Introduction

Chronic obstructive pulmonary disease (COPD) is a common, preventable and treatable but incurable lung disease. COPD causes a significant economic burden primarily due to hospital admissions [1–6]. In the European Union the costs associated with COPD are estimated to be about 40 billion euros, i.e. 3.4% of the total healthcare spending. The estimated worldwide prevalence for COPD is 7.6% [7], but varies from study to study [8–11]. Apart from the economic burden, COPD also causes a social burden, increases morbidity and mortality for the patient [3, 5, 12]. According to the World Health Organization it is the third leading cause of death with more than 3 million people dying from COPD each year.

Several risk factors can contribute to the development of COPD. Studies have shown that the primary cause of COPD is exposure to tobacco smoke, followed by other risk factors such as indoor or outdoor air pollution as well as occupational dust and fumes [12–16]. In developed countries the prevalence of COPD is often directly related to the prevalence of smoking [1], whereas in more rural areas of developing countries exposure to biomass fumes also needs to be taken into account [17, 18]. Apart from the exposure to these noxae, there are also genetic characteristics and early life events involved in the pathogenesis of COPD [19, 20]. In a study investigating early life origins of COPD, Svanes and colleagues reported that “the impact of childhood disadvantage was as large as that of heavy smoking” [19]. Lange and collaborators suggested that a substantial proportion of persons who develop COPD had a low forced expiratory volume in 1 second (FEV₁) level in early adulthood [20]. COPD is often associated with clusters of comorbidities, which can worsen the outcome of patients [21]. Data from several studies showed that co-existing comorbidities affect mortality strongly [8, 9, 22]. It is now well established that a significant number of patients with COPD die from cardiovascular diseases especially in the early disease stages.

Gender is also an important factor to take into account in the development of COPD [23]. The pathophysiology of COPD is influenced by gender-specific differences [24]. Studies show that women are more susceptible to tobacco smoke [25]. There is more and more evidence showing that gender-specific approaches to COPD are crucial.

If diagnosed and treated early, the progression of COPD can be decelerated [12, 13]. Consequently, it is crucial to prevent COPD effectively and find strategies to detect and treat it at an early stage. Notwithstanding, many patients with COPD remain undiagnosed in the early stages of the disease [6, 7, 26–28]. In another study focusing on the early detection of COPD, Vandervoorde and colleagues found that underdiagnosis of COPD was more frequent in the younger age categories [29].

Apart from classical screening, there is a growing body of literature that recognizes the importance of disease associations, multimorbidity and temporal disease trajectories [30]. Along these lines, Jensen et al. used registry data of 6.2 million patients to descry temporal disease trajectories [31], which they claim to be useful for predicting and preventing future diseases of individual patients.

Thus, analyzing antecedent hospitalizations occurring prior to the first-time diagnosis of COPD could improve the understanding on how the disease presents itself in an early stage and could reveal key factors that are associated with a future onset of COPD. Using a nation-wide hospitalization database, the aim of the current study was to investigate the nature and the time evolution of the comorbidities prior to the onset of COPD in both genders, assuming that this can unveil patterns that help diagnosing COPD before it is clinically manifest.

Materials and methods

Results

Hospitalizations preceding the onset of COPD

Between 2002 and 2018, 697,714 hospitalizations coded with a first- or co-diagnosis of COPD were recorded, corresponding to 257,164 unique patients. In 145,271 unique patients, 569,889 antecedent hospitalizations were identified ranging from 1 month until 17 years prior to COPD (Fig 1). The median number of hospitalizations per patient was 3 (IQR: 1 to 5). Fifty-nine percent of cases were males and the median age category in both genders was 65–69. The median length of stay was 5 days (IQR: 2 to 12 days) compared to 9 days (IQR: 4 to 15 days) after the first diagnosis of COPD. The median number of comorbidities was 3 (IQR: 1 to 6) in both genders compared to 8 (IQR: 5 to 10) after diagnosis of COPD.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Number of hospitalizations preceding the initial diagnosis of COPD in both genders.

https://doi.org/10.1371/journal.pone.0288237.g001

Conditions preceding the onset of COPD

Overall, 11,003 comorbidities were coded in the hospitalizations preceding the initial COPD diagnosis. Sixty-two were significantly over-represented in COPD compared to the controls (adjusted p-value <0.05 and prevalence ≥ 0.5%) (see details in Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. List of comorbidities significantly over-represented in the population of patients prior to COPD onset compared to a matched-control population.

Rows are grouped according to the time to onset of COPD using a gradient color code from light red (former comorbidity) to dark red (recent comorbidity). Within each time category, rows are ordered according to odds-ratios. The stars next to the ICD-10-GM code indicate that the comorbidity was part of a confirmed trajectory from the Danish disease trajectory browser.

https://doi.org/10.1371/journal.pone.0288237.t001

Fig 2a and 2b shows the temporal evolution of the significant comorbidities prior to the onset of COPD. Diagnoses preceding the onset of COPD over 8 years included nicotine and alcohol abuse (F171, F172, F102) together with alcohol abuse-associated conditions such as hyponatremia (E871) or liver cirrhosis (K703). Mental pathologies including depression (F329, F339) and schizophrenia (F200) were also diagnosed at an early stage. Conditions over-represented 4 to 6 years before COPD diagnosis included obesity (E660, E669), sleep apnea (G473) and chronic viral hepatitis C (B182). At a later stage, conditions including chronic ischaemic heart disease (I259) and atherosclerotic heart disease (I251) were significantly over-represented. The use of artificial heart devices, implants, graft and prosthesis (Z95) were frequently observed in patients prior to developing COPD. Diagnoses of hypertensive heart disease (I119) and atherosclerosis (I702, I739) were generally identified shortly before the onset of COPD. Comorbidities occurring at a stage very close to the initial COPD diagnosis included right heart failure (I5001), vitamin D deficiency (E559), chronic kidney disease (N183) and lung cancer (C341).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Conditions preceding the onset of COPD.

Panel a shows the between-class principal component analysis (first two axes) of the time evolution of comorbidities preceding the initial COPD diagnosis. The 2-year time ranges prior to the onset of COPD are indicated by circled numbers colored from white/pale red (former comorbidities) to dark red (recent comorbidities). Preceding comorbidities are displayed using ICD-10-GM codes. The disease trajectories leading to COPD form a time gradient along the first PCA axis shown in panel b. Panel c displays the disease-network of the 36 most prevalent pre-diagnosed comorbidities involved in at least 400 trajectories in our data set. Each comorbidity is represented by a circle whose size is proportional to its prevalence and the color refers the time prior to COPD onset when the comorbidity was most prevalent. The size and the direction of the arrows joining 2 diagnoses is proportional to the number of patients diagnosed consecutively with both conditions.

https://doi.org/10.1371/journal.pone.0288237.g002

Fig 2c provides an overview of the networks and directionality between the conditions preceding COPD. The flow of diagnoses includes early alcohol/tobacco abuse together with atherosclerosis followed by heart diseases (hypertensive heart disease, chronic ischaemic heart disease) ultimately leading to COPD. The following succession of conditions were associated with a high risk of short-term onset of COPD: hypertension (I10), followed by chronic ischaemic heart disease (I25) followed by the presence of cardiac pacemaker (Z95). Several pre-conditions triggered this trajectory including nicotine dependence (F17), obesity/lipidemias (E66). Intermediate diagnoses of atherosclerosis (I70) or diabetes (E11) were also identified.

Gender-specific disease trajectories

The temporal evolution of diagnoses preceding COPD (ICD-10-GM codes summarized at the 2-digit level) was separately investigated in both genders. Compared to matched controls, 18 and 12 conditions were specifically over-represented in male and female patients with COPD, respectively. Fig 3a shows the Venn diagram of the conditions specific to each gender. Conditions that were common in both genders included tobacco/nicotine dependence (F17), hypertension (I10), hypercholesterolemia (E78) and chronic kidney diseases (N18). Conditions more frequent in females included E. coli infections (B96), hypothyroidism (E03), heart failure (I50), varicose/venous insufficiency (I83, I87), intestinal disorders (K57, K59) and personal history of breast cancer (Z85). Conditions more frequent in males included prostate cancer (C61), obesity (E66), hyperlipidemia (E78), depressive disorder (F32), hypertensive heart disease (I11), chronic ischaemic heart disease (I25), atherosclerosis (I70), pneumonia (J18), low back pain (M54) and cardiac implant (Z95).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Gender-specific diagnoses preceding COPD.

Panel a shows the Venn diagram of preceding conditions in males and females. Panel b shows the odds-ratio / prevalence plot of preceding conditions in males with the detailed list of ICD-10-GM codes (cardiovascular conditions are highlighted in red). Panels c shows the odds-ratio / prevalence plot of preceding conditions in females with the detailed list of ICD-10-GM codes (cardiovascular conditions are highlighted in red).

https://doi.org/10.1371/journal.pone.0288237.g003

Fig 3b and 3c provide a gender-specific overview of the odds-ratio / prevalence of conditions preceding the onset of COPD. In males, early atherosclerotic heart diseases (I25) was highly prevalent and constituted together with hypertension (I11) important pre-conditions triggering aortocoronary bypass, angioplasty implants (Z95), and were often followed by the use long-term coagulant therapies (Z92). In females, hypothyroidism (E03) was very prevalent and was associated with heart failure (I50), vein disorders (I87), intestine disorders (K59, K57) and osteoarthritis of knee (M17).

Discussion

The longitudinal assessment of comorbidities recorded in the hospitalization history of patients developing COPD highlights a series of health conditions intrinsically forming distinct pathogenetic paths converging to COPD. These paths may provide early indicators of a future onset of COPD. Some of these indicators were anticipated (like nicotine and alcohol dependence and chronic viral hepatitis C) whereas others were more unexpected including obesity and atherosclerosis.

In the current study, a series of comorbidities were significantly over-represented in the years preceding the diagnosis of COPD compared to a matched control group. Conforming with the current understanding of COPD pathogenesis a clearly anticipated link between nicotine and alcohol dependence was found years before the diagnosis of COPD. The association between cigarette smoking (F17), alcohol consumption (F10), and depressive / mental disorders (F20, F32, F33, F41) is well documented in the literature [33].

Another pathophysiological link uncovered by the current analysis shows a clear association between COPD and preexisting cardiovascular entities including atherosclerosis (I70), ischaemic heart disease (I25) and hypertension (I11). The link between atherosclerosis and COPD was independently confirmed in the study from Jensen and colleagues investigating temporal disease trajectories in a health registry data set covering the whole population of Denmark [31]. They found a clear convergence between a pre-diagnosis of atherosclerosis and a subsequent COPD diagnosis.

The association between chronic viral hepatitis C (B182) and COPD has been described previously. In our study, an odds-ratio of 1.9 (95% CI: 1.8 to 2.0) was found when comparing the COPD population to the matched control population. A similar association (relative risk of 1.8) was found using the DTB. Several studies evaluated the impact of HCV infections on the development of COPD, with divergent conclusions as whether HCV is directly involved in the pathogenesis of COPD [34–39]. While most studies discussed HCV as a risk factor for the development or exacerbation of COPD, Fischer and colleagues report that HCV is not an independent risk factor for obstructive lung disease [37]. Instead, the authors concluded that there is a strong correlation between HCV status, the use of injection drugs and smoking, therefore HCV alone might not be an independent contributor to the increased prevalence of obstructive lung diseases. Inversely, other authors stated that there is a direct pathogenetic link between the two conditions. In a review, Merkov and colleagues suggested that the most likely pathogenic link between COPD and HCV is systemic inflammation [34]. HCV leads to a higher production of inflammatory mediators such TNF, IL-6 and IL-8, which play an important role in the pathogenesis of COPD. Other proposed mechanisms include cirrhosis-triggered hypertension which can affect lungs [39].

Hyponatremia (E871) is another COPD-associated comorbidity typically worsening patient’s outcome during exacerbation [40, 41]. In the current study, hyponatremia was diagnosed more often in the population developing COPD compared to the controls with an odds-ratio of 1.1 (95%CI: 1.1 to 1.2). Hyponatremia is rather associated with co-occurring cardiac pathology, e.g. atrial fibrillation and its treatment [42, 43].

Although many identified comorbidities could be validated from the DTB, a few discrepancies remained. These might be explained by specificities of the German modification of the ICD-10 codes used in the current database. Same conditions can be coded differently using congruent codes (e.g. ischaemic heart disease coded by I20 vs. I25, or depressive disorders coded as F32 or F33). Coding rules can slightly differ between the two systems (e.g. the use of codes for pneumonia vs. bacterial pneumonia). This can further explain other subtle inconsistencies. Obesity was identified in our database as a very early preceding diagnosis before the onset of COPD. The link between obesity and COPD has been described elsewhere [44, 45]. The absence of confirmed link between obesity and COPD in the Danish database might be explained by the fact that the DTB did not track back information as long as our database (14.9 years vs. 17 years).

Finally, gender-specific differences in preceding conditions were identified. A series of conditions typically associated with females such as history of breast cancer, hypothyroidism, varicose, E. coli infections were identified. Somewhat surprisingly, heart failure was also more common in women as a preceding condition. Overall, the incidence of heart failure is lower in women than in men. Still, it remains possible that there is a common predisposition for heart failure and COPD specifically in women, e.g. via inflammatory pathways leading to both endothelial dysfunction (and heart failure with preserved ejection fraction even without disease of the epicardial coronary arteries) and on bronchial inflammation. In contrast, usual male conditions included prostate cancer and prostatic hyperplasia. However, cardiovascular diseases were also found in particular coronary artery disease. The latter is much more common in men, at least in part mediated by sex hormones and cigarette smoking. The higher prevalence of smoking in men in combination with a common susceptibility to develop atherosclerosis and bronchial destruction may represent the basis for the association between coronary artery disease and COPD in men. Depression and anxiety contribute to the burden of COPD-related morbidity [46]. Our data showed that anxiety-related disorders were more frequently diagnosed in females, whereas depression was more often diagnosed in the male population. Literature findings previously reported a higher level of both anxiety and depression in the female population [47, 48]. Thyroid gland dysfunction had been described as a comorbid condition in COPD [49], with possible higher prevalence in females compared to males [50].

Our study has several limitations. The current study is in essence retrospective. The data analyses are based on hospitalizations occurring prior to the onset of COPD and our population derives from patients with a coded COPD. The correctness of the COPD diagnosis depends on the coding skills of the healthcare worker which may impact on the completeness of the identified COPD cases. In addition, detailed COPD-related diagnostic information was not available in the current hospitalization database.

Conclusion

By investigating the hospitalization antecedents preceding the onset of COPD, the current study enumerates a series of comorbidities, which define distinct pathogenetic paths consequently leading to COPD. The main highlight of the current study is that distinct gender-specific trajectories could be identified reflecting differential patho-mechanisms involved in females and males. Cardiovascular diseases were frequently diagnosed prior to the onset of COPD with heart failures and vein disorders constituting an early COPD indicator in females whereas vascular and artherosclerotic diseases constituted an early COPD indicator in males. Finally, the temporal evolution of these complex disease networks provide indicators of a future onset of COPD, and as such should be carefully scrutinized in the standard clinical check-up in individuals at risk for the development of COPD.

Supporting information

Acknowledgments

We would like to thank the Lung League St. Gallen—Appenzell for their support.