The mean age of the patients (70.5 years) was rather similar to that in other cohort studies that have reported the mean age being between 61.0 and 73.5 years at the time of diagnosis (56,68,107,146). The gender distribution was similar to other studies with approximately 75
% of the patients being males (21,56,107,146). PFT results at baseline were also mostly in line with other studies with the majority of the patients being diagnosed in mild-to-moderate stages of the disease i.e. FVC % over 50 % (56,107).
The mean CPI was 39.9, which is mostly in line with other reports, but also higher baseline values have been reported (16,21,131,146). Extended analyses from the clinical trial INFIGENIA reported CPI values of 49.2 in patients completing and 57.0 in those patients not completing the study (131). Antoniou et al. evaluated the outcome in relation to smoking status; they reported a mean CPI 55.3 in the whole cohort (21). However, a study from Korea reported baseline CPI of 38.6 and 55.3 % of their patients had CPI less than 41 (16). The numbers of patients categorized here into different GAP stages were mostly in line with other studies, which have applied GAP staging (143,149,232).
Similarly to our results, most studies describe a history of smoking in more than half of the patients (18,19,21,117). The proportion of current smokers in the whole cohort was 12.8
%, being very close to the previously reported 4 – 17 % (18-21). There were 35.2 % non-smokers, a smaller proportion than in the Finnish IPF register study, in which 44 % of the patients were non-smokers (56). The registry includes only patients who have provided consent to participate and for that reason, all patients with a rapidly progressing disease or with cognitive impairment may have not been registered. Most likely due to cultural differences, the percentage of non-smokers was higher in Finnish studies as compared to Denmark where only 19 % of the patients were non-smokers (173). In agreement with many other reports, females were more likely to be non-smokers than males (21,22,114).
The median survival of 42 months is comparable to values reported in the literature (1,4).
In this study, males had worse survival than females confirming the results of a study that focused on gender differences in disease progression (114). In this present cohort, 3 (2.3 %) patients received a lung transplant and 2 (66.7 %) were still alive more than 6 years after the transplantation, which is more than the median survival time reported by ISHLT (217).
Since this study included patients from KUH diagnosed between 2002 and 2012, the two new anti-fibrotic drugs, pirfenidone and nintedanib, were not in common use. At the beginning of the 21st century, the recommended treatment for selected patients was prednisone combined with azathioprine or cyclophosphamide and in this cohort these were the most commonly used therapies (53). In a few cases, both azathioprine and
cyclophosphamide combined with prednisone were prescribed. One third of the patients did not receive any medical treatment. In addition, no pharmacological treatment was shown to favor longer or shorter survival. For these reasons, the impact of medication on disease progression can be considered as having been insignificant in this study population.
6.3 LUNG FUNCTION IN MORTALITY PREDICTION
In this study, DLco %, but not FVC % at baseline predicted mortality and similar results have been reported (114,146,153,233). Often both values have been utilized for predicting mortality (4,109,234,235). Since the 1982 reference values were used in this study and the patients were mostly older than 65 years, the values of FVC % predicted and DLco % predicted may have shown lower values than what they would have actually been (227).
However, the possible error would be similar in all patients and therefore the comparisons between patient groups may be considered as reliable. On the contrary, comparisons to other studies using different reference values might be less reliable.
It has been reported that a change in FVC % and DLco % in 6 months predicted mortality (235). In this study, the 6 months’ change in FVC % and in DLco % did not differ between the different disease course subgroups. However, patients in the rapid group experienced a greater decline in both FVC % and DLco % in 12 months when compared to moderate and slow groups. In our multivariate analysis, model including age, DLco % at baseline and a more than 10 % decline in DLco in 12 months was most significantly related to the risk of death. This confirms the results of a prospective patient cohort from the UK where the best performing multivariate model contained gender, DLco % at baseline and DLco % change in 12 months (146). In addition, it has been reported previously that a 10 % decline in DLco was associated with an increased risk of death (121). Similarly, in this present study, a more than 10 % decline in DLco % in 6 or 12 months more than doubled the risk of death.
Moreover, a more than 10 % decline in FVC % in 6 months, but not in 12 months, was significantly related to the risk of death.
When patients were categorized into groups of mild (FVC % > 80 %), moderate (FVC 80 – 50 %) and severe (FVC < 50 %) disease, significant differences were observed in median survival times (52, 42 and 4 months, respectively) as well as in mortality at 1, 2 and 3 years after diagnosis. Nathan et al. categorized patients with different thresholds (mild FVC ≥ 70
% predicted, moderate (FVC 55 % to 69 % predicted) and severe (FVC < 55 % predicted) but reported significant differences in median survival times i.e. 55.5, 37.7 and 27.4 months, respectively (4). However, in that particular study, only 13.7 % of patients had FVC % over 80 %, probably since they were applying the year 2000 diagnostic criteria for IPF diagnosis, which demanded that there was an abnormality in lung function tests (4,53). In this present study, 42.8 % of the patients had FVC % over 80 % and their 3-year mortality was 29 %. The results were similar to a recent Australian study in which 52 % of the patients had FVC > 80
% and their 3-year mortality was 25 % (103). Here, only 7 % of patients were categorized into the severe group with FVC % less than 50 %. Their survival was only 4 months.
When DLco % has been applied in distinguishing the mild stage of IPF, the limits between 50 % and 60 % have been previously used (4,103). Significant survival differences have also been reported between the groups, when mild disease was categorized with DLco % ≥ 50
%, moderate as 49 % to 35 % and severe disease with DLco % < 35 % (4). Since the limits are different from this study, the results are not directly comparable. In this present study, DLco
> 60 % was categorized as mild, DLco % values between 60 % and 40 % were categorized as moderate stage and DLco < 40 % was estimated to be the severe stage. These thresholds were able to separate groups with significantly different survival times. Similar to previous findings, DLco % less than 40 % distinguished patients with more advanced disease and
increased risk of mortality (1,4,233). In this study, patients with DLco % less than 40 % had more CAD as compared to patient with DLco % over 60 % and more heart failure as compared to patients with DLco 60 % – 40 %. These comorbidities may partially explain why the patients with lower DLco % values had shorter survival times.
In the abovementioned Australian study, 3-year mortality rate was approximately 10 % in patients with DLco % over 55 %, which is lower than 23.1 % in patients with DLco % over 60 % in this study (103).
6.4 GAP AND CPI IN MORTALITY PREDICTION
In this study, the mortality observed at 1 year and 2 and 3 years in patients with GAP stages I and II was comparable to the original GAP cohort (14). However, in GAP stage III, the mortality was significantly higher than in the original GAP cohort (14).
Some studies have reported that patients in GAP stage III, but not those in GAP II, had a higher risk of death as compared to GAP stage I patients (123,146). Here, both GAP stage II and III patients exhibited a significantly increased risk of death in the first year after the IPF diagnosis when compared to patients in GAP stage I. The researchers of the original study proposed that the patients belonging to GAP stage II should be candidates for lung transplantation if appropriate (14). The results of this study support the view, that the eligibility for lung transplantation should be considered in stage GAP II, since GAP stage III patients had a median survival of only 4 months. In addition, 40 % of patients with a rapid disease course (survival less than 2 years) were allocated to GAP stage II (Table 22).
This also indicates that the consideration of palliative care should not be delayed to GAP III.
The results of this study also confirmed that higher CPI values were significantly associated with a higher risk of death (16,21,103,131,146). It has been previously reported that CPI > 41 significantly and independently predicted 3-year mortality (105). Very similar results were obtained in this present study, i.e. a CPI value over 42 predicted mortality at 2 years.
In our study, GAP and CPI seemed to be equally good in the assessment of severity, however, GAP staging had a somewhat better accuracy in predicting survival less than 2 years as well as survival for more than 5 years. Some other studies have also made comparisons between CPI and GAP. A similar size prospective study (N=137) from the UK reported that CPI was best performing multivariate model when compared to GAP as well as to the duBois index and DSP (146). Another study comparing CPI and GAP models indicated that CPI was better than GAP in predicting 1-year, 2-year and 3-year mortality (16). In that particular study, the proportions of patients in GAP stages I and II were rather similar to this study (16). However, only 3.4 % were allocated to GAP stage III as compared to 9.5 % in the present study (16). The mean CPI values in GAP stages I, II and III, 33.2, 46.8 and 63.7, respectively, were very similar to this study (16). However, in our study, GAP stage III had a 23 times higher risk of death than GAP I as compared to the only 2.5 times higher risk in GAP III in their study cohort (16). In that study, the mean follow-up time was 22.5 months i.e. approximately 2 years and during that time, the majority, i.e. 57.1 % of GAP stage III patients died (16). In this present study, the 2-year mortality of GAP stage III patients was 100 % (16). Since they had only 28 GAP III patients and 430 patients were lost in follow-up, it is possible that many GAP III patients were lost (16). It might also be an explanation why the risk of death in GAP III was lower than in this study.
Even though CPI has recently been evaluated to be better than GAP in the prediction of patient survival, we noted that the accuracy of GAP staging was somewhat better than CPI in predicting mortality (16,146). GAP staging seems to be more useful in clinical work, due
to the clear instructions about how to apply the system provided by the authors of the original study (14). According to our results, a careful monitoring of GAP II patients seems important since 40 % of the patients tended to survive less than 2 years. The difficulty with CPI is that there are no clear instructions on what an individual value of CPI actually means or how to determine the threshold for separating the more advanced disease from its less advanced stages (15). However, CPI may be useful in assessing disease progression if it is performed repeatedly during follow-up.
6.5 COURSE OF DISEASE
In this study, patients were divided into groups according to their observed lifetime i.e.
rapid (survival less than 2 years), moderate (survival 2 – 5 years) and slow (survival more than 5 years). Only 34.3 % of the patients with a rapid disease course were allocated to GAP stage III and 40 % to GAP stage II.
A study comprising 106 IPF patients presented a similar division of phenotypes; they named rapid progressors (survival less than 2 years), usual survivors (survival 2 – 5 years) and long-term survivors (survival more than 5 years) (93). Here, approximately one third of patients were allocated to each survival group, but in that study, 40 % were usual survivors whereas 28 % and 29 % were rapid progressors and long-term survivors, respectively (93).
Similarly to this study, fewer ex-smokers and males were evident in the slow disease course group and their PFT values were better preserved at baseline as compared to the groups with a more rapid course of disease (rapid and moderate group) (93). They reported that more (11 %) of these individuals in the long-term survival group suffered an AEx in comparison with rapid progressors (3.7 %) (93). In this study, significantly more patients with a rapid disease course suffered an AEx before death, but the total numbers of AEx were not investigated.
The study of Selman et al. examined rapid progressors with a shorter duration of symptoms before diagnosis of IPF (92). Similar to this study, there were significantly more males and ever-smokers than females and never-smokers in the rapid progressor group as compared to the other groups (92). The same approach to disease progression was utilized in a Portuguese cohort (107), but since in our study cohort, the data of the time from symptoms to diagnosis was inadequately documented in the medical records, it is very difficult to make comparisons between that investigation and this present study.
Other comparisons of clinical variables between IIP patients living less than 2 years or more than 2 years have been reported (129). That study included IIP patients with radiological UIP and NSIP patterns (129). Similar to the results emerging from this study, patients with shorter survival times had higher CPI and lower baseline FVC % and DLco % (129). However, there were no differences in age between the groups in that study (129).
Since patients with a NSIP pattern were included, then direct comparisons between studies cannot be made (129).
Here, cut-off values in clinical variables as well as in GAP and CPI and their accuracy in distinguishing patients with different survival times (survival less or more than 2 years as well as survival less or more than 5 years) were investigated. Similar to our results, Sharp et al. reported that DLco % had a better accuracy than FVC % and CPI at baseline in predicting 2-year mortality (146). However, in this present study, FVC % at baseline was as good as CPI in separating groups with survival less or more than 2 years (accuracy 0.64).
GAP staging had better accuracy (0.67) than CPI in separating groups with survival less or more than 2 years. Furthermore, age alone displayed a better accuracy in separating these groups than either of the PFT values, CPI or GAP stage at baseline. Sharp et al. reported that age outperformed FVC % at baseline, but not CPI or DLco % in predicting 2-year mortality
(146). It has also been reported that the ROC derived cut-off value that predicted 2-year mortality was DLco 39 % predicted (236). In this present study, the cut-off value separating groups with survival less and more than 2 years was DLco 54.5 % predicted.
Even though the groups of various disease courses could be differentiated in terms of clinical values, the accuracy of any single factor in predicting that the patient will live less than 2 years or over 5 years remained generally less than 0.80. This confirms the difficulty of predicting the course of disease in IPF with baseline clinical variables. As a conclusion from the evaluation of the course of disease in this present cohort, a patient with a higher age (over 72.5 years), with a history of smoking, with a FVC % less than 73.5 % predicted and DLco % of less than 54.5 % predicted is more likely to die in 2 years as compared to younger patients with better preserved lung function. Similarly, a patient with a lower age (less than 70.5 years), without a history of smoking, with FVC more than 74.5 % and DLco
% more than 58.5 % predicted is more likely to survive over 5 years when compared to an older patient with less preserved lung function.
6.6 SMOKING HISTORY
In this study, current smokers and non-smokers had significantly longer survival times than ex-smokers. In 2001, King et al. reported that current smokers had longer survival compared both to non-smokers and ex-smokers (20). Their study was conducted at the end of the 20th century i.e. before the appearance of the first diagnostic criteria for IPF (20). However, all the patients included in the prospective study cohort underwent SLB i.e. the IPF diagnoses were histologically proven (20). Similar to our results, they reported current smokers as being significantly younger (mean 54.7 years) than non-smokers and ex-smokers, with a difference of approximately 8 years (20). In the present study, the mean age of current smokers was 58.1 years with an age difference of approximately 13 years to ex-smokers and non-smokers. In contrast to our results, they reported higher FVC % in patients with a smoking history as compared to non-smokers and also higher FVC % in current smokers as compared to ex-smokers, but here, no significant difference in PFT was found between the groups with different smoking histories (20).
In 2008, Antoniou et al. studied the outcome in IPF in relation to smoking status (21).
They used the year 2000 diagnostic criteria, but only 39 (16 %) of the IPF diagnoses were histologically proven (21,53). Compared to this study, their cohort included smaller proportions of non-smokers and current smokers (21). However, in agreement with this study, they reported longer survival times in current smokers and non-smokers as compared to ex-smokers (21). They also detected a statistically significant age difference between current smokers and ex-smokers as well as with non-smokers, i.e. current smokers being approximately 3 years younger (21). They also reported significant differences in FVC
% and CPI, but not in DLco % between the groups with different smoking histories (21).
Similar to this study, Antoniou et al. compared ex-smokers separately with non-smokers and current smokers in unadjusted and severity-adjusted analyses (21). When they used CPI in severity adjustment, the survival difference between current smokers and ex-smokers was eliminated (21). In this study, the same effect was seen when both CPI and DLco % were used in severity assessment (21). They also reported that the difference in survival between non-smokers and ex-smokers was retained in both DLco %- and CPI- adjusted analyses, but that was not the case in our study (21). They used the term “healthy smoker effect” to explain why the survival difference between current and ex-smokers disappeared in severity-adjusted analyses (21,23). This could not be confirmed in this study, since the PFT and CPI did not differ between the groups with different smoking histories i.e. current smokers did not have less advanced disease at baseline.
A more recent study from Kishaba et al. using the 2011 diagnostic guidelines, included 32 (32.7 %) never-smokers and 66 ever-smokers i.e. current smokers and ex-smokers (1,22).
They confirmed our results that the non-smokers were mostly female (22). They reported no difference in age between the groups as the mean age of both groups was 73 years (22).
In their study, the combined smoker group had a longer survival time, a shorter time to exacerbation, higher DLco values and a lower CPI compared to never-smokers (22). In addition, the poorer prognosis of non-smokers was retained even in severity-adjusted analyses (22). Since they applied a different study approach, their results are not directly comparable to ours (22).
In this study, gender was significantly related to survival only when non-smokers and ex-smokers were compared. Similar results have been reported before, when the effect of gender was investigated while taking into account age, DLco % and smoking history adjusted analyses (114). In that particular study, smoking history was considered as non-smoker or ever-non-smokers i.e. current and ex-non-smokers (114). However, the difference between males and females in the proportion of non-smokers was comparable to the results emerging from this study (114). In the present study, the results may have been influenced by the small number of ex-smoker females (6 compared to 60 ex-smoker males).
Despite the younger age of current smokers in this study, they had smoked significantly
Despite the younger age of current smokers in this study, they had smoked significantly