Pneumococcal pneumonia infection
is associated
with end-stage renal disease in adult
hospitalized
patients
Shih-Ting Huang
1,2,8, Cheng-Li Lin
3, Yen-Jung Chang
4,8, Yuh-Pyng
Sher
5, Ming-Ju Wu
1, Kuo-Hsiung Shu
1,
Fung-Chang Sung
2,3and Chia-Hung Kao
2,6,7Streptococcus pneumoniae is the most common pathogen in community-acquired pneumonia (CAP) cases. CAP caused by S. pneumoniae has a poorer long-term prognosis compared with those of other forms of CAP, with a high all-cause mortality rate and cardiovascular-related mortality rate in elderly patients.1,2 Pneumococcal pneumonia (PP) might also contribute to extrapulmonary complications, including bacteremia with metastatic infection, endocarditis, and meningitis.3–5 An association between PP and acute cardiac events, such as myocardial infarction, arrhythmia, and congestive heart failure (HF), has also been reported.6,7
The proposed pathogenesis of the cardiac events associated with PP includes inflammation, hypoxia, and hypotension6; moreover, high circulating levels of interleukin 6 and tumor necrosis factor-a in pneumonia patients suggest that inflammatory cytokines might affect the myocardium.8,9
Pneumococcal disease also leads to renal complications, ranging from persistent proteinuria to end-stage renal disease (ESRD), in pediatric patients. Pneumococcal-associated
hemolytic-uremic syndrome (P-HUS) has emerged as the most common form of atypical P-HUS in children.10 P-HUS often results in acute renal failure that requires dialysis. Recent studies have reported variability in the long-term
outcomes of P-HUS.10–13 Investigators have reported chronic renal sequelae and have recommended that children receive long-term follow-up to reduce the risk of permanent renal injury.12,13 Although studies on P-HUS and invasive
pneumococcal disease have characterized the long-term renal effects in pediatric patients, the long-term renal effects in adult patients diagnosed with PP remain unclear.
The aim of our current study was to determine the relationship between PP and the subsequent risk for ESRD.
We conducted a retrospective cohort study in Taiwan, and the incidence of ESRD in adult patients at the time that PP diagnosis was recorded. The demographic characteristics, comorbidities, and long-term outcomes of PP patients were examined and compared with those of control group patients.
RESULTS
During the study period, there were 44,621 hospitalized patients diagnosed with PP (Figure 1). Among them, patients with ESRD before the index date, under 20 years of age, and with missing information were excluded.
The remaining 18,733 PP cases without ESRD at baseline were included in the study cohort. The comparison cohort included all people covered within the Taiwan National Health Insurance Program whether or not the people had ever been admitted to the hospital for other causes. For the non-PP comparison cohort, 73,409 subjects were randomly selected in a 1:4 ratio and matched with the PP cases on age, sex, comorbidities, and index year. The proportion of adult PP patients declined from 0.104 per thousand in 1998 to 0.071 per thousand in 2004, and increased to 0.084 per thousand in 2010. The average duration of hospitalization in the PP cohort was 21.9±48.5 days (mean¼9 days, interquartile range¼1,265).
The characteristics of the participants’ frequency matched by sex, age, and comorbidities are displayed in Table 1. The PP group comprised more men than women (65.9% vs. 34.1%). The average ages of the non-PP and PP participants were 65.8 and 66.1 years, respectively. Among the baseline
comorbidities, hyperlipidemia, chronic kidney disease
(CKD), peripheral vascular disease (PVD), HF, and malignancy were slightly more prevalent in the PP cohort than in
the non-PP cohort. No significant difference was observed in the prevalence of hypertension, diabetes mellitus (DM), and
chronic obstructive pulmonary disease (COPD) between the two cohorts. Adults at the time of PP diagnosis carried more comorbidity risks compared with the non-PP cohort. During the observation periods, the proportion of death and withdrawal from the insurance system (death events) was higher in the PP cohort than in the non-PP cohort.
The ESRD incidence densities for the PP and non-PP cohorts and the ESRD crude (relative) hazard ratios (HRs) for the patients in the PP cohort compared with those in non-PP cohorts are summarized in Table 2. The overall incidence of ESRD was 23% higher in the PP cohort than in the non-PP cohort (5.26 vs. 3.10 per 1000 person–years),
with an adjusted HR of 1.14 (95% confidence interval (CI)¼1.01–1.29). The incidence rates were 5.08 and 5.53 per
1000 person–years for the male and female PP patients, respectively. The crude HRs of ESRD for the male and female PP patients were 1.21 and 1.26, respectively, compared with the male and female non-PP subjects. The highest agespecific ESRD incidence rates for patients with PP were
observed in patients aged 50–64 years (7.68 per 1000 person–years), with an adjusted HR of 1.61 (95% CI¼1.28–2.02) compared with the non-PP cohort in the same age subgroup. The age-specific hazard ratio of ESRD in the PP cohort was greatest for patients aged p34 years, with an adjusted HR of 4.15 (95% CI¼1.55–11.1) compared with the non-PP cohort in the same age subgroup.
The increased adjusted HRs of ESRD in the PP cohort were observed in the comorbid subgroups of hypertension (adjusted HR¼1.19; 95% CI¼1.01–1.39) and malignancy (adjusted HR¼2.15; 95% CI¼1.05–4.40) compared with that of the non-PP cohort. However, the case numbers in the malignancy subgroup were small in both cohorts, and the interpretation of the
increased adjusted HRs in the malignancy subgroup should be limited. In the other aspect, the decreased adjusted HR of ESRD in the PP cohort was observed in patients with COPD (adjusted HR¼0.71; 95% CI¼0.52–0.96) compared with those without COPD. In the non-CKD subgroup, the adjusted HR of ESRD was higher in patients with PP compared with those without PP (adjusted HR¼1.20; 95% CI ¼1.05–1.37).
The results of the univariate and multivariate competingrisk regression analyses of the incidence of ESRD are shown in Table 3. In the univariate analysis, the measured HR of developing ESRD for PP patients was 1.23 (95% CI¼
1.09–1.38, Po0.001). After adjusting for age, sex, comorbidities, and calendar year, the adjusted HR of ESRD in the PP
cohort was 1.14 (95% CI¼1.01–1.29, Po0.05). The risk of developing ESRD among patients older than 75 years of age was 3.04-fold higher than that for patients who were younger than 34 years of age (95% CI¼2.64–7.15, Po0.001). The risk of developing ESRD was also greater for patients with CKD (adjusted HR¼9.59, 95% CI¼8.20–11.2, Po0.001), DM
(adjusted HR¼3.56, 95% CI¼3.15–4.02, Po0.001), hypertension (adjusted HR¼1.88, 95% CI¼1.65–2.14, Po0.001),
hyperlipidemia (adjusted HR¼1.39, 95% CI¼1.20–1.62, Po0.001), and HF) adjusted HR¼1.20, 95% CI¼1.03–1.40, Po0.05), whereas the decreased risk of ESRD was observed in COPD patients (adjusted HR¼0.57, 95% CI¼0.50–0.66, Po0.001). The significant interaction term between PP and comorbidities including hypertension, PVD, HF, and malignancy were observed (Supplementary Table 1 online).
Compared with the patients without PP and hypertension (adjusted HR¼1, Reference), patients with PP and hypertension had the highest adjusted HR (adjusted HR¼4.10, 95%
CI¼3.45–4.87, P¼0.04; adjusted HR¼3.32, 95% CI¼ 2.92–3.77 for those with hypertension only; adjusted HR¼1.20, 95% CI¼0.99–1.45 for those with PP only). The significant increased risks were also observed in the subgroup of patients with concomitant PP and PVD, HD, or malignancy.
The ESRD cumulative incidence curve after accounting for death and withdrawal from the insurance system as the competing risks (Figure 2) showed that the PP participants had a higher risk of ESRD than the patients with no history of PP in the observation periods. DISCUSSION
Summary of the main results
On the basis of our research, this study is the first to demonstrate a correlation between PP and ESRD, two seemingly
unrelated medical conditions. The incidence rate of ESRD following admission for PP was 5.26 per 1000 person–years, and the adjusted HR of ESRD was 1.14 (95% CI¼1.01–1.29, Po0.05). According to the latest annual report of the United States Renal Data System,14 the incident rates of ESRD in the United States, Taiwan, and Japan were 369, 361, and 288 per million people, respectively, and the prevalence of ESRD in Taiwan and the United States reached 2584 and 1870 per million people, respectively, in 2010. Owing to the financial burden caused by treating ESRD worldwide, our findings have major implications. First, clinicians should actively monitor declining renal function in patients with a history of hospitalized PP infection. Second, growing evidence shows that the incidence of non-PCV7-serotype–invasive pneumococcal disease has increased, and that the prevalence of
antibiotic-resistant S. pneumoniae has increased.15–18 Thus, public health officials should promote vaccination among high-risk patients, and a more broadly effective pneumococcal vaccine is required.
An explanation for the findings
To control the possible confounding factors affecting ESRD or PP, the control group patients were matched by age, sex, and comorbidities. The two cohorts were homogeneous with regard to sex, age, and comorbidities including hypertension, DM, and COPD. The other comorbidities, including hyperlipidemia, CKD, PVD, HF, and malignancy did not differ
substantially among cases and comparisons. After conducting a competing-risk model analysis, hypertension, DM, hyperlipidemia, CKD, and HF were identified as predictors of
our study, the patients with CKD carried the highest risk of developing ESRD (adjusted HR¼9.59, 95% CI¼8.20–11.20, Po0.001). After conducting a multivariate analysis, it was determined that PP infections still have a significant association with ESRD, with an adjusted HR of 1.14 (95% CI¼1.01–1.29).
The natural outcomes of CKD include elevated cardiovascular disease risk and infectious complications such as
pneumonia. Naqvi and Collins24 showed that CKD is associated with major infectious complications, which occurred three to four times more frequently than in the general population; their study also indicated the role of preventive vaccination for improving the prevention of PP. Later, James et al.25 demonstrated that patients with lower estimated glomerular filtration rate carry a higher risk of pneumonia-related death, particularly in younger patients. Recently, Viasus et al.26 also illustrated that Streptococcus pneumonia was the most frequent pathogen of pneumonia in CKD patients, and he suggested the protective role of
pneumococcal vaccination. One question that might be raised is whether PP infection should be considered as a competing risk for CKD as ESRD would be. However, based on these studies, the range of CKD outcomes is difficult to evaluate because kidney failure or CKD might lead to early death because of pneumonia, without undergoing dialysis or transplantation. In addition, the extent to which CKD marks or causes an elevated pneumonia risk cannot be determined using observational data. Thus, it was difficult to eliminate CKD as a confounding factor for ESRD in our study.
Nevertheless, the vigilance for pneumonia prevention and the subsequent morbidity and mortality apply to both the
general population and CKD patients.
Although we found that in the COPD subgroup the
coexistence of PP decreased the subsequent risk of ESRD, this result remains uninterpretable according to current knowledge. In the malignancy subgroup, the interpretation of the
increased adjusted HRs in the malignancy subgroup should be made cautiously because of the limited case number in
both cohorts.
To reveal the mechanisms of persistent renal sequelae following pneumococcal infection, we reviewed all the relevant studies and propose that acute sepsis causes the initial renal injury and the host immune response, and that the additional underlying comorbid diseases may contribute to secondary renal insults (Figure 3).
First, systemic hypotension and hypoxemia may result in peritubular hypoxia. Tissue hypoxia in the presence of various cytokines induces fibrosis-related processes.27 Second, the cell wall of pneumococci, a major virulence determinant, triggers host inflammatory responses and induces the production of cytokines.28 Fried et al.29 showed that an increased level of inflammation or prothrombotic markers, such as those of C-reactive protein, fibrinogen, and factor VII, are predictors of the deterioration of renal function in elderly patients.30 ClpL chaperonin, a member of the HSP100/Clp family, induces the secretion of tumor necrosis factor-a in the early stages of S. pneumoniae infection.31 Studies have suggested that tumor necrosis factor-a is a contributing factor for acute renal failure in sepsis,32 and it also induces renal interstitial fibrosis
mediated through the increased production of transforming growth factor-b1.33 Finally, a recent study indicated the role of the soluble urokinase-type plasminogen activator receptor (suPAR) in the initiation of focal segmental glomerulosclerosis.
34 Enhanced suPAR deposits in the glomeruli
resulted in the activation of podocyte b3 integrin. An earlier study suggested that an increased suPAR level reflects an increased expression of inflammatory cells in vessels during pneumococcal sepsis.35 Thus, an elevated suPAR level during PP might reflect ongoing inflammation that contributes to subsequent podocyte damage and proteinuria. Certain serotypes of S. pneumoniae might be nephritogenic and should be considered among patients with concurrent pneumonia and acute glomerulonephritis.36
Limitations and future research
The strengths of our study include the use of populationbased
studies of general populations,
and identifying the association between two comorbid diseases with potential prevention implications. Our findings should be interpreted in the context of the inherent
limitations in the observational study using administrative databases. First, the National Health Insurance Research Database (NHIRD) does not contain detailed information regarding smoking habits, socioeconomic status, and a family history of renal diseases. The relevant clinical variables, including blood pressure, pathology findings, and laboratory data such as estimated glomerular filtration rate and
albuminuria level, as important measures of CKD severity with prognostic implications, were unavailable. These variables cannot be adjusted in the analysis. However, the claims
data regarding the diagnoses of PP, ESRD, and the
comorbidities including CKD from the NHIRD were nonetheless reliable.37,38 Second, the evidence derived from a
retrospective cohort study is generally of lower statistical quality than that derived from randomized trials because of potential biases. Although we matched the controls to balance the demographic characteristics of the study cohorts and multivariate analysis were used to robustness of our results, bias resulting from residual confounders might have affected the results. However, residual confounders are unlikely to explain the association observed between PP and ESRD. Finally, we use the discontinuity of national health insurance as a proxy measure of death for the mortality calculation. The possible reasons for the discontinuity of national health insurance include the following: (1) death (major proportions, with a high 30-day mortality rate among hospitalized PP patients ranging from 11 to 21%39,40); (2) prison sentence and immigration (a few cases); and (3) withdrawal from the insurance system (a few cases, as the health insurance coverage rate wasB99% in 2012). Although we believe that death would be the major reason, the
mortality would still be overestimated. A prospective patient registration survey would be needed to monitor the change of estimated glomerular filtration rate and the degree of
proteinuria in patients following high-risk PP infection.
Future research to evaluate the effectiveness of vaccination in patients to reduce ESRD risk, particularly in the CKD or
elderly populations, would be promising.
PATIENTS AND METHODS Data sources
In March 1995, the Taiwanese Department of Health integrated 13 health insurance agencies into a nationwide, universal health insurance system. By the end of 2009, B99% of the 23.74 million residents of Taiwan were enrolled in the National Health Insurance (NHI) system. The NHI also includes a catastrophic illness program that exempts patients from co-payments for the corresponding medical services, and the registry for catastrophic illness data set includes dialysis patients. The NHIRD contains comprehensive claims records of the outpatient and inpatient care provided through the NHI.41 Data used in our study were extracted from the NHIRD.
We used the following three data sources: the registry of
beneficiaries, the catastrophic illness data set, and inpatient claims records. Patient information is linked among these databases by
encrypted and unique personal identification numbers, which we used to obtain a longitudinal medical history for each study participant. Our study was approved by the ethics review board of the China Medical University (CMU-REC-101-012).
Participants
Figure 1 shows the selection process of the participants in the two study cohorts. We reviewed 18,733 hospitalized cases of PP
infection. On the basis of the diagnostic criteria of the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), we identified patients newly diagnosed with PP with bacterial isolates (ICD-9-CM 481) between 1998 and 2010 who were at least 20 years old. The pneumonia severity index, a clinical prediction rule, was used by clinical physicians for risk stratification and as a decision aid to determine the initial site of treatment for high-risk adults with CAP,42 with brief observation stays recommended
for patients with class III risk, and traditional inpatient care for patients with class IV and V risks. For the non-PP comparison cohort, 73,409 subjects without a history of PP were selected and matched with PP-cohort patients by age (5-year increments), sex,
and comorbidities, with a control-to-case ratio of 4:1.
The index date was defined as either the date of PP diagnosis (for the PP-cohort patients) or the 15th day of the same month (for the non-PP cohort patients). We excluded patients with ESRD (ICD-9-CM 585) diagnosed before the index date, and those whose records did not contain patient sex and age information. Patients with a subsequent diagnosis of ESRD (ICD-9-CM 585) were identified from the catastrophic illness data set. The patients in both cohorts were followed up until one of the following conditions occurred: an initial diagnosis of ESRD, death, withdrawal from the NHI, or 31 December 2010. Hypertension 9-CM 401-405), DM (ICD-9-CM 250), hyperlipidemia (ICD-(ICD-9-CM 272), CKD (ICD-(ICD-9-CM 585), COPD (ICD-9-CM 490-492, 494, and 496), PVD (ICD-9-CM 441, 443.9, 785.4, and V43.4), HF (ICD-9-CM 428), and malignancy (ICD-9-CM 140-208) diagnoses were recorded to establish the baseline comorbidity history of each participant.
Statistical analysis
The number of adult PP patients diagnosed annually was measured chronologically from 1998 to 2010. The demographic characteristics and the prevalence of comorbidities in the PP and non-PP cohorts were compared using a w2 test for categorical variables and a t-test
for continuous variables; the incidence densities by sex, age, and comorbidity were also calculated for each cohort. After accounting for the competing risks of death, we used the Fine and Gray model43
(which extends the standard Cox proportional hazard regression model) to estimate the cumulative incidence of ESRD, after accounting for the competing risk of deaths. The identification of death events was based on hospital discharge because of death and withdrawal from the NHI, as indicated in the NHIRD. Univariate and multivariate competing-risks regression models were proposed to calculate the subhazard ratios and the 95% CIs of the risk of
ESRD associated with PP. The multivariate models were simultaneously adjusted for the demographic characteristics and baseline
comorbidities. We plotted a 13-year cumulative incidence of ESRD, using the cumulative risk method, which considers death events as a competing risk.44 All statistical analyses were performed using the
SAS statistical software package, version 9.1 (SAS Institute, Cary, NC). The cumulative incidence survival curve was plotted using
Stata version 11.1 (Stata LP, College Station, TX). Results with a two-tailed P-value o0.05 were considered to indicate a statistically significant relationship.
CONCLUSION
PP not only reflects the underlying comorbid conditions but is also associated with ESRD in adult patients. The underlying pathophysiological mechanisms contributing to this relationship might be multifactorial. One episode of PP might exert
clinically substantial long-term renal effects. Therefore, the long-term follow-up of renal function is recommended in adult patients with a history of PP.