Secondary Logo

Journal Logo

Original Articles: Gastroenterology

Proton Pump Inhibitors and Infant Pneumonia/Other Lower Respiratory Tract Infections: National Nested Case-control Study

Blank, Mei-Ling; Parkin, Lianne; Zeng, Jiaxu; Barson, David

Author Information
Journal of Pediatric Gastroenterology and Nutrition: September 2018 - Volume 67 - Issue 3 - p 335-340
doi: 10.1097/MPG.0000000000001984

Abstract

What Is Known

  • Uncertainty remains regarding the possible association between the use of proton pump inhibitors and community-acquired pneumonia or other lower respiratory tract infections.
  • Methodological issues, including selection bias and information bias, may help explain the heterogeneous results reported in previous studies.

What Is New

  • This is the first nationwide study to examine the risk of community-acquired pneumonia or other lower respiratory tract infections among infant users of proton pump inhibitors.
  • We were able to address a number of methodological limitations found in previous observational studies.

Of concern, given the startling increases in proton pump inhibitor (PPI) use observed among children (1–5), there is very little information regarding the risk of community-acquired pneumonia (CAP) or other lower respiratory tract infections (LRTI) among infant users of PPIs, and uncertainty also prevails in the few studies conducted with children (6,7). An analysis of randomized clinical trial data reported that LRTIs were more frequent among participants aged 28 days to <12 months who were randomized to receive lansoprazole compared to placebo, although the difference was not statistically significant (7). Uncertainty also persists in studies investigating these outcomes among adults (8–11).

Methodological issues may explain the heterogeneity of reported findings. Although chiefly centered on the possible impact of unmeasured and uncontrolled confounders (12,13), researchers have also highlighted how differences between studies in the selection of cases and controls, level of case validation, and potential misclassification of PPI exposure may impact on observed associations (14).

We conducted a case-control study nested in a national cohort of infants (<12 months) in New Zealand (born between January 1, 2005 and December 31, 2012) who were dispensed the PPIs omeprazole, lansoprazole, or pantoprazole (the study PPIs) on at least one occasion before their first birthday. We aimed to estimate the risk of CAP, and all LRTIs, resulting in hospitalization or death in current, and recent, users of the study PPIs compared with past users of these drugs.

METHODS

Study Setting

During the study period, New Zealand's universal healthcare system was characterized by free hospital treatments, heavily subsidized prescription drugs ($NZ3 per dispensing), and free primary care consultations for children under 6 years old.

Data Sources

We used New Zealand's National Collections, population-based repositories of routinely collected demographic, health, and dispensing data covering approximately 4.2 million citizens and permanent residents (15). National Health Index identifiers (NHIs) assigned to approximately 95% of the population (16) enable patient-level linkage across the datasets. NHIs have been allocated at birth for all babies born in New Zealand since 1992, or are otherwise assigned at the very first encounter the patient has with the health system (17). We identified our cohort of infant PPI users from the Pharmaceutical Collection (Pharms), which records government-subsidized prescription drugs which are prescribed in primary care, hospital outpatient clinics, and on discharge from hospital, and are subsequently dispensed from community pharmacies (18). NHIs have been reliably recorded with Pharms data from 2005. Cases were identified from the National Minimum Dataset of hospital discharges (19), and the Mortality Collection (hospital and community deaths) (20). Diagnoses and underlying causes of death in these datasets are recorded using relevant editions of the International Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification codes (ICD-10-AM). Data were also obtained from the New Zealand Cancer Registry (21) and the National Immunization Register (data available from 2006 to 2011) (22).

Identification of the Study Cohort

All patients born between January 1, 2005 and December 31, 2012 and dispensed at least one course of treatment for a study PPI (omeprazole, lansoprazole, or pantoprazole) before their first birthday, were identified from Pharms by the Ministry of Health. The Ministry provided us with datasets containing the patients’ demographic data, details of all dispensings of the study PPIs and all other medicines dispensed during the first year of life, details of all public and privately funded hospital admissions during the first year of life, underlying causes of death (where applicable), cancer registration details (where applicable), and immunization data. Patients’ NHIs were replaced with unique patient identifiers generated by the Ministry; for patients identified by the Ministry as potential cases (see below), NHIs were also provided so that the patients’ clinical information could be requested from the treating hospitals. Cohort entry was the date of the first dispensing of a study PPI during the patient's first year of life.

We excluded all linked records for “patients” in which the demographic, health, and dispensing data could not have referred to the same person (eg, “patients” who supposedly received medicines before their recorded date of birth). We also excluded patients with a recorded history of any LRTI, including CAP, before their cohort entry date using the ICD-10-AM rubrics (determined in consultation with a professional clinical coder) under which LRTI, including CAP, may be coded (A15(0–3), A16(0–2), J09–J18, J20–J22, or J82) (Supplemental Digital Content 1, https://links.lww.com/MPG/B343).

Case Identification

To identify all patients in the cohort who were potentially diagnosed with CAP or another LRTI after cohort entry, we asked the Ministry to search the hospital discharge and mortality data of the cohort members using the ICD-10-AM rubrics listed above. For patients whose underlying causes of death had not yet been coded to ICD-10-AM rubrics, we searched the free text causes of death for “respiratory tract infection,” “lung infection,” “pneumonia,” “influenza,” “bronchitis,” “bronchiolitis,” “pulmonary eosinophilia,” and “tuberculosis.” We excluded patients who had ever been diagnosed with congenital pneumonia or neonatal aspiration syndromes. To verify the diagnoses of the remaining potential cases, hospital discharge letters, postmortem reports, and any available chest radiography reports were requested and independently reviewed by M-LB and LP (blinded to the patients’ PPI exposure status). Patients who had insufficient clinical information to validate the diagnosis were excluded, as were patients with hospital-acquired pneumonia or LRTI. In order to minimize potential confounding by underlying medical conditions, we also excluded patients with a major congenital abnormality or illness, including cancer, which may have increased the risk of CAP or another LRTI (Supplemental Digital Content 2, https://links.lww.com/MPG/B344), and patients whose CAP or other LRTI was secondary to a systemic disease.

Definite cases of CAP were patients whose diagnosis was verified by discharge letter or postmortem report, and chest radiography. Cases of LRTI (including CAP) were patients whose diagnosis was verified by discharge letter or death record, with or without chest radiography confirmation. The diagnosis date was designated as the index date for each case and their matched controls (see below).

Control Selection

Using risk set sampling (23), we randomly selected up to 10 controls (ie, patients who had not developed CAP or another LRTI) from the study cohort for each case (blinded to PPI exposure status), matched by sex and birthdate (±14 days). The same exclusion criteria were applied to both potential cases and potential controls. Potential controls were excluded if they had ever been diagnosed with congenital pneumonia or neonatal aspiration syndromes, or one of the congenital abnormality or major illness (including cancer) ICD-10-AM codes in Supplemental Digital Content 2 (https://links.lww.com/MPG/B344) before the index date of the case. Thus, eligible controls were study cohort members who were at risk of developing CAP or another LRTI on the matched case's index date. Patients were eligible to be controls for more than 1 case, and cases were eligible to be controls for other cases whose index date preceded their own.

Proton Pump Inhibitor Exposure

For each case and their matched controls, we determined the date a dispensed supply of PPI treatment would have extended to using an algorithm based on the “date dispensed,” “days supply,” and “quantity prescribed” fields recorded in Pharms, in conjunction with New Zealand guidelines for the maximum dispensed supply (15 days) of a compounded PPI (ie, a PPI capsule or tablet that has been made into a suspension with sodium bicarbonate and water) (24), and maximum dispensed supply (90 days) of any prescription drug (Supplemental Digital Content 3, https://links.lww.com/MPG/B345). In the Pharms database, compounded PPI preparations are recorded with a “days supply” value of zero.

Current use of a PPI was defined as a dispensed supply that extended to within 30 days before the index date. Recent use was a supply that extended within 31 to 90 days before the index date. Past use was a supply that ended more than 90 days before the index date. Because cases and their matched controls shared the same index date, the PPI exposure status for each matched case-control set was determined at the same calendar time.

Statistical Analysis

Sample size calculations conducted during the study design phase suggested we would need approximately 75 cases (assuming type I error probability of 0.05, power 0.80, probability of exposure among controls 0.5, and 10 matched controls for each case) to detect a weak association (odds ratio 2.0), should one exist. Conditional logistic regression was used to estimate the association between CAP or other LRTIs, and PPI use (reference category: past use). We considered the following potential confounders in the analysis: age at index date, ethnicity, socioeconomic status, pneumococcal conjugate vaccine (PCV) exposure (at least 1 recorded completed dose before the index date compared to no completed or recorded doses), dispensings of nontopical antibiotics within 30 days before the index date (1 or more dispensings compared to no dispensings), and hospital admissions for any diagnosis (excluding birth event) within 30 days before the index date (1 or more admissions compared to no admissions) (Supplemental Digital Content 4, https://links.lww.com/MPG/B346). To deal with missing data in the analysis, we used logistic, ordered logistic and multinomial logistic imputation models based on chained equations to impute the missing values for PCV exposure, socioeconomic status, and ethnicity, respectively. Variables included in the imputation models were case/control status, definite CAP case status, PPI exposure status, age at cohort entry, age at index date, birth year, sex, area-level deprivation, ethnicity, PCV exposure status, nontopical antibiotic dispensed within the 30 days before the index date, and any hospital admission (excluding birth event) within the 30 days before the index date. All statistical analyses were performed using Stata version 14 (College Station, TX).

Ethical approval came from the Southern Health and Disability Ethics Committee (14/STH/2). Funding was provided by The New Zealand Pharmacovigilance Centre and Medsafe, and a Strategic Research Grant from the Department of Preventive and Social Medicine, University of Otago.

RESULTS

The Ministry identified 22,661 patients from Pharms who were born between January 1, 2005 and 31 December 2012 and who were dispensed at least one course of PPI treatment before their first birthday (Fig. 1). The study cohort comprised 21,991 patients who had correctly linked demographic, health, and dispensing data and who did not have a hospital-based diagnosis of pneumonia or another LRTI before cohort entry. After excluding 91 patients who had ever been diagnosed with congenital pneumonia or neonatal aspiration syndromes, we identified 995 patients as potential cases, and requested discharge letters, postmortem reports, and chest radiography reports, receiving sufficient clinical information to validate the coded diagnoses for 978 patients (98.3%). On the basis of this information, an additional 412 patients were excluded. The study included 566 validated cases of LRTI (5650 matched controls), including 65 cases of radiography-confirmed CAP (650 matched controls).

F1
FIGURE 1:
Flowchart showing the process of selecting the study cohort and cases.

The characteristics of the cases and controls are shown in Table 1. Cases (definite CAP and all LRTI) were more likely than controls to be of Māori or Pacific ethnicities, to live in more socially deprived areas, and to have been dispensed a nontopical antibiotic, or to have been hospitalized for any diagnosis (excluding nonproblematic birth event), in the 30 days before the index date. Similar proportions of cases and controls had at least 1 recorded completed dose of PCV before the index date.

T1
TABLE 1:
Characteristics of cases and matched controls at index date

The associations between CAP, and LRTI, and PPI use are shown in Table 2. The unadjusted odds ratios were obtained based on all of the available CAP, or LRTI, cases and matched controls. Ninety-three patients in the CAP analysis (9 of whom were cases), and 698 patients in the LRTI analysis (64 of whom were cases), were, however, excluded from the adjusted analyses due to missing values for ethnicity, socioeconomic status, and/or PCV status. Missing PCV status (because the child was born after October 31, 2011 and hence may not have had the opportunity to have had their vaccinations recorded in the National Immunization Register) accounted for virtually all of the missing data (Table 1). Neither the unadjusted nor the adjusted analyses showed an association between PPI use and CAP or LRTI. Similarly, estimates obtained after imputing missing values for ethnicity, socioeconomic status, and PCV status showed no association between PPI use and CAP or LRTI (Table 3). As 5228 participants (84.1%) had at least 1 PPI dispensing before the index date where the “days supply” field was recorded as zero (indicating a compounded preparation), we conducted a sensitivity analysis based only on the “date dispensed” of the last PPI dispensing before the index date. In this analysis, current users were those who had a dispensing date for a PPI within the 30 days before the index date, recent users were those who had a PPI dispensed within 31 to 90 days before the index date, whereas past users’ last PPI dispensing occurred more than 90 days before the index date. As in the main analyses, there were no associations between current, or recent, PPI use and definite CAP or LRTI, compared to past use, in either the unadjusted or adjusted analyses (Supplemental Digital Content 5, https://links.lww.com/MPG/B347).

T2
TABLE 2:
Risk of community-acquired pneumonia or other lower respiratory tract infection (including community-acquired pneumonia) among children in New Zealand born between January 1, 2005 and December 31, 2012 who were dispensed at least 1 course of proton pump inhibitor treatment (omeprazole, lansoprazole, or pantoprazole) before their first birthday
T3
TABLE 3:
Risk of community-acquired pneumonia or lower respiratory tract infection (including community-acquired pneumonia) among children in New Zealand born between January 1, 2005 and December 31, 2012 who were dispensed at least 1 course of proton pump inhibitor treatment (omeprazole, lansoprazole, or pantoprazole) before their first birthday using multiply imputed values for missing ethnicity, socioeconomic status, and pneumococcal conjugate vaccine exposure status

DISCUSSION

In this nationwide study of infants in New Zealand who were dispensed a PPI (omeprazole, lansoprazole, or pantoprazole) on at least 1 occasion before their first birthday, we found that relative to past use, neither current nor recent use of a PPI was associated with an increased risk of CAP or LRTI resulting in hospitalization or death.

We searched MEDLINE to October 2017 to identify studies that have examined the association between use of PPIs and CAP or LRTI among infants younger than 12 months. An analysis of a small clinical trial observed a greater occurrence of LRTIs among infants with symptoms of gastroesophageal reflux disease randomized to receive lansoprazole (n = 81) compared to placebo (n = 81) (7). Among the patients randomized to receive lansoprazole, 4 (5%) developed a serious LRTI compared to 1 patient (1%) randomized to receive placebo; however, the finding was not statistically significant. A cohort study conducted among children aged 4 to 36 months old who were recruited from Italian pediatric gastroenterology clinics reported that children receiving omeprazole or ranitidine for symptoms attributed to gastroesophageal reflux disease were more likely to develop pneumonia compared to healthy children unexposed to these drugs (odds ratio 6.39, 95% confidence interval 1.38–29.70, P = 0.03) (6); however, confounding by indication cannot be excluded (ie, the underlying condition for which the PPI was prescribed, rather than the drug itself, may have been responsible for the elevated risk of pneumonia).

Our investigation, however, has a number of methodological strengths compared to these previous studies. We used a broad search strategy to preliminarily identify potential cases, and validated diagnoses with clinical information (including chest radiography reports when available). Patients were only included as cases if their CAP or LRTI was severe enough for them to be admitted to hospital or to die, minimizing the potential for diagnostic and referral bias (25) whereby current PPI users could be more likely to be diagnosed or referred because clinicians were aware of a possible association between PPI use and the outcome. We also included only patients with a first-ever diagnosis of CAP or another LRTI, avoiding any possible bias arising from the inclusion of recurrent cases (26). Within our nationwide cohort of infant users of PPIs, we identified 566 validated cases of LRTI, giving us sufficient power (80%) to detect a weak association (odds ratio 1.3) between PPI use and LRTI, if one exists. We used routinely collected, patient-level linkable demographic, dispensing and health data from a country with universal access to healthcare (27), and thus avoiding potential issues with selection bias that may arise in studies that rely on participant enrolment. The data used in this study are likely to be consistent and virtually complete due to a pharmaceutical reimbursement system that mandates the recording of key dispensing details, and national clinical coding standards for recording hospital diagnoses and causes of death. We likely identified virtually all infant users of the study PPIs during the study period because New Zealand has dispensing co-payments ($NZ3 during the study period) that are substantially lower than retail prices, and free access to primary care for all infants, thereby minimizing potential selection bias due to differential access to healthcare. Importantly, we minimized the potential for confounding by indication by restricting the study to ever-users of PPIs, comparing current, and recent, users with past users (28). Because patients who are prescribed a particular drug are likely to be systematically different from patients who are not prescribed the drug, observational studies comparing users with nonusers risk introducing bias. By following our cohort from birth, we included only new users of the study PPIs and our results are not biased by including prevalent users of PPIs (29).

Our study also had a number of limitations. A very large proportion of PPI dispensings had the “days supply” field recorded as zero, indicating a compounded preparation. However, sensitivity analyses based on the “date dispensed” field (a mandatory reporting field for reimbursement) of the last PPI dispensing before the index date did not change the interpretation of the findings. We also lacked information on drugs dispensed while in hospital or those bought over-the-counter as these data are not recorded in Pharms. In addition, we lacked information on diagnoses made in primary care, including possible indications for PPI treatment, and information on important confounders such as exposure to household tobacco smoke.

In conclusion, our study, which was able to address a number of methodological limitations found in previous observational studies, suggests that use of the PPIs omeprazole, lansoprazole, or pantoprazole by infants during the first year of life is not associated with significantly increased risk of CAP or LRTI resulting in hospitalization or death.

REFERENCES

1. Barron JJ, Tan H, Spalding J, et al. Proton pump inhibitor utilization patterns in infants. J Pediatr Gastroenterol Nutr 2007; 45:421–427.
2. Blank ML, Parkin L. National study of off-label proton pump inhibitor use among New Zealand infants in the first year of life (2005-2012). J Pediatr Gastroenterol Nutr 2017; 65:179–184.
3. Chen I-L, Gao W-Y, Johnson AP, et al. Proton pump inhibitor use in infants: FDA reviewer experience. J Pediatr Gastroenterol Nutr 2012; 54:8–14.
4. De Bruyne P, Christiaens T, Stichele RV, et al. Changes in prescription patterns of acid-suppressant medications by Belgian pediatricians: analysis of the national database (1997-2009). J Pediatr Gastroenterol Nutr 2014; 58:220–225.
5. Illueca M, Alemayehu B, Shoetan N, et al. Proton pump inhibitor prescribing patterns in newborns and infants. J Pediatr Pharmacol Ther 2014; 19:283–287.
6. Canani RB, Cirillo P, Roggero P, et al. Therapy with gastric acidity inhibitors increases the risk of acute gastroenteritis and community-acquired pneumonia in children. Pediatrics 2006; 117:e817–e820.
7. Orenstein SR, Hassall E, Furmaga-Jablonska W, et al. Multicenter, double-blind, randomized, placebo-controlled trial assessing the efficacy and safety of proton pump inhibitor lansoprazole in infants with symptoms of gastroesophageal reflux disease. J Pediatr 2009; 154:514.e4–520.e4.
8. Estborn L, Joelson S. Occurrence of community-acquired respiratory tract infection in patients receiving esomeprazole: retrospective analysis of adverse events in 31 clinical trials. Drug Saf 2008; 31:627–636.
9. Sultan N, Nazareno J, Gregor J. Association between proton pump inhibitors and respiratory infections: a systematic review and meta-analysis of clinical trials. Can J Gastroenterol 2008; 22:761–766.
10. Giuliano C, Wilhelm SM, Kale-Pradhan PB. Are proton pump inhibitors associated with the development of community-acquired pneumonia? A meta-analysis. Expert Rev Clin Pharmacol 2012; 5:337–344.
11. Johnstone J, Nerenberg K, Loeb M. Meta-analysis: proton pump inhibitor use and the risk of community-acquired pneumonia. Aliment Pharm Ther 2010; 31:1165–1177.
12. Jena AB, Sun E, Goldman DP. Confounding in the association of proton pump inhibitor use with risk of community-acquired pneumonia. J Gen Intern Med 2012; 28:223–230.
13. Othman F, Crooks CJ, Card TR. Community acquired pneumonia incidence before and after proton pump inhibitor prescription: population based study. BMJ 2016; 355:i5813.
14. De Groot MCH, Klungel OH, Leufkens HGM, et al. Sources of heterogeneity in case-control studies on associations between statins, ACE-inhibitors, and proton pump inhibitors and risk of pneumonia. Eur J Epidemiol 2014; 29:767–775.
15. Statistics New Zealand. 2013 Census Usually Resident Population Counts. Wellington: Statistics New Zealand; 2013.
16. Ministry of Health. National Health Index questions and answers. http://www.health.govt.nz/our-work/health-identity/national-health-index/nhi-information-health-consumers/national-health-index-questions-and-answers. Updated February 9, 2017. Accessed November 8, 2017.
17. Ministry of Health. National Health Index Data Dictionary Version 5.3. Wellington: Ministry of Health; 2009.
18. Ministry of Health. Pharmaceutical Claims Data Mart (PHARMS) Data Mart—Data Dictionary Version 4.1. Wellington: Ministry of Health; 2012.
19. Ministry of Health. National Minimum Dataset (hospital events) Data Dictionary version 7.8. Wellington: Ministry of Health; 2015.
20. Ministry of Health. Mortality Collection Data Dictionary Version 1.3. Wellington: Ministry of Health; 2009.
21. Ministry of Health. New Zealand Cancer Registry Data Dictionary Version 1.2. Wellington: Ministry of Health; 2010.
22. Ministry of Health. Questions and answers—National Immunisation Register. http://www.health.govt.nz/our-work/preventative-health-wellness/immunisation/national-immunisation-register/questions-and-answers-national-immunisation-register. Updated April 9, 2014. Accessed November 8, 2017.
23. Rothman KJ, Greenland S, Lash TL. Rothman KJ, Greenland S, Lash TL. Case-control studies. Modern Epidemiology 3rd edPhiladelphia: Lippincott Williams & Wilkins; 2008. 111–127.
24. BPAC. Irritable infants, reflux and GORD. Best Pract J 2011; 40:30–37.
25. Heinemann LAJ, Garbe E, Farmer R, et al. Venous thromboembolism and oral contraceptive use: a methodological study of diagnostic suspicion and referral bias. Eur J Contracept Reprod Health Care 2000; 5:183–191.
26. Heaton B, Applebaum KM, Rothman KJ, et al. The influence of prevalent cohort bias in the association between periodontal disease progression and incident coronary heart disease. Ann Epidemiol 2014; 24:741–746.
27. Thygesen LC, Ersbøll AK. When the entire population is the sample: strengths and limitations in register-based epidemiology. Eur J Epidemiol 2014; 29:551–558.
28. Psaty BM, Siscovick DS. Minimizing bias due to confounding by indication in comparative effectiveness research: the importance of restriction. JAMA 2010; 304:897–898.
29. Ray WA. Evaluating medication effects outside of clinical trials: new-user designs. Am J Epidemiol 2003; 158:915–920.
Keywords:

epidemiology; lansoprazole; lower respiratory tract infection; omeprazole; pantoprazole

Supplemental Digital Content

Copyright © 2018 by European Society for Pediatric Gastroenterology, Hepatology, and Nutrition and North American Society for Pediatric Gastroenterology, Hepatology, and Nutrition