Development and Internal Validation of a Retrospective Prediction Model for Fungal Infections in Patients with Chronic Obstructive Pulmonary Disease

Abstract

Background

Patients with chronic obstructive pulmonary disease (COPD) are at an elevated risk of fungal infections, yet reliable tools for early risk stratification are lacking. This study aimed to analyze the risk factors associated with fungal infections in patients with COPD and to develop a predictive model.

Methods

This study retrospectively analyzed patients with COPD who were admitted to Shanghai Pulmonary Hospital, between June 2021 and June 2023. Demographic and clinical data were extracted from electronic medical records. The risk factors were explored by univariate logistic regression and stepwise multivariate logistic regressions. Model performance was assessed using the receiver operating characteristic curve and area under the curve (AUC), along with calibration and decision curve analyses. The internal validation was performed using Bootstrap (n=1000) method.

Results

A total of 498 patients with COPD were enrolled, with 78 patients in the fungal infection group. Multivariable logistic regression analysis revealed that PaO2 within the normal range (OR=2.04, 95% CI: 1.12–3.71, P=0.02) and mold mix specific IgE above the normal threshold (OR=2.37, 95% CI: 1.04–5.36, P=0.04) were independently associated with an increased risk of fungal infections, whereas lymphocyte counts within the normal range (OR=0.37, 95% CI: 0.21–0.63, P<0.01) were associated with a reduced risk. Procalcitonin, albumin, and medical history of COPD duration were also included in predictive model based on the regression analysis. The AUC was 0.74 (95% CI: 0.68–0.79), and the Hosmer-Lemeshow goodness-of-fit test and decision curves indicated good differentiation and calibration.

Conclusion

This study identifies elevated PaO2, mold mix specific IgE and lymphocyte counts as risk factors for fungal infections. The predictive model incorporating PaO2, mold mix specific IgE, lymphocyte counts, procalcitonin, albumin, and medical history of COPD duration demonstrated good predictive performance, offering a potential tool for early identification of high-risk patients in clinical practice.

Introduction

Chronic obstructive pulmonary disease (COPD) is a condition in which chronic inflammation and swelling cause the inside of the breathing tubes (airways) to be smaller than normal.¹ The disease is a leading cause of morbidity and mortality worldwide, with a substantial impact on public health and healthcare systems. Globally, COPD accounted for approximately 212.3 million prevalent cases and 3.3 million deaths in 2019.² In China, it affected 45 million cases, with a significant increase in incidence compared with 1990.³

Patients with COPD need frequent visits and long-term treatment, which affects the quality of life. At the same time, patients with COPD are vulnerable to climate, environment, infection and other factors, leading to acute exacerbation.^4,5 Patients with COPD are at an increased risk of pulmonary fungal infections, which can exacerbate their respiratory symptoms and lead to significant morbidity and mortality.⁶ It has been reported that 70–80% of COPD patients may be related to lower respiratory tract fungal infection,⁷ with Aspergillus species being the most commonly implicated pathogens, and the treatment of deep fungal infection has always been a clinical difficulty, which is one of the main causes of infection leading to the death of critically ill patients. Invasive pulmonary aspergillosis (IPA), for example, is a severe form of fungal infection that can occur in COPD patients with compromised immunity. If not diagnosed and treated early, IPA can be almost always fatal.⁸ Moreover, fungal infections can lead to a worsening of COPD symptoms, an increase in the frequency of exacerbations, and a decline in lung function.⁹ Fungal infections can also cause significant inflammation, which can further damage the lung tissue and contribute to the progression of COPD. Fully revealing risk factors of fungal infections in COPD patients is important to help clinicians identify infection.

Previous research has identified several risk factors for fungal infections in COPD patients, and the susceptibility of COPD patients to fungal infections is multifactorial, primarily involving impaired mucociliary clearance due to chronic inflammation,⁹ immunosuppression from long-term use of inhaled or oral corticosteroids,⁶ and the presence of comorbidities such as diabetes and cardiovascular diseases that further compromise immune responses. In addition, increasing patient age, smoking habit, and severity of COPD were related to an increasing frequency of bacterial and fungal infections.¹⁰ Several predictive models have been developed for fungal infections in COPD patients. Han et al⁷ identified independent risk factors for lower respiratory tract fungal infection but did not construct a quantifiable predictive tool. Gu et al¹¹ developed a risk model for invasive pulmonary aspergillosis (IPA) in AECOPD patients based on factors such as GOLD stage III–IV, corticosteroid use, and hypoalbuminemia; however, this model focused on a single fungal pathogen and lacked inflammatory and allergic biomarkers. However, existing studies have largely focused on individual risk factors, and a comprehensive predictive model that incorporates these risk factors into a practical clinical tool remains an unmet need. Such a model could significantly reduce the risk of infection and prevent nosocomial infections, ultimately improving patient outcomes. To address this gap, this study will further complement the association between fungal infection and COPD and develop a predictive model for COPD patients.

Materials and Methods

Study Design and Patients

This retrospective study included patients with COPD who visited Shanghai Pulmonary Hospital between June 2021 and June 2023, a period selected to ensure consistent diagnostic criteria and clinical management protocols, thereby supporting the relevance of the findings to current clinical practice while allowing sufficient enrollment of patients with confirmed fungal infections for analysis. The inclusion criteria were: 1) Meeting the diagnostic criteria for COPD as defined by the Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines;¹² 2) Complete clinical data. The exclusion criteria were: 1) Patients with mental disorders or cognitive communication impairments; 2) Patients with malignant tumors; 3) Patients with hematological diseases or systemic infectious diseases; 4) Patients with severe hepatic or renal dysfunction; 5) Patients with other pulmonary diseases. This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Shanghai Pulmonary Hospital (Ethical Approval Number: K24-575).

According to the European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group (EORTC/MSG Study Group),¹³ we diagnosed fungal infection status using clinical, radiological, and/or microbiological findings. Patients categorized as “probable” or “proven” were judged to be fungal infections.

Data Collection

Demographic and clinical data from patients were systematically extracted from electronic medical records. Key demographic details included age, height, weight, and Body Mass Index (BMI). Sex was recorded as male or female, and occupation was categorized as worker/employee, military, retired/unemployed/other. Smoking and drinking habits were recorded in cigarette-years and median values, respectively, with current habits noted as yes/no.

Clinical data encompassed COPD medical history, exacerbation duration, aggravating factors, and frequency of exacerbations. Medical history of COPD was recorded in years as a continuous variable. Arterial Oxygen Tension (PaO2), Arterial Carbon Dioxide Tension (PaCO2), Pondus Hydrogenii (pH) were measured to assess respiratory function. The normal range for PaO2 is 80–100 mmHg. A broad range of inflammatory markers, including Serum amyloid A (SAA), White Blood Cells (WBCs), neutrophils, lymphocytes, monocytes, eosinophils, platelets, and hemoglobin, were quantified with median values and range categorization. The normal range for lymphocytes is 0.8–4.0×10⁹/L. SAA was measured using the rate method on an AU5800 automated biochemistry analyzer, with results interpreted according to the adult reference ranges of our hospital. Hepatic function was evaluated by albumin levels, with a normal range of 32–48 g/L, and cardiac health was assessed through Cardiac Enzyme Lactate Dehydrogenase (LDH), Cardiac Enzyme Creatine Kinase Isoenzyme (CK-MB), D-Dimer, and fibrin degradation products.

Cardiac function was further evaluated with B-Type Natriuretic Peptide (BNP) and troponin T, while systemic inflammation was gauged by Procalcitonin (PCT), Erythrocyte Sedimentation Rate (ESR), Interleukin-6 (IL-6), Interleukin-8 (IL-8), Soluble Interleukin-2 Receptor (sIL-2R), and Tumor Necrosis Factor-alpha (TNF-α). PCT was measured using a sandwich immunoassay on an automated hospital analyzer, with a normal value of ≤ 0.05 ng/mL. Humoral immunity was assessed via immunoglobulins including Immunoglobulin A (IgA), Immunoglobulin M (IgM), Immunoglobulin E (IgE), complement components (C3, C4), transferrin, and C-Reactive Protein (CRP). Reference ranges were as follows: IgG 7.0–16 g/L, IgA 0.7–4 g/L, IgM 0.4–2.3 g/L, IgE 0–100 g/L, complement C3 0.9–1.8 g/L, complement C4 0.1–0.4 g/L, transferrin 2–3.6 g/L. Cellular immunity was determined by CD3+, CD4+, and CD8+ T cell counts, and the CD4+/CD8+ ratio. Finally, fungal β-glucan levels and specific IgE responses to inhalant, food allergens, and molds (A. fumigatus, A. alternata) were measured to evaluate susceptibility to fungal infections and allergies. The normal threshold for mold mix specific IgE is ≤ 0.35 kUA/L. Fungal β-D-glucan (G test) was measured using the dynamic turbidimetric method. Galactomannan (GM test) was detected by enzyme-linked immunosorbent assay. Specific IgE was measured using the ImmunoCAP system, with positivity thresholds of <0.35 kUA/L for food allergens, <0.35 kUA/L for inhalant allergens, <0.35 kUA/L for mold mix, and 0.10 kUA/L for Aspergillus fumigatus. This comprehensive dataset provides a holistic view of health status in COPD patients, with a focus on immune and respiratory parameters.

Statistical Analysis

All statistical analyses were performed using R statistical software (Chicago, IL, USA 4.2.3). A p value < 0.05 was considered as statistically significant. A total of 78 patients with fungal infections were identified, yielding an event-per-variable ratio exceeding the recommended threshold of 10 for logistic regression analysis, thereby supporting the stability and reliability of the multivariable model. Additionally, bootstrap resampling with 1000 iterations was employed for internal validation to further mitigate overfitting and provide robust estimates of model performance. The continuous variables with a normal distribution was expressed as Mean±SD, and compared by independent sample t-test. The continuous variables with a skewed distribution was reported as median (M) and quartile (P25, P75), and compared by Mann–Whitney test. The categorical variables were reported as n (%), and compared by the Fisher exact test or χ² test.

Taking fungal infection as the dependent variable, a univariate logistic regression model was constructed using the autoReg package in R, incorporating patients’ demographic characteristics, blood gas analysis, infectious indicators, liver function, cardiac enzymes, coagulation profiles, cardiac function, serum humoral immunity, allergen serum, and mold mix as influencing factors. Factors with a P-value < 0.05 in the univariate logistic regression were included in a backward stepwise multivariate analysis to eliminate collinearity and to explore the potential influencing factors of fungal infection.

From the multivariate logistic regression model, we selected significant factors to construct a predictive model. The pROC package, rms package, and ggplot2 were used to evaluate the model’s predictive performance with the area under the curve (AUC) as the criterion. Receiver operating characteristic (ROC) curves, calibration curves, and decision curves for the model were plotted, and a nomogram for the model was constructed. For internal validation of the constructed predictive model, we employed the boot package, caret package, foreign package, and rms package in R language to perform 1000 times of Bootstrap resampling internal validation on the constructed predictive model. Bootstrap (n=1000) was used to determine the 95% confidence intervals of diagnostic efficiency. Furthermore, bootstrap (n=1000) was also used for internal validation. The AUC was used as the criterion for evaluating the model’s predictive performance. The ROC curve for the Bootstrap resampling internal validation was plotted, and the accuracy, Kappa value, sensitivity, and specificity were calculated.

For the missing data, we used Original dataset for baseline table analysis, and we have included all missing data details as a Supplementary Material. Nevertheless, in the regression, we used Multiple interpolation method to process the missing data.

Results

Clinical Characteristics of Patients with COPD

A total of 867 patients with COPD were treated during this period. There were 546 patients eligible for inclusion, 48 were excluded, and 498 were finally enrolled (Figure S1). Among 498 patients, 420 were divided into No fungal infection group, 78 were divided into Fungal infection. Specific demographic and clinical characteristics were compared as shown in Table 1. Compared with patients without fungal infection, those with fungal infection had a longer COPD medical history (P=0.018), higher levels of serum amyloid A, procalcitonin, ESR, BNP, and CRP (all P<0.001), and a higher proportion of above-normal mold mix specific IgE (P=0.009). Conversely, the fungal infection group showed lower hemoglobin, albumin, and lymphocyte counts (P=0.01 for each). The detailed normal ranges for the laboratory indicators are presented in Supplementary Table 1. There were no significant differences in age, BMI, Smoking history and Drinking history between No fungal infection group and Fungal infection group.

Table 1.

Clinical Characteristics of Patients

Variable	Overall (N=498)	Missing	No fungal Infection (N=420)	Fungal Infection (N=78)	p
Age	68.00 [63.00, 74.00]	0 (0.00)	68.00 [63.00, 74.00]	70.00 [64.00, 77.00]	0.35
Length of stay days	7.00 [6.00, 10.00]	8 (0.02)	7.00 [6.00, 9.00]	8.00 [6.00, 11.00]	0.07
Height (cm)	169.00 [165.00, 172.00]	42 (0.08)	169.00 [163.00, 172.00]	168.00 [165.00, 175.00]	0.16
Weight (kg)	60.00 [51.00, 68.00]	43 (0.09)	60.00 [51.00, 68.00]	60.00 [50.75, 65.00]	0.46
BMI	21.44 (3.77)	43 (0.09)	21.56 (3.83)	20.73 (3.32)	0.07
Smoking History cigarette years	500.00 [0.00, 800.00]	7 (0.01)	400.00 [0.00, 800.00]	600.00 [0.00, 875.00]	0.65
Medical History of COPD years	10.00 [5.00, 12.00]	2 (0.00)	10.00 [5.00, 10.00]	10.00 [6.25, 20.00]	0.02
Time to Disease Exacerbation days	20.00 [7.00, 30.00]	18 (0.04)	20.00 [7.00, 30.00]	30.00 [8.50, 30.00]	0.37
Time with More than 2 Exacerbations per Year years	0.00 [0.00, 1.00]	7 (0.01)	0.00 [0.00, 1.00]	0.00 [0.00, 2.00]	0.64
PaO2	79.30 [68.45, 94.60]	2 (0.00)	78.45 [68.10, 94.15]	82.00 [70.00, 98.80]	0.17
PaCO2	43.90 [39.20, 51.92]	2 (0.00)	43.75 [39.40, 51.80]	44.05 [37.38, 54.95]	0.97
Arterial pH	7.40 [7.38, 7.43]	2 (0.00)	7.41 [7.38, 7.43]	7.40 [7.37, 7.43]	0.38
Serum amyloid A	12.15 [6.75, 41.56]	88 (0.18)	10.92 [6.65, 35.52]	35.37 [9.50, 222.66]	<0.01
WBCs	7.29 [5.54, 9.71]	4 (0.01)	7.17 [5.46, 9.39]	8.46 [6.19, 12.24]	0.01
Neutrophils	4.53 [3.20, 6.93]	3 (0.01)	4.47 [3.19, 6.71]	5.04 [3.50, 8.54]	0.06
Lymphocytes	1.15 [0.80, 1.65]	3 (0.01)	1.18 [0.85, 1.64]	0.87 [0.66, 1.65]	0.01
Monocytes	0.53 [0.38, 0.70]	3 (0.01)	0.53 [0.38, 0.70]	0.56 [0.38, 0.77]	0.43
Eosinophils	0.10 [0.01, 0.22]	3 (0.01)	0.10 [0.02, 0.23]	0.05 [0.01, 0.15]	0.03
Platelets	205.00 [162.00, 258.50]	3 (0.01)	205.00 [162.00, 258.00]	213.00 [158.00, 259.25]	0.94
Hemoglobin	135.00 [124.00, 145.75]	4 (0.01)	136.00 [126.00, 146.00]	131.00 [109.75, 142.50]	0.01
Albumin	39.00 [37.00, 42.00]	5 (0.01)	40.00 [37.00, 42.00]	38.00 [35.00, 42.00]	0.01
Blood Glucose	5.80 [4.90, 7.27]	64 (0.13)	5.60 [4.80, 7.00]	6.40 [5.15, 8.10]	0.03
LDH	170.00 [151.00, 203.25]	58 (0.12)	169.00 [151.00, 199.75]	176.50 [145.75, 214.00]	0.26
CK MB	12.00 [9.00, 16.00]	44 (0.09)	12.00 [10.00, 16.00]	13.00 [8.00, 20.00]	0.22
Coagulation D Dimer	139.00 [80.50, 307.00]	11 (0.02)	135.00 [79.00, 279.00]	180.50 [98.25, 483.00]	0.01
Fibrin Degradation Products	1.19 [0.72, 3.24]	25 (0.05)	1.17 [0.69, 3.00]	2.18 [0.91, 3.79]	0.01
BNP	122.90 [49.60, 383.80]	13 (0.03)	106.50 [40.30, 335.20]	286.75 [98.50, 1395.50]	<0.01
Cardiac troponin T	0.01 [0.01, 0.02]	22 (0.04)	0.01 [0.01, 0.02]	0.01 [0.01, 0.02]	0.05
Procalcitonin	0.04 [0.04, 0.05]	212 (0.43)	0.04 [0.04, 0.05]	0.04 [0.04, 0.12]	<0.01
ESR	28.00 [15.00, 54.00]	138 (0.28)	27.00 [13.00, 50.00]	52.00 [21.00, 84.00]	<0.01
Serum IgA	2.35 [1.69, 3.05]	94 (0.19)	2.34 [1.69, 3.04]	2.42 [1.69, 3.22]	0.51
Serum IgM	0.77 [0.54, 1.12]	94 (0.19)	0.77 [0.55, 1.12]	0.74 [0.53, 1.09]	0.82
Serum IgE	56.35 [19.02, 152.50]	92 (0.18)	59.50 [19.40, 143.00]	51.90 [17.80, 186.50]	0.80
Serum Complement C3	1.08 [0.96, 1.24]	93 (0.19)	1.09 [0.96, 1.25]	1.07 [0.94, 1.17]	0.28
Serum Complement C4	0.23 [0.19, 0.29]	93 (0.19)	0.23 [0.20, 0.30]	0.22 [0.17, 0.28]	0.08
Serum Transferrin	1.93 [1.69, 2.24]	94 (0.19)	1.95 [1.72, 2.27]	1.85 [1.56, 2.01]	<0.01
Serum CRP	3.30 [3.10, 19.32]	68 (0.14)	3.30 [3.10, 14.80]	17.20 [3.30, 50.00]	<0.01
Fungal β Glucan	54.20 [37.27, 114.25]	242 (0.49)	51.40 [37.40, 111.20]	55.60 [34.20, 120.60]	0.65
Total IgE	67.00 [19.73, 173.50]	64 (0.13)	68.15 [21.83, 173.50]	55.25 [14.93, 177.25]	0.89
Inhalant Allergens	0.06 [0.02, 0.15]	63 (0.13)	0.06 [0.02, 0.15]	0.05 [0.02, 0.16]	0.57
Food Allergens	0.03 [0.02, 0.07]	63 (0.13)	0.03 [0.02, 0.07]	0.03 [0.01, 0.07]	0.70
Aspergillus fumigatus	0.01 [0.00, 0.05]	63 (0.13)	0.01 [0.00, 0.04]	0.02 [0.00, 0.09]	0.32
Alternaria alternata	0.01 [0.00, 0.03]	63 (0.13)	0.01 [0.00, 0.02]	0.02 [0.01, 0.03]	0.02
Mold mix specific IgE	0.04 [0.02, 0.09]	65 (0.13)	0.04 [0.02, 0.08]	0.05 [0.03, 0.17]	0.04
Sex		0 (0.00)			0.80
Male	465 (93.4)		391 (93.1)	74 (94.9)
Female	33 (6.6)		29 (6.9)	4 (5.1)
Profession		0 (0.00)			0.47
Worker/employee	18 (3.6)		15 (3.6)	3 (3.8)
Active duty military	7 (1.4)		5 (1.2)	2 (2.6)
Retired/unemployed/other	473 (95.0)		400 (95.2)	73 (93.6)
Drinking history					0.20
No	464 (93.7)		390 (93.1)	74 (97.4)
Yes	31 (6.3)		29 (6.9)	2 (2.6)
Aggravating reason this time					0.10
No obvious cause	86 (17.6)		71 (17.1)	15 (20.0)
Seasonal change	58 (11.9)		51 (12.3)	7 (9.3)
Cold/flu/bacteria/virus infection	301 (61.6)		260 (62.8)	41 (54.7)
Others	44 (9.0)		32 (7.7)	12 (16.0)
PaO2 range					0.07
Below normal	256 (51.6)		225 (53.8)	31 (39.7)
Normal	139 (28.0)		111 (26.6)	28 (35.9)
Above normal	101 (20.4)		82 (19.6)	19 (24.4)
PaCO2 range					0.21
Below normal	471 (95.0)		399 (95.5)	72 (92.3)
Normal	17 (3.4)		14 (3.3)	3 (3.8)
Above normal	8 (1.6)		5 (1.2)	3 (3.8)
Arterial pH range					0.27
Below normal	65 (13.1)		51 (12.2)	14 (17.9)
Normal	390 (78.6)		334 (79.9)	56 (71.8)
Above normal	41 (8.3)		33 (7.9)	8 (10.3)
Serum amyloid A range					0.01
Normal	181 (44.1)		164 (47.0)	17 (27.9)
Above normal	229 (55.9)		185 (53.0)	44 (72.1)
WBCs range					0.03
Below normal	12 (2.4)		11 (2.6)	1 (1.3)
Normal	352 (71.3)		305 (73.3)	47 (60.3)
Above normal	130 (26.3)		100 (24.0)	30 (38.5)
Neutrophils range					0.04
Below normal	65 (13.1)		54 (12.9)	11 (14.1)
Normal	334 (67.5)		290 (69.5)	44 (56.4)
Above normal	96 (19.4)		73 (17.5)	23 (29.5)
Lymphocytes range					<0.01
Below normal	124 (25.1)		88 (21.1)	36 (46.2)
Normal	369 (74.5)		327 (78.4)	42 (53.8)
Above normal	2 (0.4)		2 (0.5)	0 (0.0)
Monocytes range					0.23
Below normal	9 (1.8)		9 (2.2)	0 (0.0)
Normal	397 (80.2)		337 (80.8)	60 (76.9)
Above normal	89 (18.0)		71 (17.0)	18 (23.1)
Eosinophils range					0.04
Below normal	183 (37.0)		146 (35.0)	37 (47.4)
Normal	238 (48.1)		203 (48.7)	35 (44.9)
Above normal	74 (14.9)		68 (16.3)	6 (7.7)
Platelets range					0.66
Below normal	52 (10.5)		42 (10.1)	10 (12.8)
Normal	405 (81.8)		344 (82.5)	61 (78.2)
Above normal	38 (7.7)		31 (7.4)	7 (9.0)
Hemoglobin range					0.02
Below normal	174 (35.2)		136 (32.7)	38 (48.7)
Normal	309 (62.6)		270 (64.9)	39 (50.0)
Above normal	11 (2.2)		10 (2.4)	1 (1.3)
Albumin range					<0.01
Below normal	26 (5.3)		15 (3.6)	11 (14.1)
Normal	461 (93.5)		394 (94.9)	67 (85.9)
Above normal	6 (1.2)		6 (1.4)	0 (0.0)
Blood Glucose range					0.01
Below normal	23 (5.3)		17 (4.7)	6 (8.3)
Normal	212 (48.8)		188 (51.9)	24 (33.3)
Above normal	199 (45.9)		157 (43.4)	42 (58.3)
LDH range					0.70
Below normal	17 (3.9)		14 (3.8)	3 (4.3)
Normal	360 (81.8)		305 (82.4)	55 (78.6)
Above normal	63 (14.3)		51 (13.8)	12 (17.1)
CK MB range					0.35
Below normal	450 (99.1)		378 (99.2)	72 (98.6)
Normal	2 (0.4)		1 (0.3)	1 (1.4)
Above normal	2 (0.4)		2 (0.5)	0 (0.0)
Fibrin Degradation Products range					0.19
Normal	63 (13.3)		57 (14.2)	6 (8.5)
Above normal	410 (86.7)		345 (85.8)	65 (91.5)
BNP range					<0.01
Normal	244 (50.3)		218 (53.3)	26 (34.2)
Above normal	241 (49.7)		191 (46.7)	50 (65.8)
Cardiac troponin T range					0.02
Normal	269 (56.5)		236 (58.7)	33 (44.6)
Above normal	207 (43.5)		166 (41.3)	41 (55.4)
Procalcitonin range					<0.01
Normal	225 (78.7)		196 (82.4)	29 (60.4)
Above normal	61 (21.3)		42 (17.6)	19 (39.6)
ESR range					0.04
Normal	95 (26.4)		86 (28.5)	9 (15.5)
Above normal	265 (73.6)		216 (71.5)	49 (84.5)
Serum IgA range					0.60
Below normal	399 (98.8)		333 (98.8)	66 (98.5)
Normal	3 (0.7)		2 (0.6)	1 (1.5)
Above normal	2 (0.5)		2 (0.6)	0 (0.0)
Serum IgM range					0.32
Below normal	36 (8.9)		32 (9.5)	4 (6.0)
Normal	354 (87.6)		295 (87.5)	59 (88.1)
Above normal	14 (3.5)		10 (3.0)	4 (6.0)
Serum IgE range					0.34
Normal	257 (63.3)		218 (64.3)	39 (58.2)
Above normal	149 (36.7)		121 (35.7)	28 (41.8)
Serum Complement C3 range					0.38
Below normal	70 (17.3)		55 (16.3)	15 (22.4)
Normal	331 (81.7)		279 (82.5)	52 (77.6)
Above normal	4 (1.0)		4 (1.2)	0 (0.0)
Serum Complement C4 range					0.05
Below normal	5 (1.2)		3 (0.9)	2 (3.0)
Normal	384 (94.8)		319 (94.4)	65 (97.0)
Above normal	16 (4.0)		16 (4.7)	0 (0.0)
Serum Transferrin range					0.02
Below normal	233 (57.7)		184 (54.6)	49 (73.1)
Normal	169 (41.8)		151 (44.8)	18 (26.9)
Above normal	2 (0.5)		2 (0.6)	0 (0.0)
Serum CRP range					<0.01
Normal	273 (63.5)		242 (67.0)	31 (44.9)
Above normal	157 (36.5)		119 (33.0)	38 (55.1)
Fungal β Glucan range		65 (0.13)			0.38
Normal	362 (83.6)		304 (84.4)	58 (79.5)
Above normal	71 (16.4)		56 (15.6)	15 (20.5)
Total IgE range					0.55
Normal	209 (48.2)		172 (47.5)	37 (51.4)
Above normal	225 (51.8)		190 (52.5)	35 (48.6)
Inhalant Allergens range					0.59
Normal	379 (87.1)		314 (86.7)	65 (89.0)
Above normal	56 (12.9)		48 (13.3)	8 (11.0)
Food Allergens range					1
Normal	426 (97.9)		354 (97.8)	72 (98.6)
Above normal	9 (2.1)		8 (2.2)	1 (1.4)
Aspergillus fumigatus range					0.17
Normal	358 (82.3)		302 (83.4)	56 (76.7)
Above normal	77 (17.7)		60 (16.6)	17 (23.3)
Alternaria alternata range					0.07
Normal	408 (93.8)		343 (94.8)	65 (89.0)
Above normal	27 (6.2)		19 (5.2)	8 (11.0)
Mold mix specific IgE range					0.01
Normal	397 (91.7)		338 (93.4)	59 (83.1)
Above normal	36 (8.3)		24 (6.6)	12 (16.9)

Univariate and Multivariate Analysis of Fungal Infections in Patients with COPD

Our univariate analysis identified significant correlations between fungal infections in COPD patients and several physiological and biochemical parameters, including partial PaO2, albumin, BNP, troponin T, and PCT (P < 0.05). Inflammatory markers such as lymphocyte, eosinophil, serum transferrin, and Hb levels were inversely associated with fungal infection risk, suggesting a protective role. Conversely, elevated SAA levels were identified as a risk factor for fungal infections. Additionally, increased CRP levels and allergen serum reactivity to mold mixtures were linked to a higher risk of fungal infection. A comprehensive overview of these associations, along with detailed statistical data, is presented in Table 2.

Table 2.

Univariate Logistic Regression Analysis for Independent Risks for Fungal Infection in Patients with COPD

Variable	OR (95% CI)	P
Age	1.010 (0.980, 1.040)	0.40
Sex
Male	1 00 (Reference)
Female	0.730 (0.250, 2.140)	0.56
Profession
Worker/employee	1 00 (Reference)
Active duty military	2.000 (0.250, 15.700)	0.51
Retired/unemployed/other	0.910 (0.260, 3.240)	0.89
Length of stay days	1.030 (0.980, 1.070)	0.24
Height (cm)	1.030 (0.990, 1.070)	0.12
Weight (kg)	0.990 (0.960, 1.010)	0.34
BMI	0.940 (0.880, 1.010)	0.09
Smoking History cigarette years	1.000 (1.000, 1.000)	0.30
Drinking history	0.390 (0.090, 1.680)	0.20
Aggravating reason this time
No obvious cause	1 00 (Reference)
Seasonal change	0.670 (0.250, 1.780)	0.42
Cold/flu/bacteria/virus infection	0.770 (0.400, 1.480)	0.44
Others	1.780 (0.740, 4.270)	0.20
Medical History of COPD years	1.030 (1.010, 1.050)	0.01
Time to Disease Exacerbation days	1.000 (1.000, 1.000)	0.94
Time with More than 2 Exacerbations per Year years	1.080 (0.960, 1.210)	0.19
PaO2 range
Below normal	1 00 (Reference)
Normal	1.840 (1.050, 3.220)	0.03
Above normal	1.680 (0.900, 3.140)	0.10
PaCO2 range
Below normal	1 00 (Reference)
Normal	1.190 (0.330, 4.270)	0.79
Above normal	3.340 (0.780, 14.340)	0.10
Arterial pH range
Below normal	1 00 (Reference)
Normal	0.610 (0.320, 1.180)	0.14
Above normal	0.880 (0.330, 2.340)	0.80
Serum amyloid A range
Normal	1 00 (Reference)
Above normal	2.260 (1.210, 4.230)	0.01
WBCs range
Below normal	1 00 (Reference)
Normal	1.670 (0.210, 13.320)	0.63
Above normal	3.290 (0.410, 26.670)	0.26
Neutrophils range
Below normal	1 00 (Reference)
Normal	0.740 (0.360, 1.520)	0.41
Above normal	1.550 (0.690, 3.450)	0.29
Lymphocytes range
Below normal	1 00 (Reference)
Normal	0.310 (0.190, 0.520)	<0.01
Above normal	0.000 (0.000, Inf)	0.98
Monocytes range
Below normal	1 00 (Reference)
Normal	2763632.380 (0.000, Inf)	0.99
Above normal	3956866.730 (0.000, Inf)	0.98
Eosinophils range
Below normal	1 00 (Reference)
Normal	0.680 (0.410, 1.130)	0.14
Above normal	0.350 (0.140, 0.860)	0.02
Platelets range
Below normal	1 00 (Reference)
Normal	0.750 (0.350, 1.570)	0.44
Above normal	0.960 (0.330, 2.800)	0.94
Hemoglobin range
Below normal	1 00 (Reference)
Normal	0.510 (0.310, 0.840)	0.01
Above normal	0.350 (0.040, 2.860)	0.33
Albumin range
Below normal	1 00 (Reference)
Normal	0.230 (0.100, 0.520)	<0.01
Above normal	0.000 (0.000, Inf)	0.98
Blood Glucose range
Below normal	1 00 (Reference)
Normal	0.390 (0.140, 1.090)	0.07
Above normal	0.780 (0.290, 2.140)	0.63
LDH range
Below normal	1 00 (Reference)
Normal	0.800 (0.220, 2.870)	0.73
Above normal	1.140 (0.290, 4.570)	0.85
CK MB range
Below normal	1 00 (Reference)
Normal	4.710 (0.290, 77.770)	0.28
Above normal	0.000 (0.000, Inf)	0.98
Fibrin Degradation Products range
Normal	1 00 (Reference)
Above normal	1.910 (0.790, 4.620)	0.15
BNP range
Normal	1 00 (Reference)
Above normal	2.310 (1.390, 3.840)	<0.01
Cardiac troponin T range
Normal	1 00 (Reference)
Above normal	1.820 (1.100, 3.020)	0.02
Procalcitonin range
Normal	1 00 (Reference)
Above normal	2.800 (1.460, 5.360)	<0.01
ESR range
Normal	1 00 (Reference)
Above normal	1.990 (0.950, 4.170)	0.07
Serum IgA range
Below normal	1 00 (Reference)
Normal	2.300 (0.210, 25.700)	0.50
Above normal	0.000 (0.000, Inf)	0.99
Serum IgM range
Below normal	1 00 (Reference)
Normal	1.510 (0.480, 4.750)	0.48
Above normal	3.000 (0.630, 14.230)	0.17
Serum IgE range
Normal	1 00 (Reference)
Above normal	1.370 (0.820, 2.310)	0.23
Serum Complement C3 range
Below normal	1 00 (Reference)
Normal	0.640 (0.310, 1.280)	0.20
Above normal	0.000 (0.000, Inf)	0.98
Serum Complement C4 range
Below normal	1 00 (Reference)
Normal	0.330 (0.050, 2.090)	0.24
Above normal	0.000 (0.000, Inf)	0.98
Serum Transferrin range
Below normal	1 00 (Reference)
Normal	0.560 (0.320, 0.970)	0.04
Above normal	0.000 (0.000, Inf)	0.98
Serum CRP range
Normal	1 00 (Reference)
Above normal	2.430 (1.460, 4.050)	<0.01
Fungal β Glucan range
Normal	1 00 (Reference)
Above normal	1.490 (0.780, 2.850)	0.23
Total IgE range
Normal	1 00 (Reference)
Above normal	1.000 (0.610, 1.640)	0.99
Inhalant Allergens range
Normal	1 00 (Reference)
Above normal	0.840 (0.380, 1.890)	0.68
Food Allergens range
Normal	1 00 (Reference)
Above normal	0.600 (0.070, 4.870)	0.63
Aspergillus fumigatus range
Normal	1 00 (Reference)
Above normal	1.630 (0.880, 3.030)	0.12
Alternaria alternata range
Normal	1 00 (Reference)
Above normal	2.260 (0.950, 5.400)	0.06
Mold mix specific IgE range
Normal	1 00 (Reference)
Above normal	2.900 (1.380, 6.070)	<0.01

Multivariable logistic regression was used to investigate the risk factor of fungal infection for patients with COPD. Adjusted odds ratios (AOR) and 95% CI were calculated and the results are shown in Table 3. Our findings reveal that among COPD patients, PaO2 within the normal range (OR=2.04, 95% CI: 1.12–3.71, P=0.02) and mold mix specific IgE above the normal threshold (OR=2.37, 95% CI: 1.04–5.36, P=0.04) were significantly and independently associated with an increased risk of fungal infections, whereas lymphocyte counts within the normal range (OR=0.37, 95% CI: 0.21–0.63, P<0.01) were associated with a reduced risk. In addition, procalcitonin, albumin, and medical history of COPD duration were also included in the predictive model based on the regression analysis.

Table 3.

Multivariate Logistic Regression Analysis for Independent Risks for Fungal Infection in Patients with COPD

Variables	OR (95% CI)	P
Medical History of COPD years	1.02 (1.00, 1.04)	0.09
PaO2 range
Below normal	1 00 (Reference)
Normal	2.04 (1.12, 3.71)	0.02
Above normal	1.28 (0.65, 2.50)	0.48
Lymphocytes range
Below normal	1 00 (Reference)
Normal	0.37 (0.21, 0.63)	<0.01
Above normal	0.00 (0.00, Inf)	0.99
Albumin range
Below normal	1 00 (Reference)
Normal	0.46 (0.17, 1.22)	0.12
Above norma	0.00 (0.00, Inf)	0.99
Procalcitonin range
Normal	1 00 (Reference)
Above normal	1.74 (0.83, 3.64)	0.14
Mold mix specific IgE range
Normal	1 00 (Reference)
Above normal	2.37 (1.04, 5.36)	0.04

Construction and Validation of Prediction Model

Utilizing a combination of regression analysis and multivariate Logistic regression, we identified six optimal predictors (including Media History of COPD years, PaO2 range, Lymphocytes range, Albumin range, Procalcitonin range and Mold mix specific IgE range) which were subsequently incorporated into a graphical representation (Figure 1). The composite score, derived from the sum of these individual predictors, serves as an index for the likelihood of fungal infection development in patients with COPD. ROC curves were constructed to further assess the predictive accuracy of our model (Figure 2). The AUC was 0.74 (95% CI: 0.68–0.79), which is indicative of the model’s ability to distinguish between those who will and will not develop fungal infections. The Hosmer-Lemeshow test showed good calibration with a P-values of 0.468 (Figure 3). This non-significant result suggests that there is no statistically significant difference between the predicted probabilities and the observed outcomes, affirming the model’s calibration. The clinical utility of our model is further supported by the decision curve analysis presented in Figure 4. The curve lies above both the “None” and “All” lines, signifying that the model’s predictive strategy surpasses both the strategy of treating no patients and treating all patients, thereby confirming its practical value in clinical decision-making. To substantiate the robustness of our model, we employed Bootstrap resampling for 1000 iterations. The resulting C-index for the nomogram model was 0.74 (95% CI: 0.68–0.79), indicating a strong discriminatory power. The model demonstrated an accuracy of 0.56, a Kappa value of 0.18, a sensitivity of 0.50, and a specificity of 0.86, suggesting a general consistency and reliability of the model in predicting fungal infections (Figure 5). These findings collectively endorse the utility and validity of our predictive model in the context of COPD-associated fungal infections.

Figure 1.

A nomogram of a predictive model for fungal infection of COPD.

Figure 2.

ROC curves of a predictive model for fungal infection of COPD.

Figure 3.

Calibration curve.

Figure 4.

Decision curve.

Figure 5.

ROC curves validated by bootstrap.

Discussion

In this study, we performed a population-based study of the association between COPD and adverse outcomes of fungal infections in a cohort of 498 COPD patients. Our model, which includes PaO2, mold mix specific IgE and lymphocyte counts as key predictors, demonstrates a strong discriminatory power with an AUC of 0.74. These findings are clinically significant as they provide a tool for early identification of COPD patients at higher risk of fungal infections, thereby allowing for preemptive clinical interventions and potentially improving patient outcomes.

Our research suggests that COPD patients with PaO2 within the normal range had a higher risk compared with those with below normal PaO2. Several explanations may account for this observation. First, patients with normal oxygenation may represent a subgroup with preserved respiratory function but greater exposure to other risk factors, such as frequent exacerbations or corticosteroid use. Second, normal PaO2 values in the setting of chronic lung disease may reflect a compensatory hyperventilation that does not necessarily indicate immune competence.¹⁴ Previous studies have demonstrated that the use of broad-spectrum antibiotics and systemic corticosteroids-common in COPD management are independent risk factors for invasive pulmonary aspergillosis (IPA).¹¹ Additionally, patients with near normal oxygenation may be more likely to receive inhaled corticosteroids or undergo procedures that facilitate fungal colonization.¹⁵ Thus, the association between normal PaO2 and increased fungal infection risk in our study may highlight the complex interplay between oxygen status, treatment patterns, and host susceptibility.

Lymphocyte count within the normal range was identified as a protective factor against fungal infections. This finding aligns with the well established role of lymphocytes in antifungal immunity. A study by Liang et al¹⁶ demonstrated that among IPA patients with different structural lung diseases, those with COPD had the lowest cell mediated immunity, as reflected by significantly reduced CD4⁺CD8⁺ double positive cells compared to patients with interstitial lung disease or bronchiectasis. T lymphocytes, particularly CD4⁺ T cells, orchestrate the host response against fungal pathogens by producing cytokines such as interferon‑γ and interleukin-17, which activate macrophages and neutrophils to eliminate fungal elements.¹⁷ In COPD patients, chronic inflammation and corticosteroid use frequently lead to lymphopenia, which has been associated with increased risk of both bacterial and fungal infections. The protective effect of normal lymphocyte counts observed in our study underscores the importance of preserving cellular immune function in COPD patients and suggests that monitoring lymphocyte levels may aid in identifying individuals at lower risk of fungal complications.

Our study indicated that mold mix specific IgE above the normal threshold was independently associated with an increased risk of fungal infections. This finding is consistent with the growing body of literature linking fungal sensitization to adverse outcomes in COPD. Sensitization to molds such as Aspergillus fumigatus and Alternaria alternata has been shown to correlate with more frequent exacerbations, worse lung function, and increased airway inflammation.^18,19 Adam et al reported that Aspergillus fumigatus induces eosinophil activation through specific signaling cascades involving Src and PI3K kinases, and that inhibition of these pathways prevents A. fumigatus‑induced eosinophil activation and Th2 cytokine/chemokine secretion,²⁰ while O’Dea et al demonstrated that repeated airway exposure to a high‑chitin‑expressing A. fumigatus isolate induced significant airway eosinophilia in mice, with eosinophil‑deficient animals showing reduced morbidity and fungal burden.²¹ The underlying mechanisms are likely multifaceted. First, mold allergens trigger eosinophilic inflammation and mucus hypersecretion, which can impair mucociliary clearance and facilitate fungal colonization.²² Second, persistent allergic inflammation may disrupt the epithelial barrier integrity, allowing fungal elements to invade deeper airways.²² Third, there is evidence that mold sensitization is associated with reduced expression of antimicrobial peptides in the airway epithelium, further compromising host defense against fungal pathogens.²³ Previous predictive models for IPA in AECOPD patients have identified risk factors such as GOLD stage III–IV lung function, broad‑spectrum antibiotic use >10 days, high‑dose corticosteroid exposure, and hypoalbuminemia;¹¹ however, these models did not incorporate allergic sensitization parameters. Our findings support the incorporation of mold allergen testing into routine risk assessment for COPD patients and highlight the potential value of allergen avoidance strategies as part of comprehensive infection prevention.

In our study, although Hemoglobin and Troponin-T were not identified as independent risk factors for fungal infections, they were included in the final logistic regression model to construct the nomogram. Hemoglobin levels are directly related to the oxygen-carrying capacity in COPD patients, and hypoxemia, a common pathological state in COPD, may affect the immune response and susceptibility to infections.²⁴ Additionally, Troponin-T, as a biomarker of myocardial injury, may reflect the increased cardiac burden due to chronic inflammation and cor pulmonale in COPD patients.²⁵ Due to the complexity of the pathophysiological mechanisms involved, Hemoglobin and Troponin-T may not emerge as independent risk factors in our analysis. However, their incorporation into the model contributes to the model’s acceptable predictive value, indicating that they may play a role in the intricate interplay of factors affecting the risk of fungal infections in COPD patients. This aligns with the understanding that multiple factors can influence the immune status and infection susceptibility in COPD patients.

Our predictive model provides a powerful tool for assessing the risk of fungal infections in patients with COPD. A previous study developed and validated a risk model using readily available clinical parameters to predict IPA in patients with AECOPD.¹¹ Through a retrospective cohort study, factors independently associated with IPA were identified, such as GOLD stage III–IV lung function, the use of broad-spectrum antibiotics for more than 10 days in the previous month, oral or intravenous corticosteroid (prednisone) use exceeding 265 mg in the past 3 months, and serum albumin <30 g/L. Currently, numerous studies have discussed predictive models for COPD exacerbations, which help identify patients at higher risk of exacerbations, allowing for preemptive intervention.²⁶ In contrast to these models, our model focused on fungal infections, which is a relatively new area in COPD management.²⁷ In clinical practice, our model could be implemented at the time of hospital admission or during routine follow-up visits to stratify patients into low- and high-risk categories. For high-risk patients, clinicians could consider enhanced surveillance, early mycological screening, and judicious use of antifungal prophylaxis when clinically indicated. Furthermore, the model’s predictors are routinely available or easily obtainable in most clinical settings, facilitating seamless integration into electronic medical record systems for automated risk alerts. Our model may also have potential application value in predicting the risk of other infections in COPD patients, as chronic inflammation and immunosuppressive states are common in COPD patients, making them more susceptible to various infections.

There are still some limitations to this study. A major limitation is the single-center retrospective design, which may introduce potential biases, such as recall, information, and selection biases, and limit the generalizability of our findings to broader COPD populations treated in different healthcare settings. Although we performed bootstrap resampling for internal validation to mitigate overfitting, the single-center retrospective design and the limited number of fungal infection cases remain inherent limitations. We acknowledged the need for external validation in future multi-center prospective studies to confirm the generalizability of our findings. Additionally, the single-center design may reflect specific local patient demographics and clinical practices, further underscoring the importance of validating this model across diverse populations. Additionally, the constraints of our sample size may introduce variability into our results, suggesting a need for replication studies to ascertain the stability and reliability of our findings. Moreover, the study was limited by sparse data in certain subgroups and a high proportion of missing data for certain variables. Regrettably, the study’s scope did not encompass detailed dietary and lifestyle data for the patients, which could be pivotal in understanding the complex interplay of risk factors associated with fungal infections in COPD. In this study, our medical records lacked certain variables such as pulmonary heart disease, hypoalbuminemia, and the use of antibiotics, which have been identified in previous predictive models as risk factors for fungal infections in COPD patients.⁷ Consequently, we were unable to further analyze the impact of these factors in our study. Furthermore, the term “fungal infection” used throughout this study is broad; although Aspergillus infections (particularly invasive pulmonary aspergillosis) were the predominant type observed in our cohort, other fungal pathogens were not specifically distinguished due to the retrospective design and the model’s intended scope. Future studies should validate the predictive model separately for different fungal species to enhance specificity. There is a pressing need for robust, prospective, multicenter cohort studies to refine and bolster the predictive model for fungal infection complications in COPD across diverse regions of China. These future endeavors will be instrumental in elucidating the multifaceted risk variables and enhancing the precision of our predictive capabilities in this critical area of respiratory medicine.

Conclusions

In conclusion, our study identified elevated PaO2, and mold mix specific IgE as risk factors for fungal infections, and normal lymphocyte counts as a protective factor. It also developed and internally validated a predictive model for fungal infections in COPD patients, suggesting its potential as a supportive tool for risk stratification pending further validation. The model demonstrated acceptable discriminatory ability upon internal validation; however, its clinical applicability requires confirmation in prospective, multi-center studies. Future research should focus on externally validating this model in diverse populations and exploring the potential of integrating additional biomarkers and leveraging artificial intelligence approaches to enhance its predictive accuracy. Furthermore, studies examining the cost-effectiveness of implementing such models in routine clinical practice are warranted to fully realize their potential impact on patient care and health outcomes in COPD management.

Funding Statement

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Abbreviations

COPD, chronic obstructive pulmonary disease; AUC, area under the curve; DALYs, disability-adjusted life years; IPA, Invasive pulmonary aspergillosis; GOLD, Global Initiative for Chronic Obstructive Lung Disease; EORTC/MSG Study Group, European Organization for Research and Treatment of Cancer/Invasive Fungal Infections Cooperative Group; BMI, Body Mass Index; WBCs, White Blood Cells; LDH, Lactate Dehydrogenase; CK-MB, Cardiac Enzyme Creatine Kinase Isoenzyme; BNP, B-Type Natriuretic Peptide; ESR, Erythrocyte Sedimentation Rate; IgA, Immunoglobulin A; IgM, Immunoglobulin M; IgE, Immunoglobulin E; CRP, C-Reactive Protein; ROC, Receiver operating characteristic.

Data Sharing Statement

All data generated or analysed during this study are included in this published article.

Ethics Approval and Consent to Participate

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Shanghai Pulmonary Hospital (Ethical Approval Number: K24-575). Informed consent was waived by the committee (Shanghai Pulmonary Hospital) because of the retrospective nature of the study. All patient data were anonymized and strictly confidential; no individual identifiers were disclosed at any stage of the research. Data access was limited to authorized personnel for study purposes only. All methods were performed in accordance with the relevant guidelines and regulations.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare that they have no competing interests in this work.