Published ahead of print on June 29, 2006, doi:10.1164/rccm.200602-163OC
© 2006 American Thoracic Society doi: 10.1164/rccm.200602-163OC
Incidence and Prevalence of Idiopathic Pulmonary FibrosisUniversity of Washington Medical Center, Seattle, Washington; Policy Analysis, Inc., Brookline, Massachusetts; and InterMune Pharmaceuticals, Brisbane, California Correspondence and requests for reprints should be addressed to Ganesh Raghu, M.D., Division of Pulmonary and Critical Care Medicine, Interstitial Lung Disease, Sarcoid, and Pulmonary Fibrosis and Lung Transplant Programs, University of Washington, Seattle, WA 98195-6522. E-mail: graghu{at}u.washington.edu
Rationale: Idiopathic pulmonary fibrosis is a chronic interstitial lung disease of unknown etiology; its epidemiology in the United States has not been well characterized. Objective: To estimate the annual incidence and prevalence of idiopathic pulmonary fibrosis in the United States. Methods: Retrospective cohort design utilizing a large health care claims database spanning the period January 1996 through December 2000. Measurements and Main Results: Persons with idiopathic pulmonary fibrosis were identified based on diagnosis and procedure codes. Using broad case-finding criteria, prevalence was estimated to range from 4.0 per 100,000 persons aged 18 to 34 yr to 227.2 per 100,000 among those 75 yr or older; annual incidence was estimated to range from 1.2 to 76.4 per 100,000. Using narrow case-finding criteria, prevalence ranged from 0.8 to 64.7 per 100,000 persons; comparable figures for incidence were 0.4 to 27.1 per 100,000 persons. Extrapolating these rates to the overall United States' population, prevalence was estimated to be 42.7 per 100,000 (incidence, 16.3 per 100,000) using broad criteria; with narrow criteria, prevalence was estimated to be 14.0 per 100,000 (incidence, 6.8 per 100,000). Conclusions: Our results suggest that idiopathic pulmonary fibrosis is probably more common in the United States than previously reported.
Key Words: epidemiology • lung diseases, interstitial • retrospective studies Idiopathic pulmonary fibrosis (IPF) is a progressive life-threatening disease that is characterized anatomically by scarring of the lungs and symptomatically by exertional dyspnea. Its etiology is unknown, and the pathogenic mechanisms involved in its initiation and progression are poorly understood (1). Of the over 150 recognized types of interstitial lung disease (ILD), IPF is the most common and one of the most deleterious (2). Criteria for the diagnosis of IPF were set forth in an international consensus statement formulated by members of the American Thoracic Society (ATS) and the European Respiratory Society (ERS) (3). In the absence of biopsy evidence of usual interstitial pneumonia, a constellation of typical clinical findings may be used to support the diagnosis of IPF (3). However, the extent to which these diagnostic guidelines are followed in actual clinical practice is unknown. Median survival among persons with IPF is believed to be from 3 to 5 yr (1, 4–8). Respiratory failure is the most frequent cause of death, and has been reported to account for over 80% of all fatalities; heart failure, bronchogenic carcinoma, ischemic heart disease, infection, and pulmonary embolism are also common causes of mortality (9–11). Antiinflammatory, antifibrotic, and immunosuppressive therapies are often used in the treatment of IPF; however, such treatment has not been demonstrated to improve survival or quality of life (10, 12). Surprisingly, little is known about the epidemiology of IPF in the United States. Initial estimates of its prevalence, based largely on case series from pulmonary clinics and tertiary-care hospitals, ranged from 3 to 6 cases per 100,000 persons (13). Substantially higher rates of prevalence (20 per 100,000 among men, 13 per 100,000 among women) and incidence (11 and 7 per 100,000, respectively) were reported in a study based on the adult population of Bernalillo County, New Mexico (14). It is unclear, however, whether these rates are generalizable to the present-day United States as a whole, because they are based on data more than 15 yr old and come from an area of the United States known to attract persons with chronic lung diseases (15). Although more recent estimates of disease prevalence (range, 1–24 cases per 100,000) are available from several European studies, their generalizability to the United States is not clear (16–20). This study used data from a large, geographically diverse, United States' health care claims database to estimate the prevalence and annual incidence of IPF. Some of the results from this study have been previously reported in the form of an abstract (21).
Data Source Data for this study were obtained from the health care claims processing system of a large United States' health plan, and spanned the period January 1, 1996, through December 31, 2000. The database consists of claims for services provided to plan members that were submitted by facilities (e.g., hospitals), health care professionals (e.g., physicians), and retail pharmacies. The plan provides health care services—through health maintenance organizations, preferred provider organizations, Medicare Risk, and indemnity products—to approximately 3 million persons in total residing in 20 states, mostly in the South Atlantic (28%), South Central (37%), and North Central (31%) regions of the United States. Nearly 20% of plan members are aged 65 yr or older. Data available for each facility and professional-service claim included dates and place of service, diagnoses (in International Classification of Diseases, 9th revision, Clinical Modification [ICD-9-CM] format), procedures (in Healthcare Common Procedure Coding System [HCPCS] format), provider specialty, and reimbursed amounts; professional-service claims also included information on the number of units or services provided. Data available for each outpatient pharmacy claim included the drug dispensed, dispensing date, quantity dispensed, therapy-days supplied, and reimbursed amounts. Information on demographics and eligibility was also available for all plan members. All data could be arrayed in chronologic order to provide a detailed longitudinal profile of all medical (inpatient and outpatient) and pharmacy services used by each plan member. All patient identifiers in the database had been fully encrypted, and the database is fully compliant with the Health Insurance Portability and Accountability Act of 1996 (22). This study is exempt from the U.S. Department of Health and Human Services' Federal Policy for the Protection of Human Subjects since it qualifies as "research involving the collection or study of existing data, documents, records, and the information is recorded in such a manner that subjects cannot be identified, directly or through identifiers linked to the subjects" (23). Institutional review board approval for this study therefore was neither sought nor obtained.
Case Definition
Because a diagnosis of IPF is more likely to be accurate if based on appropriate testing, we also used an alternative ("narrow") case definition in which we required that persons with IPF (1) satisfy the broad case definition set forth above and (2) who, on or before the date of their last medical encounter with a diagnosis of IPF, have one or more medical encounters with a procedure code for surgical lung biopsy (ICD-9-CM 33.28, 34.21; Current Procedural Terminology, Fourth Edition [CPT-4] 32095, 32100–32160, 32602), transbronchial lung biopsy (ICD-9-CM 33.27; CPT-4 31628, 31629), or computed tomography of the thorax (ICD-9-CM 87.41; CPT-4 71250, 71260, 71270). Persons whose first medical encounter with a diagnosis code for IPF was in CY2000 and who were continuously eligible for health benefits for at least 365 d before the date of that encounter, including those with and without surgical lung biopsy, transbronchial lung biopsy, and computed tomography of the thorax, also were identified. All such persons were designated as having "newly diagnosed" (i.e., incident) disease.
Measures and Analyses
Prevalence by Age and Sex Among the 2.2 million men and women aged 18 yr and older in the study database in CY2000, 1,943 had one or more diagnoses of IPF at any time during the 5-yr period of observation. Among these 1,943 patients, 62% (n = 1,211) did not have a diagnosis of any other type of ILD on or after the date of their last diagnosis of IPF and thus met our broad case definition (Table 2). Among these persons, 387 were found to have undergone a diagnostic procedure on or before the date of their last claim with a diagnosis of IPF (surgical lung biopsy, 10%; transbronchial lung biopsy and computed tomography of the thorax, 14%; computed tomography of the thorax only, 71%; and transbronchial lung biopsy only, 5%), and thus also met our narrow case definition.
Based on the broad case definition, the prevalence of IPF was estimated to range from 4.0 per 100,000 persons aged 18 to 34 yr to 227.2 per 100,000 among those aged 75 yr or older; based on the narrow definition, estimated prevalence ranged from 0.8 (age, 18–34 yr) to 64.7 (age  75 yr) per 100,000. Prevalence was generally higher among men than women (Figure 1).
Incidence by Age and Sex Among the 1,211 persons meeting the broad definition of IPF, 295 had newly diagnosed disease in CY2000. Among the 387 persons meeting the narrow case definition, the corresponding figure was 120. Based on the broad case definition, the annual incidence of IPF was estimated to range from 1.2 per 100,000 persons aged 18 to 34 yr to 76.4 per 100,000 among those aged 75 yr or older; based on the narrow definition, estimated incidence ranged from 0.4 (age, 18–34 yr) to 27.1 (age 75 yr) per 100,000. Like prevalence, incidence was generally higher among men than women (Figure 2).
Prevalence and Incidence in the United States Overall rates of prevalence and incidence of IPF in the United States were estimated to be 42.7 and 16.3 per 100,000 persons, respectively, based on the broad definition (Figure 3). Based on the narrow definition, overall prevalence and incidence in the United States were estimated to be 14.0 and 6.8 per 100,000 persons, respectively.
Using a large, geographically diverse United States' health care claims database, we examined the prevalence and annual incidence of IPF. Our findings suggest that IPF is probably more common in the United States than previously reported. In the New Mexico study, for example, Coultas and coworkers reported that the prevalence of IPF was approximately 17 cases per 100,000 men and women aged 18 yr and older; incidence among these persons was estimated to be 9 cases per 100,000 (14). Although these figures are similar to those from our study based on narrow case-finding criteria (14 and 7 cases per 100,000, respectively), they are substantially below those based on broader case-finding criteria (43 and 16 cases, respectively, per 100,000). Extrapolating our results based on broad case-finding criteria to the United States' adult population, we estimate there are approximately 89,000 persons aged 18 yr and older with (diagnosed) IPF in the United States, and that 34,000 persons in this age group are newly diagnosed with the disease each year. Using narrow case-finding criteria, corresponding estimates are 29,000 and 14,000, respectively. The similarity of our findings (based on narrow case-finding criteria) and those of Coultas and colleagues is not entirely surprising, given that both studies relied on physician diagnosis but also required additional evidence of IPF. In our study, an IPF diagnosis must have been preceded by appropriate testing, whereas in Coultas and colleagues' study, the medical records of patients referred by their clinicians to study investigators were reviewed by trained abstractors. Because our narrow case-finding criteria and the case ascertainment methods used by Coultas and colleagues probably sacrifice sensitivity to improve specificity, it is likely that results based on these approaches underestimate the true extent of the disease. For this reason, we also report estimates based on more liberal criteria. We note that findings based on our narrow and broad case-finding criteria are consistent with existing estimates of survival among persons with IPF (median, 3–5 yr) (4–8). (For incurable diseases with relatively high mortality rates, prevalence is approximately equal to incidence x average survival.) Interestingly, although the prevalence of IPF was higher among older men than older women, a finding consistent with current knowledge of the disease, prevalence was roughly comparable among younger men and women (24). Although the validity of the latter finding is less certain because of the (relatively) small sample sizes, it may be the case that the prevalence of IPF is indeed similar among the sexes at younger ages. Another possibility is that more women than men with other types of pulmonary fibrosis that may mimic IPF—and that some data suggest occur at a higher rate among younger women versus men—incorrectly receive diagnoses of IPF (14). Unfortunately, we could not address this issue in our study, because histology and high-resolution computed tomography data were unavailable. Ideally, an examination of the epidemiology of IPF in the United States should be based on data from a large, nationally representative sample. However, because IPF is a relatively uncommon disorder, existing surveys (e.g., the National Health and Nutrition Examination Survey [NHANES]) probably include only a few, if any, cases. A prospective study would require identification and recruitment of a large, geographically diverse sample of patients with IPF or pulmonary disease characteristic of IPF, thorough examination by physicians experienced in the diagnosis of IPF, and determination that clinical, radiologic, physiologic, and pathologic findings are consistent with the ATS/ERS diagnostic criteria. The prohibitive cost of such a study renders estimation of incidence and prevalence from large health care claims databases an attractive alternative; in fact, such an approach has been used to examine the epidemiology of many other diseases (25–31). Although we believe our results are provocative, the inherent limitations of our study must be acknowledged. For one, the study database includes only health-plan members in selected geographic regions of the United States, and some anecdotal evidence suggests that disease prevalence may vary by region (14, 15, 32). In addition, our data were gathered, for the most part, during the period before the publication of the international consensus statement on IPF and the more recent statement on the classification of the idiopathic interstitial pneumonias (which distinguishes IPF from six other forms of idiopathic interstitial pneumonias) (3, 24). It is possible that with evolving knowledge and gained clinical experience, the estimates of prevalence and incidence of IPF may change over time. Finally, our study used data from a convenience sample of private health and Medicare Risk plans. Persons with such insurance may differ systematically from the rest of the population in terms of their health status and/or levels of medical-care utilization and expenditures. These differences may be particularly pronounced among the elderly, because enrollees in Medicare Risk plans may be healthier than those in the traditional Medicare fee-for-service program (who were not represented in our database) (33). Certain aspects of our operational definitions of IPF may have imparted an upward bias to our estimates of prevalence and incidence. First, the ICD-9-CM diagnosis code for IPF is not specific to that disease; alveolar capillary block and Hamman-Rich syndrome are also included under that code. However, because the former diagnosis is seldom, if ever, used in clinical practice and the latter is quite rare, we believe that only a very few cases of these two entities may have been included in our study population. Second, some persons identified as having IPF using our broad case definition undoubtedly would not have been found to have the disease had consensus statement guidelines for establishing an IPF diagnosis been applied. Third, among persons identified using our narrow case definition (i.e., diagnosis of IPF plus record of diagnostic testing), we were unable to ascertain the results of diagnostic testing and therefore assumed that if a diagnosis of IPF persisted after testing, it was consistent with test results. Fourth, surgical lung biopsy, the traditional gold standard for the diagnosis of ILD, was performed in only a small percentage of cases (10%) (34–36). Finally, some patients who were designated as incident cases may have had symptoms suggestive of IPF in prior years but did not receive a qualifying diagnosis until the reference year (either because they did not present to the health care system during the predefined window or because their condition was initially misdiagnosed). On the other hand, it also should be noted that some true incident cases in the reference year might have been missed because a corresponding diagnosis was not recorded (for the same reasons as stated above). Thus, the magnitude of any resulting bias may be small. Despite the deficiencies in the diagnostic criteria that we used to identify persons with IPF, we believe that our estimates are conservative for several reasons. First, it is likely that many persons with IPF did not have evidence of diagnostic testing in their records (and thus were not considered to have IPF based on the narrow case definition) because such testing was done before enrollment in their current health insurance plan. Second, it is likely that some persons who had IPF, but were newly enrolled health plan members in CY2000, may not have been observed long enough in the study database to have had an encounter and/or diagnostic test for IPF. Third, to the extent that Medicare fee-for-service enrollees are more likely to have IPF than those in a Medicare Risk plan, disease prevalence and incidence may have been underestimated. Fourth, our estimates of disease epidemiology are based on data from CY2000 and disease awareness and education have increased in recent years. Finally, and most important, we believe that our estimates are conservative due to the large number (n = 12,891) of persons aged 18 yr or older in the database who had a medical claim with a diagnosis code for postinflammatory pulmonary fibrosis (ICD-9-CM 515)—which is typically used for unspecified pulmonary fibrosis—and no subsequent diagnoses of any other forms of ILD. Other studies have also found that postinflammatory pulmonary fibrosis is by far the most commonly encountered diagnosis among persons with ILD (32). The large majority (75%) of these 12,891 persons had no evidence of having undergone diagnostic testing—as was the case with IPF (68%)—and when persons with nonspecific ILD are thoroughly evaluated at referral centers, a substantial proportion (> 50%) are found to have IPF (14, 34, 37–39). Unfortunately, there was no way of determining retrospectively the proportion of these 12,891 persons who actually had IPF, and therefore we limited our focus in the study to the subset of persons with the specific diagnosis code for IPF (ICD-9-CM 516.3). However, if we assume conservatively that as few as 25% of patients with a diagnosis of postinflammatory pulmonary fibrosis actually had IPF, our estimate of disease prevalence in the United States (based on broad case-finding criteria) would increase almost fourfold (from 43 to 156 cases per 100,000). Estimates of the annual incidence and prevalence of postinflammatory pulmonary fibrosis are available in an online supplement. Given the limited sensitivity of our narrow case-finding criteria for IPF, the large gap between estimates based on narrow and broad case-finding criteria, and the conservative biases outlined in the preceding paragraph, the results of our study suggest that the prevalence and incidence of IPF in the United States are probably higher than previously reported. Further research is needed to confirm and update these findings. Additional research on the economic burden of this disease is also needed so that the cost-effectiveness of newer, more effective medical interventions may be properly evaluated.
The authors thank Loren Lidsky, M.S., and May Hagiwara, Ph.D., of Policy Analysis, Inc., for their assistance with data management and data processing.
Supported by InterMune, Inc., Brisbane, California. Originally Published in Press as DOI: 10.1164/rccm.200602-163OC on June 29, 2006 Conflict of Interest Statement: G.R. served as an advisor (2003), including for patient-related studies, and as a consultant (2004) for InterMune, Inc.; he also served as a consultant during 2003–2005 for Actelion. D.W. is employed by Policy Analysis, Inc., which received funding for this research from InterMune, Inc., in 2005. J.E. is employed by Policy Analysis, Inc., which received funding for this research from InterMune, Inc., in 2005. W.Z.B. is an employee of InterMune, Inc., who owns stock. G.O. is employed by Policy Analysis, Inc., which received funding for this research from InterMune, Inc., in 2005. Received in original form February 3, 2006; accepted in final form June 13, 2006
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
METHODS
1320-d2. 1996.