Survival in Patients With Cryptogenic Fibrosing Alveolitis* A Population-Based Cohort Study Richard Hubbard, DM; Ian johnston, MD; and john Britton, MD Study objectives: To determine the median survival of patients with cryptogenic fibrosing alveolitis, in comparison to that expected of individuals of the same age and sex from the general population. To estimate the effect of survival bias incurred by studying both incident and prevalent cases. To identify factors associated with survival. Design: Cohort study. Setting: Nine hospitals from the Trent Region of England. Patients: Two hundred forty-four cases of cryptogenic fibrosing alveolitis: 168 patients were alive on day 1 of the study (prevalent cases), and 76 patients had newly diagnosed conditions over a prospective 18-month period of patient recruitment (incident cases). Measurements: Age, sex, date of diagnosis, lung function at presentation, and details of treatment prescribed were extracted from hospital clinical records. Data on lifetime smoking habits were collected by postal questionnaire. Vital status was established from the general practitioner. The expected survival of each case was extracted from the English life tables. Results: Median survival for incident cases was 2.9 years and for prevalent cases 9 years, compared to expected values of 10 and 13 years, respectively. Incident cases had significantly worse survival than prevalent cases, even after adjustment for the effects of age, sex, smoking history, lung function at presentation, and treatment (hazard ratio, 4.53 [95% confidence interval, 2. 71 to 7.56]; p<0.001). Lower FVC at presentation and the use of corticosteroid treatment were both associated with worse survival. Conclusions: The inclusion of prevalent cases leads to significant overestimation of the median survival for patients with cryptogenic fibrosing alveoli tis. In newly diagnosed cases of cryptogenic fibrosing alveolitis, median survival is only 2.9 years, and expected life span is reduced by approximately 7 years. (CHEST 1998; 113:396-400) Key words: cryptogenic fibrosing a lveolitis; epidemiology; idiopathic pulmonary fibrosis ; survival Abbreviations: CI=confidence inte1val; HRCT = high-resolution CT; Tc o =transfer factor for carbon monoxide. cryptogenic fibrosing alveolitis is the most common of the interstitial lung diseases, with a prevalence in the United States of approximately 20/100,000.1 The registered mortality from cryptogenic fibrosing alveolitis is increasing progressively in a number of countries. 2 The median survival of patients with cryptogenic flbrosing alveolitis has been estimated from clinical case series to be approximately 5 y ears. 3-5 However, because these series included prevalent cases who by definition are survivors, these estimates of median survival may have been influenced by survival bias. Furthermore, because the cases included in these series were mainly collected at tertiary referral centers, they may be unrepresentative of the typical spectrum of disease. To estimate more accurately the prognosis of patients with cryptogenic fibrosing alveolitis and to estimate years of life lost in what is an elderly patient group, we have examined the survival of a population-based sample of prevalent and incident cases, and compared this to the expected survival in the general population. *From the University of N otingham, Division of Respiratory Medicine (Drs. Hubbard and Britton), City Hospital, and the Queens Respiratory Unit (Dr. Johnston), Queens Medical Centre, Nottingham, UK. Supported by the Medical Research Council. Manuscript received April 17, 1997; revision accepted July 14, 1997. 396 MATERIALS AND METHODS The patients were identified as part of a case-control study o f occupational causes of cryptogenic fibrosing a lveolitis. Details of patient identi fication, data collection, and diagnostic criteria have been published elsewhere. 6 We attempted t o identify all patients
with cryptogenic fibrosing alveoli tis attending nine centers within the Trent Region of England at the outset of the study (the prevalent cases), and all new patients seen during an 18-month recruitment period (the incident cases), using all available lung function laboratory records, hospital discharge data, and local diagnostic registers. The hospital clinical records of each potential case were then inspected to ascertain those who fulfilled our diagnostic criteria, which are set out in Table 1 and which are similar to those used by other researchers. 3 7 Since it is usual clinical practice in the United Kingdom to obtain histologic confirmation of the diagnosis only in the minority of patients in whom the high-resolution CT (HRCT) and other clinical findings do not support a diagnosis of cryptogenic fibrosing alveolitis, 8 we did not make histologic confirmation mandatory to our diagnostic criteria. We defined date of diagnosis as the date on which the cases first met our diagnostic criteria. Six months after the end of patient recruitment, we established the survival status of each prevalent case from their general practitioner, and confirmed dates of deaths were possible with hospital records and local health authority records. Eighteen months after completion of patient recruitment, we repeated this process for the incident cases. We collected data on lifetime smoking habit by postal questionnaire as previously described, 6 and extracted details of lung function at presentation and treatments prescribed from hospital clinical records. We estimated median actuarial survival from diagnosis for incident and prevalent cases separately using a statistical package (EGRET; Statistics and Epidemiology Research Corp; Seattle, Wash; 1988 9 ). To provide a comparison with survival in the general population, we used data from the English life tables 10 that contain the expected future survival for each sex at each year of life. We extracted the expected survival of all patients, based on their age and sex, from the tables and calculated the median expected survival for this "control population," again using the statistical package (EGRET). We then compared the survival of incident cases with that of prevalent cases, and that of our control population, using Cox's proportional hazards survival analysis (EGRET). Finally we examined the effect of age at diagnosis, sex, histmy of every having been a smoker, FVC at presentation, transfer factor for carbon monoxide (Teo) at presentation, and use of treatment with corticosteroids, cyclophosphamide, or azathioprine on survival using Cox's proportional hazards survival analysis for incident and prevalent cases separately. RESULTS A total of 168 (105 male) prevalent cases and 76 (56 male) incident cases were identified. Open lung Table!-Diagnostic Criteria Criteria Histologic confirmation of diagnosis from an open lung biopsy specimen or all of the following: l. Basal inspiratory pulmonary crackles on examination 2. Bilateral interstitial lung shadowing on chest radiograph 3. No documented history of exposure to asbestos or other recognized fibrogens, including birds 4. No clinical evidence of coexisting collagen vascular disease 5. No other coexisting cause of interstitial lung disease 6. Restrictive lung function defined as FEV,IFVC>70% together with an FVC or Tco<80% of predicted; in the absence of restrictive lung function, patients were included if there were pathognomonic changes of cryptogenic fibrosing alveolitis on an HRCT scan biopsy was performed in 31 (12.7%) cases, and the mean age of these cases was significantly lower than those who did not have this procedure (52.4 years vs 69.0 years; p<o.ool). The proportion of prevalent cases who underwent open lung biopsy was higher than that for incident cases, but the opposite was true for HRCT (Table 2). Incident cases tended to be older, and were more likely to be male, and to report ever having been a smoker than prevalent cases (Table 2). Mean percent predicted FVC and Teo at presentation were similar for incident and prevalent cases, but the use of corticosteroids and immunosuppressive drugs was more common among prevalent cases. Median duration of follow-up from diagnosis was 3.6 years for prevalent cases, during which time 43 (26%) died, and 2.3 years for incident cases, during which time 39 (51%) died. The median actuarial survival for prevalent cases was 9.0 years, and for incident cases 2.9 years (Fig. 1). This compared to an expected median survival of 13 years for prevalent cases and 10 years for incident cases. Incident cases were significantly more likely to die than prevalent cases both before (hazard ratio, 3.69 [95% confidence interval (CI), 2.34 to 5.84]; p<o.ool) and after adjustment for the effects of age at diagnosis, sex, smoking, FVC, and Teo at presentation, steroid use, azathioprine use, and cyclophosphamide use (hazard ratio, 4.53 [CI, 2.71 to 7.56]; p<o.ool). Survival for incident cases was substantially worse than that expected for the general population (hazard ratio, 100.1 [CI, 12.5 to 799.5]; p<o.ool). This increase in risk was much less marked for prevalent cases (hazard ratio, 6.21 [CI, 3.81 to 10.12]; p<o.ool). For incident cases, higher percent predicted FVC and Teo at presentation were associated with better survival, while steroid use and cyclophosphamide use were both associated with worse survival (Table 3). Survival was worse in men for each increasing year of age and for never smokers, but these effects were small and not significant at the 5% level. In contrast, for prevalent cases, age at diagnosis was significantly associated with survival (Table 3), and prognosis was worse in women, although not significantly so. Again, poorer lung function at presentation, never having been a smoker, and the use of treatment tended to be associated with worse survival, but only the effect of Teo was significant at the 5% level (Table 3). DISCUSSION In this study, we have shown that the median survival of newly diagnosed cases of cryptogenic fibrosing alveolitis is 2.9 years, and as such is not only substantially lower than that of prevalent cases, but also considerably worse than currently recognized. 3-5 CHEST I 113 I 2 I FEBRUARY, 1998 397
Table 2-Details of Age, Sex, Smoking Habit, Lung Function at Presentation, and Treatment for Incident and Prevalent Cases* Incident Cases (n= 76) Prevalent Cases (n= 168) p Value Male :female 2.8:1 Mean (SD) age, yr, at presentation 69.7 (9.6) Ever smokers 1 58 ( n=65; 89%) Lung function at presentation: Mean (SE) percent predicted FVC 78.5 (2.8) Teo 49.2 (22) Open lung biopsy 5 (6.6%) HRCT 50 (65.8%) Treatment prescribed Corticosteroids 36 (47%) Cyclophosphamide 5 (6.8%) Azathioprine 2 (2.6%) *x 2 test used f or categorical data and unpaired t test for continuous data. 1 Different denominator as data from questionnaire. 1.7:1 0.09 64.7 (12.3) 0.002 104 (n=145; 72%) 0.005 78.5 (1.6) 0.98 51.3 (1.4) 0.41 29 (17.3%) 0.042 58 (34.5%) < 0.001 109 (65%) 0.010 20 (11.9%) 0.203 22 (131%) 0.038 To establish our population of patients, we identified patients attending a range of secondary and tertiary referral centers within the Trent region of England, and thus believe that these cases are typical of patients presenting to respiratory and general physicians. To minimize survival and selection bias, we attempted to identify all cases attending these centers, and because we wanted to study survival in a general population-based sample of patients with disease diagnosed according to usual clinical practice, I,8 we did not make a histologic diagnosis an obligatory diagnostic criterion. Our diagnostic criteria were, however, more stringent than those used by others in epidemiologic studies.l 7 11 12 In fact, in our study, 13% of patients underwent open lung biopsy, 0 0- l - - - - - - - ~ - - ---~ - - - - ~ - ~ 3 FIGURE l. Actuarial and expected survival for incident and prevalent cases. Y-axis=survival probability. X-axis = years. 1 =incident cases (circles represent censored data); 2=prevalent cases (triangles represent censored data); 3=expected survival for incident cases; and 4=expected survival for prevalent cases. and the mean age of this subgroup was significantly lower than that of cases who did not have an open lung biopsy, confirming that this group represent a selected subpopulation of patients who are not typical, at least in age, of patients with cryptogenic fibrosing alveolitis in the general population. The use of open lung biopsy was less common in incident cases compared to prevalent cases, while the use of HRCT was more common with two thirds of incident cases undergoing this investigation. The likely explanation for these findings is the general increasing reliance by physicians in many countries on HRCT as a diagnostic procedure since this technique is noninvasive, well tolerated by patients, and has a low false-positive rate for diagnosing cryptogenic fibrosing alveolitis. 13 The finding of a longer median survival in prevalent cases compared to incident cases is not unexpected, since the incident case group will contain the complete spectrum of newly diagnosed cases, while the prevalent group will exclude those who died quickly from aggressive disease (survival bias). The considerable difference in median survival between incident and prevalent cases demonstrates the strong effect of survival bias in this patient group. It seems likely that it is this effect of survival bias which is the explanation for the difference in median survival between our incident cases and previous estimates of median survival derived from mixtures of prevalent and incident cases accrued as part of clinical case series. 3-5 We believe, therefore, that our results represent a more accurate estimate of typical life expectancy of patients with newly diagnosed conditions which, at <3 years, is actually worse than that of patients with stage I non-small cell lung cancer. 14 Compared to their expected survival derived from the English life tables, our patients with newly 398
Table 3-The Effects of Sex, Age, Smoking History, Lung Function at Presentation, and Treatment on Survival* Incident Cases (n= 76) Prevalent Cases (n= 168) Odds Ratio (95% CI) p Value Odds ratio (95% CI) p Value Male sex 1.50 (0.77 to 2.93) 0.23 0.71 (0.37 to 1.35) 0.29 Age 1.02 (0.98 to 1.06) 0.30 1.09 (1.06 to 1.13) < 0.001 Ever smoker 0.73 (0.28 to 1.88) 0.51 0.62 (0.32 to 1.20) 0.16 FVC 0.98 (0.96 to 0.99) 0.002 0.99 (0.98 to 1.00) 0.15 Teo 0.98 (0.96 to 0.99) 0.005 0.96 (0.94 to 0.98) 0.001 Steroids 2.01 (1.06 to 3.79) 0.032 2.08 (0.96 to 4.51) 0.063 Cyclophosphamide 5.05 (1.69 to 15.06) 0.004 l. 79 (0. 79 to 4.04) 0.16 Azathioprine 2.39 (0.57 to 10.13) 0.24 0.94 (0.37 to 2.38) 0.89 *Odds ratios are for men compared to women, for each additional year of life, ever smokers compared to never smokers, for each additional percent predkted FVC or TCO, and for treated compared to untreated. diagnosed cryptogenic fibrosing alveolitis have a much reduced survival with an average reduction in median life expectancy of approximately 7 years. One consequence of the marked effect of survival bias seen in this patient group is that when assessing the role of various factors on prognosis, ideally only data from incident cases should be used. The data from our incident cases demonstrate that although survival was worse in men, with each increasing year of age, these effects were small and not statistically significant. Previous studies have produced conflicting data with some finding positive effects for age3,5,l 5 or sex, 3 15 and others finding no effects for age 7 or sex. 5 16 Some of these previous studies 3 15 included a proportion of patients with connective tissue diseases, who tend to be younger, more often female, 17 and to have a better prognosis. 5.l 8 Inclusion of these patients will tend to bias the effects of age and sex on survival away from unity. The strong and significant effects of lung function at presentation on survival that we found have been repmted before. 5 7 19 Cigarette smoking has been identified recently as a risk factor for the development of cryptogenic fibrosing alveolitis, 6 11 and it is known to be associated with lower lung function in these patients. 20 However, we found no association between a history of ever having been a smoker and survival, and similar findings have been reported before. 3 7 21 A significant number of patients were not prescribed corticosteroids, and the fact that incident cases were treated less often than prevalent cases suggests that the decision to treat is often delayed for some time after diagnosis. This in turn probably reflects the paucity of data on the efficacy of treatments in this condition. In our study, patients who received treatment had a poorer survival than those who did not, and others have reported similar findings.16 The likely explanation for this is selection bias as physicians will tend to observe patients with stable or slowly progressive disease without initiating treatment, while patients thought to have deteriorating conditions with aggressive disease will tend to be given corticosteroids, azathioprine, or cyclophosphamide. The alternative explanation, that corticosteroids are detrimental to patients with cryptogenic fibrosing alveolitis, is less likely but cannot be excluded, because to our knowledge, there have been no controlled clinical trials of corticosteroids in this disease. In summary, unless only incident cases are included in estimates of survival in patients with cryptogenic fibrosing alveolitis, significant errors will occur because of survival bias. We estimate that median survival of patients with newly diagnosed cryptogenic fibrosing alveolitis is < 3 years and is considerably lower than previous reports. On average, the life span of patients with cryptogenic fibrosing alveolitis is reduced by 7 years. ACKNOWLEDGMENTS: The authors thank Dr. Sarah Firby (actuary) for statistical advice, the lung function technicians at the nine centers involved in this study for assisting with patient identification, and the consultant physicians who allowed us to approach their patients. We would also like to thank Nottinghamshire, Derbyshire, Leicestershire, Lincolnshire, and South Yorkshire Family Health Services Authorities for help with identification of control subjects, the General Practitioners, and the Medical Research Council who funded this study. REFERENCES 1 Coultas DB, Zumwalt RE, Black WC, et al. The epidemiology of interstitial lung diseases. Am J Respir Grit Care Med 1994; 150:967-72 2 Hubbard R, Johnston I, Coultas DB, et a!. Mortality rates from cryptogenic fibrosing alveolitis in seven countries. Thorax 1996; 51:711-16 3 Turner-Warwick M, Burrows B, Johnson A. Cryptogenic fibrosing alveolitis: clinical features and their influence on survival. Thorax 1980; 35:171-80 4 Carrington CB, Gaensler EA, Coutu RE, et a!. Natural history and treated course of usual and desquamative interstitial pneumonia. N Eng! J Med 1978; 298:801-08 CHEST /113/2/ FEBRUARY, 1998 399
5 Wells AU, Cullinan P, Hansell DM, et al. Fibrosing alveolitis associated with systemic sclerosis has a better prognosis than lone cryptogenic fibrosing alveolitis. Am J Respir Ct;t Care Med 1994; 149:1583-90 6 Hubbard R, Lewis S, Richards K, e t al. Occupational exposure to metal or wood dust and aetiology of cryptogenic fibrosing alveolitis. Lancet 1996; 347:284-89 7 Schwartz DA, Helmers RA, Galvin JR, et al. Determinants of survival in idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 1994; 149:450-54 8 Johnston IDA, Gomm SA, Kalra A, et al. The management of cryptogenic fibrosing alveolitis in three regions of the United Kingdom. Eur Respir J 1993; 6:891-93 9 Statistics and Epidemiology Research Corporation. Egret, reference manual. Seattle, Wash: SERC, 1990 10 Government Actuaty's Department. English life tables No. 14. 14th ed. London: HMSO, 1986; 1 11 Baumgartner KB, Samet JM, Stidly CA, et al. Cigarette smoking: a risk factor for idiopathic pulmonary fibrosis. Am J Respir Crit Care Med 1994; 155:242-48 12 Iwai K, Mori T, Yamada N, e t al. Idiopathic pulmonary fibrosis: epidemiologic approaches to occupational exposure. Am J Respir Crit Care Med 1994; 150:670-75 13 Tung KT, Wells AU, Rubens MR, et al. Accuracy of the typical computed tomographic appearances of fibrosing alveolitis. Thorax 1993; 48:334-38 14 Mountain CF. A new international staging system for lung cancer. Chest 1986; 89:225S-33S 15 Tukiainen P, Taskinen E, Holsti P, et al. Prognosis of cryptogenic fibrosing alveolitis. Thorax 1983; 38:349-55 16 Coultas DB, Mapel Dlvl, Hunt C, et al. Idiopathic pulmonary fibrosis (IPF): survival in a population-based cohort. Am J Respir Crit Care Med 1996; 153:A149 17 Crystal RG, Fulmer CD, Roberts WC, et al. Idiopathic pulmonary fibrosis: clinical, histologic, radiographic, physiologic, scintigraphic, cytologic, and biochemical aspects. Ann Intern Med 1976; 85:769-88 18 Agusti C, Xaubet A, Roca J, et al. Interstitial pulmonary fibrosis with and without associated collagen vascular disease: results of 2-year follow-up. Thorax 1992; 47:1035-40 19 Stack BHR, Choo-Kang YFJ, Heard BE. The prognosis of cryptogenic fibrosing alveolitis. Thorax 1972; 27:535-42 20 Schwartz DA, Merchant RK, Helmers RA, e t al. The influence of cigarette smoking on lung function in patients with idiopathic pulmonary fibrosis. Am Rev Respir Dis 1991; 144:504-06 21 de Cremoux H, Bernaudin JF, Laurent P, et al. Interactions between cigarette smoking and the natural history of idiopathic pulmonary fibrosis. Chest 1990; 98:71-76 400