| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Guest Access | Sign In via User Name/Password
|
|
|
|||||||
Guest Access | Sign In via User Name/Password |
|||||||||
* From the Division of Pulmonary, Allergy, and Critical Care (Drs. Danai, Moss, and Martin), Emory University School of Medicine, Atlanta, GA; and Division of Pulmonary and Critical Care (Dr. Mannino), University of Kentucky, Lexington, KY.
Correspondence to: Greg S. Martin, MD, MSc, FCCP, Assistant Professor of Medicine, 49 Jesse Hill Jr Dr SE, Atlanta, GA 30303; e-mail: Greg.Martin{at}emory.org
Abstract
Study objectives: To evaluate the longitudinal epidemiology of sepsis in patients with a history of cancer and to specifically examine sepsis-related disparities in risk or outcome.
Design: Sepsis cases from 1979 through 2001 using a nationally representative sample of nonfederal acute-care hospitalizations in the United States (the National Hospital Discharge Survey) integrated with cancer prevalence from the Surveillance, Epidemiology, and End Results database.
Setting: Eight hundred fifty-four million acute-care hospitalizations and 8.9 million patients with cancer.
Patients or participants: Patients with a history of cancer hospitalized with a diagnosis of sepsis.
Measurements and results: From 1979 to 2001, there were a total of 1,781,445 cases of sepsis in patients with cancer, yielding a mean annual incidence rate of 1,465 cases per 100,000 cancer patients and a relative risk [RR] of 9.77 compared to noncancer patients (95% confidence interval [95% CI], 9.67 to 9.88). In contrast to the absolute number of cases, the incidence rate of sepsis decreased over time, from a peak of 1,959 cases per 100,000 cancer patients in 1987 to 995 cases per 100,000 in 2001. The distribution of infectious sources causing sepsis was associated with the type of malignancy. White cancer patients had a lower risk for sepsis compared to nonwhites (African-American RR, 1.28; 95% CI, 1.16 to 1.40) and other races (RR, 1.47; 95% CI, 1.22 to 1.72); and male cancer patients had a higher risk for sepsis compared to female cancer patients (male RR, 1.17; 95% CI, 1.10 to 1.23). Cancer was an independent predictor of death among sepsis patients by multivariable analysis (adjusted odds ratio for death, 1.98; 95% CI, 1.97 to 1.99).
Conclusions: Patients with a history of cancer are at increased risk for acquiring and subsequently dying from sepsis, compared to the general population, although incidence and fatality rates are decreasing over time. There are significant racial and gender disparities in the incidence and outcome of sepsis among cancer patients that require explanation.
Key Words: cancer • disparity • epidemiology • malignancy • outcomes • sepsis
Sepsis is the host immune response to infection, defined clinically as the intersection between physiologic derangements known as the systemic inflammatory response syndrome and infection.1 Sepsis is a relatively common reason for acute-care hospitalization, occurring in > 700,000 people each year in the United States and occurring in 2% of all hospitalizations.23 Sepsis is known to be increasing in frequency, and certain patient populations are at higher risk than others.2 The risk for sepsis among male patients is approximate 30% higher compared to female patients, while African Americans and other races have nearly twice the risk for sepsis compared to whites. Certain subsets of sepsis, such as sepsis with acute organ dysfunction (severe sepsis) and sepsis with refractory hypotension (septic shock), have case-fatality rates > 40%.24 At present, sepsis is the tenth-leading cause of death overall in this country.5
Cancer is one of the most common diseases in the US population. It occurs with an approximate 3% prevalence in this country, yielding 9.8 million people living with the diagnosis of cancer in the year 2001.6 Cancer incidence rates and death rates are highest in African Americans, although women generally have lower rates than men irrespective of race.7 Prostate cancer incidence rates are 1.5 times higher in African Americans than whites, while breast cancer is 1.2 times more common in whites than African Americans. Invasive cancer was diagnosed in an estimated 1.4 million Americans in 2004, and this number is expected to double in the next 50 years.7 It is estimated that cancer costs society > $60 billion annually.8 Cancer is the second-leading cause of death in this country behind heart disease and accounts for one in four deaths overall.9
Chronic comorbid medical conditions are common among sepsis patients. Epidemiologic studies2410 show that chronic comorbidities are present in 54 to 65% of all sepsis patients. Cancer is the most common comorbid medical condition in patients with sepsis, reported to occur in 16.8% of US sepsis patients and in 16.7% of European and Canadian sepsis patients.211 Cancer has been found in 11.6% of patients with severe sepsis.4 Thus, the prevalence of cancer among sepsis patients is disproportionate compared to the prevalence in the US population.
Epidemiologic estimates of sepsis among cancer patients are limited. A study12 of severe sepsis reported a rate of 1,640 cases per 100,000 cancer patients in 1999. Previous studies have not provided a longitudinal examination of sepsis in cancer patients, nor have they examined differences in risk based on race or gender that may be expected based on known disparities in both sepsis and cancer. We sought to define these relationships while additionally examining clinically relevant information regarding the source and type of infection according to tumor types and providing comparisons of sepsis incidence in other common chronic comorbid medical conditions.
Materials and Methods
Data Source
The National Center for Health Statistics has conducted the National Hospital Discharge Survey (NHDS) continuously since 1965.13 Since 1979, the NHDS has conformed to the guidelines of the Uniform Hospital Discharge Data Set for consistency of reporting in records. The NHDS is composed of a sample of all nonfederal acute-care hospitals in the United States, including approximately 500 hospitals, with equal representation of all geographic regions. The database is constructed through the surveying of discharge records for inpatients from each participating hospital, representing approximately 1% of all hospitalizations, or 350,000 discharges annually in the United States. Discharge records are abstracted for demographic information (age, sex, ethnic background, geographic location, and marital status), seven diagnostic codes and four procedural codes (International Classification of Diseases, Ninth Revision-Clinical Modification [ICD-9-CM]), dates of hospital admission and discharge, sources of payment, and disposition at discharge.2
The National Cancer Institute maintains the Surveillance, Epidemiology, and End Results (SEER) database, which can be accessed through the SEER Web site.14 The SEER Program collects and publishes cancer incidence, prevalence, and survival data from 14 population-based cancer registries and 3 supplemental registries covering approximately 26% of the US population. The SEER Registries collect data on patient demographics, primary tumor site, morphology, stage at diagnosis, first course of treatment, and follow-up for vital status. The prevalence of cancer for the United States is estimated from SEER for individual years through 2001, as determined by population incidence data and a survival model that has been previously validated.15 For our purposes, this prevalence method was applied by a software package (SEER*Stat; National Cancer Institute, Statistical Research and Application Branch; Bethesda, MD).16 Annual cancer prevalence data were subclassified by race, gender, and/or according to 13 of the most common cancer subtypes: GI, lung, breast, female reproductive, prostate, urinary tract, multiple myeloma, CNS, lymphoma, leukemia, pancreas, head and neck, and skin.
For comparison purposes, prevalence data for the most common chronic comorbid medical conditions were obtained from published sources: diabetes mellitus,17 COPD,18 HIV/AIDS,19 coronary artery disease,20 and hypertension20 were obtained from several peer-reviewed sources. All data were publicly available in the biomedical literature or through the American Heart Association, the Centers for Disease Control and Prevention, the HIV InSite Knowledge database, or the National Center for Health Statistics. Statistical testing was not performed on these data for lack of access to the primary data sources. This project was considered exempt from the requirement for informed consent according to Federal Regulations of Human Subjects Protection 45 CFR 
46.101(b).
Definitions
Cases of sepsis in patients with malignancy were identified from discharge records in the NHDS during the years 1979 through 2001 that included ICD-9-CM codes for both sepsis and for cancer. Sepsis was defined as previously validated221: 038.x (septicemia), 20.0 (septicemic), 790.7 (bacteremia), 117.9 (disseminated fungal infection), 112.5 (disseminated candida infection), and 112.81 (disseminated fungal endocarditis). Cancer was defined by the range of ICD-9-CM codes 140.x–239.x. Other comorbid conditions were identified by standard ICD-9-CM codes as modified from Charlon et al and Deyo et al222324: diabetes mellitus (250.x), COPD (490.x–496.x), HIV/AIDS (042.x), coronary artery disease (410.x–414.x, 429.2), and hypertension (401.x). Cancer prevalence was determined from SEER (as above) from 1979 to 2001, and reporting categories were matched to tumor-specific sets of ICD-9-CM codes to carry out tumor type-specific analyses.
Statistical Analysis
Incidences were normalized to the population distribution of the year-specific cancer population prevalence, as calculated from the SEER Registries and the SEER*Stat prevalence software. All estimates are presented according to accepted guidelines for the accuracy of NHDS data, restricting use to absolute, unweighted samples of > 60 patients with relative SE (RSE) measures of < 30% for data analysis. The RSE was calculated as a first-order Taylor-series approximation, as outlined in the RSE tables of the 2000 NHDS documentation. The SE was calculated by multiplying the RSE by the estimated incidence or mortality rate, and 95% confidence intervals (CIs) were calculated from these SEs (Excel; Microsoft Corporation; Redmond, WA). Data for continuous variables were compared by analysis of variance, and data for categorical variables were compared by the 
2 test or Fisher exact test, as appropriate for the size of the sample (SAS 9.1 for Windows; SAS Institute; Cary, NC). Differences were considered significant when the 95% CIs did not overlap and/or when two-sided p values were < 0.05. When stated race was missing for a given observation (ranging from 1 to 20% for any given year), these persons were excluded from the calculations of race specific rates but were included in all other calculations of rates.
Predictors of death with sepsis were assessed by multivariable logistic regression modeling including potentially relevant predictors (age, gender, race, source of infection [categorized as respiratory, genitourinary, GI, skin, soft tissue and bone, and other]; chronic comorbid medical conditions [categorized as cancer, chronic renal failure, COPD, congestive heart failure, coronary artery disease, diabetes mellitus, hepatic cirrhosis, HIV/AIDS, hypertension]; and severity of illness [by cumulative number of organ dysfunctions]) using a backward elimination technique described by Kleinbaum and Klein.25 Effect modification was tested in stratified analyses according to clinical relevance and using the Breslow-Day test for homogeneity, which showed that interaction terms were not necessary in the regression model. Variable collinearity was assessed by model stability and by a multiple covariance matrix. Goodness-of-fit was assessed in deciles of observed and expected values as per Hosmer et al.26 Regression models reported adjusted odds ratios with Wald confidence intervals.
Results
Using the NHDS, from the years 1979 to 2001, there were 854,000,000 hospitalizations, of which 76,650,000 involved a coexisting diagnosis of malignancy and 10,989,533 involved sepsis. From these populations, there were 1,784,445 cases of sepsis in patients with cancer; sepsis occurred in 2.3% of all hospitalizations among cancer patients. Based on the SEER database, the prevalence of cancer increased during the study period, by an average of 4.7% during the last 10 years, to a complete prevalence of 8,907,000 people living with cancer in the United States in the year 2001.
Demographics and Causes of Sepsis
The characteristics of the cancer and sepsis population are presented in Table 1
. The average age at hospitalization with sepsis was 62.8 years for patients with cancer and sepsis, while in the noncancer sepsis population the average age was 58.1 years (p < 0.001). The mean age of noncancer sepsis patients increased during the study period by 9.1 years (from 53.0 to 62.1 years), compared to 1.3 years (from 62.2 to 63.5 years) for sepsis patients with cancer (p < 0.001). Of the patients with a history of cancer who acquired sepsis, GI malignancies were most common at 24.4%, followed by lung (20.0%), lymphoma (14.1%), prostate (9.3%), and breast cancers (8.8%). Over the study period, there were no changes in the representation of malignancy types in patients with sepsis.
|
|
Incidence
During the 23-year study period, the number of sepsis cases in patients with cancer increased from 24,150 in 1979 to 87,160 in 2001, representing an increase of 261% during the study period. There were a total of 1,781,445 cases of sepsis in patients with cancer during the study period, yielding a mean annual incidence rate of 1,465 cases per 100,000 cancer patients. This, when compared to a mean annual incidence rate of 150 cases of sepsis per 100,000 of the noncancer US population, makes cancer patients at 9.8 times increased risk for sepsis (relative risk [RR] 9.77; 95% CI, 9.67 to 9.88). In contrast to the absolute number of cases, the incidence rate of sepsis decreased over time (Fig 1
), with 1,280 cases per 100,000 cancer patients in 1979, rising to a peak of 1,959 cases per 100,000 in 1987 and then falling to 979 cases per 100,000 in 2001. This represented a 24% cumulative decrease in incidence and an average annual decline of 1.0% during the study period despite temporal increases in the incidences of both sepsis and cancer in the US population.
|
Male patients accounted for a disproportionate share of sepsis cases among patients with cancer, with a mean incidence rate of 1523 per 100,000 male patients with cancer, compared to 1,320 per 100,000 female patients with cancer, yielding a RR of sepsis of 1.17 (95% CI, 1.10 to 1.23). However, the annual incidence rate in male patients with cancer decreased at a faster rate than for female patients during the study period (mean, – 1.4%/yr vs – 0.7%/yr, p = 0.01), reducing the magnitude of the disparity. Data for specific cancer subtypes showed similar disparities and are displayed in Table 3 . The mean annual incidence rate for sepsis increased most among women with lung cancer at 12.4%/yr, contributing to a decline in the gender disparity to a mean RR of 1.17 (95% CI, 0.65 to 1.68), while sepsis rates with breast cancer and prostate cancer both declined over time (by 2.9%/yr and 3.8%/yr, respectively).
|
Comparison With Other Comorbidities
For comparison, the incidences of sepsis for other common medical comorbidities were also calculated. The incidences of sepsis in patients with HIV/AIDS, COPD, coronary artery disease, diabetes mellitus, and hypertension were determined for the year 2000 and are shown in Table 4
.
|
The case-fatality rate for sepsis was highest for patients with lung cancer at 39.7%, significantly higher than for patients with cancers of the breast, prostate, and skin, which each averaged approximately 26% (each p < 0.05 compared to lung cancer). Intermediate case-fatality rates were observed for cancers of the head and neck, GI system, female reproductive system, and multiple myeloma at 38.9%, 34.0%, 33.4%, and 31.7%, respectively. The sepsis-related fatality rates for most tumor types decreased significantly over the study time period, with GI cancer (58 to 22%) and lung cancer (67 to 31%) having the greatest reductions (both p < 0.001).
The mean annual case-fatality rate for men with sepsis and cancer was 33.2% and for women was 31.0% (p = 0.45), declining throughout the study period (male rate decreasing from 51.2 to 27.5%, and female rate decreasing from 34.6 to 18.7%). Population-based sepsis mortality rates were consistently higher in male patients with cancer due to the greater incidence (mean, 507 male deaths vs 409 female deaths per 100,000; p = 0.04), making male patients with cancer more likely to die with sepsis than female patients with cancer (RR, 1.29; 95% CI, 1.14 to 1.44). The mean fatality rate for whites over the study period was 31.5%, while the rate for African Americans was 36.8% (p = 0.09) and other races was 35.4% (p = not significant). Population-based mortality rates were lowest in whites, at 398 deaths/100,000, compared to African Americans (580 deaths/100,000; p = 0.006) or other races (617 deaths/100,000; p = 0.02). Using a multivariable logistic regression model adjusting for race, gender, severity of illness, source of infection, and other chronic comorbid medical conditions, cancer was independently associated with a two times greater odds of death among patients with sepsis (Table 5 ).
|
Sepsis was present in 28.8% of all deaths among hospitalized cancer patients. Patients with a history of cancer who survived hospitalization with sepsis were more likely to be discharged to another medical facility than for cancer patients without sepsis, but their rate was still lower than for the noncancer sepsis population: 18.2% of sepsis survivors with cancer were discharged to another medical facility vs 8.8% for nonsepsis cancer patients and 29.8% for noncancer sepsis patients (p < 0.001 for comparison between any groups). Similarly, 61.8% of noncancer sepsis survivors were discharged directly to home, compared to 72.7% of sepsis survivors with cancer and 86.5% of nonsepsis cancer survivors (p < 0.001 for comparison between any groups).
Discussion
In our study of the epidemiology of sepsis in cancer patients, we found nearly 100,000 cases of sepsis each year in patients with a history of malignancy and found cancer patients to be nearly 10 times more likely to acquire sepsis than the noncancer US population. Sepsis occurred in 30% of all hospitalized cancer deaths, and cancer was an independent predictor of death in patients with sepsis. From this large group of patients, we also observed significant racial disparities, with nonwhite cancer patients at increased risk for acquiring and dying from sepsis. Fortunately, both the incidence rate and mortality rate of sepsis in patients with cancer have steadily decreased in recent years. This finding may reflect better or safer chemotherapy and may have been expected given the overall improvement in survival of patients with sepsis.2
This study is the first to present nationally representative longitudinal data for the incidence of sepsis among cancer patients and to examine gender and racial disparities. The finding of racial and gender disparities in the incidence of sepsis is not surprising given the disparities seen in other major diseases.2728 Nonwhite cancer patients have consistently higher rates of sepsis over the last quarter century, at a magnitude of disparity seen in other common medical conditions, including sepsis.2 Also consistent with our prior study2 of the incidence of sepsis is the higher incidence in men relative to women. The etiology for these disparities remains controversial, but factors such as differing access to care and differing likelihood of receiving aggressive care have been suggested as possible etiologies for the disparity.28
Our finding of cancer as a strong independent predictor of mortality from sepsis may be due to one of several biologically plausible reasons for increased susceptibility to sepsis.29 Patients may be immunocompromised due to the use of chemotherapy, radiation, or other immune modulating therapy employed to combat the underlying malignancy. They may have impaired leukocyte function secondary to the malignancy itself.29 Further, this group of patients is especially prone to having additional chronic comorbid medical conditions, which are difficult to quantify (eg, chronic alcohol abuse) or have been identified by others243031 and suggested to predispose to poorer sepsis outcomes. Understanding which of these factors significantly contribute to increased sepsis mortality from cancer could aid in developing new therapies to decrease sepsis-related deaths.
The findings of this study extend the observations of prior studies of sepsis in cancer patients, which have generally shown a markedly increased rate of sepsis for this population, as well as increased risk of death for cancer patients who acquire sepsis. The study by Williams et al12 reported an incidence rate of 1,640 cases per 100,000 patients and noted a 52% higher risk of mortality among cancer vs noncancer patients who acquire sepsis. In a similar study, Angus and colleagues4 reported a case-fatality rate of 43% for severe sepsis patients with metastatic neoplasms, reflecting the highest mortality of all comorbidities for sepsis.
The results of this study may impact clinical practice by increasing the oncologist’s or health-care provider’s clinical suspicion for the development of sepsis in cancer subgroups that previously had been considered low risk for sepsis. Patients with pancreatic cancer were found to have the highest incidence of sepsis, even greater than for leukemia. This study also found that the source of infection is associated with the type of malignancy, which may directly influence empiric antimicrobial therapy in these patients. Finally, cancer patients with sepsis have a much lower rate of utilization of long-term medical facilities. This may be an indication that certain services available to cancer patients, such as hospice, allow for more efficient use of health-care resources.32
This study utilized data from the NHDS and SEER databases, which may limit generalizability to the US population and relies on accurate coding of both sepsis and chronic comorbid medical conditions, although the large sample sizes and longitudinal design make administrative data important epidemiologic tools.2323334 There may be inaccuracies, such as presuming that all sepsis cases result in hospitalization, or the attribution of death to sepsis on the basis of crude mortality, rather than data on directly attributed mortality. Patient race was missing from a portion of observations and therefore could not be included in our analysis, causing a potential underestimation of race-specific sepsis rates. However, it would not influence comparative risk between races unless differential absences were present. The limited representation of nonwhite, nonblack patients restricts us from generalizing these findings to other specific racial or ethnic groups. Cancer prevalence was derived from incidence and survival data using the SEER database in conjunction with a mathematical model to determine prevalence,15 which depends on accrued patient data, thus reducing accuracy in early years of prevalence estimates. Finally, cancer patients may be more likely to have medical care limited than the general sepsis population, and therefore their risk of dying may be greater due to patient or physician directives, which may contribute to shorter hospitalizations.35
The diagnoses of malignancy and sepsis are becoming increasingly common, a trend that is likely to accelerate with the aging of the US population.36 Patients with cancer are at significantly greater risk for sepsis and are much more likely to die from sepsis compared to the general population. Within the population of cancer patients, men and nonwhite races are at especially increased risk. In addition, the sources of infection in cancer patients with sepsis are associated with the type of malignancy, making the diagnosis and treatment of these infections more predictable using these data. This information may also help health-care providers more accurately prognosticate for this group and help providers allocate resources more appropriately. Further studies are needed to evaluate the longitudinal changes in the incidence of sepsis among cancer patients and the mechanisms responsible for the disparities in susceptibility to sepsis.
Footnotes
Abbreviations: CI = confidence interval; ICD-9-CM = International Classification of Diseases, Ninth Revision-Clinical Modification; NHDS = National Hospital Discharge Survey; RR = relative risk; RSE = relative SE; SEER = Surveillance, Epidemiology, and End Results database
The authors have no financial relationship to disclose relative to this work.
Supported by grants HL K23–67739 (Dr. Martin) and AA R01–11660 (Dr. Moss) from the National Institutes of Health.
Received for publication July 6, 2005. Accepted for publication December 13, 2005.
References
This article has been cited by other articles:
|
|
 |
|
  A. P. Wheeler Recent Developments in the Diagnosis and Management of Severe Sepsis Chest, December 1, 2007; 132(6): 1967 - 1976. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Epidemiology of Sepsis in Cancer Patients Journal Watch Infectious Diseases, July 12, 2006; 2006(712): 9 - 9. [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
