This case-control study reaffirmed that the use of inhaled corticosteroids and atypical antipsychotic drugs was independently associated with an increased risk for CAP among hospitalized older adults of a rural community with the adjustment of nutritional status based on serum albumin levels and other risk factors. However, the association between current PPI exposure and the risk for CAP was not observed in our study. Our study also demonstrated that by decreasing serum albumin level by one gram per deciliter, the risk for CAP can be increased by 1.89- fold in older adults.
COPD has been demonstrated as a major risk factor for CAP. Inhaled corticosteroids, often prescribed for the treatments of COPD and/or asthma, were also identified as an independent risk factor by our study and others [20
]. Although inhaled corticosteroid therapy may improve the quality of life [26
], inappropriate prescriptions of these medications may be highly prevalent among stable COPD patients [27
]. Re-evaluating the need for a long-term corticosteroid therapy or using it at a minimally effective dose is warranted.
Our study also demonstrated that atypical antipsychotic drug use independently increased the risk of CAP, which was consistent with the report by Knol et al [14
]. Their study included all types of pneumonia in the cases and demonstrated a 2.1-fold increase in the risk for CAP among those taking atypical antipsychotic drugs. With the exclusion of aspiration pneumonia in the case group, our study still revealed a similar result with a modest risk (AOR = 2.26, 95% CI = 1.23-4.15). Effects of atypical antipsychotic drugs on the gastrointestinal system and particularly on the esophageal dysfunction were described [28
]. Though it is highly possible that the intrinsically sedative and anticholinergic side effects of this group medication could increase the risk of aspirations by decreasing peristalsis, the exact mechanism of increasing the risk for CAP is not clear.
Our study did not find an association between current PPI therapy and the risk for CAP after the adjustment for confounders. The result was consistent with a recent report of a large nested, case-control study by using the UK General Practice Research Database by Sarkar et al. [19
], in which patients with the diagnosis of aspiration pneumonia were excluded in the data analysis as we did in our study. Please note that the study by Sarkar et al. included cancers in the full adjustment when reporting the adjusted OR, while our study excluded active lung cancers and metastatic diseases in the final data analysis. Compared to previous studies that reported the association between PPI use and the risk for CAP (but not hospital-acquired pneumonia) [15
], the exclusion of aspiration pneumonia was not explicitly mentioned in their studies, which could be one of the possible explanations for the discrepancy of reported findings between ours and other studies.
Our studied patients had a higher prevalence rate of co-morbidities, and a higher percentage of patients received PPI therapy compared to other studies [15
]. This finding was expected as aging older adults have more co-morbidity and increasingly use pharmacotherapy for chronic illness, especially among hospitalized patients. As the most potent acid-suppression drugs, PPIs have substantially replaced H2 receptor antagonists for the treatment of acid-related gastroesophageal diseases. Evidence suggests that patients with multiple co-morbid conditions are most likely to receive PPI therapy [29
], and it is also this group of patients who are most vulnerable to lower respiratory tract infections. Evidence also suggests that hypoalbuminemia and malnutrition are associated with CAP in older adults [10
], which is also demonstrated by our study. However, PPI users and non-users had no difference in average serum albumin levels among hospitalized older patients as demonstrated in our previous study [30
Despite PPI therapy drastically reduces gastric acid secretion it does not prevent or inhibit the gastro-esophageal reflux episodes. Clinical observations suggest that reflux of gastric contents into the upper and lower airway may precipitate acute asthma attacks, or acute exacerbation of COPD though there is no strong evidence from clinical studies [31
]. It is quite possible that PPI therapy could also be prescribed for the early symptoms that are associated with lower respiratory tract infections or gastro-esophageal reflux disease (GERD) (i.e., protopathic bias). This may partially explain why PPI therapy started within the past 7 to 14 days of pneumonia diagnosis was found to be associated with an increased risk for pneumonia in some studies [15
Another possible explanation for the different result observed between our and other studies [15
] was the categorization of exposure history. Other studies [15
] had data available (based on prescriptions of PPIs) that employed three categories: current, recent, and past use; ours only included current use vs. non-use. Furthermore, our study was hospital-based, case-control. One may suggest that a Berkson's bias, a phenomenon in which the relationship between exposure and disease becomes indistinct in hospital-based studies, may occur in our study. Although we could not exclude the possibility, this current study also demonstrated the association between the use of inhaled corticosteroids and atypical antipsychotic drugs and the increased risk for CAP in our studied patients. Because of the small sample size compared to those population-based studies, the issue of statistical power could be one of the explanations for the lack of the association. However, other investigators [19
] of a large study also reported a finding similar to ours in the relationship between overall PPI use and the risk for CAP. Therefore, we do not believe that increasing the sample size will change the conclusions drawn from our study.
Whether age could be an explanation for the discrepancy between our study and the others [15
] is an interesting question. In the study by Gulmez et al. [16
], the association between current use of PPI and CAP became weaker in the age group of older than 60 years (adjusted OR 1.5, 95% CI = 1.3 - 1.7) when compared to the age group of less than 40 years (adjusted OR 2.3, 95% CI = 1.3 - 4.0). We speculate that the association between PPI use and the risk for CAP may become non-significant in the advanced aged group. As demonstrated by Sarkar et al. [19
], their paper also reported no association between PPI use and the risk for CAP in persons aged 60 years or older.
As gastric acid production decreases with aging and there is a high prevalence rate of chronic gastritis in people aged older than 70 years, there may be a high prevalence rate of bacterial colonization in the upper gastro-esophageal tract of older adults at baseline even in those older people not receiving PPI therapy. If this is true, use of PPI therapy may not further increase the prevalence rate of bacterial colonization in older adults, by which is the mechanism thought to be responsible for the increased risk for CAP by PPI therapy. There is no information in how PPI therapy affecting the rate of bacterial colonization in older adults compared to young adults in the literature.
One of the strengths in our study was to demonstrate that patients with CAP had a significantly higher mean white blood cell count on admission compared to the control patients (Table ). Though we could not exclude the possible use of systemic corticosteroid treatment prior to the blood tests obtained in the ED or on admission as the cause of leukocytosis, the exclusion of acute exacerbation of COPD, respiratory failure requiring a ventilator therapy and leukemia cases from our data analysis, as well as very few patients receiving oral systemic steroid therapy, should minimize such a possibility. Our study also demonstrated that lower serum albumin levels independently increased the risk for CAP 1.89-fold in older adults. This is the first of such a report to the best of our knowledge.
The limitations of our study included the lack of diversity as the study subjects were a predominately white population. Therefore, the findings may not apply to other ethnic groups. Furthermore, our data collection resources could not confirm whether patients were compliant with medication use nor the duration as the data collection was based on medical records and physician's notes. Dosage information will provide more insights into the association between a drug therapy and the risk for CAP. Another limitation was unmeasured or residual confounding. Examples are the use of NSAIDs, as readily available over-the-counter, and co-morbidity that may not be completely documented by health-care providers. Certain medical conditions that may be under- assessed and under-documented, such as cognitive impairment or dementia, lack of standardized criteria for making a diagnosis of diseases, as well as the lack of data on cognitive functions prior to and during hospital stay, in which potential limitations are all inherent to a retrospective study design.