1

[ COPD Original Research ] 56

2 57
3 58
4 59
5 60
6
7
The Effect of Inhaled Corticosteroids on 61
62
8
9
Pneumonia Risk in Patients With 63
64
10
11
COPD-Bronchiectasis Overlap 65
66
12 A UK Population-Based Case-Control Study 67
13 68
14 Q24Q1 Andrew I. Ritchie; Aran Singayagam; Sebastian Mitchell; Jadwiga A. Wedzicha; Anand Shah; and Chloë I. Bloom 69
Q2
15 70
16 71
17 72
18 BACKGROUND: Inhaled corticosteroids (ICS) increase the risk of pneumonia in COPD and 73
19 commonly are used in patients with COPD-bronchiectasis overlap. 74
20 RESEARCH QUESTION: Is the risk of pneumonia associated with ICS further heightened in 75
21 76
COPD-bronchiectasis?
22 77
23 STUDY DESIGN AND METHODS: Electronic health care records (from 2004-2019) were used to 78
24 obtain a cohort of patients with COPD and a nested case-control group (age and sex matched 79
25 1:4). Analyses were conducted to determine the risk of hospitalization for pneumonia in 80
26 COPD associated with ICS use in those with bronchiectasis. Findings were confirmed by 81
27 several sensitivity analyses. Additionally, a smaller nested case-control group containing only 82
28 patients with COPD-bronchiectasis overlap and those with recent blood eosinophil counts 83
29 (BECs) was used to determine any association with BEC. 84
30 85
RESULTS: Three hundred sixteen thousand six hundred sixty-three patients were eligible for
31 86
32
the COPD cohort; bronchiectasis significantly increased the risk of pneumonia (adjusted
87
33 hazard ratio, 1.24; 95% CI, 1.15-1.33). In the first nested case-control group of 84,316 patients 88
34 with COPD, ICS was found to increase the odds of pneumonia (adjusted OR [AOR], 1.26; 89
35 95% CI, 1.19-1.32) only if used in the previous 180 days. However, bronchiectasis was a 90
36 significant modifier such that ICS use did not augment further the already elevated 91
37 bronchiectasis-associated pneumonia risk (COPD-bronchiectasis: AOR, 1.01; 95% CI, 0.8- 92
38 1.28; no bronchiectasis: AOR, 1.27; 95% CI, 1.20-1.34). Several sensitivity analyses and a 93
39 second smaller nested case-control group confirmed these findings. Finally, we found that 94
40 BEC modified the ICS-associated pneumonia risk in COPD-bronchiectasis overlap, where 95
41 96
lower BEC was associated significantly with pneumonia (BEC # 3
109/L: AOR, 1.56;
42 97
95% CI, 1.05-2.31; BEC > 3
109/L: AOR, 0.89; 95% CI, 0.53-1.24).
43 98
44 INTERPRETATION: ICS use does not augment further the already increased risk of hospitali- 99
45 zation for pneumonia associated with concomitant bronchiectasis in patients with COPD. 100
46 CHEST 2023; -(-):--- 101
47 102
48
Q6 KEY WORDS: bronchiectasis; COPD; inhaled corticosteroid; pneumonia 103
49 104
50 105
51 106
52 107
53 108
54 ABBREVIATIONS: AOR = adjusted OR; BEC = blood eosinophil count; AFFILIATIONS: From the National Heart and Lung Institute (A. I. R., Q3
109Q4
CAP = community-acquired pneumonia; ICS = inhaled corticoste- A. Singayagam, S. M., J. A. W., and C. I. B.), the Department of In-
55 110
roids; ILD = interstitial lung disease; OCS = oral corticosteroids fectious Disease (A. Singayagam), the Department of Infectious Disease

chestjournal.org 1

FLA 5.6.0 DTD
CHEST5708_proof
4 July 2023
8:17 am
EO: CHEST-D-23-00807

111 No studies of a COPD-bronchiectasis overlap cohort 166
112 Take-home Points have compared ICS use directly with no ICS use, nor 167
113 168
Study Question: Is the use of inhaled corticosteroids addressed this question using real-world data, which
114 169
(ICS) associated with an enhanced risk of pneumonia typically have the advantages of larger sample size and
115 170
in people with COPD-bronchiectasis overlap as a wealth of prescribing data. Interestingly, a US study
116 171
compared with those with COPD alone? of patients with primary bronchiectasis used Medicare
117 172
118 Results: ICS did not augment the already elevated data to compare the effect of macrolide therapy with 173
119 pneumonia risk in those with COPD-bronchiectasis that of ICS. The study investigated respiratory 174
120 overlap, but this seemed to be found only in those infection in general and found that patients using 175
121 with an elevated eosinophil count. ICS—as compared with those using macrolide 176
122 Interpretation: An elevated blood eosinophil count antibiotics—were at an increased risk of all respiratory 177
123 seems to protect patients with COPD from any addi- infections requiring hospitalization, including 178
124 respiratory viral infections, bacterial infections, and 179
tional ICS-associated increased pneumonia risk that
125 bronchiectasis exacerbations.16 However, because the 180
might be expected with concomitant bronchiectasis.
126 181
effect of ICS was in comparison with the effect of only
127 182
the antibiotic, it is difficult from that study to
128 183
extrapolate conclusions directly regarding ICS-
129 184
Inhaled corticosteroids (ICS) are prescribed commonly associated risks. Furthermore, the study population
130 185
for COPD and currently are recommended to reduce the was patients with primary bronchiectasis, but patients
131 186
risk of exacerbations in those with severe COPD.1-3 with COPD-bronchiectasis overlap may demonstrate a
132 187
Their usefulness is debated because they have little different ICS-associated infection risk profile.
133 188
impact on lung function, are not effective in all Martinez-Garcia et al,17 in a study of 201 selected
134 189
patients,4,5 and increase the risk of pneumonia,
135 patients with COPD, of whom just more than half had 190
Pseudomonas outgrowth, and mycobacterial
136 bronchiectasis, found ICS use and bronchiectasis both 191
infections.6-9
137 to be independent risk factors for pneumonia; 192
138 ICS prescribing also is cautioned in those with COPD- however, they did not assess the interaction of the two 193
139 bronchiectasis overlap.10 Underdiagnosis of 194
risk factors.
140 bronchiectasis is common, but reports estimate that the 195
141 The latest European Respiratory Society adult 196
burden of bronchiectasis varies from 4% of the 2,164
142 bronchiectasis guidelines acknowledged this paucity of 197
Q7 patients with COPD in the ECLIPSE cohort11 to 58% in
143 198
a smaller cohort of 92 patients with COPD.12 COPD- evidence in COPD-bronchiectasis overlap and
144 199
bronchiectasis overlap is associated with more severe conclude that “in the absence of definitive data, the
145 200
146 COPD,10 a higher risk of community-acquired presence of bronchiectasis should not lead to a decision
201
147 pneumonia (CAP),13 prolonged hospitalization for to withdraw ICS from patients with established 202
148 COPD exacerbations,14 and higher mortality.15 The COPD.”18 Herein, we conducted a study that leveraged 203
149 characteristic dysregulated immunity and impaired a nationwide data source enriched with prescription 204
150 airway clearance results in mucus accumulation and data to investigate if ICS use—compared with no ICS 205
151 increased susceptibility to acute and persistent lung use—further augments the risk of hospitalization for 206
152 infections. Therefore, it may be anticipated that ICS pneumonia in those with an already elevated 207
153 further dampens host-defence mechanisms, leading to pneumonia risk because of the presence of COPD- 208
154 209
an even greater risk of pneumonia. bronchiectasis overlap.
155 210
156 211
157 212
Epidemiology (A. Shah), MRC Centre for Global Infectious Disease
158 Analysis, Imperial College London, the Royal Brompton Hospital Study Design and Methods 213
159 (A. Shah), Guys and St. Thomas’ NHS Foundation Trust, and the Study Design and Data Source 214
Imperial College Healthcare NHS Trust (C. I. B.), London, England.
160 The Clinical Practice Research Datalink Aurum is a database of 215
Q5 CORRESPONDENCE TO: Chloë I. Bloom; email: chloe.bloom06@
161 imperial.ac.uk pseudoanonymized UK primary care electronic health care records 216
162 Copyright Ó 2023 The Author(s). Published by Elsevier Inc under li- that includes approximately 20% of the UK population and has been 217
163 cense from the American College of Chest Physicians. This is an open shown to be nationally representative.19 This represents one of the 218
access article under the CC BY license (http://creativecommons.org/ largest, extensively validated longitudinal health care databases
164 219
licenses/by/4.0/). worldwide.20 Clinical Practice Research Datalink data were linked
165 individually to Hospital Episode Statistics data (hospital admissions 220
DOI: https://doi.org/10.1016/j.chest.2023.06.007

2 Original Research [ -#- CHEST - 2023 ]

FLA 5.6.0 DTD
CHEST5708_proof
4 July 2023
8:17 am
EO: CHEST-D-23-00807
221 for England), socioeconomic data (Index of Multiple Deprivation; see pneumonia as the index date. Each patient was matched on the 276
222 Q8 Covariates section) and Office of National Statistics mortality data. index date using incidence density sampling, randomly selecting 277
223 control participants who matched on year of birth and sex from the 278
Study Population and Design pool of available control participants. Patients were matched
224 279
We identified patients with COPD using validated Read codes.21 preferably to four control participants, when four were not available,
225 patients were matched to three control participants. 280
Patients’ cohort entry date occurred on the latest date of the COPD
226 diagnosis, January 1, 2004, 1 year after the general practitioner 281
227 practice registration, or their 35th birthday. Patients were censored Exposure Measurement 282
228 on the minimum date of their transfer out of their general We examined all ICS prescription records in the year before the index 283
229 practitioner practice, death, or January 1, 2019. Patients with cystic date. ICS use was categorized as (1) a binary variable (use or no use), 284
fibrosis were excluded. To create an incident ICS new-user cohort, (2) time since the most recent prescription (current, # 30 days; recent,
230 285
patients who were prescribed ICS before cohort entry were excluded. 31-90 days; past, 91-180 days; and remote, > 180 days), (3) total
231 A new-user design mitigates bias because it enforces appropriate number of ICS prescriptions, and (4) type of ICS (beclomethasone, 286
232 temporal ordering of measurements of confounders, treatment, and fluticasone, budesonide, ciclesonide, or mometasone). 287
233 outcome. Two nested case-control groups were derived from the 288
234 new-user cohort. Nested case-control group 1, included all possible Covariates 289
patients with COPD and matched control participants. As a
235 All covariates in the cohort were ascertained at entry, whereas 290
sensitivity analysis, a second, much smaller, nested case-control covariates in the nested case-control group were ascertained on the
236 group 2 was drawn including only patients and control participants 291
index date. BMI was measured using kilograms per square meter. A
237 who had COPD-bronchiectasis overlap and a recent blood eosinophil 292
history of bronchiectasis (diagnosed before cohort entry, but could
238 count (BEC) (Fig 1). Because of the nature of case-control groups— have been before or after COPD diagnosis), as well as CAP, lower 293
239 only selecting patients from the cohort and then those control respiratory tract infections, depression, anxiety, heart failure, 294
participants who match by the criteria set—these greatly reduce the
240 cardiovascular disease, diabetes mellitus, asthma, cerebrovascular 295
number of patients as compared with the original cohort from which disease, chronic renal disease, lung cancer, and interstitial lung
241 the case-control group was nested. 296
disease (ILD) was recorded using appropriate Read codes reviewed
242 297
by two physicians at the time of entry into the cohort (for the
243 Patient Identification cohort analysis) and were measured again at 1 year before the index 298
244 Patients who attended hospital with CAP, identified using date (for the nested case-control group). The covariate Index of 299
245 International Classification of Diseases, Tenth Revision, codes J15 Multiple Deprivation is the official measure of social deprivation in 300
246 (pneumonia, organism unspecified) and J18 (bacterial pneumonia) the United Kingdom and is based on seven domains: income, 301
were enrolled. We set the date of the hospital admission for employment, education, health, crime, housing, and living
247 302
248 303
249 304
Copd cohort
250 305
N = 316,663
251 306
252 307
253 Patients excluded if prescribed 308
254 ICS before cohort entry date 309
255 310
256 311
Copd new-user
257 312
cohort
258 n = 152,243 313
259 314
260 Patients with COPD 315
261 concomitant bronchiectasis 316
Nested n = 2,687
262 317
case-
263 318
control group 1
264 319
Nested
265 case- 320
266 control group 2 321
267 322
268 Patients Control participants 323
269 n = 20,195 n = 61,121 324
270 325
Patients Control participants
271 326
print & web 4C=FPO

n = 521 n = 2,006
272 327
273 COPD = Chronic Obstructive Pulmonary Disease; CPRD = Clinical Practice Research 328
274 Datalink; ICS = Inhaled Corticosteroid 329
275 330
Figure 1 – Flow diagram outlining the selection of the study cohort and the two nested case-control study populations. ICS ¼ inhaledcorticosteroids. Q20
Q25

chestjournal.org 3

FLA 5.6.0 DTD
CHEST5708_proof
4 July 2023
8:17 am
EO: CHEST-D-23-00807

331 environment deprivation. Prescription records were used to identify Obstructive Lung Disease stage, Medical Research Council dyspnea 386
332 past vaccinations, pneumococcal (ever), and influenza (in the past score, OCS courses, and history of exacerbations requiring 387
333 year). Indicators of COPD severity were FEV1 % predicted, Medical hospitalization), vaccination history (pneumonia and influenza), and 388
Research Council dyspnea score, COPD exacerbation events comorbidities (lung cancer, ILD, diabetes, cerebrovascular accident,
334 389
requiring hospitalization, and the number of short courses (# cardiovascular disease, asthma, and depression) using complete case
335 7 days) of oral corticosteroids (OCS) to treat acute COPD analysis. Interaction analyses were conducted in nested case-control 390
336 exacerbations in the year before cohort entry or index date. Absolute group 1 for bronchiectasis and in nested case-control group 2 for 391
337 blood eosinophil counts, using the maximum value during the past 2 BEC; likelihood ratio tests were used to test for significance and 392
338 years, were categorized using the cutoff of 0.3
109/L as either high models were adjusted fully. 393
Q9 or normal.22 Antibiotic use was defined as any oral antibiotic
339 Several sensitivity analyses were conducted in nested case-control 394
prescription used for lower respiratory tract infections (penicillins,
340 fluoroquinolones, cephalosporins, and macrolides) and separately, group 1. First, patients with OCS use in the prior year were removed 395
341 specifically macrolides. if the association between ICS and CAP requiring hospitalization no 396
longer was present or was reduced; this would suggest the
342 397
Statistical Analysis association was confounded by OCS use. Second, a negative
343 interaction analysis was conducted using a different chronic 398
The main analysis was the association between ICS use and pneumonia
344 respiratory condition, ILD.23 This was a suitable condition because 399
requiring hospitalization and any modification by bronchiectasis
345 (nested case-control groups). However, we first measured the risk of
ILD is not anticipated to modify the association between ICS use 400
346 and CAP requiring hospitalization; therefore, if ILD also modified 401
pneumonia requiring hospitalization associated with bronchiectasis
the association, this would suggest residual confounding in the
347 alone, after adjusting for other recognized confounders (using the 402
interaction model. Third, patients with antibiotics prescribed within
348 main cohort). For this initial cohort analysis, we applied Cox 403
14 days before CAP requiring hospitalization occurred were removed
proportional hazards for time to first pneumonia requiring
349 because it could be postulated antibiotic use may explain any 404
hospitalization using age as the timescale variable; the assumptions
350 of proportional hazards were met by plotting the Schoenfeld
modification by bronchiectasis. All statistical analyses were 405
351 conducted using Stata version 17 (StataCorp). 406
residuals and inspecting for symmetry over time.
352 Ethics 407
For the main analyses of nested case-control groups 1 and 2,
353 multivariable conditional logistic regression was used to determine The protocol for this research was approved by the Independent 408
354 effect estimates. Models were adjusted for sociodemographic Scientific Advisory Committee for MHRA Database Research 409
355 variables (BMI, tobacco use status, and Index of Multiple (Identifier: 22_001718). Data management was provided by the Big 410
356 Deprivation), COPD severity markers (Global Initiative for Chronic Data and Analytical Unit at the Institute of Global Health Innovation. 411
357 412
358 413
359
Results control group 2 showed a mean age of 74.3 years 414
360 (interquartile range, 67.5-79.5 years) and 53.6% were 415
Characteristics of the COPD Cohort
361 male. Patients and control participants showed broadly 416
362
The COPD cohort included 316,663 patients . similar BMI, comorbidities, and vaccinations, but patients 417
363 Demographic and clinical characteristics are shown in e- showed a higher proportion with greater socioeconomic 418
364 Table 1. The median age was 67.9 years (interquartile deprivation levels, more severe COPD clinical 419
365 range, 58.9-76.5 years) and 52.6% were male; 48.1% of indicators, and previous CAP (Table 1, e-Table 3); 420
366 the patients were using ICS at the time of cohort entry however, patients in nested case control group 2 also 421
367 and 2.5% had bronchiectasis (e-Table 1). The COPD 422
differed in proportions of BMI categorization, some
368 cohort showed a 24% increased risk of CAP requiring 423
comorbidities, and higher eosinophil count.
369 hospitalization associated with concomitant 424
370 bronchiectasis (adjusted hazard ratio, 1.24; 95% CI, 1.15- 425
371 Association Between ICS and Pneumonia 426
1.33; P < .0001) (e-Table 2).
372 A 26% increased odds of CAP requiring hospitalization 427
373 Characteristics of the Nested Case-Control Groups 428
was associated with using ICS (adjusted OR [AOR],
374 429
From the COPD cohort, 152,243 patients were eligible for 1.26; 95% CI, 1.19-1.32) (Table 2). Recent ICS use
375 430
the ICS new-user cohort (Fig 1, e-Table 1), from which showed the strongest association, but the association was
376 431
377
the two nested case-control groups were drawn. The first no longer significant by 180 days after ICS use 432
378 nested case-control group contained 20,195 patients and (0-30 days: AOR, 1.27; 95% CI, 1.20-1.35; 30-90 days: 433
379 61,121 matched control participants. The second group AOR, 1.32; 95% CI, 1.22-1.42; 90-180 days: AOR, 1.14; 434
380 was drawn from only those with COPD-bronchiectasis 95% CI, 1.00-1.29; 180-365 days: AOR, 1.05; 95% CI, 435
381 overlap and a recent BEC; this group contained 521 0.90-1.22) (Table 2). The cumulative number of ICS 436
382 patients and 2,006 matched control participants (Fig 1, e- prescriptions in the year before the index date did not 437
383 Table 3). Patients in nested case-control group 1 showed affect the strength of the association (Table 2). Assessing 438
384 had a mean age of 77.6 years (interquartile range, 70.0- the effect of different ICS drugs, we found an increased 439
385 odds with fluticasone (AOR, 1.42; 95% CI, 1.33-1.51) 440
84.0 years) and 58.4% were male. Patients in nested case-

4 Original Research [ -#- CHEST - 2023 ]

FLA 5.6.0 DTD
CHEST5708_proof
4 July 2023
8:17 am
EO: CHEST-D-23-00807
441 TABLE 1 ] Demographic Features, Clinical Characteristics, and Comorbidities of Nested Case-Control Group 1 496
442 (With COPD) Q21
497
443 Characteristic Patients Control Participants Total 498
444 499
Total No. 20,195 61,121 84,316
445 500
Age, y 77.8 (70.3-84.2) 77.5 (69.9-83.9) 77.6 (70.0-84.0)
446 501
447 Sex 502
448 Male 11,817 (58.5) 35,631 (58.3) 47,448 (58.4) 503
449 Female 8,378 (41.5) 25,490 (41.7) 33,868 (41.6) 504
450 Socioeconomic deprivation (IMD) 505
451 1 2,950 (14.6) 10,059 (16.5) 13,009 (16.0) 506
452 507
2 3,331 (16.5) 11,130 (18.2) 14,461 (17.8)
453 508
3 3,709 (18.4) 12,047 (19.7) 15,756 (19.4)
454 509
455 4 4,438 (22.0) 12,761 (20.9) 17,199 (21.2) 510
456 5 (most deprived) 5,767 (28.6) 15,124 (24.7) 20,891 (25.7) 511
457 BMI 512
458 Normal 6,553 (32.4) 17,963 (29.4) 24,516 (30.1) 513
459 Underweight 1,583 (7.8) 3,609 (5.9) 5,192 (6.4) 514
460 515
Overweight 7,435 (36.8) 24,266 (39.7) 31,701 (39.0)
461 516
Obese 3,834 (19.0) 12,841 (21.0) 16,675 (20.5)
462 517
Missing 790 (3.9) 2,442 (4.0) 3,232 (4.0)
463 518
464 COPD characteristics 519
465 FEV1 % predicted 520
466 $ 80 2,404 (11.9) 11,012 (18.0) 13,416 (16.5) 521
467 50-79 8,533 (42.3) 27,690 (45.3) 36,223 (44.6) 522
468 30-49 5,320 (26.3) 11,638 (19.0) 16,958 (20.9) 523
469 524
< 30 1,298 (6.4) 2,342 (3.8) 3,640 (4.5)
470 525
Missing 2,640 (13.1) 8,439 (13.8) 11,079 (13.6)
471 526
472 MRC score 527
473 1 1,391 (6.9) 7,714 (12.6) 9,105 (11.2) 528
474 2 4,076 (20.2) 16,520 (27.0) 20,596 (25.3) 529
475 3 4,869 (24.1) 13,224 (21.6) 18,093 (22.3) 530
476 4 4,603 (22.8) 8,195 (13.4) 12,798 (15.7) 531
477 532
5 1,924 (9.5) 2,570 (4.2) 4,494 (5.5)
478 533
Missing 3,332 (16.5) 12,898 (21.1) 16,230 (20.0)
479 534
Tobacco use history
480 535
481 Never 1,099 (5.4) 4,437 (7.3) 5,536 (6.8) 536
482 Former 9,681 (47.9) 28,071 (45.9) 37,752 (46.4) 537
483 Current 8,000 (39.6) 23,790 (38.9) 31,790 (39.1) 538
484 Missing 1,415 (7.0) 4,823 (7.9) 6,238 (7.7) 539
485 Previous CAP 2,956 (14.6) 4,586 (7.5) 7,542 (9.3) 540
486 541
Previous LRTI 13,981 (69.2) 36,197 (59.2) 50,178 (61.7)
487 542
In year before index date
488 543
489 ICS use 544
490 Yes 10,068 (49.9) 21,376 (35.0) 31,444 (38.7) 545
491 No 10,127 (50.1) 39,745 (65.0) 49,872 (61.3) 546
492 Last ICS prescription, d 547
493 None 10,127 (50.1) 39,745 (65.0) 49,872 (61.3) 548
494 549
(Continued)
495 550

chestjournal.org 5

FLA 5.6.0 DTD
CHEST5708_proof
4 July 2023
8:17 am
EO: CHEST-D-23-00807

551 TABLE 1 ] (Continued) 606
552 607
Characteristic Patients Control Participants Total
553 608
0-30 6,079 (30.1) 12,152 (19.9) 18,231 (22.4)
554 609
555 30-90 2,694 (13.3) 5,741 (9.4) 8,435 (10.4) 610
556 90-180 783 (3.9) 1,925 (3.1) 2,708 (3.3) 611
557 180-365 512 (2.5) 1,558 (2.5) 2,070 (2.5) 612
558 Total ICS prescriptions 613
559 None 10,127 (50.1) 39,745 (65.0) 49,872 (61.3) 614
560 615
1-3 inhalers 2,174 (10.8) 5,240 (8.6) 7,414 (9.1)
561 616
3-8 inhalers 3,835 (19.0) 8,406 (13.8) 12,241 (15.1)
562 617
9þ inhalers 4,059 (20.1) 7,730 (12.6) 11,789 (14.5)
563 618
564 Type of ICS 619
565 None 10,127 (50.1) 39,745 (65.0) 49,872 (61.3) 620
566 Beclomethasone 2,055 (10.2) 5,464 (8.9) 7,519 (9.2) 621
567 Budesonide 2,078 (10.3) 5,182 (8.5) 7,260 (8.9) 622
568 623
Fluticasone 5,927 (29.3) 10,711 (17.5) 16,638 (20.5)
569 624
Ciclesonide/mometasone 8 (0.0) 19 (0.0) 27 (0.0)
570 625
COPD exacerbations requiring
571 626
hospitalization
572 627
0 17,038 (84.4) 58,307 (95.4) 75,345 (92.7)
573 628
574 1 2,165 (10.7) 2,202 (3.6) 4,367 (5.4) 629
575 2 992 (4.9) 612 (1.0) 1,604 (2.0) 630
576 OCS courses 631
577 0 11,866 (58.8) 46,177 (75.6) 58,043 (71.4) 632
578 1 2,971 (14.7) 6,634 (10.9) 9,605 (11.8) 633
579 634
2 5,358 (26.5) 8,310 (13.6) 13,668 (16.8)
580 635
Comorbidities
581 636
Bronchiectasis 984 (4.9) 1,989 (3.3) 2,973 (3.7)
582 637
583 Lung cancer 777 (3.8) 1,411 (2.3) 2,188 (2.7) 638
584 Asthma 3,887 (19.2) 11,156 (18.3) 15,043 (18.5) 639
585 Interstitial lung disease 668 (3.3) 1,344 (2.2) 2,012 (2.5) 640
586 Diabetes 5,022 (24.9) 14,613 (23.9) 19,635 (24.1) 641
587 CVA 2,310 (11.4) 5,824 (9.5) 8,134 (10.0) 642
588 643
Cardiovascular disease 14,682 (72.7) 43,076 (70.5) 57,758 (71.0)
589 644
Depression 4,292 (21.3) 11,331 (18.5) 15,623 (19.2)
590 645
591 Vaccinations 646
592 Pneumonia vaccine (ever) 2,380 (11.8) 6,880 (11.3) 9,260 (11.4) 647
593 Influenza vaccine (prior year) 3,931 (19.5) 12,367 (20.2) 16,298 (20.0) 648
594 649
Data are presented as No. (%) or median (interquartile range). CAP ¼ community-acquired pneumonia; CVA ¼ cerebrovascular accident; ICS ¼ inhaled
595 650
corticosteroids; IMD ¼ Index of Multiple Deprivation; LRTI ¼ lower respiratory tract infection; MRC ¼ Medical Research Council; OCS ¼ oral corticosteroids.
596 651
597 652
and beclomethasone (AOR, 1.15; 95% CI, 1.06-1.25), but of the association between ICS and CAP requiring
598 653
not with budesonide (AOR, 1.04; 95% CI, 0.95-1.12). hospitalization (P < .01, likelihood ratio test). In
599 654
patients with COPD-bronchiectasis overlap, ICS was
600 655
601
Association of ICS With CAP Requiring not associated significantly with pneumonia (AOR, 656
Hospitalization in COPD With Concomitant 1.01; 95% CI, 0.8-1.28), unlike in COPD without
602 657
Bronchiectasis bronchiectasis (AOR, 1.27; 95% CI, 1.20-1.34). In the
603 658
604 In nested case-control group 1, concomitant negative exposure analysis, interstitial lung disease 659
605 bronchiectasis was found to be a significant modifier was not found to be an effect modifier, suggesting that 660

6 Original Research [ -#- CHEST - 2023 ]

FLA 5.6.0 DTD
CHEST5708_proof
4 July 2023
8:17 am
EO: CHEST-D-23-00807
661 TABLE 2 ] Association Between Community-Acquired Pneumonia and ICS, Time Since Last ICS Prescription, No. of 716
662 ICS Prescriptions, and Type of ICS Drug in Nested Case-Control Group 1 (With COPD; n ¼ 51,335) Q22
717
663 Variable Adjusted OR P Value 95% CI 718
664 719
ICS use in past year
665 720
No . Reference .
666 721
667 Yes 1.26 < .001 1.19-1.32 722
668 Time since last ICS, d 723
669 None . Reference . 724
670 0-30 1.27 < .001 1.20-1.35 725
671 30-90 1.32 < .001 1.22-1.42 726
672 727
90-180 1.14 > .05 1.00-1.29
673 728
180-365 1.05 .530 0.90-1.22
674 729
675 No. of ICS inhalers in past year 730
676 None . Reference . 731
677 1-3 1.26 < .001 1.16-1.37 732
678 4-8 1.25 < .001 1.17-1.34 733
679 $9 1.25 < .001 1.17-1.34 734
680 735
Type of ICS used in past year
681 736
None . Reference .
682 737
Beclomethasone 1.15 < .01 1.06-1.25
683 738
684 Budesonide 1.04 .377 0.95-1.12 739
685 Fluticasone 1.42 < .001 1.33-1.51 740
686 Ciclesonide/mometasone 0.55 .364 0.15-1.98 741
687 742
Adjusted for socioeconomic deprivation (Index of Multiple Deprivation), BMI, FEV1 % predicted, Medical Research Council dyspnea score, tobacco use
688 743
history, history of pneumonia, history of lower respiratory tract infections, hospital exacerbations in the past year, courses of oral corticosteroids in the
689 past year, bronchiectasis, lung cancer, asthma, interstitial lung disease, diabetes mellitus, cerebrovascular accidents, depression, pneumonia vaccine 744
690 (ever), and influenza vaccine (year prior). ICS ¼ inhaled corticosteroids. 745
691 746
692 747
693 significant residual confounding in the model was The Association With BEC, ICS Use, and CAP 748
694 unlikely (P > .05, likelihood ratio test). Requiring Hospitalization in COPD-Bronchiectasis 749
695 Overlap 750
696 Two other sensitivity analyses were applied. First, Nested case-control group 2 also was used to investigate 751
697 removing patients with OCS use in the previous any modification effect from BEC. An interaction 752
698 year had a negligible impact on the effect estimates 753
analysis found eosinophils were a significant modifier of
699 (e-Table 4). Second, assessing the influence of recent 754
the association between ICS use and CAP requiring
700 antibiotic use, antibiotic use was prevalent in all 755
hospitalization (P < .05, likelihood ratio test). High BEC
701 patients, especially those with bronchiectasis: 756
(> 0.3
109/L) was not associated with CAP requiring
702 757
70.8% of bronchiectasis patients had been prescribed hospitalization (AOR, 0.89; 95% CI, 0.53-1.24), but a
703 758
antibiotics in the year prior, although the use of normal eosinophil count (# 0.3
109/L) was associated
704 759
macrolides occurred in only around 2% of patients with CAP requiring hospitalization (AOR, 1.56; 95% CI,
705 760
706
(e-Table 5). Excluding patients with antibiotic 0.78-1.41). 761
707 prescription in the 14 days before the index date 762
708 had a negligible impact on the effect estimates 763
709 (e-Table 6). As a further sensitivity analysis, nested Discussion 764
710 case-control group 2 (containing only patients with In this study, we identified a cohort of > 300,000 765
711 COPD-bronchiectasis overlap) was drawn. Again, patients with COPD and investigated the association 766
712 ICS use was not associated with CAP requiring of ICS use and pneumonia requiring hospitalization in 767
713 hospitalization in those with COPD-bronchiectasis those with concomitant bronchiectasis. Using a cohort 768
714 overlap (AOR, 1.05; 95% CI, 0.78-1.41; P ¼ .75). design, we initially determined that concomitant 769
715 770

chestjournal.org 7

FLA 5.6.0 DTD
CHEST5708_proof
4 July 2023
8:17 am
EO: CHEST-D-23-00807

771 bronchiectasis was associated with a 24% increased People with bronchiectasis may receive antibiotics for 826
772 risk of CAP requiring hospitalization, replicating respiratory infections earlier than those without 827
773 828
findings from one earlier observational study in US bronchiectasis. Additionally, they may receive long-term
774 829
data and landmark trials.24-26 Next, we used a nested prophylactic macrolides.18 Differential early use of
775 830
case-control design to determine that ICS use was antibiotics potentially could have explained the relatively
776 831
associated with 26% increased odds of CAP requiring reduced risk of ICS-associated pneumonia in those with
777 832
778
hospitalization and that this occurred in a time- bronchiectasis. Indeed, in our cohort, those with 833
779 dependent manner. Finally, by using an interaction bronchiectasis were more likely to receive antibiotics, 834
780 analysis, we were able to determine that in people with but the overall use of macrolides was very low in this 835
781 COPD-bronchiectasis overlap, ICS use did not primary care population. Furthermore, excluding 836
782 augment further the already elevated bronchiectasis- patients prescribed antibiotics had negligible 837
783 associated pneumonia risk. All models were adjusted consequences on the overall effect estimates. 838
784 for multiple potential confounders including COPD 839
785 Neutrophils are the most abundant airway inflammatory 840
severity and vaccination status. Furthermore, to
786 cell in bronchiectasis, but the importance of eosinophils 841
explore if additional residual confounding was
787 recently was recognized.32 In a post hoc analysis of 842
present, we conducted a negative control exposure
788 bronchiectasis trial data from 86 participants, inhaled 843
analysis. The negative exposure chosen was another
789 fluticasone reduced bronchiectasis exacerbations in 844
chronic respiratory condition, ILD, that would not be
790 those with BEC of $ 150 cells/mL.33 A large European 845
791
expected to modify the association between ICS use 846
multicenter cohort of 1,007 patients with bronchiectasis
792 and pneumonia. Therefore, if ILD was found also to 847
reported that approximately one-fifth of patients
793 modify the association, this would imply that the case- 848
exhibited sputum eosinophil counts of > 3%.34 They
794 control model was confounded; however, ILD did not 849
showed that elevated BECs (> 300 cells/mL) were an
795 modify the association. Another sensitivity analysis 850
effective surrogate marker of sputum eosinophils, and
796 was to draw a second nested case-control group, but 851
797
these were associated with infective exacerbations, 852
this time containing only those with COPD-
798 suggesting that in bronchiectasis, eosinophilic 853
bronchiectasis overlap, resulting in a much smaller
799 inflammation and microbial dysbiosis are interrelated. 854
sample size. Again, we found the same effect, that ICS
800 In contrast, in COPD, a lack of eosinophilic 855
use did not further increase the already elevated
801 inflammation is associated with chronic bacterial 856
pneumonia risk. Other sensitivity analyses conducted,
802 infection. Martinez-Garcia et al17 demonstrated in a 857
including removing patients with antibiotic
803 microbiology study of 201 carefully characterized 858
804
prescriptions or OCS use, also did not change the 859
patients with COPD (57% also had bronchiectasis) that
805 effect estimate. 860
those with low BECs had an increased incidence of
806 861
Our data demonstrated an ICS association with CAP bacterial colonization and pneumonia and that ICS use
807 862
that is dependent on the time since last ICS use, but no further increased the risk. A pooled analysis of 10
808 863
association with the total number of ICS inhalers used. randomized control trials of ICS-containing
809 864
810
These findings imply that pneumonia risk relates to combination treatment in patients with COPD found 865
811 the proximity of ICS use, rather than the cumulative that the risk of pneumonia was 31% higher in patients 866
812 effect of ICS exposure. Postulated biological with a BEC of < 2% vs $ 2%.35 Additionally, using real- 867
813 mechanisms include ICS-induced alterations of the world observational data, patients with COPD using 868
814 innate and adaptive immune system,27,28 which may ICS-containing inhalers with a BEC of < 150 cells/mL 869
815 increase bacterial load and change the microbial again exhibited a higher risk of pneumonia.36 870
816 composition in the airways.29 The odds of pneumonia Correspondingly, the authors of the Global Initiative for 871
817 were increased 1.42-fold with fluticasone and 1.15-fold Chronic Obstructive Lung Disease Science Committee 872
818 873
with beclomethasone, but were not increased with recently published a review article that concludes that
819 874
budesonide, which enters the lungs with a lower low BECs are associated with an increased risk of
820 875
lipophilicity and dissolves more quickly into recurrent bacterial infections and pneumonia in COPD
821 876
pulmonary fluids, leading to a reduced local effect and that these risks likely are increased by ICS use.22
822 877
823
because of a more rapid passage into the systemic Therefore, our study findings parallel these earlier 878
824 circulation,30 supporting findings in the Canadian studies that low BEC is associated with higher 879
825 observational study by Suissa et al.31 pneumonia risk, albeit using a different methodologic 880

8 Original Research [ -#- CHEST - 2023 ]

FLA 5.6.0 DTD
CHEST5708_proof
4 July 2023
8:17 am
EO: CHEST-D-23-00807
881 approach and the first study to study this directly in a of misclassification), or neurologic conditions managed 936
882 population with COPD-bronchiectasis overlap. 937
in secondary care that may cause oropharyngeal. Fifth, Q10
883
Furthermore, a recent publication from a large Spanish we assessed only for increase in pneumonia requiring 938
884 939
bronchiectasis registry found that ICS use reduces the hospitalization and did not assess nonpneumonia
885 940
risk of a bronchiectasis exacerbation only in those with requiring hospitalization, which many of the
886 941
elevated BEC of > 300 cells/mL.37 Overall, these findings randomized controlled trials in COPD reported as
887 942
888
suggest that ICS may be beneficial with low risk of increasing.6 Finally, although biological plausibility 943
889 pneumonia in patients with elevated BEC and COPD- exists and we have endeavored to identify and minimize 944
890 bronchiectasis overlap, in keeping with the recent bias, we have not made causal statements because of the 945
891 recommendations in the European Respiratory Society observational nature of the study. 946
892 adult bronchiectasis guidelines. However, other potential 947
893 side-effects from ICS use should be considered. 948
894 Interpretation 949
895 Study Limitations Our main finding is that although ICS use and 950
896 This observational study has several limitations. First, as concomitant bronchiectasis independently increase the 951
897 with all routinely collected database studies, we did not risk of pneumonia requiring hospitalization, no further 952
898 953
have information on patient adherence to the ICS augmentation from ICS use occurs in patients with
899 954
prescriptions. This may have biased effect estimates COPD-bronchiectasis overlap. Additionally, in our
900 955
towards null. Second, the diagnosis of bronchiectasis smaller cohort of only those with COPD-bronchiectasis
901 956
requires radiologic imaging, a limited resource that overlap, we found that an elevated BEC seems to protect
902 957
903
likely led to underdiagnosis in this cohort, a well- patients from the ICS-associated increased pneumonia 958
904 recognized problem in bronchiectasis prevalence risk, reflecting findings found in previous cohorts with 959
905 studies.11,12 In addition, the increase of non-tobacco COPD. 960
906 users in the case-control groups may suggest that several 961
907 of these patients in fact had primary bronchiectasis with 962
908 airflow obstruction, but were labelled as having Funding/Support Q11
963
909 comorbid COPD, which could impact on the effect of 964
Supported by the National Institute for Health Research Q23
910 ICS. Third, we did not have access to microbiology data Imperial Biomedical Research Centre. 965
911 966
or data on severity of bronchiectasis for each patient.
912 Financial/Nonfinancial Disclosures 967
Q12
Fourth, we did not have information on pneumonia risk
913 968
factors, alcohol use (this is often not recorded and at risk None declared.
914 969
915 970
916 971
917 Acknowledgments 2. Halpin D, Celli B, Criner G, et al. The Database Syst Rev. 2014;2014(3): 972
Q14 GOLD Summit on chronic obstructive Cd010115.
918 Author contributions: C. I. B. takes full 973
pulmonary disease in low-and middle-
919 responsibility for the content of the article income countries. Int J Tuberc Lung Dis. 7. Brassard P, Suissa S, Kezouh A, et al. 974
including the data and analysis. A. I. R. 2019;23(11):1131-1141. Inhaled corticosteroids and risk of
920 tuberculosis in patients with respiratory 975
drafted and revised the manuscript. A. I. R.,
921 3. Bloom CI, Elkin SL, Quint JK. Changes diseases. Am J Respir Crit Care Med. 976
S. M., and C. I. B performed the analyses.
in COPD inhaler prescriptions in the 2011;183(5):675-678.
922 C.I.B designed the study. A. Singayagam, A. United Kingdom, 2000 to 2016. Int J 977
923 Shah, J. A. W., and C. I. B. conceptualized the Chron Obstruct Pulmon Dis. 2019;14: 8. Andréjak C, Nielsen R, Thomsen VØ, 978
study. All authors contributed to data 279-287. et al. Chronic respiratory disease, inhaled
924 interpretation and manuscript revision. corticosteroids and risk of non- 979
925 Q13 4. Bafadhel M, Peterson S, De Blas MA, et al. tuberculous mycobacteriosis. Thorax. 980
Role of sponsors: The sponsor had no role in Predictors of exacerbation risk and 2013;68(3):256-262.
926 the design of the study, the collection and response to budesonide in patients with 981
927 analysis of the data, or the preparation of the chronic obstructive pulmonary disease: a 9. Eklöf J, Ingebrigtsen TS, Sørensen R, et al. 982
manuscript. post-hoc analysis of three randomised Use of inhaled corticosteroids and risk of
928 trials. Lancet Respir Med. 2018;6(2): acquiring Pseudomonas aeruginosa in 983
929 Additional information: The e-Tables are 117-126. patients with chronic obstructive 984
available online under “Supplementary pulmonary disease. Thorax. 2022;77(6):
930 Data.” 5. Magnussen H, Disse B, Rodriguez- 573-580. 985
931 Roisin R, et al. Withdrawal of inhaled 986
glucocorticoids and exacerbations of 10. Dente FL, Roggi MA, Del Cesta R, et al.
932 References COPD. N Engl J Med. 2014;371(14): COPD and asthma overlap with 987
933 1285-1294. bronchiectasis. Eur Respir Monograph. 988
1. Hurst JR, Vestbo J, Anzueto A, et al.
2018;81:167-185.
934 Susceptibility to exacerbation in chronic 6. Kew KM, Seniukovich A. Inhaled steroids 989
obstructive pulmonary disease. N Engl J and risk of pneumonia for chronic 11. Agusti A, Calverley PM, Celli B, et al.
935 Med. 2010;363(12):1128-1138. obstructive pulmonary disease. Cochrane Characterisation of COPD heterogeneity 990

chestjournal.org 9

FLA 5.6.0 DTD
CHEST5708_proof
4 July 2023
8:17 am
EO: CHEST-D-23-00807

991 in the ECLIPSE cohort. Respir Res. 21. Quint JK, Müllerova H, 29. Contoli M, Pauletti A, Rossi MR, et al. 1030
992 2010;11(1):122. DiSantostefano RL, et al. Validation of Long-term effects of inhaled 1031
12. Martínez-García M, Soler-Cataluña JJ, chronic obstructive pulmonary disease corticosteroids on sputum bacterial and
993 recording in the Clinical Practice Research viral loads in COPD. Eur Respir J. 1032
Donat Sanz Y, et al. Factors associated
994
Q17
with bronchiectasis in patients with Datalink (CPRD-GOLD). BMJ Open. 2017;50(4). 1033
2014;4(7):e005540.
995 COPD. Chest. 2011;140(5):1130-1137. 30. Hübner M, Hochhaus G, Derendorf H. 1034
22. Singh D, Agusti A, Martinez FJ, et al. Comparative pharmacology,
996 13. Lin SH, Ji BC, Shih YM, et al. Comorbid
bioavailability, pharmacokinetics, and 1035
pulmonary disease and risk of Blood Eosinophils and Chronic
997 community-acquired pneumonia in Obstructive Pulmonary Disease: A GOLD pharmacodynamics of inhaled 1036
998 COPD patients. Int J Tuberc Lung Dis. Science Committee 2022 Review. Am J glucocorticosteroids. Immunol Allergy 1037
Q16
2013;17(12):1638-1644. Respir Crit Care Med. 2022. Clin North Am. 2005;25(3):469-488.
999 1038
23. Lipsitch M, Tchetgen Tchetgen E, 31. Suissa S, Patenaude V, Lapi F, et al.
1000 14. Patel IS, Vlahos I, Wilkinson TM, et al. Cohen T. Negative controls: a tool for Inhaled corticosteroids in COPD and the 1039
Bronchiectasis, exacerbation indices, and
1001 inflammation in chronic obstructive detecting confounding and bias in risk of serious pneumonia. Thorax. 1040
pulmonary disease. Am J Respir Crit Care observational studies. Epidemiology. 2013;68(11):1029-1036.
1002 1041
Med. 2004;170(4):400-407. 2010;21(3):383-388. 32. Keir HR, Chalmers JD. Bronchiectasis
1003 1042
15. Goeminne PC, Nawrot TS, Ruttens D, 24. Ernst P, Gonzalez AV, Brassard P, et al. enters the inflammation era. Respirology.
1004 et al. Mortality in non-cystic fibrosis Inhaled corticosteroid use in chronic 2022;27(7):488-489. 1043
1005 bronchiectasis: a prospective cohort obstructive pulmonary disease and the 33. Aliberti S, Sotgiu G, Blasi F, et al. Blood 1044
analysis. Respir Med. 2014;108(2):287-296. risk of hospitalization for pneumonia. Am eosinophils predict inhaled fluticasone
1006 J Respir Crit Care Med. 2007;176(2): 1045
16. Henkle E, Curtis JR, Chen L, et al. response in bronchiectasis. Eur Respir J.
1007
Q18 162-166. 2020;56(2). 1046
Comparative risks of chronic inhaled
1008 corticosteroids and macrolides for 25. Crim C, Calverley P, Anderson J, et al. 34. Shoemark A, Shteinberg M, Soyza AD, 1047
Q15
1009 bronchiectasis. Eur Respir J. 2019;54(1). Pneumonia risk in COPD patients et al. Characterization of eosinophilic 1048
17. Martinez-Garcia MA, Faner R, Oscullo G, receiving inhaled corticosteroids alone or bronchiectasis: a European multicohort
1010 in combination: TORCH study results. 1049
et al. Inhaled steroids, circulating study. Am J Respir Crit Care Med.
1011 eosinophils, chronic airway infection, and Eur Respir J. 2009;34(3):641-647. 2022;205(8):894-902. 1050
1012 pneumonia risk in chronic obstructive 26. Janson C, Larsson K, Lisspers KH, et al. 35. Pavord ID, Lettis S, Anzueto A, et al. 1051
pulmonary disease. A network analysis. Pneumonia and pneumonia related Blood eosinophil count and pneumonia
1013 Am J Respir Crit Care Med. 2020;201(9): 1052
mortality in patients with COPD treated risk in patients with chronic obstructive
1014 1078-1085. with fixed combinations of inhaled pulmonary disease: a patient-level meta- 1053
1015 18. Polverino E, Goeminne PC, corticosteroid and long acting b2 analysis. Lancet Respir Med. 2016;4(9): 1054
McDonnell MJ, et al. European agonist: observational matched cohort 731-741.
1016 study (PATHOS). BMJ. 2013;346:f3306. 1055
Respiratory Society guidelines for the 36. Ashdown HF, Smith M, McFadden E,
1017 management of adult bronchiectasis. Eur 1056
27. Singanayagam A, Glanville N, et al. Blood eosinophils to guide inhaled
1018 Respir J. 2017;50(3):1700629. Cuthbertson L, et al. Inhaled maintenance therapy in a primary care 1057
1019 19. Herrett E, Gallagher AM, Bhaskaran K, corticosteroid suppression of COPD population. ERJ Open Res. 1058
et al. Data resource profile: Clinical cathelicidin drives dysbiosis and 2022;8(1):00606-02021.
1020 Practice Research Datalink (CPRD). Int J bacterial infection in chronic 1059
37. Martínez-García MA, Méndez R,
1021 Epidemiol. 2015;44(3):827-836. obstructive pulmonary disease. Sci
Olveira C, et al. The U-shaped 1060
Transl Med. 2019;11(507):eaav3879.
1022 20. Wolf A, Dedman D, Campbell J, et al. relationship between eosinophil count and 1061
1023 Data resource profile: Clinical Practice 28. Barnes P. Corticosteroid effects on cell bronchiectasis severity: the effect of 1062
Research Datalink (CPRD) Aurum. Int J signalling. Eur Respir J. 2006;27(2): inhaled corticosteroids. Chest. 2023. Apr
1024
Q19 Epidemiol. 2019;48(6). 1740-1740g. 413-426. 21;S0012-3692(23)00604-9. 1063
1025 1064
1026 1065
1027 1066
1028 1067
1029 1068

10 Original Research [ -#- CHEST - 2023 ]

FLA 5.6.0 DTD
CHEST5708_proof
4 July 2023
8:17 am
EO: CHEST-D-23-00807