Description of the condition
Cystic fibrosis (CF) is the most common recessively inherited genetic condition affecting Caucasian populations. It is estimated that, among Northern Europeans, one person out of 25 is a carrier of a CF gene and that one child out of every 2500 born is affected, with other populations affected to a lesser extent (Rosenstein 1998). The most important clinical manifestation in CF is progressive pulmonary deterioration, which is associated with recurrent lower respiratory tract illnesses, respiratory failure and death. As morbidity and mortality among those with CF is largely a consequence of these respiratory complications, the pathogenesis of respiratory disease in CF and the prevention of pulmonary deterioration are important therapeutic goals.
The role of airway inflammation in mediating lung injury among those with CF has received increasing attention (Konstan 1997). Persistent infection of the airways with Pseudomonas aeruginosa (P. aeruginosa), Staphylococcus aureus or Haemophilus influenzae is thought to contribute to a hyperactive inflammatory response, with cytokines, proteases, oxygen radicals and elastase being released profusely once infection is established (Chmiel 2002; Gibson 2003). These mediators of inflammation are likely to make an important contribution to airway damage. In addition, persistent stimulation of the immune system leads to high serum and tissue IgG levels, accompanied by extensive neutrophil infiltration. This combination of cellular and biochemical events form part of a cascade of inflammation which is associated with progressive and irreversible airway damage in CF.
Description of the intervention
Non-steroidal anti-inflammatory drugs (NSAIDs) are agents that affect the cyclo-oxygenase pathway, resulting in inhibition of prostaglandin synthesis and, at high doses, ibuprofen also affects the lipoxygenase pathway. However, uncertainty remains regarding the exact mode of action of ibuprofen and it has been suggested that, at low doses, ibuprofen may be pro-inflammatory (Konstan 2003). Mean peak plasma concentrations of greater than 50 micrograms per milliliter (μg/mL) have been associated with maximal anti-inflammatory effect. A potentially narrow therapeutic window for ibuprofen may exist, as experimental work in animals and humans suggests that low plasma concentrations may be associated with a paradoxical increase in inflammation. In view of this, the importance of maintaining high plasma levels in the therapeutic range based on pharmacokinetic trials and careful pharmacological monitoring to devise optimal individual dosing schedules has been stressed.
Difficulties in access to pharmacological assays with which to optimise individual dosing schedules, combined with concerns regarding potential adverse effects of NSAIDs, may be important considerations. Recent debates in literature about safety of low-dose NSAIDs and cardiovascular disease further highlights the need for careful monitoring of pharmacokinetic levels. The likelihood of adverse events among people with CF may be increased in the presence of impaired liver or renal function, and these in turn may influence plasma levels of potentially toxic co-therapy such as, for example, with aminoglycosides. This emphasises the importance of assessing these consequences adequately within the context of randomized trials.
How the intervention might work
As pulmonary damage in CF may occur as a consequence of inflammation, it has been hypothesized that prolonged use of NSAIDs may prevent progressive pulmonary deterioration and respiratory morbidity.
Previous trials have provided circumstantial evidence in support of this hypothesis. In the Wheeler trial, the authors observed that, among young children with CF, respiratory status was better in those with very low serum IgG levels compared to age-matched CF children with normal or elevated serum IgG (Wheeler 1984). Konstan infected a group of rats with P. aeruginosa and then treated them with either ibuprofen or placebo for 14 days (Konstan 1990). The active treatment group had 30% less lung inflammation than the placebo-treated group. Although the rat model for pulmonary infection with P. aeruginosa is different from human CF in many ways, the above results offer further support for the hypotheses that:
1. there is an immunological or inflammatory component to lung function decline in CF; and
2. treatment with NSAIDs may delay progression of Pseudomonas-related lung disease in people with CF.
Why it is important to do this review
This systematic review aims to establish whether administering oral high-dose NSAIDs to people with CF alters the pattern of change in specific clinical markers of CF-related disease progression. Specifically, it aims to explore the extent to which existing evidence supports a beneficial effect of NSAIDs on lung function and respiratory morbidity in people with CF. The potential negative effects of such treatment, especially when given over prolonged periods, will also be considered. Recognized adverse effects of NSAIDs include: minor and major haemorrhage (life-threatening gastrointestinal bleed, haemorrhagic stroke); gastrointestinal upset (dyspepsia, nausea); diarrhoea; allergic reactions (rash, fever); tinnitus; bronchospasm; and fluid retention.
The original review summarising three trials found preliminary evidence to suggest that non-steroidal anti-inflammatory drugs may prevent pulmonary deterioration in people with mild lung disease due to CF, but that their routine use could not be recommended (Dezateux 1999).
The objective of this review is to determine the effectiveness of treatment with NSAIDs in preventing pulmonary deterioration and maintaining an optimal level of pulmonary function among those with CF. Specifically, we wished to examine the hypotheses that treatment with NSAIDs for a period of two months or more:
- improves or prevents deterioration in objective measures of lung function;
- reduces the need for intravenous antibiotic therapy for respiratory exacerbations;
- reduces the need for hospital admissions for respiratory exacerbations;
- reduces the number of days spent in hospital for respiratory exacerbations;
- improves quality of life;
- improves survival;
- is associated with adverse effects (including major haemorrhage such as a life threatening gastrointestinal bleed or haemorrhagic stroke, gastrointestinal upset including dyspepsia or nausea, diarrhoea, allergic reactions (rash, fever), tinnitus, bronchospasm or fluid retention).
Criteria for considering studies for this review
Types of studies
Randomized controlled trials, published and unpublished.
Types of participants
Children and adults, of any age, with defined CF, diagnosed clinically and by quantitative sweat chloride testing or genetic testing or both. People with CF at all stages of lung disease were included.
Types of interventions
Non-steroidal anti-inflammatory drugs (NSAIDs) administered orally at any dose for a period of at least two months compared to placebo or existing conventional therapy.
Types of outcome measures
- Objective measures of lung function (measured as absolute change in per cent predicted from baseline compared to control and post-treatment change in per cent predicted compared to control):
- forced expiratory volume in one second (FEV
- forced vital capacity (FVC)
- forced expiratory flow 25-75% (FEF
- Number of days of intravenous antibiotic therapy for respiratory exacerbations
- Number of hospital admissions for respiratory exacerbations
- Number of hospital days for respiratory exacerbations
- Quality of life measures
- Proportion surviving at six months and one year
- Nutritional status as noted by weight gain, body mass index, z score or other indices of nutritional state
- Chest X-ray scores
- Adverse effects (such as minor haemorrhage, major haemorrhage (life-threatening gastrointestinal (GI) bleed, haemorrhagic stroke), gastrointestinal upset (dyspepsia, nausea), diarrhoea, allergic reactions (rash, fever), tinnitus, bronchospasm, fluid retention)
- Dropout rates
Search methods for identification of studies
Relevant trials were identified from the Group's Cystic Fibrosis Trials Register using the terms: anti-inflammatory AND (non-steroidal OR other) AND (oral OR not stated).
The Cystic Fibrosis Trials Register is compiled from electronic searches of the Cochrane Central Register of Controlled Trials (CENTRAL) (updated each new issue of The Cochrane Library), quarterly searches of MEDLINE, a search of EMBASE to 1995 and the prospective handsearching of two journals - Pediatric Pulmonology and the Journal of Cystic Fibrosis. Unpublished work is identified by searching the abstract books of three major cystic fibrosis conferences: the International Cystic Fibrosis Conference; the European Cystic Fibrosis Conference and the North American Cystic Fibrosis Conference. For full details of all searching activities for the register, please see the relevant sections of the Cystic Fibrosis and Genetic Disorders Group Module.
Date of the most recent search of the Group's Cystic Fibrosis Trials Register: 15 May 2013.
Searching other resources
The reference lists of identified publications were searched for additional references. The pharmaceutical companies which manufacture NSAIDs were contacted by letter with requests for information on any published or unpublished trials using NSAIDs in CF.
Data collection and analysis
Selection of studies
At the time of the initial review two authors (CD and AC) independently applied inclusion criteria to all potential reports. Subsequently during the 2007 update, two different authors (SS and LL) assessed the inclusion criteria of all identified trials. The authors resolved any disagreements by discussion.
Data extraction and management
The authors independently extracted data using a standard data extraction form.
In order to directly compare pulmonary function outcomes in the pediatric population as defined in the Konstan trial (Konstan 1995), data from the Trans-Canadian trial (Lands 2007) were re-analyzed by stratifying the data at 13 years of age.
Assessment of risk of bias in included studies
In order to establish a risk of bias for each included study, the authors assessed the methodological quality of the included studies. The authors recorded details of the following: concealment and generation of the randomisation sequence; whether the trial was blinded; whether intention-to-treat analyses were possible from the available data; and whether the report documented the number of participants lost to follow up or subsequently excluded from the trial.
If the generation and concealment of allocation were judged to be adequate, this corresponded to a low risk of bias; if they were judged to be inadequate, this corresponded to a high risk of bias and if they were judged to be unclear, the risk of bias too was judged to be unclear.
For blinding, the risk of bias decreased if more people (participants, clinicians or outcome assessors) were blinded to the intervention. The risk of bias was unclear if the studies did not describe the degree of blinding used.
Regarding incomplete outcome data, there was a risk of bias if a number of participants withdrew or were lost to follow up with no good reason supplied or if the reasons differed across groups. The risk of bias was low if the amount of missing data was approximately equal and balanced across groups. If the numbers randomized into each intervention group are not clearly reported, the risk of bias from missing data was judged to be unclear.
Measures of treatment effect
We proposed a pooled estimate of the treatment effect for each outcome measure across trials, specifically the odds of an outcome among treatment-allocated participants to the corresponding odds among controls. We considered improvement in any of those outcomes that demonstrated efficacy of treatment to be beneficial.
For continuous outcomes, we calculated the mean difference (MD) and the 95% confidence intervals (CI) and for dichotomous outcomes, the Peto odds ratio (OR) and the 95% CI.
Unit of analysis issues
We do not feel that cross-over trials are appropriate for assessing the treatment effects of NSAIDs in CF, as CF is a progressive disease and requires a longer follow-up period. If any cross-over trials are published in the future, we will consider whether they should be included or not on an individual basis.
Dealing with missing data
In order to allow an intention-to-treat analysis, the authors sought data on the number of participants with each outcome event, by allocated treated group, irrespective of compliance or subsequent exclusion from treatment or follow up.
We contacted the primary authors of the trials for any information required for our analysis which did not appear in the published trial reports. We were unable to obtain original data from Dr. M. Konstan for the 13- to 18-year old participants from the 1995 study to make the Konstan and Lands studies less heterogeneous (Konstan 1995; Lands 2007).
Assessment of heterogeneity
Heterogeneity between trial results was examined using a chi-squared test and the I
Assessment of reporting biases
Reporting bias was assessed to ensure all outcomes listed in the methods were reported in the results. Where possible, the original study protocol were used. If a sufficient number of studies had been identified, a funnel plot would have been used to quantify any publication bias.
We have analysed these data using a fixed-effect model. If in future we include further studies and identify significant heterogeneity, we plan to analyse the data using a random-effects model.
Subgroup analysis and investigation of heterogeneity
As data were available we analysed the effect of NSAIDs in a subgroup of young participants (under 13 years of age) with mild lung disease and a separate subgroup of older participants (over 13 years of age) with moderate or severe lung disease. The age-stratified data were not published in the Canadian Trial and were obtained directly from the authors.
If in future sufficient data are available we plan to conduct a sensitivity analysis to test the robustness of our results excluding outlying centres or centres identified with high risk of bias or both.
Description of studies
Results of the search
The original review in 1999 identified five trials, of which three, reporting data from a total of 145 participants met the inclusion criteria (Konstan 1991; Konstan 1995; Sordelli 1994). One trial (with two reports) was ineligible for inclusion (Noritake 1982). At the time of the original review, the remaining trial (a large Canadian trial comparing ibuprofen to placebo in children aged 6 to 18 years) had completed recruitment but results were not available. This trial, including a total of 142 children followed for a period of two years, has since been published and was included in this review in 2007 (Lands 2007). In 2005 a further trial was identified (n = 47), which has been reported in abstract form at the 2004 annual congress of the European Respiratory Society (Shmarina 2004). An additional reference to this trial was identified in the 2010 search and following correspondence with the authors the trial has been included in the full review. The search in June 2008 identified one trial which was presented in abstract form at the 2007 North American CF Conference; this has been excluded (Chmiel 2007). The most recent search in 2010 identified another trial which has also been excluded (Kovaleva 2000).
The participants included in the five eligible trials ranged from 5 to 39 years of age. In two separate trials from the same center, eligibility was defined according to an objective measure of lung function (Konstan 1991; Konstan 1995). In the first trial, participants with mild to moderate lung disease were included, defined as those in whom FEV
In the third trial by Sordelli, eligibility was defined by a combination of microbiological, anthropometric and clinical features; namely, previous sputum culture positive for P. aeruginosa, a Schwachman score less than 50, body weight greater than 15 kg, an ability to swallow medication and an ability to comply with pulmonary function tests (Sordelli 1994).
The Canadian multicenter trial recruited children with a sweat chloride diagnosis of CF and an FEV
In the most recently included study, 47 participants (mean age 12.3 years) with CF were randomized to receive either cycloxygenase-2 inhibitor nimesulide or clarithromycin (Shmarina 2004). Diagnosis of CF was confirmed by sweat chloride test, clinical symptoms and/or detection of mutations in both CFTR alleles.
Three of the trials examined ibuprofen compared with placebo (Konstan 1991; Konstan 1995; Lands 2007), while a fourth reported the use of piroxicam (Sordelli 1994). In the first Konstan trial reporting ibuprofen, children were enrolled in a randomized double-blind placebo-controlled three-month dose escalation trial (Konstan 1991). This trial examined the pharmacokinetics of ibuprofen, to identify the optimal dose and to evaluate adverse effects of ibuprofen treatment in children treated over a three-month treatment period (Konstan 1991). This first trial contributed to the design of the second trial from this center, which examined the long-term use of ibuprofen given over a four-year period (Konstan 1995). The most recent trial from Canada is the biggest trial of NSAIDs in a CF population, following 142 children aged 6 to 18 years over a period of two years (Lands 2007). In the trial of piroxicam, participants were assigned to piroxicam for a period of between 12 and 19 months (Sordelli 1994).
In each trial the NSAID was compared with placebo. In the first Konstan trial, 13 children were randomized to receive ibuprofen and five to receive placebo, both of which were given for between one and three months (Konstan 1991). The dose of ibuprofen was 300 mg twice daily in the first month, increasing to 400 mg in the second month and 600 mg in the third month, if peak plasma concentrations of ibuprofen were less than 50 µg/mL (Konstan 1991). In the second Konstan trial, 42 participants were randomized to ibuprofen and 43 to placebo (Konstan 1995). All participants underwent a baseline pharmacokinetics trial employing 20 mg tablets at a dose of 20 to 30 mg/kg to a maximum of 1600 mg. The number of assigned pills was adjusted to provide a peak plasma concentration of 50 to 100 µg/mL. The participants took the prescribed dose twice daily and were followed for four years. This dose was revised if body weight increased by more than 25% during the trial and was reviewed in all participants during the third year to take account of increased body weight (Konstan 1995). In the double-blind Canadian trial, participants were allocated treatment by a centralized pharmacy, using a predefined block randomization schedule (Lands 2007). Like Konstan's second ibuprofen trial, all participants underwent a baseline pharmacokinetics trial employing 200 mg tablets at a dose of 20 to 30 mg/kg to a maximum of 1600 mg. The number of assigned pills was adjusted to provide a peak plasma concentration of 50 to 100 μg/mL for each participant in the trial. The participants took the prescribed dose twice daily and were followed bi-annually for two years (Lands 2007). In the piroxicam trial, a single morning dose of between 5 to 20 mg, determined by participant's body weight, was given (Sordelli 1994). One trial compared cycloxygenase-2 inhibitor (3 mg/kg body weight) nimesulide to clarithromycin (250 mg every other day) in 47 participants (Shmarina 2004).
Of the outcomes specified in the prior hypotheses of the review FEV
Chest X-ray (Brasfield) scores were reported in three trials (Konstan 1995; Lands 2007; Sordelli 1994). In two trials these were summarized as the number showing improvement or deterioration (Konstan 1995; Lands 2007), rather than a specific score as in the piroxicam trial (Sordelli 1994). Chest x-ray was not reported by Shmarina (Shmarina 2004).
Change in weight was reported in two trials (Konstan 1995; Sordelli 1994). Konstan reported per cent ideal body weight and its change over four years (Konstan 1995). Although weight gain was evaluated in the piroxicam trial, this was grouped into a clinical status summary; and summaries of weight change were not provided (Sordelli 1994). The Canadian trial reported nutritional status as weight and BMI z scores (Lands 2007). Change in weight was not reported by Shmarina (Shmarina 2004).
Antibiotic use was reported in two trials (Konstan 1995; Sordelli 1994), but the data were summarized differently: as average days oral antibiotic treatment per participant per year in the piroxicam trial (Sordelli 1994); and as change in the percentage using oral, inhaled or intravenous antibiotics over the period of the Konstan trial (Konstan 1995). The trans-Canada trial reported use of oral or inhaled antibiotics (Lands 2007). Antibiotic use was not reported by Shmarina (Shmarina 2004).
Of the adverse outcomes specified, abdominal pain was reported in four trials; epistaxis (nose bleeds) in two trials (Konstan 1991; Konstan 1995); conjunctivitis in two trials (Konstan 1995; Lands 2007) occult blood, decrease in stool frequency in one trial (Konstan 1995); nausea, diarrhoea, gastritis, epigastric pain, reflux esophagitis, hepatitis, vomiting tinnitus, reactive arthritis, elevated liver enzymes and gastro-intestinal bleeds in one trial (Lands 2007). None of the trials reported haemorrhagic stroke, dyspepsia, rash, fever, bronchospasm or fluid retention. Adverse events were not reported by Shmarina (Shmarina 2004).
Dropout rates and number of participants completing the trial were reported in four trials. Compliance was only reported in the ibuprofen trials, in which it was monitored by pill counts or plasma levels or both (Konstan 1991; Konstan 1995; Lands 2007). Deaths were reported in the piroxicam trial only (Sordelli 1994). Dropout rate was not reported by Shmarina (Shmarina 2004).
One trial (with two reports) was ineligible for inclusion as it reported a single-dose challenge with aspirin (Noritake 1982). A second trial identified in 2008 was ineligible as it only followed participants for a period of four weeks (Chmiel 2007). A third trial identified in 2010 was excluded as it did not employ an oral NSAID (Kovaleva 2000).
Risk of bias in included studies
The method used to generate the random sequence and the concealment of allocation were assigned to one of three categories: adequate; unclear; or inadequate; these corresponded to a low, unclear or high risk of bias respectively.
Generation of sequence
The method used to generate the random sequence was adequate in all three ibuprofen trials (Konstan 1991; Konstan 1995; Lands 2007). In the earlier Konstan trial, randomisation was based upon a computer-generated randomisation sequence (Konstan 1991). In the later Konstan trial, randomisation was carried out with permuted blocks of four participants each stratified by age (under 13 years, 13 to 18 years and 19 years or over) (Konstan 1995). In the Lands trial, participants were allocated using a predefined block-randomisation schedule (Lands 2007). This corresponded to a low risk of bias for generation of the randomisation sequence in each of these trials.
In the piroxicam trial, while the process was described as random in the paper, the randomisation method was unclear (Sordelli 1994). This led to the risk of bias for generation of sequence being judged to be unclear.
In the trial by Shmarina, participants were randomized, but the generation of sequence was unclear; therefore we judged the risk of bias for generation of sequence to be unclear (Shmarina 2004).
Concealment of allocation
Concealment of allocation was adequate in all three ibuprofen trials (Konstan 1991; Konstan 1995; Lands 2007). In the earlier Konstan trial, the randomisation sequence was provided by the pharmaceutical company (Konstan 1991). In the later Konstan trial, only the pharmacologist and pharmacist were privy to the allocation (Konstan 1995). In the Canadian trial, a central pharmacy coded and shipped the tablets to the participating centers; the code was broken by the central pharmacy only on request from the Safety and Monitoring Committee (Lands 2007). This corresponds to a low risk of bias in these trials due to the concealment of allocation.
Reporting of double blinding was recorded as either present or absent.
Four trials were described as double blinded. In the earlier Konstan trial the pharmaceutical company provided the clinics with identical-appearing placebo tablets (Konstan 1991). In the later Konstan trial, the placebo tablets were again identical in appearance to the ibuprofen tablets (Konstan 1995). Lands states that participants, care-givers and study personnel were all blinded to treatment assignment (Lands 2007). Sordelli describes the placebo tablets as being "indistinguishable" from the piroxicam tablets (Sordelli 1994). Therefore we judged the risk of bias from blinding to be low in four of the five included trials.
The trial by Shmarina was not blinded, and we judge the risk of bias to be high (Shmarina 2004).
Incomplete outcome data
The reported 'intention-to-treat' analysis was defined as complete; or analysed with less than 15% of participants excluded; or analysed with more than 15% of participants excluded.
In two of the ibuprofen trials, analysis was based on intention-to-treat (Konstan 1995; Lands 2007), while in the other there were less than 15% of participants excluded (three participants) (Konstan 1991). These three withdrawals were due to poor venous access, behavioural problems and difficulty in transport to follow up trial visits. We therefore judged the risk of bias due to incomplete outcome data to be low in these three trials.
More than 15% of participants were excluded from the intention-to-treat analysis (n = 8) (Sordelli 1994). Four participants from the treatment group and four from the control group did not complete the study. Reasons for exclusion included abdominal pain, hematemesis, hepatic dysfunction and acute respiratory exacerbation. Risk of bias was also low as withdrawals described and were equal across groups.
It was not clear whether analysis was done on the basis of intention-to-treat for the Shmarina trial, therefore we defined the risk of bias from incomplete outcome data unclear for this trial (Shmarina 2004).
In two studies we did not identify any selective reporting as all the listed outcomes measured were reported in the results section of the paper (Lands 2007; Sordelli 1994). Thus we judged these trials to be at a low risk of bias from selective reporting. It should be noted that in Konstan's ibuprofen trials, the trial investigators monitored a large number of potential adverse effects of ibuprofen but reporting was confined to those considered to be most important and findings which were not statistically significant were not reported; we therefore judge there to be a high risk of bias from selective reporting in these two trials (Konstan 1991; Konstan 1995). Due to a lack of information, selective reporting was judged unclear in the trial by Shmarina (Shmarina 2004).
Other potential sources of bias
The later Konstan trial was funded by the Cystic Fibrosis Foundation and the National Institutes of Health (Konstan 1995) and the Canadian trial specifically states that the funders did not have a role in the analysis or publication of results (Lands 2007).
Effects of interventions
Oral non-steroidal anti-inflammatory drug versus placebo
1. Objective measures of lung function
In the second Konstan trial and the Canadian trial measures of lung function were summarized as the within-participant annual rates of change over the trial period (Konstan 1995; Lands 2007). In the Konstan trial differences in rates of change for FVC, FEV
a. Forced expiratory volume in one second (FEV
We were able to combine data from two ibuprofen trials in the meta-analysis (Konstan 1995; Lands 2007). Overall, the meta-analysis results show a statistically significant positive influence of high-dose ibuprofen on the rate of change of per cent predicted FEV
Overall, the post-hoc analysis of data from two trials split by age showed a statistically significant slower rate of annual decline of per cent predicted FEV
b. Forced vital capacity (FVC)
Again, data from two trials could be combined in the meta-analysis (Konstan 1995; Lands 2007). Overall there was a significant difference in annual rate of change per cent predicted FVC in favour of NSAIDs, MD 1.27% (95% CI 0.26 to 2.28) ( Analysis 1.3). Individually, in the second Konstan trial the mean annual decline favoured the ibuprofen group versus the placebo group, but the result was not significant, MD 0.99% (95% CI -0.43 to 2.41) (Konstan 1995). The Canadian trial found a significant difference in the annual rate of change of per cent predicted FVC in favour of the ibuprofen group, MD 1.55% (95% CI 0.12 to 2.98) (Lands 2007).
As with the FEV
c. Forced expiratory flow 25-75% (FEF
Similar trends were observed for the annual rate of decline for per cent predicted FEF
1. Treatment with intravenous antibiotics
Intravenous antibiotic usage was reported in only one trial (Konstan 1995). In this trial, a higher percentage of participants in the ibuprofen arm had received intravenous antibiotics in the 12 months preceding randomisation compared with the placebo group (27% versus 14% respectively). In the final 12 months (Year Four) of the trial, this percentage was similar in the ibuprofen arm (29%), but markedly increased in the placebo arm (37%). Within the trial report, changes in the frequency of concomitant therapy with intravenous antibiotics were compared using categorical-data modelling, a method which cannot be replicated here without access to individual patient data. Thus a data table was not compared for this outcome, although from this one trial long-term use of high dose ibuprofen appears to be associated with less use of intravenous antibiotic usage over a 12-month period. Inhaled antibiotic use was not recorded in the Canadian Trial, and no differences were observed in the concomitant use of inhaled and oral antibiotics (Lands 2007). Participants treated with piroxicam did not require more antibiotics than those treated with placebo in the Sordelli trial (Sordelli 1994). Antibiotic use was not reported in the first Konstan trial (Konstan 1991).
2. Number of hospital admissions for respiratory exacerbations
Hospital admissions for respiratory exacerbations were only reported in the Canadian trial: 26% (placebo) versus 20% (ibuprofen); however, the difference between the two groups was not statistically significant Peto OR 0.75 (95% CI 0.34 to 1.65) (Lands 2007) ( Analysis 1.7).
The percentage of participants requiring at least one hospital admission for any reason during the trial period were reported for all four trials (Konstan 1991 Konstan 1995; Lands 2007; Sordelli 1994). In the first Konstan trial, one participant in the placebo group was hospitalised (Konstan 1991). In the second ibuprofen trial by Konstan, 21 participants (51%) in the ibuprofen arm were hospitalised during the course of the trial, compared with 26 (60%) in the control arm (Konstan 1995). In the Canadian trial admission rates were lower in the ibuprofen group 18 (27%) versus the placebo group 25 (36%), but the difference was not statistically significant (Lands 2007). In the piroxicam trial, seven participants (35%) in the piroxicam arm were admitted to hospital compared with 11 (52%) in the control arm (Sordelli 1994). Combining these results in the analysis, overall there was no significant difference in the number of admissions in the ibuprofen group compared to the placebo group, Peto OR 0.64 (95 % CI 0.39 to 1.05) ( Analysis 1.8). However, the studies were heterogeneous with respect to type of NSAIDs assessed, criteria for inclusion, treatment duration (and hence period at risk for hospital admission), as well as methodological quality. Data were not reported by age group.
This analysis does not take into account the number of admissions per participant or the duration of each admission. These data were presented in three trials but in a different format in each, so that the information provided could not be combined (Konstan 1995; Lands 2007; Sordelli 1994). In one ibuprofen trial, the mean number of hospital days per participant and mean number of admissions per hospital patient were summarised by treatment arm (Konstan 1995). Both were lower in the ibuprofen compared with the control group, but were statistically non-significant, Peto OR 0.62 (95% CI 0.31 to 1.25) (Konstan 1995). In the piroxicam trial, median duration of admission was reported to be greater among the control group compared with the intervention group (18 days versus one day respectively). However, this analysis did not take account of individual variation in number and duration of admissions (Sordelli 1994).
3. Number of days admitted to hospital for respiratory exacerbations
Similarly the number of days spent in hospital for respiratory exacerbations was not reported separately within any trial. Although the number of days spent in hospital for any reason was reported in all four trials, data could not be combined due to the differing summary measures used. In the second Konstan trial, there was no difference reported in number of hospitalisations or days spent in hospital between the groups (Konstan 1995). This did not change when divided into subgroups by age. The Canadian trial reported the absolute number of hospital days in each group 561 (placebo) versus 248 (ibuprofen), as well as the hospitalization rate adjusted for overdisposition; 4.1 days per year (placebo) compared to 1.8 days per year (ibuprofen). Post-hoc analysis of hospitalization data split by age revealed a significant age trend with older participants having more hospital admissions than younger participants (Lands 2007). In the Sordelli trial, the number of hospital days per patient days at risk was based on a completed treatment analysis (Sordelli 1994). The placebo-treated group required 301 hospital days of a possible 8609 patient days at risk, while the piroxicam group had 192 of 7920 risk days hospitalised (42% decrease), including one participant admitted four times for a total of 90 days.
4. Quality of life measures
These were not reported in any trial.
5. Proportion surviving at six months and one year
Deaths were explicitly reported in only one trial which recruited participants with more severe pulmonary disease than in the other two trials (Sordelli 1994). In this trial, two participants (one in each treatment group) died as a consequence of respiratory exacerbations with resistant strains of P. aeruginosa (Sordelli 1994). The timing of these deaths in relation to enrolment in the trial was not stated: hence data tables for the proportion surviving to six months and one year in this study could not be reported.
Although not explicitly reported in the ibuprofen trials, the principal author of two of the trials has confirmed that no death occurred and that all participants were accounted for in these two trials (Konstan 1999). No deaths were reported in the Canadian Trial (Lands 2007) ( Analysis 1.9).
6. Nutritional status
Only the second ibuprofen trial reported change in percentage ideal body weight (Konstan 1995). Ideal body weight was determined on the basis of age, sex and height. Body weight as a percentage of ideal declined among those treated with placebo, whereas those treated with ibuprofen showed no or very little decline, OR 0.99 (95% CI 0.17 to 1.81) ( Analysis 1.10). This effect was most marked among those aged less than 13 years; for each year the percentage ideal body weight declined by 1.45% (95% CI 0.33, 2.57) more among control children compared to those treated with ibuprofen ( Analysis 1.11).
Body mass index and z scores of weight-for-height were only measured in the Canadian trial; neither of these outcomes changed in the two group over the study period and absolute numbers were not reported (Lands 2007).
Although weight gain was incorporated into an overall assessment of clinical status, it was not reported separately in the piroxicam trial (Sordelli 1994).
7. Chest X-ray scores
Chest X-ray scores were reported in two trials (Konstan 1995; Lands 2007). Overall, the Brasfield chest X-ray score declined more among the placebo relative to the ibuprofen group MD 0.37 (95% CI -0.08 to 0.81), but this decline was not statistically significant ( Analysis 1.12). The results for the individual trials were also not significant; MD 0.53 (95% CI -0.02 to 1.08) (Konstan 1995) and MD 0.07 (95% CI -0.68 to 0.82) (Lands 2007). When the chest X-ray score was stratified by age in the Konstan trial, there was no statistical difference between the ibuprofen and control group in either age group, MD 0.45 (95% CI -0.24 to 1.14) and MD 0.63 (95% CI -0.30 to 1.56) for participants under 13 years and 13 years and over respectively (Konstan 1995) ( Analysis 1.13).
8. Adverse effects
Of the adverse outcomes specified, abdominal pain was reported in all trials; epistaxis (nose bleeds) in two trials (Konstan 1991; Konstan 1995); conjunctivitis in two trials (Konstan 1995; Lands 2007) occult blood, decrease in stool frequency in one trial (Konstan 1995); nausea, diarrhoea, gastritis, epigastric pain, reflux esophagitis, hepatitis, vomiting tinnitus, reactive arthritis, elevated liver enzymes and gastro-intestinal bleeds in one trial (Lands 2007). None of these data were reported by age group, so it is not possible to assess whether age is an important factor in determining the frequency of adverse effects attributable to NSAIDs. Of the adverse effects listed in the protocol, data were either not available or were not explicitly reported for any of the four eligible trials for the following outcomes: major haemorrhage (i.e. life-threatening gastrointestinal bleeding or acute haemorrhagic stroke); allergic reactions such as rash, fever or fluid retention.
Abdominal pain is a recognised complication of CF as well as of treatment with NSAIDs. Overall, in all three ibuprofen trials, a greater proportion in both treatment groups reported a decrease in abdominal pain (Konstan 1991; Konstan 1995; Lands 2007); while more participants in the piroxicam trial reported an increase in abdominal pain (Sordelli 1994). We were able to combine data from two trials looking at the increase in abdominal pain and the analysis showed a non-significant result in favour of ibuprofen, Peto OR 0.54 (95% CI 0.20 to 1.48) ( Analysis 1.14).
Stool frequency was reported in one trial and shown to be less among those treated with ibuprofen, Peto OR 2.92 (95% CI 0.44 to 19.25); however the confidence intervals were very wide (Konstan 1991) ( Analysis 1.17).
Other adverse events
There were no significant differences between treatment and control groups in the proportion of participants with occult blood in the stools, Peto OR 0.36 (95% CI 0.04 to 3.62) (Konstan 1991) ( Analysis 1.18); or in the proportions reporting epistaxis, increase in epistaxis Peto OR 1.27 (95% CI 0.39 to4.10) ( Analysis 1.19) and decrease in epistaxis Peto OR 2.08 (95% CI 0.21 to 20.57) (Konstan 1995) ( Analysis 1.20); or conjunctivitis, increase in conjunctivitis Peto OR 0.72 (95% CI 0.22 to 2.40) ( Analysis 1.21) and decrease in conjunctivitis Peto OR 0.69 (95% CI 0.11 to 4.17) (Konstan 1995; Lands 2007) ( Analysis 1.22).
The Canadian trial reported on the increase in nausea, Peto OR 2.02 (95% CI 0.21 to 19.78) ( Analysis 1.23) and the increase in diarrhoea, Peto OR 1.03 (95% CI 0.06 to 16.62) ( Analysis 1.24), neither of which were significant (Lands 2007).
Prophylactic use of medications to protect intestinal mucosa was not routinely recommended in any of the studies, except in the latter part of the Canadian study (Lands 2007). The use of antacids or H2-receptor antagonists was similar in both groups during the second Konstan trial (Konstan 1995).
In Konstan's ibuprofen trials, the trial investigators monitored a large number of potential adverse effects of ibuprofen: reporting was confined to those considered to be most important and findings which were not statistically significant were not reported (Konstan 1991; Konstan 1995).
Oral non-steroidal anti-inflammatory drug versus macrolide antibiotics
1. Objective measures of lung function
a. Forced expiratory volume in one second (FEV
In the trial by Shmarina the changes in per cent predicted FEV
b. Forced vital capacity (FVC)
In the cycloxygenase-2 inhibitor nimesulide group the mean (SD) pre-intervention changes in FVC were -4.3 (1.6) and 5.9 (2.1) post intervention (Shmarina 2004). In the clarithromycin group the mean (SD) pre-intervention changes for FEV
There is increasing biological and scientific evidence to suggest that inflammation is a major contributor to declining lung function in people with CF. The effectiveness of long-term treatment with anti-inflammatory agents such as NSAIDs in preventing this decline is therefore of major clinical relevance. Given that measures of lung function, particularly FEV
Based on data from 226 participants from these two trials, there is moderate evidence to support the use of ibuprofen twice daily for slowing the progression of CF lung disease. The use of ibuprofen can slow the rate of decline in lung function in people with mild CF lung disease. The additional results do not allow for further comment on an effect on nutritional status, other than what is reported in the Konstan trial (Konstan 1995). Overall there were fewer hospitalisations and improved chest x-ray scores in the ibuprofen group, but the results were not significant. The rate of change of percentage of ideal body weight was greatest in the under 13 year group receiving NSAIDs, and for the cohort as a whole. Unfortunately, these trials are too small and follow up is too short to determine whether treatment with NSAIDs increases survival. Nonetheless, pulmonary function as assessed by FEV
The trials reporting treatment with ibuprofen employed a high dose, designed to achieve mean peak plasma concentrations of greater than 50 µg/mL, which have been associated with maximal anti-inflammatory effect. A potentially narrow therapeutic window for ibuprofen may exist as experimental work in animals and humans suggests that low plasma concentrations may be associated with a paradoxical increase in inflammation (Konstan 1990; Konstan 2003). Careful pharmacological monitoring was employed in the ibuprofen trials to maintain peak plasma concentrations in the desired range. Further studies are required to determine whether other NSAIDs are associated with this biphasic response.
One additional point concerning ibuprofen is the need to repeat pharmacokinetic analyses. The current programs using ibuprofen suggest repeating these assays at least every two years but more frequent follow-up is necessary if there is a weight change of more than 25%. It is suggested that hematologic, renal, and hepatic status are monitored annually.
Monitoring the effects of therapeutic treatment relies on appropriate detection of important clinical changes in lung function. Since a biological marker for identifying lung function changes has yet to be identified, follow-up studies require recruitment of a sufficiently large sample, followed repeatedly over a period deemed adequate to detect important clinical changes which are a direct result of the therapeutic intervention.
Conventional therapy for CF is an evolving practice that requires constant evaluation of new treatments and therapies. To date, the study of NSAIDs to slow down the progression of lung deterioration has relied solely on the comparison of a placebo group on a background of other accepted therapies. To our knowledge, there have been no comparisons between different anti-inflammatory therapies or other medications, such as azithromycin, which may have beneficial anti-inflammatory effects (Southern 2004).
Since NSAIDs are considered a safe medication in children, they have become increasingly available as an over-the-counter preparation to young children; however, there is little known about the effects of using high doses for a prolonged period of time. Confirming previous studies, the additional information provided by the most recent Canadian trial suggests that high-dose ibuprofen did not result in an increase in abdominal pain, nausea, or diarrhoea (Lands 2007). The most recent trial did, however, report one case of significant gastrointestinal haemorrhage which led to the introduction of routine prophylactic prescription of H2-blockers or proton ion pump inhibitors to protect intestinal mucosa. The effectiveness of this protective strategy has not been documented. Furthermore, four cases of transient renal failure among children being treated concurrently with ibuprofen and intravenous aminoglycosides have been reported, emphasising the importance of adequately assessing adverse effects in the context of other drug therapies (Kovesi 1998). More recently, Bertenshaw et al have retrospectively shown acute renal failure among individuals who were treated concurrently with aminoglycosides and NSAIDs (Bertenshaw 2007), which provides further evidence of an interaction between NSAIDs and intravenous aminoglycosides that may have important clinical implications. Given this evidence, the use of NSAIDs should be temporarily discontinued during intravenous administration of aminoglycosides and possibly colymycin (colistin) (Bertenshaw 2007). Furthermore, individuals with esophagitis or peptic ulcer disease should not use high-dose ibuprofen. Individuals with recurrent significant hemoptysis should also not be placed on this treatment.
A recent review of people on the US Cystic Fibrosis Foundation Patient Registry was undertaken to evaluate the effect of high-dose ibuprofen (Konstan 2007). Data was evaluated from individuals with an initial age of 6 to 17 years and an FEV
The mechanisms for the beneficial effects of high-dose ibuprofen have not been clearly elucidated. Commonly the effects have been attributed to the anti-inflammatory properties of ibuprofen. However, a recent cellular study, using clinically relevant doses, suggests that ibuprofen may exert some of its effect through the activation of the cystic fibrosis transmembrane regulator (Li 2008).
An important consideration is that the existing trials have recruited participants from across a wide age span, and have included young children as well as adults with CF. The results of this review suggest that the beneficial effects of anti-inflammatory treatment with NSAIDs may be greatest among young children with mild disease, in whom fixed structural damage to the lungs has not yet occurred. The second Konstan trial used the age of 13 years as an arbitrary cut off point and found similar results with different age cut off points (Konstan 1995). The effects of NSAID treatment in children aged less than five years has not been reported. Further trials examining the effectiveness and safety of long-term use of NSAIDs and other anti-inflammatory therapies are required in young symptomatic children as well as very young pre-symptomatic children.
The current management of CF requires those affected and their families to devote a significant effort to complying with physiotherapy, and oral, inhaled or intravenous medication. The extent to which treatment with oral NSAIDs may reduce the need for concomitant treatment or hospital admission is of interest, since this may improve quality of life and reduce treatment costs for the individual and the health service. Evidence from this review suggests that there may be important reductions in days spent in hospital for all causes, but these findings require confirmation in other studies, which ideally should identify hospital admissions for respiratory exacerbations separately. Although there are methodological problems in measuring quality of life for very young children, it is feasible to measure quality of life in older children and in their parents or carers. Future trials should begin to assess these outcomes as, given the current median survival for CF, quality of life assumes greater relevance.
Implications for practice
The results of this review suggest that high-dose ibuprofen can slow the progression of lung disease in children with CF. There may also be a beneficial effect on the number of days spent in hospital and ibuprofen appears to be relatively well-tolerated. However, the long-term effects of prolonged use of high doses of NSAIDs have yet to be determined. Furthermore, gastrointestinal protection is required as are regular pharmacokinetic trials and safety profiles.
Implications for research
This review suggests that non-steroidal anti-inflammatory therapies can be beneficial for slowing down the rate of lung function decline in people with CF. Further work is required to understand the mechanism of action so that therapies with enhanced safety profiles can be developed. Future trials need to address the age at which such therapies are most effective. While pulmonary function and nutritional status are the primary outcomes of interest, evidence for a reduction in concomitant therapy, specifically intravenous antibiotic use, as well as in hospital admissions for respiratory exacerbations, should be sought, as these secondary outcome measures may have important implications for cost-effectiveness of treatment as well as quality of life. All trials should be designed to be of adequate power to reliably identify important adverse events such as, for example, major gastrointestinal haemorrhage. Multicenter trials will add to the validity of findings by enhancing their generalisability.
The clinical variability of CF suggests that it may be helpful for future updates of this systematic review to be based on individual patient data rather than on the published trial reports, as this will allow appropriate meta-analysis of within-participant changes from baseline.
We are grateful to Professor Rosalind Smyth, Jill Motley, Nikki Jahnke and the Cochrane Cystic Fibrosis and Genetic Disorders Editorial Group for their invaluable support; to Professor Carol Dezateux and Anna Crighton for their input into previous versions of this review; to Dr Michael Konstan for his helpful comments on an earlier draft of the original review; and to Dr Somnath Mukhopadhyay for his encouragement.
Data and analyses
- Top of page
- Authors' conclusions
- Data and analyses
- What's new
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Last assessed as up-to-date: 23 May 2013.
Protocol first published: Issue 2, 1997
Review first published: Issue 2, 1999
Contributions of authors
Anna Crighton had the original idea for the review, drafted the protocol, extracted data and co-authored the review. Carol Dezateux drafted the protocol and data extraction forms, extracted and analyzed data and was responsible for writing the review.
Anna Crighton and Carol Dezateux stepped down from the review as from February 2006.
Larry Lands took over as lead reviewer in May 2004. He led on the update of the review and acts as guarantor of the review.
Sanja Stanojevic joined the review in June 2006. She extracted and analyzed data for the subsequent updates and co-authored the review.
Declarations of interest
Dr Larry Lands was the lead investigator in the Trans-Canadian Trial (Lands 2007).
Sources of support
- Institute of Child Health, Ninewells Hospital and Medical School, Dundee DD1 9SY, UK.
- Tenovus, Scotland, UK.
Differences between protocol and review
There was a post-hoc subgroup analysis of the lung function data split by age (two groups: 'under 13 years at randomisation' and '13 years and over at randomisation').
Medical Subject Headings (MeSH)
Administration, Oral; Anti-Inflammatory Agents, Non-Steroidal [*administration & dosage]; Clarithromycin [administration & dosage]; Cystic Fibrosis [*drug therapy]; Ibuprofen [administration & dosage]; Piroxicam [administration & dosage]; Randomized Controlled Trials as Topic; Sulfonamides [administration & dosage]
MeSH check words
* Indicates the major publication for the study