Respiratory Infections and Antibiotic Usage in Common Variable Immunodeficiency.

BACKGROUND: Patients with common variable immunodeficiency (CVID) suffer frequent respiratory tract infections despite immunoglobulin replacement and are prescribed significant quantities of antibiotics. The clinical and microbiological nature of these exacerbations, the symptomatic triggers to take antibiotics, and the response to treatment have not been previously investigated.
OBJECTIVES: To describe the nature, frequency, treatment, and clinical course of respiratory tract exacerbations in patients with CVID and to describe pathogens isolated during respiratory tract exacerbations.
METHODS: We performed a prospective diary card exercise in 69 patients with CVID recruited from a primary immunodeficiency clinic in the United Kingdom, generating 6210 days of symptom data. We collected microbiology (sputum microscopy and culture, atypical bacterial PCR, and mycobacterial culture) and virology (nasopharyngeal swab multiplex PCR) samples from symptomatic patients with CVID.
RESULTS: There were 170 symptomatic exacerbations and 76 exacerbations treated by antibiotics. The strongest symptomatic predictors for commencing antibiotics were cough, shortness of breath, and purulent sputum. There was a median delay of 5 days from the onset of symptoms to commencing antibiotics. Episodes characterized by purulent sputum responded more quickly to antibiotics, whereas sore throat and upper respiratory tract symptoms responded less quickly. A pathogenic virus was isolated in 56% of respiratory exacerbations and a potentially pathogenic bacteria in 33%.
CONCLUSIONS: Patients with CVID delay and avoid treatment of symptomatic respiratory exacerbations, which could result in structural lung damage. However, viruses are commonly represented and illnesses dominated by upper respiratory tract symptoms respond poorly to antibiotics, suggesting that antibiotic usage could be better targeted.

Even with immunoglobulin replacement, respiratory tract infections remain the commonest clinical feature in common variable immunodeficiency (CVID) and impair quality of life. Encapsulated bacteria are thought to be the most common pathogens.

This is the first detailed description of respiratory exacerbations in CVID, capturing 6210 days of data. Viruses are commonly represented. There is a delay in commencing antibiotic therapy and the response to antibiotic therapy depends on the symptomatic presentation.

Because viral infections are common in CVID, antibiotic therapy should be considered with caution. However, self-administered antibiotic therapy should be started more promptly with symptoms of cough and purulent sputum.

Common variable immunodeficiency (CVID) is a heterogeneous primary immunodeficiency in which patients fail to produce adequate levels of immunoglobulins. With a prevalence between 1 in 10,000 and 1 in 50,000, it is the most common symptomatic primary immunodeficiency.1, 2, 3, 4

Despite adequate immunoglobulin replacement, recurrent respiratory tract infections are the commonest clinical feature in CVID2, 5 and can result in progressive bronchiectasis.6, 7, 8, 9 Respiratory tract infections were thought to be caused largely by encapsulated bacteria.6, 10 However, recent evidence shows that there may be a significant contribution from viral infections.11, 12

Despite the high incidence of respiratory tract infections and their negative influence on quality of life in primary antibody deficiency syndromes, the nature of symptoms during these episodes remains unknown. Patients are often prescribed antibiotics to mitigate respiratory tract infections, both as “rescue” courses to promptly self-administer for acute events and as prophylaxis to reduce infection frequency. However, the symptomatic triggers for taking breakthrough antibiotics and the clinical response to these treatments are not known.

In this prospective study, we sought to answer these questions by systematically recording daily symptoms and treatment in a cohort of patients with CVID over a winter period. In a parallel analysis, we also explored bacterial and viral pathogens encountered during acute respiratory symptoms in patients with CVID.

Methods

Participants

Patients were recruited from the joint Immunology-Respiratory service at the Royal Free Hospital, London. Patients had a diagnosis of CVID made by a clinical immunologist following the definitions of the Pan-American Group for Immunodeficiency and the European Society for Immunodeficiencies. All were receiving immunoglobulin replacement and were under regular (at least 6-monthly) clinical review. The only exclusion criterion was inability to provide informed consent. All participants provided written informed consent (REC 04/Q0501/119).

Study design

For this observational, prospective cohort study, patients completed daily checkbox symptom diaries for 90 days between December 2014 and February 2015, covering the UK winter season. Participants were asked to report new or increased respiratory symptoms from a predefined list (Table I ). Chronic or stable symptoms were not to be reported. Definitions of symptoms and instructions for diary completion were clearly explained; further details are provided in this article's Online Repository at www.jaci-inpractice.org. We have previously used such methodology in other chronic respiratory diseases. Participating patients were also asked to complete the St George's Respiratory Questionnaire (SGRQ), a validated measure of respiratory health status scored between 0 (best) and 100 (worst) quality of life.

Table I

List of symptoms collected in diaries and variables used for analysis

Variable	Values	Analysis group (all dichotomous)
Blocked nose	Present, not present	Upper respiratory tract symptoms
Nasal discharge	Present, not present
Sinus pain	Present, not present
Sore throat	Present, not present	Sore throat
Cough	Present, not present	Cough
Shortness of breath	Present, not present	Shortness of breath
Wheeze	Present, not present	Wheeze
Sputum color	White, yellow, green, not present	White sputumPurulent sputum
Sputum volume	Equivalent to teaspoon, egg cup, cup, not present	Increased sputum volume

Note. “Upper respiratory tract symptoms” are generated by a combination (inclusive disjunction) of blocked nose, nasal discharge, and sinus pain. Sputum color with 4 possible values was separated into 2 binary variables. Sputum volume with 4 possible values was reduced to a binary variable.

Simultaneously, but independently from the described study, we conducted a cross-sectional study in which patients experiencing acute respiratory symptoms provided samples (nasopharyngeal swabs and spontaneously expectorated sputum) for bacterial and viral testing. Sputum was considered purulent when more than 10 granulocytes per hpf were found. Samples were either collected by clinic staff or, after careful instruction on sampling, submitted directly from patients by mail. Figure 1 shows the 2 investigations undertaken on the cohorts.

Figure 1

Study design and analysis flow chart. Out of 134 patients with CVID, 69 completed a symptom diary for investigation 1 and 41 provided microbiological samples for investigation 2. Details of further analyses and the numbers of participants included for each are provided.

Definition of exacerbations and variables

For preanalysis, we grouped clinically related symptoms as presented in Table I. For calculation of total symptom count, each symptom was counted individually for each patient and each day. Cumulative total symptom count is the sum over all days of an exacerbation period.

We used 2 definitions of exacerbation, based either on symptoms or health care utilization. Similar methodology has been reported and validated in chronic obstructive pulmonary disease (COPD). For the first definition, we identified a symptomatic exacerbation as an event of 2 or more new symptoms lasting for 2 or more consecutive days as recorded by the patient in their diary, whether or not they received additional treatment. The start of a symptomatic exacerbation episode was the first day of 2 or more new symptoms lasting for 2 or more consecutive days. The end of the episode was the last consecutive day with 2 or more symptoms (allowing symptoms to change over time). If oral antibiotic therapy (OAT) was used during a symptomatic exacerbation episode, this was considered a treated symptomatic exacerbation (TSE). If not, it was an untreated symptomatic exacerbation (USE).

We defined a health care utilization exacerbation as use of OAT for worsening respiratory symptoms. We call this a treated exacerbation (TE) event, and if it coincided with diary-defined symptoms it would be a TSE. The episode was considered to last from the first day on which a symptom occurred until recovery, defined as the last day of any symptom that was present when OAT was started. Additional details regarding exacerbation and variable definitions are provided in this article's Online Repository at www.jaci-inpractice.org.

Data handling and statistical analysis

Statistical analysis was performed using Stata version 14.0 (StataCorp LP, College Station, Tex). Continuous variables are presented as median and first and third quartiles or by mean and SD as appropriate. For categorical and binary variables, we present frequencies.

Missing data were not imputed. Results were considered statistically significant at a P value of less than .05. Data were analyzed with logistic regression for trigger symptom analysis, Cox regression for antibiotic response analysis, Pearson correlation, Wilcoxon rank sum test, and t tests as indicated. Further details are provided in this article's Online Repository at www.jaci-inpractice.org.

Analysis of microbial samples

A multiplex real-time PCR (RT-PCR) for adenovirus, coronavirus (HKU, NL63, OC43, and 229E), enterovirus, human metapneumovirus, influenza virus (A and B), parainfluenza virus (1, 2, 3, and 4), parechovirus, respiratory syncytial virus, and rhinovirus was performed in the National Health Service Virology Laboratories at the Royal Free Hospital.

Sputum samples were examined by microscopy and culture for bacteria and mycobacteria plus in-house multiplex RT-PCR for Chlamydia pneumoniae, Legionella pneumophila, and Mycoplasma pneumoniae. Further details are provided in this article's Online Repository at www.jaci-inpractice.org.

We included multiple samples from a single patient if separated by at least 2 weeks and the patient was asymptomatic between episodes. Airway colonization by pathogenic bacteria was diagnosed when the same organism had been isolated more than twice within the 2 years before our study.

Results

Study population

A total of 134 patients with CVID were given a diary. Out of these, 69 (51%) patients returned a diary after completion of the study period, providing 6210 days of data (Figure 1). Demographic and clinical characteristics of included patients are presented in Table II . Patients who completed a diary were older (median [interquartile range, IQR], 59.36 [46.74-68.22] vs 45.02 [36.33-53.79] years; P < .001) and had a higher bronchiectasis severity index (median [IQR], 3.5 [2-6] vs 2 [1-4]; P = .01) than those who did not.

Table II

Patients' characteristics at study enrollment

Characteristic	Patients who completed symptom diaries (n = 69)	Patients who did not complete symptom diaries (n = 65)	P value
Age (y), median (IQR)	59.36 (46.74-68.22)	45.02 (36.33-53.79)	<.001
Female patients, n (%)	41 (59)	36 (55)	.73
IgG trough level∗ (g/L), median (IQR)	9.0 (8.0-10.0)	9.0 (7.8-10.6)	.99
Prophylactic antibiotic, n (%)	45 (65)
Amoxicillin	7 (10)
Azithromycin	22 (32)
Ciprofloxacin	3 (4)
Clarithromycin	3 (4)
Co-amoxiclav	2 (3)
Cotrimoxazole	2 (3)
Doxycycline	4 (6)
Lymecycline	1 (1)
Penicillin	1 (1)
Smoking status, n (%)
Current smoker	6 (9)	4 (6)	.92
Past smoker	15 (22)	15 (23)
Never a smoker	48 (70)	46 (71)
Bronchiectasis on CT, n (%)	37 (57.81)	29 (49.15)	.22
BSI score†, median (IQR)	3.5 (2-6)	2 (1-4)	.01
FEV1 (L), median (IQR)	2.24 (1.80-3.23)	2.63 (2.12-3.36)	.11
FEV1 predicted‡ (%), median (IQR)	93.2 (73.3-102.9)	93.5 (74.6-105.4)	.95
SGRQ§ total score, median (IQR)	24.47 (8.41-45.54)
SGRQ symptoms score, median (IQR)	39.28 (23.76-58.56)
SGRQ activity score, median (IQR)	29.31 (5.96-59.46)
SGRQ impact score, median (IQR)	14.90 (1.98-29.90)

CT, X-ray computed tomography.

Note. P values were calculated using the Wilcoxon rank sum test for continuous variables and the Fisher exact test for categorical variables.

Serum IgG level measured immediately before the next immunoglobulin replacement is administered.

Bronchiectasis severity index, ranging from 0 (best) to 25 (worst), is a validated multicomponent score in bronchiectasis that predicts the future risk of exacerbations, hospitalizations, and mortality.

FEV1 predicted is the proportion of actual FEV1 vs predicted FEV1 in accordance with the European Respiratory Society guidelines of 1993.

The SGRQ is a validated measure of respiratory health status scored between 0 (best) and 100 (worst) quality of life.

Patients with CVID suffer frequent respiratory exacerbations and often use antibiotics

During the study period, there were 170 symptomatic exacerbation events (mean, 0.82 per patient month). Of these events, 75 (mean, 0.36 per patient month) were treated by OAT but 95 (mean, 0.46 per patient month) were not. Nine patients had no symptomatic exacerbations during the period. Published literature suggests that 106 courses of antibiotics were prescribed per 1000 men and 155 per 1000 women for respiratory tract infections by general practitioners in the United Kingdom in 2014. This corresponds to 2.3 courses of antibiotics in total (0.01 per patient month) prescribed to a group similar to our cohort in the general population during 3 months.

TSE episodes were more severe than USE in terms of cumulative total symptom count (median [IQR], 40 [24-82] vs 12 [6-30] symptoms; P < .001) and episode duration (median duration, 10 vs 4 days; hazard ratio [HR], 0.54; P < .001).

A total of 76 TEs were covered within our study period. One TE did not meet the criteria of TSE. The median (IQR) duration of TE episodes was 14 (9-19) days; median (IQR) time from the start of symptoms until OAT was 5 (2-7) days; and median (IQR) time until recovery was 6.5 (5-14) days. The median (IQR) duration of therapy was 14 (7-14) days. A detailed description of symptom prevalence is shown in Figure 2 . As treatment, patients used co-amoxiclav for 23 exacerbations (30%), amoxicillin for 20 exacerbations (26%), doxycycline for 12 exacerbations (16%), ciprofloxacin for 10 exacerbations (13%), clarithromycin for 7 exacerbations (9%), and azithromycin, erythromycin, flucloxacillin, and levofloxacin for 1 exacerbation each (1%).

Figure 2

Characterization of respiratory exacerbations treated with antibiotics. Symptom prevalence (%) and TSC are displayed over time (d) for 76 antibiotic-treated respiratory exacerbations in patients with CVID. Bar diagram reflects mean (SD) TSC. Day 1 is defined as start of OAT. TSC, Total symptom count; URTS, upper respiratory tract symptom.

Cough, shortness of breath, and purulent sputum are the strongest triggers for patients to initiate antibiotic therapy

We compared 76 days on which OAT was started with 5370 days without OAT. The 764 days comprising the remainder of the antibiotic courses were ignored. In univariate analysis, all symptoms were positively and significantly associated with start of OAT. Cough (odds ratio [OR], 48.70; 95% CI, 24.02-111.47), purulent sputum (OR, 25.26; 95% CI, 15.25-42.49), increased sputum volume (OR, 23.85; 95% CI, 13.85-42.70), and shortness of breath (OR, 17.27; 95% CI, 10.54-28.22) showed the highest ORs (Figure 3 , A). In multivariate analysis, only cough (OR, 13.00; 95% CI, 5.93-28.47), purulent sputum (OR, 6.30; 95% CI, 1.19-33.40), and shortness of breath (OR, 2.41; 95% CI, 1.31-4.46) remain significant when adjusted for other symptoms (Figure 3, A).

Figure 3

Trigger symptom analysis for patients with CVID to commence antibiotic therapy. A, Prospective diary data of respiratory symptoms and OAT usage were collected from 69 patients with CVID. The forest plot displays ORs and 95% CI as a measure of effect size for individual symptoms to trigger the start of OAT (higher ORs imply a strong association between the symptom and starting OAT). Results are derived from univariate and multivariate logistic regression on the basis of 5446 observations (d). B, The bar graph shows the proportion of patients initiating OAT on each of the first 14 d of consecutive symptoms. The time since start of symptoms is defined as the number of days for which 2 or more symptoms were present. The OAT initiation proportion is the proportion of OAT that was started after a specific time since start of symptoms. ISV, increased sputum volume; SoB, shortness of breath; URTS, upper respiratory tract symptom.

In univariate analysis, time since start of symptoms was not positively associated with start of OAT, and instead patients started OAT at a fairly constant rate over the first 12 days of symptoms (Figure 3, B). There was, however, a significant positive association between total symptom count and start of OAT (OR, 2.19; 95% CI, 1.96-2.43), suggesting an approximate doubling of the odds to start OAT for each additional symptom. The mean number of symptoms on days on which OAT was started was 4.97 ± 1.94 versus 0.86 ± 1.55 symptoms on days when antibiotics were not taken (P < .001).

Exacerbations characterized by purulent sputum respond rapidly to antibiotics, whereas those characterized by upper respiratory tract symptoms and sore throat respond more slowly

Median (IQR) time until recovery after start of OAT in all TEs was 6.5 (5-14) days (Figure 4 , A). In 56% of TEs, time until recovery was 7 days or less; in 81% it was 14 days or less.

Figure 4

Antibiotic response analysis of predictor symptoms. Kaplan-Meier plots display time until recovery based on 76 antibiotic-treated respiratory exacerbations in patients with CVID: A, for all exacerbations; B, according to presence or absence of upper respiratory tract symptoms (URTSs); C, according to presence or absence of sore throat (ST); and D, according to presence or absence of purulent sputum (PS). E, Forest plot displays HRs for time until recovery after start of OAT depending on the presence of specific symptoms. A multivariate Cox model was used for all symptoms that proved to be significant in univariate analysis (only multivariate data are shown for these variables). HR reflects the “risk” for earlier complete symptomatic remission over time. ISV, increased sputum volume; SoB, shortness of breath.

In univariate analysis, time until recovery was longer in the presence of upper respiratory tract symptoms (median, 8 vs 5 days; HR, 0.50; P = .03; Figure 4, B), sore throat (12 vs 6 days; HR, 0.54; P = .007; Figure 4, C), or white sputum (12 vs 6 days; HR, 0.63; P = .03) on the day before commencing OAT. However, time until recovery was shorter in exacerbations in which purulent sputum was present (6 vs 13 days; HR, 1.98; P = .02; Figure 4, D). In multivariate analysis, upper respiratory tract symptoms, sore throat, and purulent sputum were significant independent predictors for response to OAT (Figure 4, E).

There was no statistically significant correlation between time until starting OAT and time until recovery nor between total symptom count on the day before OAT was started and subsequent time until recovery. However, a longer time until starting OAT was associated with a longer episode duration (HR, 0.92; P < .001).

Patients taking prophylactic antibiotics have more untreated exacerbations and wait longer from the onset of symptoms to initiate breakthrough antibiotics

We proceeded to investigate whether the frequency and nature of exacerbations were affected by antibiotic prophylaxis or by the presence of bronchiectasis. The mean numbers of symptomatic exacerbations (total, treated, and untreated) were analyzed with t tests and interaction was tested with 2 (prophylactic antibiotics) by 2 (bronchiectasis) analyses of variance.

There were more symptomatic exacerbation events in patients on prophylactic antibiotics than in patients not on prophylaxis (mean, 2.87 ± 2.21 vs 1.71 ± 1.33; P = .02). However, there was no significant difference in the number of TSE events (mean, 1.09 ± 1.02 vs 1.08 ± 1.32; P = .98) and the difference was explained by more USEs in patients on prophylactic antibiotics (mean, 1.78 ± 2.14 vs 0.63 ± 0.88; P = .01). In contrast, there was no significant difference in the numbers of symptomatic exacerbations (mean, 2.46 ± 1.74 vs 2.63 ± 2.48; P = .75), USEs (mean, 1.19 ± 1.63 vs 1.81 ± 2.24; P = .20), and TSEs (mean, 1.27 ± 1.15 vs 0.81 ± 1.08; P = .11) between patients with or without bronchiectasis.

The higher numbers of symptomatic exacerbations and USEs in patients on prophylactic antibiotics did not depend on the presence or absence of bronchiectasis. We observed a mean difference of 0.93 in the number of symptomatic exacerbations between patients on and off prophylaxis in those with bronchiectasis and of 1.61 in those without. For USEs, the differences in mean values were 1.27 and 0.98, respectively. Interaction effects were nonsignificant in all analyses.

Regarding the impact of antibiotic prophylaxis on exacerbation severity, there was no significant difference in episode duration or cumulative total symptom count during symptomatic exacerbations between patients on or off prophylactic antibiotics. Patients taking prophylactic antibiotics waited longer before starting OAT for breakthrough infections (median, 6 vs 3 days; HR, 0.55; P = .03). However, time until recovery after commencing OAT was not significantly different between patients on or off prophylactic antibiotics.

There were no significant differences between patients with and those without bronchiectasis in exacerbation severity, time until OAT, and time until recovery.

Prospective symptoms correlate modestly with cross-sectional analysis of quality of life

There was moderate correlation between the SGRQ total score and the number of days on which new cough (r = 0.29; P = .02), sore throat (r = 0.29; P = .02), shortness of breath (r = 0.38; P = .002), and wheeze (r 0.32; P = .01) were present. The cumulative total symptom count or cumulative number of days of symptomatic exacerbation episodes over the study period also correlated with SGRQ symptom score (r = 0.36; P = .004) and SGRQ total score (r = 0.28; P = .03).

Respiratory exacerbations in CVID demonstrate a high frequency of viral and bacterial pathogens

A total of 54 nasopharyngeal swabs were obtained from 41 patients with acute respiratory symptoms. Viruses were detected in 30 (56%) exacerbations (Figure 5 ). Rhinovirus was the most common virus detected (in 18 [33%] exacerbations), including 2 (4%) co-infections with respiratory syncytial virus, 2 (4%) co-infections with adenovirus, and 1 (2%) co-infection with human metapneumovirus.

Figure 5

Pathogenic viruses and bacteria analysis. Viral and bacterial pathogens are frequently isolated in CVID-related respiratory exacerbations. A, Viral PCR was performed on nasopharyngeal swabs in 54 symptomatic respiratory exacerbations in patients with CVID. No pathogen (gray) was found in 24 (44%) exacerbations. A pathogenic virus was found in 30 (56%) exacerbations. Rhinovirus was found in 18 (33%) exacerbations, including 2 co-infections with adenovirus (Adeno), 2 with respiratory syncytial virus (RSV), and 1 with human metapneumovirus (hMPV). B, Bacterial culture was performed on spontaneously expectorated sputum in 43 symptomatic respiratory exacerbations in patients with CVID. No pathogen (gray) was found in 29 (67%) exacerbations. A pathogenic bacterium was found in 14 (33%) exacerbations. Pseudomonas aeruginosa was isolated in 3 (7%) exacerbations; among those was 1 co-infection with Streptococcus pneumoniae. Two patients (accounting for 4 exacerbations) were probably colonized with Haemophilus influenzae.

A total of 43 spontaneously expectorated sputum samples were obtained from 34 patients with acute respiratory symptoms. Pathogenic bacteria were isolated in 14 (33%) exacerbations (Figure 5). The most common bacteria were Haemophilus influenzae in 8 (19%), Streptococcus pneumoniae in 2 (5%), and Pseudomonas aeruginosa in 2 (5%) exacerbations. Two patients accounting for 4 exacerbations were colonized with H. influenzae as defined earlier. All samples were negative for mycobacterial culture and PCR for atypical pneumonia organisms.

There was bacterial and viral co-infection in 25% of exacerbations; in 27.5% no pathogen was found. Microscopic evidence of purulence as measured by more than 10 granulocytes per hpf was found on microscopy in 41% of exacerbations positive for a pathogenic virus (whether or not patients produced sputum), in 69% of exacerbations positive for a pathogenic virus where contemporaneous sputum was collected, and in 64% of exacerbations positive for pathogenic bacteria.

Discussion

This is the first prospective cohort study describing symptoms and treatment of respiratory tract infection in CVID. We discovered a clinically important delay in commencing antibiotic therapy and that many symptoms are untreated, especially in patients taking prophylactic antibiotics. Episodes characterized by purulent sputum respond more quickly to antibiotics, whereas sore throat and upper respiratory tract symptoms respond less quickly; perhaps correspondingly, in many respiratory exacerbations we detected a pathogenic virus.

Patients with CVID are frequently prescribed antibiotics and educated to promptly take them if they suffer “breakthrough” infections. However, their actual behaviors in relation to this therapy have not previously been documented. Here, across 6210 days of data, we discovered that individual “warning” symptoms (cough, shortness of breath, and purulent sputum) are the most important triggers for patients to start OAT. Time since start of symptoms is a less important trigger, and the proportion of patients starting therapy each day is fairly constant across the first 12 days of symptoms. Consequently, and despite the fact that all patients should have antibiotics available for immediate usage, there is a median delay of 5 days in starting OAT. We are investigating whether delays to commencing treatment are explained more by patient choice or by access to health care.

A longer time to commencing therapy did not adversely impact subsequent response to antibiotics (measured by time until recovery), but inevitably increased the total length of an infectious episode. Because infections in CVID can lead to structural lung damage,2, 13 this delay may be clinically significant. Similarly, many exacerbations (95 across the study period) were untreated and may not have been reported without prospective data collection. Indeed, it is well documented that USEs often go unreported in COPD, with up to 3 times more exacerbations collected by symptom diaries than by interview; patients with COPD also treat only half of all exacerbations recorded in diaries.

Response to OAT, judged by time until recovery, did not correlate with delay to commencing therapy or total symptom count, but depended on individual symptoms. There was a slower response in patients with upper respiratory tract symptoms and sore throat, which we hypothesize may be explained by a purely viral etiology for some of these episodes. Conversely, exacerbations with purulent sputum resolved more quickly on antibiotics, perhaps indicating a dominant bacterial component.

The number of USEs, and delay to commencing OAT, was higher in patients on prophylactic antibiotics. This could imply a reluctance to start OAT in this group because of over-reliance on prophylaxis or as an increased tolerance of symptoms (generally prophylaxis is instituted only in patients with a background of high incidence of exacerbations). We found no difference in severity or duration of individual symptomatic exacerbations with or without prophylaxis; this could indicate effectiveness of prophylaxis, but conversely there is no evidence that prophylaxis attenuates the severity of breakthrough exacerbations.

There was a modest correlation between some acute symptoms reported in diaries and the SGRQ, which measures the impact of symptoms on health-related quality of life. Cumulative total symptom count and cumulative number of days of symptomatic exacerbation over the study period also correlated with SGRQ scores, confirming that symptomatic exacerbations have a significant impact on patients. However, our study design included only new or worsening symptoms rather than chronic symptoms, which presumably explains only a moderate correlation between diary-derived parameters and SGRQ scores.

In our analysis of pathogens isolated during symptomatic exacerbations, we detected a virus in 56% of patients' samples. This is similar to other reports; for example, Kainulainen et al reported positive viral PCR in 54% of 65 exacerbations in 12 patients. Bacterial pathogens, most commonly encapsulated organisms, were found in 33% of symptomatic exacerbations.

Interestingly, in exacerbations positive for a pathogenic virus but in which the patient also expectorated sputum, there was evidence of purulence as measured by high microscopic granulocyte count in 69% of samples. Although this may be partly explained by underlying bronchiectasis in some patients, we frequently observed co-infection with bacteria. Although this may represent simply colonizing bacteria in the presence of an acute viral exacerbation, there is evidence from COPD that rhinovirus infections adversely affect microbiome and the prevalence of pathogenic bacteria.21, 22 Our earlier results suggest that new or worsening purulent sputum predicts rapid response to antibiotic therapy, regardless of the organism isolated. Further research is required to investigate how the pathogens identified here influence the balance of other organisms in the respiratory tract and thereby the response to antibiotic therapy. However, our current recommendation would be to promptly treat exacerbations characterized by purulent sputum irrespective of virology results, not least because neutrophil elastase is significantly implicated in bronchiectasis pathogenesis.22, 23

Our study has some limitations. It was performed at a single tertiary care center during the winter, when respiratory tract infections are more frequent.24, 25 The true incidence of symptomatic exacerbations and antibiotic use throughout the year thus cannot be extrapolated. We cannot exclude that factors particular to our geographic location and particular to the brief study period have influenced our results.

Because of its design, this study lacks a healthy control group. We therefore cannot discuss differences in quality or quantity of exacerbations between CVID and nonimmunocompromised patients, but available data from other sources suggest that the usage of antibiotics in our cohort is many times higher than in the general population.

We have data only from patients who agreed to complete a diary (69 patients) and not the entire CVID cohort (134 patients). This may result in a selection bias, especially because these patients are older and have more clinically severe bronchiectasis. Although exacerbations did not differ in number or severity between patients with or without bronchiectasis, generalizability may be affected by variation in the prevalence of bronchiectasis throughout centers.

The symptomatic definition of a respiratory exacerbation in CVID is not standardized and we therefore operated with a simplified definition, which has been validated in COPD.

Although patients were carefully instructed to record only new or worse symptoms, we cannot exclude the possibility that some reported chronic morbidity. We note that the mean number of “new” symptoms even on days without antibiotic therapy was 0.86; however, this includes the period before and after antibiotic therapy in exacerbations and may also reflect a genuinely high frequency of acute symptoms.

Because many patients resided at a significant distance from the hospital, we were unable to perform microbiology and virology tests on diary-defined exacerbations and thus performed 2 parallel studies (Figure 1).

Conclusions

We have demonstrated that respiratory exacerbations are extremely common in CVID, but that patients delay starting antibiotics and ignore symptoms. Although viruses were identified commonly, patients should nevertheless be educated to take antibiotics promptly if they develop purulent sputum.