Background Lung function is an important indicator of cystic fibrosis disease status and those with better forced expiratory volume in 1 s (FEV1)% predicted have tended to report a better health-related quality of life (HRQoL) in cross-sectional studies. The relationship between lung function and HRQoL over time is unknown. This work assesses the natural progression of HRQoL reporting over many years and compares assessments across a whole decade and evaluates the relationship between lung function and HRQoL longitudinally.
Methods Demographic (age, gender), clinical (FEV1% predicted, body mass index, diabetes, Burkholderia cepacia complex, intravenous access device and nutritional status) and HRQoL (Cystic Fibrosis Quality of Life Questionnaire) variables were obtained every 2 years over a 12-year period (seven time points from 1998 to 2010).
Results HRQoL and lung function declined slowly over time and significant decade changes were observed for FEV1% predicted and the nine domains of the Cystic Fibrosis Quality of Life Questionnaire. The results of random coefficient modelling indicated that, at the population level, decreasing FEV1% predicted was associated with decreasing HRQoL after adjusting for confounding variables. However, the percentage of patients for whom a decrease in lung function was associated with a decrease in HRQoL differed according to the quality of life domain.
Conclusions HRQoL and FEV1% predicted decline slowly; nevertheless, a decrease in lung function predicted a decrease in HRQoL over time.
- Cystic Fibrosis
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
What is the key question?
Health-related quality of life (HRQoL) measures are increasingly being used to evaluate the effectiveness of interventions but there are scant data as to how HRQoL scores naturally vary over time. This work evaluates the relationship between HRQoL and lung function longitudinally and addresses the question 'Does HRQoL reporting follow the trend of FEV1% predicted?'
What is the bottom line?
FEV1% predicted and HRQoL decline slowly, nevertheless, a decrease in lung function predicts a decrease in HRQoL over time.
Why read on?
This work provides the first longitudinal HRQoL data for cystic fibrosis that assesses the natural progression of HRQoL reporting over many years, providing a benchmark to inform data interpretation and clinical relevance.
Cystic fibrosis (CF) is the most common lethal autosomal recessive condition in the white population.1 Improvements in care have led to median survival reaching the late 30s.2 With increasing age, however, a high proportion of people develop diabetes mellitus and some patients develop a variety of complications including pneumothorax, haemoptysis and osteoporosis. The daily treatment regimen of chest physiotherapy, nutrition, exercise and oral, inhaled and intravenous medication is burdensome, complex and time-consuming. The progressive nature of CF disease and the demanding treatment burden3 ,4 have the potential to encroach on daily living and to impact on the person's health-related quality of life (HRQoL). HRQoL measurement allows the inclusion of the patient's perspective in research and clinical practice.5 The insights that patients have concerning their health and how they report these are important, given that aspects of HRQoL have been shown to be independent predictors of survival.6
Forced expiratory volume in 1 s (FEV1) % predicted is an important indicator of disease status, and those with better lung function tend to report a better HRQoL. The most convincing data come from analyses of known groups undertaken during the evaluation of the CF-specific HRQoL instruments.7–9 Differences between disease severity groups (mild: FEV1% predicted ≥70%; moderate: FEV1% predicted 40–69%; severe: FEV1% predicted <40%) were observed across many HRQoL domains. Those with mild disease generally reported a better HRQoL than those with severe disease. However, studies have demonstrated weak to moderate correlations between HRQoL and lung function in children and adults,7–12 with weaker correlations often found in studies employing generic rather than CF-specific scales. Even though FEV1% predicted was positively associated with HRQoL domains in multivariate analyses, it accounted for only a small proportion of variability in HRQoL scores.11–14 This cross-sectional work suggests that very large differences in FEV1% predicted would be required to predict even modest differences in HRQoL scores.
HRQoL instruments are increasingly being used to evaluate the effectiveness of interventions and their use in routine clinical management has been shown to be feasible,15 but there are scant data as to how HRQoL scores naturally vary over time. A few studies have assessed changes in HRQoL and lung function over 1–2 years. The most salient finding was the stability of HRQoL over this time and its independence from the course of FEV1% predicted.15 ,16 Similarly, data from the US Epidemiologic Study of Cystic Fibrosis showed that, over a 1-year period, changes in both clinical variables and HRQoL were small.17
The cross-sectional nature of most studies does not allow the assessment of cause and effect. The initial aim of this work was to provide the first true prospective longitudinal HRQoL data for CF, to assess the natural progression of HRQoL reporting over many years and to compare assessments across a whole decade. A further objective was to evaluate the relationship between HRQoL and lung function longitudinally and to address the question: ‘Do HRQoL domains follow the trend of FEV1% predicted?’ Our hypothesis, based on previous work, was that HRQoL is independent of FEV1% predicted.
Subjects and procedure
All patients aged 16 years or older who attended a large Adult Cystic Fibrosis Centre in the UK were approached to take part in the study. Demographic, clinical and HRQoL variables were initially obtained during the patient's routine annual CF assessment. The centre followed standard treatment protocols and annual reviews were undertaken as close to the patient's birthday as possible, predominantly when the patient was clinically stable. They were followed-up every 2 years over a 12-year period (seven time points from 1998 to 2010). At each time point patients provided consent for their participation in accordance with ethical committee approval. The Cystic Fibrosis Quality of Life Questionnaire (CFQoL) was mailed out for completion prior to clinic appointments at which demographic and clinical variables were recorded. Not all patients entered the study at time point 1 (T1), but joined the study at later time points (T2–T7).
At each time point demographic, clinical and HRQoL variables were collected. Age and gender, FEV1% predicted, body mass index (BMI), whether the person had diabetes, Burkholderia cepacia complex or an intravenous access device were recorded. Nutritional status (no oral calorie supplements, prescribed oral calorie supplements or prescribed enteral tube feeds) and lung transplantation were also documented. These variables describe the characteristics of the sample and include known predictors of decline in FEV1% predicted.18 ,19 They have also been shown to relate to HRQoL, but only in cross-sectional studies.14 ,15 ,20–22
Quality of life was measured using the CFQoL.9 The CFQoL was developed and validated in the UK and was therefore the most appropriate CF-specific HRQoL measure for this UK sample. The instrument measures nine domains of functioning: Physical functioning, Social functioning, Emotional responses, Treatment issues, Chest symptoms, Body image, Interpersonal relationships, Career concerns and Concerns for the future. The psychometric properties of the instrument are good. Internal reliability of the domains was demonstrated using Cronbach's alpha coefficients (range 0.72–0.92, median 0.89) and item to total domain score correlations. Concurrent validity with three appropriate SF-36 domains (range r=0.64–0.74), known groups validity between different levels of disease severity, sensitivity across transient changes in health (effect size range, d=0.56–1.95) and test-retest reliability (r=0.74–0.96, median 0.91) were found to be robust.9 ,10 Each domain has a minimum score of 0% and a maximum score of 100%, with higher scores reflecting a better quality of life.
Patient assessments were described by summary statistics. Long-term changes were summarised at time points separated by 10 years; mostly T2 and T7, otherwise T1 and T6. Long-term changes were tested for significance using t tests.
The longitudinal relationships between the nine domains of CFQoL and FEV1% predicted were explored using regression models with fixed and random coefficients. These models are an established methodology in longitudinal studies23 ,24 since they allow the observations on one individual to be explained by an individual regression coefficient which arises from a population of normally distributed coefficients across individuals. For each HRQoL domain, CFQoL=100 × S/N where S was the domain score and N was the maximum score. Maximum scores were: Physical functioning (50), Social functioning (20), Emotional responses (40), Treatment issues (15), Chest symptoms (20), Body image (15), Interpersonal relationships (50), Concerns for the future (30) and Career concerns (20). The relationship between CFQoL domain and FEV1% predicted was adjusted for the confounding effects of age, gender, BMI, diabetes mellitus and nutritional status by including these variables in the models. Statistical analysis of HRQoL presents challenges because of the well-understood ‘ceiling effects’ for HRQoL measures which are limited above by 100%25 and because domain scores are discrete measures only taking values corresponding to whole numbers. There is particular difficulty when the maximum score is low—for example, for Chest symptoms which can take only the values 100%, 95%, 90% etc.
Binomial regression with fixed and random coefficients was chosen as an appropriate analysis because the binomial is a discrete distribution for scores having a predetermined maximum. Binomial regression predicts within the range 0–100% because the covariates predict the logistic transformation of HRQoL, similar to logistic regression analysis. This approach was assessed initially by fitting binomial regression models for individual participants and then exploring the distribution of the coefficients, and the approach was found to be suitable. Models were fitted using the software MLwiN.26
The models estimated the means and variances of the random coefficients for FEV1% predicted, BMI and the intercept. Gender, diabetes mellitus and nutritional status were included as fixed effects. Age was included as a fixed effect because age changes deterministically with time. The random coefficients were tested to check whether each should be retained as random or could be included as fixed.
The normal distribution for the random coefficient for FEV1% predicted provided evidence of whether HRQoL was related to FEV1% predicted longitudinally. If the mean of the normal distribution was significantly greater than zero, then the participant population showed overall a positive mean coefficient for the relationship between the CFQoL domain and FEV1% predicted after adjusting for the confounding variables. A positive coefficient implied that a declining HRQoL domain was associated with declining FEV1% predicted. The proportion of the participant population exhibiting a positive coefficient was estimated from the fitted normal distribution and 95% CIs were obtained by computer simulation.
Thus, the random coefficients approach permitted an informative answer to the question: ‘Do CFQoL domains follow the trend of FEV1% predicted?’ The answer was both a ‘yes’ or ‘no’ for the whole cohort, and also an estimate of the proportion of patients with CF for whom the answer was in the affirmative.
Demographic and clinical characteristics of the sample for the seven time points are presented in table 1, together with the clinic total population size at each time point. A total of 234 patients entered the study. The numbers of time points at which patients were assessed are given in table 2. The length of time patients remained in the study (time between first and last assessment) are provided in table 3 together with the number of patients who died during the study period. The median age of death was 28 years (IQR 24–35). Patients with only one time point were still included in the analyses because this improved the estimation of the random intercept and hence contributed to the precision of the longitudinal coefficients.
Changes in FEV1% predicted and CFQoL domains over a decade
The database contained 91 patients (40 men) with assessments separated by precisely 10 years. Table 4 presents the mean changes for age, clinical variables and CFQoL domains for men and women over one decade. As a visual example, figures 1 and 2 illustrate the change in FEV1% predicted and Chest symptoms for men and women, respectively, and for most patients the change is a decline. Significant decade changes were observed for FEV1% predicted and the nine domains of the CFQoL, but not for BMI (men, p=0.087; women, p=0.305). There were no significant differences between men and women except for BMI (p=0.046), but the actual difference was small and not deemed clinically relevant.
Relationship between FEV1% predicted and HRQoL over time
Table 5 presents the results of the random coefficient modelling. The FEV1% predicted variance-covariance estimates were significant for all domains except Body image, and this confirmed that FEV1% predicted was a random effect for all the domains except possibly for Body image. The FEV1% predicted coefficient mean was positive and highly significantly different from zero for all domains except Body image. Hence, at the population level, decreasing FEV1% predicted was associated with decreasing HRQoL even after adjusting for all the other confounding variables in the models.
The statistical analysis assumed patients had a personal coefficient of change in HRQoL with changing FEV1% predicted. The analysis estimated the mean coefficient of the CF population together with the variability (variance) of the patients’ coefficients; the method did not estimate individual patient coefficients and hence could not categorise individuals as those having a positive or negative coefficient. However, because the method estimated the statistical distribution of coefficients, it did allow for the computation of the percentage of patients for whom a decrease in lung function was associated with a decrease in HRQoL, and this percentage differed according to the HRQoL domain. For over 80% of adults, a decrease in lung function predicted a decrease in Chest symptom score. For approximately 70% of patients a decline in lung function predicted a decline in Physical functioning, Social functioning, Treatment issues, Emotional responses and Career concerns. This relationship held true for over 60% of patients for Concerns for the future and Interpersonal relationships. Hence, for all domains except Body image, the majority of patients showed a relationship between declining FEV1% predicted and declining HRQoL in the longer term, indicating that HRQoL reporting followed the trend of FEV1% predicted. Approximate CIs for the percentages are given in table 5.
This work provides the first longitudinal HRQoL data in CF that follows patients over many years, providing a natural progression of HRQoL reporting over time. Both FEV1% predicted and HRQoL decline slowly and, contrary to our hypothesis, a decrease in lung function was associated with a decrease in HRQoL. The changes in lung function over a decade were representative of the UK population.2 ,18 BMI remained constant, and this is attributed to the exceptional longstanding gastroenterology expertise within this centre.27 The increase in CF-related diabetes (CFRD) over the decade is likely to reflect the ageing of the cohort since there has been a decrease in CFRD in this centre.2
Although HRQoL declined slowly for young adults with CF, it is noteworthy that HRQoL population norms remain constant over a comparative age span.28 With declining lung function it was not the physical aspects of HRQoL that decreased the most but the psychosocial domains of Career concerns, Concerns for the future and Treatment issues. This may be expected as treatment becomes more burdensome and employment and the future more uncertain with advancing CF disease. Interestingly, Physical functioning and Chest symptoms decreased by 8% and 9.4%, respectively, over a decade, a similar decrease to FEV1% predicted. However, it would be an erroneous assumption to consider FEV1% predicted a proxy for the clinical/physical HRQoL domains. HRQoL can provide ‘added value’ as it can supply information not captured by other outcomes. Indeed, the Physical functioning domain of the CFQoL is an independent predictor of survival6 and, in a recent trial, only HRQoL provided additional information to lung function from several secondary endpoints.29 It is also noteworthy that the standard deviations for FEV1% predicted and the CFQoL domains are large so, although there was a similar percentage decrease at the population level, this was not necessarily true at the individual level.
The cross-sectional data from the literature suggested that large differences in lung function were required to predict differences in HRQoL.11–14 There was also a lack of association reported over 12–18 month periods.15–17 This is not surprising since decreases in FEV1% predicted and HRQoL scores are approximately 1% per year and are highly variable across individuals, as evidenced in figures 1 and 2. It is noteworthy that similar short-term results were reported with the use of different HRQoL instruments.15–17 Whether HRQoL instruments containing different domains yield similar long-term data remains to be evaluated. HRQoL changes in response to acute clinical situations are not explicitly considered here (eg, new symptoms or improvements expected with treatments), but knowing the natural rate of change in HRQoL domains provides a benchmark against which changes due to complications or interventions can be compared and should help to inform the clinical relevance of such data.
The decade changes (table 4) were for individuals who survived the decade, but the modelling of the relationship between HRQoL and FEV1% predicted was carried out using all patients who entered the study. Some of these individuals died or dropped out during the 12 years, but their available data were still included in the analyses. HRQoL reporting followed the decreasing trend of FEV1% predicted over time. However, the percentage of patients for whom a decrease in lung function was associated with a decrease in HRQoL differed according to the HRQoL domain. For 82% of adults a decrease in lung function was associated with a decrease in their Chest symptom score, whereas this relationship was true for only 58% of people concerning their Body image score. For a significant minority, a decrease in lung function seemed not to be associated with a decrease in HRQoL domains. What differentiates these patients is unknown, but is potentially very important as some individuals had a decrease in lung function but maintained a high level of HRQoL over time, indicating that they were psychologically resilient.30 Understanding the determinants for sustaining a good HRQoL with advancing disease may help to ensure the best possible HRQoL for people with CF.
There is some evidence in the CF literature that the role of coping and mood may be important. Optimistic coping (focused, determined and optimistic beliefs about the future) and distraction coping (doing things to forget CF) emerged as significant predictors of how adults with CF report their HRQoL. Optimism was consistently associated with a superior quality of life while distraction was consistently associated with a poorer quality of life.31 Conversely, depressive symptoms were associated with a poorer HRQoL across all CFQ domains.32 Screening for and treating depression may improve HRQoL, and further robust studies using multivariate modelling are required to assess the impact of mood on HRQoL and clinical status. We await results from the International Depression and Anxiety Epidemiological Study to clarify these relationships.
A limitation of the work is that it was a single-centre study. Nevertheless, this is the first reported longitudinal data set that has enabled the evaluation of population and individual level clinical and HRQoL changes over many years. Additionally, several variables now considered to be important because they increase the rate of decline in FEV1% predicted were not collected. These included microbiological status (eg, Pseudomonas aeruginosa, MRSA and members of the B cepacia complex18 ,33 ,34) and the frequency of pulmonary exacerbations.35 It is also possible that patients receiving more therapies report different trajectories in HRQoL. When undertaking longitudinal work in a disease that is changing (improving specialist centre care, treatment intensity and survival), it is difficult to predict the factors that will be important a decade later. However, if these variables contributed to an increased rate of decline in FEV1% predicted, it does not invalidate these models as the focus is on the relationship between HRQoL and FEV1% predicted. The more FEV1% predicted declines (for whatever reason), the easier it becomes to detect the relationship between HRQoL and FEV1% predicted.
The future entry of HRQoL data into registries is important, although the usefulness and challenges of including HRQoL data in registries (eg, choice of HRQoL tool and domains, time interval, patient burden) requires consideration. Such data could provide evidence to support our conclusion that a decrease in lung function predicts a decrease in HRQoL over time.
Contributors JA: conception and design, analysis and interpretation of data, drafting, revising and final approval of the article. MAH: analysis and interpretation of data, drafting, revising and final approval of the article. AMM: conception and design, acquisition of data, critical revision and final approval of the article. SPC: conception and design, interpretation of data, critical revision and final approval of the article.
Competing interests None.
Ethics approval Ethics approval was obtained from Leeds research ethics committee.
Provenance and peer review Not commissioned; externally peer reviewed.