© 2002 American Thoracic Society
Detecting Stenotrophomonas maltophilia Does Not Reduce Survival of Patients with Cystic FibrosisDivision of Pulmonary and Critical Care Medicine, Department of Medicine, University of Washington Medical Center, Seattle, Washington Correspondence and requests for reprints should be addressed to Christopher H. Goss, M.D., M.S., Assistant Professor of Medicine, University of Washington Medical Center, Campus Box 356522, 1959 N.E. Pacific, Seattle, WA 98195. E-mail: goss{at}u.washington.edu
Stenotrophomonas maltophilia is a gram-negative bacterium that has been cultured with increasing prevalence from the sputum of patients with cystic fibrosis (CF). We conducted a cohort study, using the Cystic Fibrosis Foundation Registry, to assess the effect of S. maltophilia on survival. We studied all patients in this registry from 1991 to 1997 who were older than 6 years of age, were S. maltophilia negative in the first year of enrollment, and had their CF diagnosed before the age of 45 years (n = 19,255 in the study). We compared patients who acquired S. maltophilia with those who did not, using survival analysis. A total of 1,673 (8.7%) had at least one sputum sample positive for S. maltophilia. Compared with patients without S. maltophilia, those patients positive for S. maltophilia had the following baseline characteristics before detection: lower FEV1 % pred (p < 0.001); older (p = 0.001); more likely to be female (p = 0.003); and more pulmonary exacerbations (p < 0.001), outpatient visits (p < 0.002), and total hospitalizations (p < 0.001). After controlling for differences in severity of disease and coinfection with Pseudomonas aeruginosa, the hazard ratio associated with S. maltophilia detection was 0.89 (95% confidence interval, 0.75–1.05). Although patients with CF who acquire S. maltophilia have more advanced disease than those who do not acquire this organism, detection does not independently affect short-term survival (3 years).
Key Words: cystic fibrosis • outcome • Stenotrophomonas maltophilia • survival • outcome
Cystic fibrosis (CF) is one of the most common inherited fatal diseases in white individuals with a reported incidence of CF from 1 in 2,000 to 1 in 3,200 live births (1). Over the past 30 years, the median age of survival has improved from 14 years in 1969 to 32 years in 1998 in the United States (2). The improvement in survival has been attributed to improved nutritional support and aggressive treatment of pulmonary infections. With improved survival, physicians caring for these patients have witnessed the emergence of new pathogens in the CF lung (Burkholderia cepacia, Stenotrophomonas maltophilia, Alcaligenes xylosoxidans, Mycobacterium avium complex). Of these organisms, only Burkholderia cepacia has clearly been associated with a worse prognosis (3, 4). Stenotrophomonas maltophilia (previously Pseudomonas and Xanthomonas maltophilia) is a nonfermenting gram-negative rod, which represents one of the more prominent emerging pathogens in the CF airway. This organism has significant resistance to broad-spectrum antibiotics, leading to only limited treatment options. S. maltophilia has been isolated from a wide range of nosocomial sources, including central venous catheters, deionized water dispensers, dialysis machines, nebulizers and inhalational equipment, and oxygen humidifier reservoirs (5). In patients without CF, the organism has been clearly defined as a nosocomial pathogen with documented cases of bacteremia, endocarditis, pneumonia, meningitis, and sepsis (6, 7). Risk factors for detecting this organism in the non-CF population are often seen in patients with CF, namely prior antibiotic therapy, the presence of a central venous catheter, prolonged hospitalization, and steroid therapy (5). However, little is known about the role of S. maltophilia in patients with CF. The importance of S. maltophilia colonization in patients with CF is still unclear. Three small case control studies have tried to address this issue and have come to different conclusions regarding the clinical significance of the presence of S. maltophilia in the sputum of patients with CF (8–10). Two studies found no evidence of clinical decline associated with S. maltophilia (8–10). The third study suggested that in patients with an FEV1 % pred less than 40%, 5-year survival was worse (8–10). For this reason, we analyzed data from the National Cystic Fibrosis Registry to assess the clinical characteristics of patients who acquire S. maltophilia, the prevalence of S. maltophilia, and the impact of detection on survival of patients with CF.
Participants The CF Foundation National Registry contains demographic and clinical data collected at U.S. CF Foundation-accredited centers, using an annual standardized entry form. A full description of the database has previously been published (11). We studied all patients in the registry between January 1, 1991 and December 31, 1997. Patients aged 6 years or younger or patients older than 45 years at the time of diagnosis of CF were excluded because the former group cannot reliably perform spirometry and the latter group represents an unusual subgroup. Patients infected with S. maltophilia in the first year of observation or with only 1 year of observation were excluded.
Design and Procedures FEV1 and FVC used were the highest values recorded during the year, with percent predicted value calculated with Knudson equations (12). Pancreatic insufficiency was defined as use of pancreatic enzymes. Other clinical measures included total hospital admissions, courses of home intravenous antibiotic treatments, outpatient visits, and number of CF pulmonary exacerbations. CF exacerbation was defined as a CF-related admission to the hospital or use of home intravenous antibiotics (data regarding hospitalization for acute respiratory exacerbation were not available until 1994). Microbiology available for the study included Pseudomonas aeruginosa, Haemophilus influenzae, Staphylococcus aureus, B. cepacia, and S. maltophilia. All microbiology was assessed 1 year before the index date. Patients with unknown sputum culture or pancreatic sufficiency status were assumed to be negative.
Statistical Analysis
We identified all patients in the Registry from 1991 to 1997 (years in which S. maltophilia data were collected) who were older than 6 years of age (n = 21,964). There were 392 patients who were excluded from the analysis because their sputum was positive for S. maltophilia in the first year of enrollment. Patients were also excluded if their CF was diagnosed after the age of 45 years (n = 41). A total of 2,276 additional patients were excluded because they had only 1 year of observation during the study period (Figure 1) .
The median follow-up for the cohort was 3 years (interquartile range [IQR], 2 to 5 years). Of the eligible patient population, 1,673 (8.7%) had at least one sputum sample positive for S. maltophilia. The annual prevalence of patients with S. maltophilia in their sputum increased from 2.1% in 1991 to 5.5% in 1997. The 1,673 patients positive for S. maltophilia were compared with 15,159 patients who were negative for S. maltophilia (Figure 1). Full 10:1 sampling for the comparison patients was not achieved in all patients given the lack of potential matches (102 S. maltophilia-positive patients were matched at 10:1 and 1,571 were matched at 9:1). Their baseline characteristics were assessed 1 year before the index date to reflect clinical status before detection of S. maltophilia. Those patients with at least one sputum sample positive for S. maltophilia had the following baseline characteristics compared with the patients without S. maltophilia: had lower FEV1 % pred (63 versus 72%, p < 0.001); were older (18.9 versus 16.6 years of age, p = 0.001); were more likely to be female (50 versus 46%, p < 0.003); and were more likely to be positive for P. aeruginosa (62 versus 52%, p = 0.001) and B. cepacia (5 versus 3%, p = 0.001) (Table 1) . Patients with S. maltophilia were also more likely to have S. aureus (p = 0.001) in their sputum, although at much lower rates than P. aeruginosa. The patients who were positive for S. maltophilia were more likely to be hospitalized (p < 0.001) and had more frequent outpatient visits (p < 0.002) and pulmonary exacerbations (p < 0.001) 1 year before detection. No significant difference in median follow-up in survivors in both cohorts was found (Table 1). Overall, these clinical features suggest more advanced disease in the patients who went on to acquire S. maltophilia in the subsequent year.
During the study period, 175 (10.5%) patients positive for S. maltophilia died compared with 1,096 (7.2%) who were negative for S. maltophilia. Thirty-four (2.6%) S. maltophilia-positive patients underwent solid organ transplantation during the follow-up compared with 167 (1.1%) S. maltophilia-negative patients (p < 0.01). Fewer patients with S. maltophilia were lost to follow-up during the study compared with patients who were in the S. maltophilia-negative group (42 [2.5%] compared with 1,290 [8.5%], p < 0.01). Those patients lost to follow-up had better lung function (FEV1 % pred of 73% compared with 63%, p < 0.005), were older (19.3 years compared with 16.6 years, p < 0.01), had fewer hospitalizations (mean number of hospitalizations: 0.59 compared with 0.8, p < 0.01), and had fewer outpatient visits (mean number of office visits: 3.1 compared with 4.7, p < 0.01) than those patients with full follow-up. The unadjusted relative risk of death associated with S. maltophilia detection was 1.35 (95% confidence interval [CI], 1.15 to 1.58). Patients who acquired S. maltophilia had a significantly worse unadjusted survival (p = 0.002). When the patients were stratified according to FEV1 % pred, however, no significant difference in survival of those patients who were S. maltophilia positive compared with S. maltophilia negative was found, using the log rank statistic (Figure 2) . No difference was seen in survival when the cohorts were stratified on age (Figure E1; see online data supplement). In a restricted subset of patients with FEV1 < 50% of predicted, with coexistent pancreatic insufficiency and P. aeruginosa colonization, no significant difference in survival was seen in those patients who had acquired S. maltophilia compared with those who had not acquired the organism (p = 0.09) (Figure E2; see online data supplement). The cutoff point of FEV1 < 50% of predicted was used instead of 40% of predicted because of the small number of patients fulfilling this latter cutoff with coexistent pancreatic insufficiency and P. aeruginosa colonization in both the patients negative for S. maltophilia and positive for S. maltophilia.
Multivariate Cox proportional hazard modeling was used as an alternate method to adjust for baseline differences between the two cohorts. Initial evaluation of the variables revealed that FEV1 % pred violated the proportional hazard assumption. For this reason, FEV1 % pred was used as a stratification variables (strata for FEV1 % pred were in deciles). S. maltophilia was not associated with worse survival (hazard ratio [HR], 0.89; 95% CI, 0.75 to 1.05) after adjusting for the effects of sex, age, P. aeruginosa colonization, and lung function. Other variables assessed in the model included calendar year, colonization with B. cepacia or S. aureus, weight, height, genotype, comorbidities, total number of hospitalizations per year, total number of exacerbations per year, total number of office visits per year, and insurance status (Table 2) . Importantly, both P. aeruginosa infection and B. cepacia infection were associated with a higher adjusted relative hazard of death, despite our inability to control for the affect of treatment. None of the interactions identified a priori affected the model. Results of the stratified Cox proportional model were similar to the nonstratified model with FEV1 % pred modeled as a continuous variable, but to avoid violating the proportional hazard assumption the stratified model has been presented.
To test the hypothesis that S. maltophilia does not affect survival when assessing infections at multiple time points, a Cox proportional hazard model with time-dependent variables was evaluated. This analysis was performed on all patients in the initial cohort of patients who fulfilled the inclusion and exclusion criteria (n = 19,255). In this model, both S. maltophilia status and P. aeruginosa status were assessed annually during the follow-up period. After adjusting for age, sex, weight, height, lung function, sputum microbiology, comorbidities, hospitalization rate, and insurance status, S. maltophilia was not associated with worse survival (HR, 0.92; 95% CI, 0.74 to 1.14) (Table 3) . To test the hypothesis that changes in culturing decisions or techniques affected the results, the final model was reanalyzed, restricting the data to the years 1995 to 1997. Again, after adjusting for baseline differences, S. maltophilia was not associated with worse survival, albeit with wider confidence intervals given the decrease in sample size (HR, 0.82; 95% CI, 0.62 to 1.09).
Patients with CF who acquire S. maltophilia were older, had lower FEV1 % pred, and had higher rates of coinfection with P. aeruginosa and B. cepacia in the year before detection than a randomly selected sample of patients with CF who were negative for S. maltophilia. As might be expected, patients who acquired S. maltophilia, given their differences at baseline, had worse unadjusted survival than uninfected patients. However, we have also shown by three different methodologies (stratification, restriction, and multivariate adjustment) that after adjusting for these initial demographic and physiologic differences, detection of S. maltophilia was not associated with worse survival. In addition, by testing various interaction terms in the multivariate model, we found no evidence for a specific subgroup of patients in whom S. maltophilia is particularly deleterious. In CF populations, the prevalence of S. maltophilia appears to have increased over the last 15 years (15–17). Bauernfeind and coworkers (15) performed a qualitative and quantitative microbiologic analysis of sputum from 102 patients with CF over a 22-month period and found 6.8% of the patients had positive sputum culture for S. maltophilia. The prevalence of S. maltophilia in this study was much higher than previous rates noted in the literature. Ballestero and coworkers (18) studied a population among 52 of their patients with CF during a 5-year period in Spain. S. maltophilia was cultured in the sputum at least once in 30.7% of their patients. Burns and coworkers (17) evaluated the microbiology of sputum samples of a Phase III national collaborative study of aerosolized tobramycin in 595 patients at 69 CF centers. They found a prevalence of 10.3% of S. maltophilia. In the analysis of the CF Foundation National Registry, the prevalence of this organism has increased, but at a lower overall prevalence than previously reported. We cannot delineate whether this increase in prevalence was due to an increase in culturing frequency, changing culturing techniques, or a true increase prevalence of the organism in the CF airways. The finding that the prevalence differed in the study by Burns and coworkers compared with the data in this study may be due specifically to misclassification of patients in the CF Foundation National Patient Registry. It must be noted, however, that in the study by Burns and coworkers, patients needed to be positive for P. aeruginosa, be 6 years of age or older, and have an FEV1 value between 25 and 75% of predicted to be included in the study. In our analysis, all patients with CF who were 6 years of age or older were included, giving a different denominator. A more formal long-term longitudinal assessment would need to be performed with a central laboratory to truly clarify this issue. Prior investigations have found that patients have worse clinical status at the time of isolation of S. maltophilia compared with those patients with CF who have not acquired the organism (10, 16). Characteristics associated with S. maltophilia detection have included lower mean weight percentile, lower Shwachman-Kulczycki score, lower FEV1 % pred, and more advanced age. Denton and coworkers (16) had noted a rapid rise in S. maltophilia prevalence with a corresponding drop in Pseudomonas prevalence and were concerned that this phenomenon was related to aggressive antipseudomonal therapy. In their study, prior use of intravenous antipseudomonal antibiotics, prior use of inhaled aminoglycosides, and prior frequent and longer hospitalizations all predicted S. maltophilia colonization in patients with CF. In a separate analysis of 58 S. maltophilia-positive patients matched by age to 116 control patients, a multivariate analysis found that chronic suppressive antibiotics and number of days of intravenous antibiotics appeared to be independent predictors of later colonization (19). We found similar results when comparing our S. maltophilia-positive cohort with the negative cohort. One year before detection of S. maltophilia, positive patients were more likely to be older, have more frequent exacerbations and hospitalizations, have worse lung function, and were more likely to be colonized with both P. aeruginosa and B. cepacia. Reports have suggested that patients colonized with S. maltophilia are more likely also to have S. aureus (19). Our analysis has confirmed these findings. However, in our model, the presence of S. aureus in the sputum of patients with S. maltophilia did not affect survival. The clinical relevance of this increase in prevalence of S. maltophilia has been unclear. Three case control studies have come to different conclusions regarding the clinical significance of the presence of S. maltophilia in the sputum of patients with CF (8–10). Gladman and coworkers monitored 216 children at a CF center between 1983 and 1990 (8). S. maltophilia was recovered from 23 of these children, with the prevalence of S. maltophilia in 1990 being 10%. The authors noted no definable clinical deterioration secondary to S. maltophilia colonization. Karpati and coworkers (9) monitored 150 patients from 1983 to 1992. S. maltophilia was isolated from 25 patients. Using controls matched according to age of first colonization with P. aeruginosa, the authors noted a significant decline in FEV1 at 2 years of follow-up in patients with S. maltophilia, but no clear evidence of clinical decline. The actual degree of lung function decline was not specified. Demko and coworkers monitored 773 patients seen at the Cleveland CF Center from 1982 to 1994 (10). S. maltophilia was recovered from 211 patients who were monitored for at least 1 year. Patients were stratified by spirometry into mild, moderate, and severely decreased FEV1 and FVC. In the severe group, S. maltophilia had a slightly higher mortality at 2 years (29 versus 22%). The 5-year mortality was 60 and 28%, respectively. The authors felt this data corroborated their theory that S. maltophilia was a marker for more severe lung disease and not a risk factor for early mortality. However, they concede that their data were not suitable for formal survival analysis and could not rule out the possibility that S. maltophilia colonization is associated independently with a higher mortality and more rapid decline in pulmonary function. These studies were single-center studies, with relatively few patients who had S. maltophilia in their sputum. None of these studies used survival analysis to assess the impact of S. maltophilia on clinical outcome. In our analyses, we have shown that initial detection of S. maltophilia is not associated with worse survival, unlike either P. aeruginosa or B. cepacia, which were both associated with a higher relative hazard of death. Patients who acquired S. maltophilia had more advanced disease, and thus had decreased unadjusted survival. The question that remained was whether this difference in disease state at the time of S. maltophilia detection explained the differences in unadjusted survival that we found. Several methods allow one to account for these differences. The simplest method is to divide the patient population into specific strata based on age and lung function, two important predictors of survival in CF, and compare survival of patients within the same strata. When this was done, no difference could be found in the survival curves between those patients who acquired S. maltophilia and those without the organism. One important limitation of such an approach is the risk of having a Type II error based on the smaller sample sizes in each strata. Evidence in the literature pointed to the possibility that a subset of patients may have been adversely affected by S. maltophilia, those with more advanced disease. By restricting the analysis to a subset of patients with more advanced disease defined by an FEV1 less than 50% of predicted, pancreatic insufficiency, and coinfection with P. aeruginosa, this possibility could be addressed. Even in this "sickest" subset, S. maltophilia had no clear association with worse survival. Unfortunately, neither of the methods can simultaneously adjust for the multitude of variables that differed in these two groups. To address this, multivariate survival analysis was performed to more completely adjust for baseline differences. Again, there was no clear evidence that initial detection of S. maltophilia was associated with worse survival, despite finding both P. aeruginosa and B. cepacia clearly associated with worse survival. Because differential follow-up in survivors can bias the results, we compared median follow-up between those with S. maltophilia and those without the organism and found no significant difference. Important limitations of these analyses relate to the assessment of S. maltophilia status. Differences in the number of sputum samples requested, the specific organisms requested, and sputum culturing techniques may have changed through time, leading to misclassification of S. maltophilia status and ascertainment bias. However, no clear temporal effects were noted when adjusting for calendar year of S. maltophilia detection. Looking only from 1995 to 1997, S. maltophilia was not associated with significantly decreased survival during a time period when more consistent culturing of this organism is likely to have occurred. We could not control for the exact observation time before being classified as culture positive in this analysis. This may lead to variable time periods when assessing clinical variables before colonization. The values of lung function and anthropomorphic values were mean values for the year before detection, which will decrease the importance of each variable's predetection time. In these analyses, patients with a single positive culture per year are treated the same as patients with multiple positive cultures per year, thus limiting our ability to assess a dose effect. By using time-varying covariates, we have addressed to some extent varying exposure to S. maltophilia through time, but have not specifically addressed whether persistent colonization leads to worse outcome. We also could not control for treatment of S. maltophilia-positive patients. Thus, we cannot rule out a distinct dose effect or treatment effect that led to the conclusion regarding the effect of S. maltophilia detection on survival. Last, our initial analysis only addressed the effect of the initial positive culture for S. maltophilia. Episodic infection with S. maltophilia has clearly been identified in the literature, with a small subset of patients remaining persistently culture positive (10, 20). To evaluate this, we assessed S. maltophilia infection through time, allowing for repeated infection and again found no clear evidence to implicate this organism with worse survival. Other important limitations of this analysis include the fact that the data are annualized anonymous data and, because of this, cannot be validated. However, Fitzsimmons estimated that the CF Foundation National Registry captured 95% of all deaths of patients with CF in the United States (11). Although we did not note differential median follow-up in the patients who survived the study period, we did note differential rates of transplantation and patients lost to follow-up between the two groups. If anything, higher rates of transplant during follow-up in the S. maltophilia-positive group likely reflect their sicker baseline status. The patients lost to follow-up were older and had better lung function compared with those with full follow-up. Last, the median follow-up time in this study was relatively short in the two cohorts. We cannot exclude the possibility that the initial detection of S. maltophilia will lead to increased long-term mortality differences. Further study of this subject is warranted. In conclusion, we have shown, using several different techniques, that the apparent increase in mortality associated with S. maltophilia disappears after adjusting for differences in the severity of illness of patients before S. maltophilia detection. Thus, S. maltophilia is not associated with decreased survival after adjusting for severity of illness. This is not to say that S. maltophilia is not a pathogen. Clearly, one encounters clinical scenarios in which S. maltophilia appears to be the primary pathogen and should be treated. Further studies are needed to address what effect treatment has on outcome and which patients should be treated for S. maltophilia.
The authors are indebted to Preston Campbell III, M.D., Vice President for Medical Affairs, Cystic Fibrosis Foundation, and the Cystic Fibrosis Foundation for their support of this project and for making the Cystic Fibrosis Registry data available to us. C. H. G. is a Leroy Matthew Fellow of the Cystic Fibrosis Foundation. Supported by the Cystic Fibrosis Foundation and the Firlands Foundation.
This article has an online data supplement, which is accessible from this issue's table of contents online at www.atsjournals.org Received in original form September 21, 2001; accepted in final form April 8, 2002
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
INTRODUCTION
2 tests were used to compare nominal data. Survival was assessed by Kaplan–Meier product limit estimator and the log rank statistic. The Cox proportional hazard model was used to look at differences in survival after controlling for confounders and to assess potential interactions (
value of 0.05 and power of 0.80. This correlates with an absolute change in mortality of 0.5% per year. We used SAS 6.12 for Windows (SAS, Cary, NC) and Stata 6.0 software (Stata, College Station, TX). This study was approved by the institutional review board at the University of Washington. 
cohorts used for Cox proportional survival model without time varying covariates. The original 10:1 matching was not fully achieved in all patients (102 S. maltophilia-positive patients were matched at 10:1 and 1,571 were matched at 9:1).
70% and < 90% and FEV1 % pred 
