Serum IgG subclass levels and risk of exacerbations and hospitalizations in patients with COPD

This is a retrospective study which used samples from participants of two COPD cohorts: Macrolide Azithromycin for Prevention of Exacerbations of COPD (MACRO – First cohort; ClinicalTrials.​gov number: NCT00325897) [19] and Simvastatin for the Prevention of Exacerbations in Moderate-to-Severe COPD (STATCOPE – Replication cohort; ClinicalTrials.​gov number: NCT01061671) [20]. Briefly, the MACRO trial tested azithromycin (250 mg daily), a macrolide antibiotic with immunomodulatory and antibacterial effects, whereas the STATCOPE trial investigated simvastatin (40 mg daily), compared to placebo for the prevention of COPD exacerbations. Both trials were double-blinded and placebo-controlled intervention studies. All participants had a diagnosis of COPD, aged 40 years or older, were current or former smokers (≥ 10 pack-year smoking history), and had a forced expiratory volume in 1 s (FEV₁) < 80% of predicted values after bronchodilator use. Additionally, all subjects had to meet at least one of the following inclusion criteria: use of supplemental oxygen, previous treatment with systemic glucocorticoids or antibiotic agents for respiratory problems, or a history of an emergency department visit or hospitalization for COPD exacerbation.

In both cohorts, the number and severity of exacerbations that occurred during follow-up were captured. The following scale was used to grade the severity of exacerbations: 1) mild – treatment conducted at home, with or without contacting a health care provider; 2) moderate – treatment requiring a visit to the emergency department; 3) severe – treatment requiring hospitalization; and 4) very severe – in which patients needed intensive care unit (ICU) admission. COPD-related hospitalizations were identified by capturing all events graded as severe and very severe exacerbations.

Baseline blood samples were collected from clinically stable participants upon entry into both studies. A total of 978 and 654 samples from the MACRO and STATCOPE cohorts, respectively, were available. Two samples from the MACRO trial (azithromycin arm) and one sample from the STATCOPE trial (placebo arm) were of insufficient quantity to allow measurements of all IgG subclasses and were thus not included in the analysis. In total, IgG levels were measured in 1629 samples, 976 from participants of MACRO (468 in the azithromycin arm; 508 in the placebo arm) and 653 from participants of STATCOPE (322 in the simvastatin arm; 331 in the placebo arm). Serum IgG subclass levels were measured using Binding Site IgG subclass (IgG1–4) reagents (Birmingham, UK) by immunonephelometry on a Siemens BNII nephelometer (BNII; Erlangen, Germany). All samples were processed in the clinical laboratory at St. Paul’s Hospital, Vancouver, British Columbia, Canada. The following normal ranges for IgG subclasses in adults were used: IgG1, 2.80–8.00 g/L; IgG2, 1.15–5.70 g/L; IgG3, 0.24–1.25 g/L; IgG4, 0.052–1.25 g/L [21]. These ranges were obtained from a healthy population and include mean levels ±2 standard deviations [22]. IgG subclass deficiency was diagnosed when IgG levels were below their respective lower limit of normal for each subclass: IgG1 < 2.80 g/L, IgG2 < 1.15 g/L, IgG3 < 0.24 g/L, and IgG4 < 0.052 g/L.

IgG subclass levels were not normally distributed, and thus are presented as median and interquartile range. Categorical variables are shown as number and percentages. We used the Mann-Whitney U test and the Pearson′s Chi-square test to compare the IgG subclass levels and the frequency of IgG subclass deficiencies between the two cohorts, respectively. The Mann-Whitney U test was also applied to analyze the IgG subclass levels according to the exacerbation status (participants with zero or one exacerbation vs. participants with two or more exacerbations) and the occurrence of hospitalizations during follow-up. Additionally, we compared the rates of COPD exacerbations in person-years (expressed as the number of acute exacerbations divided by the duration of follow-up in years) between participants with IgG subclass levels below and above the lower limit of normal (LLN) using the Mann-Whitney U test.

To investigate the effect of IgG levels on time to first COPD exacerbation or hospitalization, we applied Cox proportional-hazards regression models to obtain a hazard ratio (HR) and its 95% confidence intervals (CI) associated with these outcomes for each IgG subclass. In our Cox models, we adjusted for covariates that were considered biologically relevant to the outcomes of interest (exacerbations and hospitalizations) or that presented with a p < 0.2 at the univariate analysis. The following variables were included in our multivariate models: IgG subclass levels (log-transformed), study drug (azithromycin in the derivation cohort, or simvastatin in the validation cohort), age, gender, ethnicity, smoking status, FEV₁ (% of predicted value), long-term oxygen therapy, inhaled corticosteroid therapy at baseline, and history of systemic steroids in the year prior to enrollment. To investigate any possible effect of steroids (as described above) on IgG subclass levels, we also included an interaction term in our multivariate models. For IgG subclasses that demonstrated significant associations in our time to event analyses, we also obtained the adjusted predicted probabilities of COPD exacerbations or hospitalizations (using logistic regression models) and calculated the Pearson’s correlation coefficient between these probabilities and IgG-levels (log-transformed). SPSS version 22.0 for Windows (IBM Corp., Armonk, NY, USA) was used for data analysis. The level of statistical significance was set at p < 0.05 for all tests.

The details of participant characteristics at baseline are presented in Table 1. In MACRO, the participants were older (65.5 ± 8.6 vs. 62.8 ± 8.4 years, p < 0.001), were more likely to be Caucasians (82.6% vs. 77.9%, p = 0.02) and had slightly lower FEV₁ (39.8 ± 15.7% vs. 41.5 ± 17.8%, p = 0.049), and a higher rate of long-term domiciliary oxygen use (60.1% vs. 48.7%, p < 0.001) compared with participants enrolled in STATCOPE. Conversely, the prevalence of current smokers was higher in STATCOPE (28.8% vs. 21.2%, p < 0.001).

Total IgG and IgG subclass levels were similar between both cohorts (Table 2). Overall, 306 (18.8%) participants in the merged cohort (n = 1629) presented with at least one IgG subclass deficiency (173 (17.7%) and 133 (20.4%) in MACRO and STATCOPE, respectively)(Table 2). IgG3 deficiency was the most frequent, whereas IgG1 deficiency was the least common. We also identified cases showing two or more IgG subclasses deficiencies simultaneously. The majority of participants identified with an IgG subclass deficiency were also diagnosed with hypogammaglobulinemia (defined as total IgG < 7.0 g/L), especially participants presenting with IgG1 (72/74, n = 97.3%) or IgG2 (69/93, 74.2%) deficiencies. No differences regarding the frequency of IgG subclass deficiencies between both MACRO and STATCOPE cohorts were observed (Additional file 1: Table S1). The median and interquartile range for each IgG subclass according to the presence or absence of the related IgG subclass deficiency are shown in Additional file 2: Tables S2 (First and Replication cohorts) and Additional file 3: Table S3 (merged cohort).

In both cohorts, we detected significantly lower IgG1 and IgG2 levels among participants who developed two or more exacerbations during follow-up versus those who did not have any exacerbations or developed only one exacerbation (Additional file 4: Figure S1). Similar results were also observed in both cohorts according to hospitalization status, with participants who required hospitalizations during follow-up presenting with lower IgG1 and IgG2 subclass levels (Additional file 5: Figure S2).

Given the low frequency of IgG subclass deficiency (< 10%) for all IgG subclasses in both cohorts, we merged both cohorts (n = 1629) and compared the rates of COPD exacerbations per person-year between participants according to the presence or not of IgG1–4 deficiencies. We observed significantly higher rates of exacerbations per person-year among participants with IgG1 deficiency (2.9 ± 5.46 vs. 1.48 ± 1.86, p < 0.001) and IgG2 deficiency (2.10 ± 1.99 vs. 1.51 ± 2.18, p = 0.001) compared to their counterparts with normal IgG1 and IgG2 levels (Fig. 1). Additionally, the frequency of IgG1 and IgG2 deficiency was also significantly higher among recurrent exacerbators compared to participants who remained stable or presented with only one exacerbation during the follow-up: IgG1–43/644 (6.7%) vs. 31/985 (3.1%), p = 0.001; IgG2–47/644 (7.3%) vs. 46/985 (4.7%), p = 0.025.

The adjusted HRs for having a COPD exacerbation increased significantly with progressively lower IgG1 and IgG2 levels in MACRO (HR: 1.30, 95% CI: 1.10–1.54, p = 0.002; HR: 1.19, 95% CI: 1.05–1.35, p = 0.006, respectively) (Table 3). Since the IgG subclass levels in our models were expressed using a negative base-2 logarithmic scale, for every one-unit increase on this log scale, the IgG subclass levels decreased by 50% (negative log₂2 = − 1). Thus, in MACRO, a 50% reduction of IgG1 levels was associated with a 30% higher adjusted HR for COPD exacerbations. A similar reduction in IgG2 levels resulted in a 19% higher adjusted HR for this outcome. These findings were replicated in STATCOPE, as the adjusted HRs related to COPD exacerbations were 1.25 (95% CI: 1.02–1.54, p = 0.035) and 1.17 (95% CI: 1.01–1.36, p = 0.046) for IgG1 and IgG2, respectively. IgG1 and IgG2 levels were also associated with hospitalizations in MACRO: the adjusted HR for every 50% reduction of IgG1 level was 1.52 (95% CI: 1.15–2.02, p = 0.004) and for IgG2 was 1.33 (95% CI: 1.08–1.64, p = 0.007) (Table 4). In STATCOPE, only decreased IgG2 levels resulted in a higher risk of hospitalizations (adjusted HR = 1.43, 95% CI: 1.12–1.83, p = 0.004). We also detected an inverse relationship between IgG levels (log-transformed) and the adjusted predicted risk of COPD exacerbations (Fig. 2) and hospitalizations (Fig. 3) in MACRO and STATCOPE cohorts (p < 0.05 for both outcomes).

In our multivariate Cox models, IgG3 and IgG4 levels did not result in significant associations for both outcomes in either MACRO or STATCOPE cohorts. For these two IgG subclasses, we also calculated the HRs using both cohorts combined (n = 1629) and obtained similar HRs compared to analyzing the cohorts separately: IgG3 (exacerbations: HR = 0.97, 95% CI: 0.89–1.05, p = 0.53; hospitalizations: HR = 0.97, 95% CI: 0.84–1.12, p = 0.70); IgG4 (exacerbations: HR = 1.01, 95% CI: 0.96–1.06, p = 0.71; hospitalizations: 1.02, 95% CI: 0.94–1.12, p = 0.59).

We observed a trend for lower IgG1 subclass levels among participants who were treated with systemic steroids in the year prior to enrollment (MACRO - outcome: time to first exacerbation, p = 0.057; STATCOPE - outcome: time to first hospitalization, p = 0.07) (Additional file 6: Table S4). No statistically significant interactions were observed between IgG subclass levels and inhaled steroid therapy in both cohorts for either exacerbations or hospitalizations.