Factors associated with FEV₁ decline in cystic fibrosis: analysis of the ECFS Patient Registry

Introduction

Despite considerable improvement in median survival, cystic fibrosis (CF) remains a life-shortening disease with pulmonary insufficiency as the main cause of death. Several methods are used to assess pulmonary disease severity in CF, including spirometry, chest imaging and quality-of-life questionnaires. For prognosis, forced expiratory volume in 1 s (FEV₁) compared to a reference population (FEV₁ % predicted), is regarded as the best generally available measure for assessing CF lung disease [1]. FEV₁ is currently still an influential driver for the definition of disease stage, for decisions on treatment [1, 2], for comparison between centres and countries [3–5], as a primary outcome in clinical studies [6, 7] and in the regulatory approval of respiratory therapies in CF [8].

The European Cystic Fibrosis Society Patient Registry (ECFSPR) collects data on an annual basis according to agreed definitions and formats of a common set of variables [9] from individual CF centres and national CF registries in Europe and neighbouring countries. Data collection in 2007 included >20 000 CF patients from 16 countries [10]. Such data collection is a reflection of the reality of CF across Europe, offering a unique basis for epidemiological analyses due to its international nature. It offers the possibility to evaluate factors potentially associated with FEV₁ in a population that shows a wide variability of both lung disease and potential risk factors. The aim of this study was to review the association between FEV₁ and factors previously described in the literature as major risk factors of CF lung disease.

Results

Study population

Data referring to 20 204 patients with CF seen in a CF clinic during 2007 were registered in the ECFSPR. The 4420 children aged <6 years were excluded from the analyses due to their unreliable ability to perform spirometry and to the lack of valid reference values for lung function. The 446 patients who had had a lung transplantation were excluded, as their FEV₁ did not represent their disease stage (the UK registry gave information only on transplants performed during 2007, so patients from the UK that had had a lung transplant before 2007 are included). For an additional 606 (3.9%) patients, FEV₁ % pred could not be computed because of missing values for FEV₁ or height. Our final study population included 14 732 patients (52.8% males).

table 1 describes the main demographic and clinical characteristics of the included patients. Information on genotype was available for 13 613 (92.4%) patients and 8.6% of the alleles remained unidentified after DNA analysis. When classified according to McKone et al. [11], 63.3% of patients carried two alleles belonging to class I or II or III, 5.4% of patients carried one allele belonging to class I or II or III and the other allele belonging to class IV or V, and 0.2% of the patients carried two alleles belonging to class IV or V. The remaining patients carried at least one allele that could not be classified. When classified according to F508del genotype, 86.7% of patients carried the F508del allele (48.3% in homozygosity and 38.4% in compound heterozygosity). The proportion of patients with low BMI (SDS ≤ -2) increased from children to young adults but decreased again in older patients (>30 years of age). 86.7% of the patients had pancreatic insufficiency. Prevalence of chronic infection with P. aeruginosa gradually increased with age until it reached a plateau of 55% at 25 years of age (fig. 1). CFRD was found in 12.4% of the patients (11.2% of males and 13.7% of females); its prevalence increased with age, from almost absent in children aged <10 years to 22.7% in patients aged ≥30 years.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1–

Prevalence of chronic Pseudomonas aeruginosa infection and total number of patients by age.

FEV₁ characteristics

FEV₁ % pred was negatively associated with age (fig 2): mean FEV₁ % pred decreased from 91.2% (95% CI 90.4–91.9) in the 6–9-year-olds to 55.5% (95% CI 53.4–57.6) in the 40–44-year-olds (table 1). As shown in figure 2, the decline starts slowly, becomes sharper at age 12 years, continues until the age of 20 years and then stays fairly stable. In contrast, the number of patients (fig. 2) is quite uniform up to age 18 years, when it steadily decreases across the remaining years, although with a less steep decrease from the age of 30 years. Thus, we noticed a gap of roughly 6 years between the sharper decline in FEV₁ and the subsequent decline in the number of patients.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2–

Forced expiratory volume in 1 s (FEV₁) and number of patients by age. Data are presented as mean (95% CI) unless otherwise stated. % pred: % predicted.

table 2 shows the differences in the estimated marginal means of FEV₁ % pred computed from the regression models: they represent the estimated effect of each factor on FEV₁, after adjusting for age and country only and after adjusting for age, country, genotype, sex, pancreatic status, CFRD, P. aeruginosa infection and BMI. Genotype did not show a clinically relevant effect on FEV₁: the highest estimated difference after adjusting for effect of age and country was between patients with the two severe alleles and patients with one severe and one mild allele, the difference in FEV₁ being -10.9 percentage points (95% CI -12.6– -9.1). This difference decreased to -3.8 percentage points (95% CI -5.9– -1.6) after adjusting for the other covariates.

View this table:

• View inline
• View popup

Table 2– Differences in the estimated marginal means of forced expiratory volume in 1 s (FEV₁) % predicted computed from the linear regression models; representing the estimated effect of each factor on FEV₁

Similarly, males had higher FEV₁ % pred values than females by 1.1 percentage points (95% CI 0.3–1.8) after adjusting for age and country effects, and this difference increased to 2.7 percentage points (95% CI 1.8–3.6) after additional adjusting for genotype, BMI, pancreatic status, P. aeruginosa infection and presence of CFRD. Since sex did not reach the threshold of clinical relevance, it was not included in the final model.

Pancreatic status was associated with FEV₁ after adjusting for age and country (p<0.0001), but the difference between pancreatic-sufficient and -insufficient patients decreased from 12.9 percentage points (95% CI 11.8–14.2) to 6.7 percentage points (95% CI 5.2–8.2) after adjusting for sex, genotype, BMI, P. aeruginosa infection and CFRD.

Patients with CFRD showed FEV₁ % pred values 11.6 percentage points (95% CI 10.4–12.8) lower than patients without CFRD, but this difference decreased to 8.0 percentage points (95% CI 6.6–9.5) after additionally adjusting for sex, genotype, BMI, P. aeruginosa infection and pancreatic status.

Chronic infection by P. aeruginosa was associated with FEV₁ after adjusting for age, country, sex, genotype, BMI, pancreatic status, and presence of CFRD (p<0.0001). Infected patients showed mean values of FEV₁ % pred 13.0 percentage points (95% CI 12.0–14.0) lower than uninfected patients.

BMI was positively associated with FEV₁ after adjusting for age, country, sex, genotype, P. aeruginosa infection, pancreatic status and presence of CFRD (p<0.0001). Patients showing a BMI within the normal range (i.e. between -2 and +2 SDS) had, on average, FEV₁ % pred 21.6 percentage points (95% CI 20.0–23.2) higher than patients with poor BMI (≤ -2 SDS).

Based on these results, we selected age, BMI and chronic P. aeruginosa as potential modifiers of FEV₁ in our final model due to their clinically relevant and their statistically significant effect. We also included pancreatic status and CFRD because, although their adjusted effect did not reach the clinically relevant threshold of 10 percentage points, from our data there was strong evidence of a non-negligible effect (∼8 percentage points). Age at diagnosis had a statistically significant effect on FEV₁, after adjusting for all the previously described factors (p<0.0001), but it was not clinically relevant and therefore was not included in the final model: for each additional year of age at diagnosis, patients showed an increased FEV₁ value of 0.16 percentage points (95% CI 0.10–0.22). We carried out a sensitivity analysis of the selection of the covariates in the final model changing the minimal threshold of clinical relevance, and any threshold >5 percentage points difference in FEV₁ (i.e. the variation that other authors claim exist in FEV₁ due merely to different timing of FEV₁ measurement) does not change the covariates that were selected in the final model, as can be seen from table 2.

In this final model we tested for interaction between variables. We found a statistically significant interaction between P. aeruginosa infection and pancreatic status (p=0.012). After adjusting for country, age, BMI and presence of CFRD, pancreatic-sufficient patients not infected with P. aeruginosa had mean FEV₁ values 16.5 percentage points higher than infected ones (95% CI 13.8–19.2). Among pancreatic-insufficient patients, this difference reduced to 12.9 percentage points (95% CI 11.8–13.9). Described from another perspective, pancreatic-sufficient patients not infected with P. aeruginosa had, on average, FEV₁ values 9.0 percentage points higher than pancreatic-insufficient patients (95% CI 7.6–10.5). Among patients infected with P. aeruginosa, this difference was only 5.4 percentage points (95% CI 2.9–7.9). There was also evidence of interaction between P. aeruginosa infection and CFRD (p=0.021). After adjusting for country, age, BMI and pancreatic status, the difference in FEV₁ % pred between infected and uninfected patients without CFRD was 13.8 percentage points (95% CI 12.7–14.8) in favour of those not infected, whereas among patients with CFRD the difference was 10.5 percentage points (95% CI 7.9–13.1), in favour of those not infected.

While these differences are statistically significant, they are clinically negligible and therefore we did not include any interaction in the final model.

In order to express the magnitude of the effect of the risk factors identified in the previous analyses on severe lung disease we carried out a multiple logistic regression analysis (based on 8650 patients with information on all covariates). Severe lung disease was defined as having FEV₁ <40% pred, a widely accepted clinical threshold. As shown in figure 3, the odds of having severe lung disease significantly increases with age (test for quadratic trend p<0.0001). Of the other factors, BMI had the biggest impact on lung disease: patients with poor BMI (i.e. ≤ -2 SDS) experienced a six-fold increased odds of severe lung disease (OR 6.0, 95% CI 5.0–7.3) compared to patients with normal BMI (between -2 and +2 SDS). Chronic P. aeruginosa infection was the factor that had the second biggest impact: being infected increases the odds of severe lung disease by 2.4 (95% CI 2.0–2.7). Furthermore, we estimated that pancreatic-insufficient patients experienced an odds of severe lung disease twice that of pancreatic-sufficient patients (95% CI 1.6–2.5) and patients affected by CFRD had a 1.8-fold increased odds of severe lung disease compared to patients not affected by CFRD (OR 1.8, 95% CI 1.6–2.2). When we inspected the variation of odds ratios between countries to check for interaction, we concluded that the odds ratios are similar across countries (results not shown).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3–

Multiple logistic model: effect of age, body mass index (BMI), chronic Pseudomonas aeruginosa infection, pancreatic status and cystic fibrosis-related diabetes (CFRD) on the odds of having forced expiratory volume in 1 s (FEV₁) <40% predicted. Estimates are controlled for country. Model estimated on 8650 patients from Austria, Belgium, Bulgaria, France, Greece, Italy, the Netherlands, Slovenia and UK. % pred: % predicted; SDS: standard deviation scores. ^#: chosen as reference as most numerous group.

Discussion

The heterogeneous European CF population showed a wide variability of factors potentially associated with FEV₁. We aimed at identifying such factors after adjusting for potential confounders. In this population five factors were associated with poor pulmonary function in patients with CF aged >5 years. Three of the five factors are either partially preventable and/or potentially treatable (poor nutritional status, chronic P. aeruginosa infection and CFRD). The fourth, pancreatic function, indirectly represents the degree of CFTR dysfunction and it might be correctable in the future by therapies that increase the level of functional CFTR. The fifth factor, age, is challenging since it reflects the continuing exposure to additional deleterious factors that might be preventable or treatable with adherence to the current available therapies.

Low BMI was the strongest potentially preventable factor found in the current study of ECFSPR data. The odds of severe lung disease were six times higher for patients with very poor nutrition than patients with BMI within “normal” range, the highest odds ratio of all the preventable factors. Although a causal relationship cannot be proved by our study (severe lung disease by itself can cause poor nutrition through increased energy needs, reduced appetite and gastrointestinal involvement such as reflux and nausea), we showed that poor BMI was significantly associated with poor lung function, even after taking into account the effect of age, pancreatic status or infection with P. aeruginosa. The association between better nutritional status, better pulmonary status and survival has been already demonstrated [15], and aggressive intervention early in life aimed at growth and nutrition may positively affect pulmonary function [16]. However, the current study shows that, despite the available guidelines for prevention and correction of nutritional deficiencies in CF [17, 18], 9.5% of the patients with CF in Europe still suffer from severe malnutrition (i.e. with a BMI ≤ -2 SDS).

Chronic infection with P. aeruginosa was the second factor negatively associated with FEV₁. This infection is well recognised as a cause of morbidity and mortality in CF [19]. In the current study, chronic P. aeruginosa infection was significantly associated with lower FEV₁ after adjusting for age and other potential confounding factors and increased the likelihood of severe lung disease 2.4-fold. This association was also found by both VanDevanter et al. [1] and Konstan et al. [20]. The association seems to be stronger in our study, which could be attributed to the use by both authors of “Pseudomonas at least once” as their definition of infection, whereas we used “chronic Pseudomonas”, which usually implies longer exposure and very often mucoid bacteria. Additionally, both authors investigated children/teenagers, who will have been exposed to the bacteria for a shorter time on average, especially in the chronic form, which would also tend to show a weaker association. However, it is noteworthy that even in young patients with only intermittent or early chronic infection the association is still significant. The importance of the association between P. aeruginosa infection and poor lung function is that although most P. aeruginosa infections are not acquired in hospital, acquisition by cross-infection can be prevented using infection-control measures. Furthermore, newly acquired infections can be eradicated in >80% of the cases by antibiotic therapy immediately after the onset of P. aeruginosa infection [21]. Once chronically colonised, efforts should be directed towards suppression of the infection. Clinical studies have shown that inhaled antibiotics improved pulmonary function and reduced the rate of exacerbations in patients already infected with P. aeruginosa [22].

CFRD was found in the current study to be associated with lower lung function, including after controlling for age and BMI. It is the most frequent comorbidity diagnosed at present, occurring in up to 40% of adults, 25% of adolescents and 9% of children [23]. CFRD is associated with a rapid decline in lung function and increased risk of respiratory failure [24]. Given that CFRD is often clinically silent in the initial stages, it may be valuable to study the potential impact on FEV₁ of annual CFRD screening for early diagnosis [17]. The current study confirms the association between CFRD and poor lung function. Since CFRD is currently not preventable, it should be diagnosed as early as possible. Treatment with insulin enhances nutritional status, temporarily improves pulmonary function and delays the decline in FEV₁ [25].

Pancreatic status in CF is genetically determined and it is associated with CFTR genotype [26]. Pancreatic sufficiency in CF is associated with carrying at least one CFTR allele from class IV or V CFTR mutations that are associated with the presence of sufficient functional CFTR chloride channels over the apical membrane of the exocrine epithelial cells to maintain adequate pancreatic exocrine function. Compared to patients with pancreatic insufficiency, patients with pancreatic sufficiency are diagnosed later, have lower sweat chloride levels and better nutritional status [20, 26]. The current study showed that pancreatic sufficiency is associated with higher lung function even after controlling for nutrition, infection with P. aeruginosa and CFRD. The importance of this finding is that patients with pancreatic sufficiency have better pulmonary function independent of other measured variables. A previous study by Corey et al. [27] showed that patients with pancreatic sufficiency had a lower rate of decline of FEV₁. Therefore, for epidemiological studies and benchmark analysis, when comparing pulmonary function, control for pancreatic status is required, since the prevalence of pancreatic insufficiency may differ between cohorts. Pancreatic status (or the associated genotype) is not a preventable or treatable factor; however, emerging new therapies are under development that may recover some of the lost CFTR function, thereby shifting a “severe” CFTR mutation towards a “mild” partially functional mutation [6, 7], which is expected to be associated with better pulmonary function.

The ECFSPR data do not show evidence of an association between genotype and FEV₁. This is in concordance with previous studies that showed variable lung disease in patients with CF regardless of their genotypes, raising the importance of the other variables previously indicated [26, 28].

Since pancreatic status and genotype are strongly related, an association between FEV₁ and genotype could be expected as well, but since pancreatic status and genotype convey very similar information, once one covariate is present in the model, the other does not additionally explain much more variation of FEV₁.

Although we did not find evidence in our data of the need to control for genotype for epidemiological and comparative studies, more detailed studies are needed to evaluate the importance of genotype on lung function.

In our study, as in other cross-sectional studies, FEV₁ seems to be maintained in the advanced age groups (fig. 2). This is potentially a survivor effect: patients with the most severe lung disease are not included in cross-sectional studies because they died or underwent lung transplant, leaving an increased proportion of “less severely ill” patients in the eldest age groups.

In conclusion, the current study on a large patient population with CF shows the independent effect of nutrition, chronic P. aeruginosa infection and CFRD on FEV₁, which is currently the gold standard measure of disease severity. Since these factors are to some degree preventable or are potentially treatable, this emphasises the importance of early identification of these modifiers of CF lung disease through adequate routine tests. Lowering the rates of new acquisition of P. aeruginosa together with methods to suppress its chronic infection and routine screening for CFRD should be additional therapeutic targets in CF care and should be used as benchmark comparisons for standards of care. Furthermore, for epidemiological and clinical studies when comparing groups of patients with CF, it is important to standardise the patients groups not only according to age, but also according to pancreatic and nutritional status, chronic infection by P. aeruginosa and CFRD. Since this was an observational study based on registry data, there is the possibility of residual confounding effect of factors that were not measured in the registry.