated.
Methods
In ILD patients including idiopathic pulmonary fibrosis (IPF), idiopathic nonspecific interstitial pneumonia (iNSIP), and collagen vascular disease-related interstitial pneumonia (CVD-IP) treated between April 2013 and April 2017, the relationships between baseline clinical parameters including age, sex, CCIS, ILD diagnosis, pulmonary function test results, and 3-year ILD-related events, including cause-specific death and first acute exacerbation (AE), were retrospectively assessed, and the ability to predict prognosis was compared between ILD-GAP and ILD-GAPC.
Results
A total of 174 patients (mean age, 74 years), all of whom underwent pulmonary function testing including percentage predicted diffusion capacity for carbon monoxide, were assessed. ILD diagnosis consisted of IPF in 57 cases, iNSIP, and CVD-IP in 117 cases. ILD-GAPC provided a greater area under the 3-year time dependent receiver operating characteristic curve (0.743) for predicting 3-year ILD-related events than ILD-GAP (0.710). In addition, log-rank tests showed that survival curves differed significantly among low, middle, and high ILD-GAPC scores (P = 0.003), unlike ILD-GAP scores (P = 0.240).
Conclusions
ILD-GAPC could provide more accurate information for predicting prognosis in patients with ILD than ILD-GAP alone.
Interstitial lung disease (ILD) is characterized by alveolar inflammation leading to progressive fibrosis. The clinical course and rate of progression of ILD are extremely variable among patients due to various radiological and pathological-morphological patterns, such as usual interstitial pneumonia (UIP), non-specific interstitial pneumonia (NSIP), organizing pneumonia, respiratory bronchiolitis, desquamative interstitial pneumonia, diffuse alveolar damage, and their combinations.1 An official statement of the American Thoracic Society/the European Respiratory Society/the Japanese Respiratory Society/the Latin American Thoracic Association (ATS/ERS/JRS/ALAT) proposed various clinical parameters associated with an increased risk of mortality, such as clinical symptoms, pulmonary function, and the extent of UIP on high-resolution computed tomography; however, clinical parameters for accurately predicting the prognosis of ILD have not been established.2
To provide more accurate prognostic information in patients with ILD, various composite approaches have been reported using peripheral blood biomarkers and physiological and radiographic measurements.3–7 Ley et al. proposed the GAP index as a mortality prediction model for idiopathic pulmonary fibrosis (IPF) patients, consisting of four parameters including gender (G), age (A), percent predicted forced vital capacity (%FVC), and diffusion capacity of carbon monoxide (%DLco) (P).3 In addition, to predict mortality in major chronic ILD subtypes including IPF, idiopathic NSIP (iNSIP), collagen vascular disease-related interstitial pneumonia disease (CVD-IP), and chronic hypersensitivity pneumonia (CHP), ILD-GAP has been reported to be useful.4 Both GAP and ILD-GAP have been widely used in the clinical setting, but these mortality prediction models do not take into account the presence or severity of comorbidity despite the previous research showing that comorbidity, such as cardiovascular disease, arteriosclerosis, and cancer, affects the long-term prognosis of ILD.8,9
The present study retrospectively investigated the accuracy for predicting ILD prognosis of the ILD-GAP scoring system combined with the Charlson Comorbidity Index score (CCIS) (ILD-GAPC), which has been widely used as a prognostic indicator for patients with colorectal cancer, advanced non-small cell lung carcinoma, acute myocardial infarction, and so on.10–13
This retrospective, observational study was performed using data from patients treated at Yokohama City University Hospital between April 2013 and April 2017. The medical records of all patients with ILD who met the following inclusion criteria were reviewed: patients with IPF, iNSIP, or CVD-IP in a stable condition who were able to perform pulmonary function tests including DLco. ILD patients in a stable condition were defined as patients who had not experienced acute respiratory deterioration including an acute exacerbation (AE) until pulmonary function testing. Pulmonary sarcoidosis, cryptogenic organizing pneumonia, drug or radiation-induced lung injuries, CHP, and other unclassified ILDs were excluded.
The relationships between baseline clinical parameters including age, sex, CCIS, ILD diagnosis, blood biomarker, pulmonary function test results, and 3-year ILD-related events including cause-specific death or first AE were assessed. Three-year ILD-related events were collected mainly from medical records. For patients who did not die in our hospital, the disease outcomes were confirmed by telephone. In addition, only one patient (0.5%), who was transferred to another hospital for best supportive case due to severe deterioration of respiratory status, was lost to follow-up; therefore, the transfer date of that patient was selected as the decision date of disease outcome.
A diagnosis of idiopathic interstitial pneumonias (IIPs) was confirmed by physical findings, serological testing, findings from high-resolution computed tomography (HRCT), and lung biopsy specimens, based on the official statement for IIP.1,2 However, patients from whom a lung biopsy could not be obtained were diagnosed based on the radiological classification.1,2 The diagnosis of CVD-IP was confirmed by physical findings, serological testing, and HRCT findings consistent with ILD. An AE of ILD was defined as: unexplained worsening of dyspnoea; hypoxaemia or severely impaired gas exchange; new alveolar infiltrates on radiography; and absence of an alternative explanation such as infection, pulmonary embolism, pneumothorax, or heart failure.14–16
Data were statistically analysed using JMP12 (SAS Institute, Cary, NC) and R software, version 3.5.1 (The R Foundation for Statistical Computing, Vienna, Austria), and are expressed as means ± standard deviation. Groups were compared using chi-square test and Wilcoxon rank-sum tests. To determine the primary predictors of 3-year ILD-related events, univariate analyses were performed. The predictive performance of the scoring systems was investigated using the areas under the time-dependent receiver operating characteristic curve (ROC) analysis (AUC), the concordance index (C-index), and Akaike’s information criterion (AIC). When comparing 3-year ILD-related events including cause-specific death or first AE among groups depending on the scoring system, Kaplan-Meier curves were used. Log-rank testing was also performed with strata based on the identified predictors. Values of P < 0.05 were considered significant.
Table 1 shows the clinical characteristics of the 174 patients evaluated, including 57 patients with IPF and 117 patients with non-IPF including iNSIP and CVD-IP. CVD-IP included rheumatoid arthritis, anti-neutrophil cytoplasmic antibody-associated vasculitis, polymyositis/dermatomyositis, and Sjögren’s syndrome. Especially in the IPF group, the incidence of males was the highest (86%), %DLco was the lowest (80.5 ± 28.2%), and the ILD-GAP score was the highest (3.2 ± 0.6 points).
Characteristic | Overall | IPF | CVD-IP/iNSIP | P value |
---|---|---|---|---|
Total number, n (%) | 180 (100) | 57 (32) | 117 (65) | |
Age, y | 71.9±9.2 | 73.3±7.2 | 71.4±9.2 | 0.212 |
Male sex, n (%) | 119 (66) | 49 (86) | 66 (56) | < 0.001 |
CCIS | 2.5±2.2 | 2.6±2.0 | 2.5±2.2 | 0.442 |
Blood biomarker | ||||
KL-6, U/mL | 879.3±1123.5 | 748.3±456.5 | 839.4±1071.6 | 0.090 |
Pulmonary function tests | ||||
FVC, %predicted | 93.9±18.7 | 93.5±18.5 | 94.1±19.2 | 0.994 |
DLco, %predicted | 91.7±30.2 | 80.5±28.2 | 97.6±30.2 | < 0.001 |
ILD-GAP score | 2.5±2.1 | 3.2±0.6 | 0.6±0.7 | < 0.001 |
Outcome | ||||
Follow-up, days | 780±483 | 722±459 | 807±501 | 0.395 |
Incidence of events within 3 y, n (%) | 20 (11) | 10 (18) | 10 (9) | 0.081 |
Abbreviations: | ||||
CCIS, Charlson Comorbidity Index Score; CI, confidence interval; CVD-IP, collagen vascular disease-related interstitial pneumonia; GAP, gender/age/physiology; ILD, interstitial lung disease; IPF, idiopathic pulmonary fibrosis; KL-6, Krebs von den Lungen; %DLco, percentage predicted diffusion capacity of lung for carbon monoxide, %FVC, percentage predicted forced vital capacity. |
To determine the primary predictors of 3-year ILD-related events, univariate analysis was performed with the following parameters: age, sex, CCIS, diagnosis of ILD (IPF vs. non-IPF), ILD-GAP score, %FVC, and %DLco (Table 2). This showed that CCIS, diagnosis of ILD, ILD-GAP score, %FVC, and %DLco were significant predictors of 3-year ILD-related events.
Variable | Hazard ratio | 95% CI | P |
---|---|---|---|
Age | 1.027 | 0.973–1.084 | 0.334 |
Sex (Male) | 1.261 | 0.517–3.080 | 0.610 |
CCIS | 4.576 | 2.304–9.089 | < 0.001 |
Diagnosis of ILDs | 2.260 | 1.082–5.594 | 0.032 |
ILD-GAP score | 1.364 | 1.099–1.694 | 0.005 |
%FVC | 0.975 | 0.953–0.997 | 0.024 |
%DLco | 0.984 | 0.969-1.000 | 0.049 |
Abbreviations: | |||
CCIS, Charlson Comorbidity Index Score; CI, confidence interval; GAP, gender/age/physiology; %DLco, percentage predicted diffusion capacity of lung for carbon monoxide, %FVC, percentage predicted forced vital capacity; ILD, interstitial lung disease. |
It was hypothesized that the ILD-GAP model combined with CCIS (ILD-GAPC model) is more accurate for predicting 3-year ILD-related events than the ILD-GAP model. Table 3 shows the details of ILD-GAPC scoring. That is, in the ILD-GAPC model, the score of CCIS (0–1: 0 points, 2–3: 1 point, ≥ 4: 2 points) was added to the ILD-GAP score (Table 3). To investigate the accuracy of the ILD-GAP and ILD-GAPC models for 3-year ILD-related events, AUCs, C-index values, and AIC values for these models were calculated. All of AUCs, C-index values, and AIC values were more favourable with the ILD-GAPC model than those with the ILD-GAP model (Table 4).
ILD-GAP model | ILD-GAPC model | ||
---|---|---|---|
Point | Point | ||
ILD diagnosis | IPF | 0 | 0 |
CVD-IP / iNSIP | -2 | -2 | |
Sex | female | 0 | 0 |
male | 1 | 1 | |
Age (y) | ≤ 60 | 0 | 0 |
61–65 | 1 | 1 | |
༞65 | 2 | 2 | |
%FVC | ༞75 | 0 | 0 |
50–75 | 1 | 1 | |
༜50 | 2 | 2 | |
%DLco | ༞55 | 0 | 0 |
36–55 | 1 | 1 | |
≤ 35 | 2 | 2 | |
Cannot perform | 3 | 3 | |
CCIS | 0–1 | 0 | |
2–3 | 1 | ||
≥ 4 | 2 | ||
Abbreviations: | |||
CCIS, Charlson Comorbidity Index Score; CI, confidence interval; CVD-IP, collagen vascular disease-related interstitial pneumonia; GAP, gender/age/physiology; ILD, interstitial lung disease; iNSIP, idiopathic nonspecific interstitial pneumonia; IPF, idiopathic pulmonary fibrosis; %DLco, percentage predicted diffusion capacity of lung for carbon monoxide, %FVC, percentage predicted forced vital capacity. |
AUC | C-index | AIC | |
---|---|---|---|
ILD-GAP model | 0.710 | 0.625 | 206.8 |
ILD-GAPC model | 0.743 | 0.724 | 198.3 |
Abbreviations: | |||
AIC, Akaike’s information criterion; AUC, areas under the time-dependent receiver operating characteristic curve; GAP, gender/age/physiology; ILD, interstitial lung disease |
Survival curves of patients with ILD were compared according to the ILD-GAP score (low score ≤ 1 point vs. moderate score 2, 3 points vs. high score ≥ 4 points) (Fig. 1 (A)). The log-rank test showed that the Kaplan-Meier survival curves of these groups did not differ significantly (P = 0.240). On the other hand, survival curves of patients with ILD were compared according to the ILD-GAPC score (low score ≤ 1 point vs. moderate score 2–4 points vs. high score ≥ 5 points) (Fig. 1 (B)). The log-rank test showed that the Kaplan-Meier survival curves of these groups differed significantly (P = 0.003).
Although the clinical course and rate of progression of ILD are extremely variable among patients, clinical parameters for accurately predicting the prognosis of ILD have not been established.1,2 From the viewpoint of clinical simplicity and versatility, various composite approaches such as GAP or ILD-GAP including age, sex, ILD diagnosis, and physiological measurements have been widely used to provide more accurate prognostic information in clinical settings.3,4 However, these mortality prediction models do not take into account the presence or severity of comorbidities. In the present study, the ILD-GAPC model was found to better predict 3-year ILD-related events than the ILD-GAP model.
FVC is widely used as a biomarker in patients with ILD for predicting prognosis or evaluating treatment efficacy.3,4,17−23 Longitudinal variation of FVC is reported to be more reliable than baseline FVC, since baseline FVC may oversimplify the staging process because disease activity in patients with ILD does not always progress in a linear pattern.2,23 Actually, in the present study, the most influential prognostic factor was CCIS, not baseline FVC. The CCIS is a summed score of 19 comorbidities weighted according to severity.10 The CCIS was developed to assess the risk of death from comorbidities and has been widely used as a prognostic indicator for patients with colorectal cancer, advanced non-small cell lung carcinoma, and acute myocardial infarction.11–13 Furthermore, in patients with ILD, both stable and with an AE, the CCIS has recently been reported to be a prognostic indicator.6,7,24,25 From the above, the CCIS could accurately predict disease prognosis in ILD patients with both stable or AE conditions. A further study is needed to determine whether the management of comorbidities improves the prognosis of ILD.
The ILD-GAP model has been reported to accurately predict mortality in major chronic ILD subtypes such as IPF, iNSIP, CVD-IP, and CHP.4 In the present study, the ILD-GAPC model was a better predictor of 3-year ILD-related events than the ILD-GAP model, though there was a significant correlation between these models (Supplement table). The reason for this was considered to be that the previously reported ILD-GAP model is a model for ILD patients with higher severity than in the present study.4 In fact, the enrolled patients in the original research on the ILD-GAP model had much lower %FVC and %DLco than those in the present study. Although all patients in the high ILD-GAP score group were included in the high ILD-GAPC score group, the patients in the moderate ILD-GAP score group were divided into the moderate and high ILD-GAPC score groups, and the patients in the low ILD-GAP score group were divided into the low and moderate ILD-GAPC score groups. Based on the above, the ILD-GAP model is considered to be a prognosis prediction model for severe cases, while the ILD-GAPC is considered to be a highly versatile model for patients with a wide range of severity from mild to severe.
This retrospective study of a small number of patients from only one institution has some limitations. To verify the utility and reproducibility of this composite scoring system, large-scale, multi-institutional, prospective, collaborative research is essential. The majority of patients enrolled were not so severely ill that pulmonary function tests including DLco could not be tolerated, which suggests a possible source of bias in the present research. A treatable traits approach has been proposed as a new paradigm for the management of chronic lung diseases such as chronic airway disease, bronchiectasis, and ILD.26–28 Especially in ILD, with the recent reports of the clinical efficacy of antifibrotic agents, evaluation of lung fibrosis and therapeutic interventions have become more important.21,22,29 As in the present research in which the CCIS proved to be an important prognostic indicator in patients with ILD, comorbidities, such as lung cancer, cardiovascular disease, gastro-oesophageal reflux disease, and pulmonary hypertension, have been reported to have prognostic impact; thus, not only lung fibrosis and inflammation but also CCIS seemed to be important treatable traits for patients with ILD. Furthermore, it is unclear whether treatment for these comorbidities will improve the prognosis of ILD patients, and this can be considered a topic for future research.
ILD-GAPC could provide more accurate information for predicting prognosis in patients with ILD than ILD-GAP alone.
acute exacerbation
Akaike’s information criterion
American Thoracic Society/the European Respiratory Society/the Japanese Respiratory Society/the Latin American Thoracic Association
areas under the time-dependent receiver operating characteristic curve
concordance index
Charlson Comorbidity Index score
chronic hypersensitivity pneumonia
collagen vascular disease-related interstitial pneumonia
gender / age / physiology
high-resolution computed tomography
idiopathic interstitial pneumonia
interstitial lung disease
idiopathic nonspecific interstitial pneumonia
interstitial pneumonia
idiopathic pulmonary fibrosis
lactate dehydrogenase
percentage predicted diffusion capacity of lung for carbon monoxide
percentage predicted forced vital capacity
receiver operating characteristic curve
usual interstitial pneumonia
Ethics approval and consent to participate
This study followed the guidelines of the Declaration of Helsinki and was approved by the institutional review board at Yokohama City University Hospital approved this study (approval number B190300005).
The need for Informed Consent was waived by the institutional review board at Yokohama City University due to the retrospective nature of the study.
Consent for publication
Not applicable.
Availability of data and materials
All data generated or analyzed during this study are included in this published article and its supplementary information files.
Competing interests
The authors declare that they have no competing interests.
Funding
Not funding.
Authors' contributions
(I) Conception and design
(II) Administrative support
(III) Provision of study materials or patients
(IV) Collection and assembly of data
(V) Data analysis and interpretation
(VI) Manuscript writing
(VII) Final approval of manuscript
FH: First author. (I: Conception and design), (II: Administrative support), (III: Provision of study materials or patients), (IV: Collection and assembly of data), (V: Data analysis and interpretation), (VI: Manuscript writing), and (VII: Final approval of manuscript).
HY: Corresponding author. (I), (II), (III), (IV), (V), (VI), and (VII).
SY: Professional statistician. (I), (II), (V), (VI), and (VII).
TY, MK, NR, AA, IA, and SK: (I), (III), (V), (VI), and (VII).
WK, HN, KN, and KT: (I), (II), (III), (VI), and (VII).
All authors have read and approved the final manuscript.
Acknowledgements
Not applicable.