Summary of findings
Description of the condition
Treating and managing loss of weight, inadequate weight gain and failure to thrive can be a challenging problem in cystic fibrosis (CF). Weight loss is a complex problem contributed to, in part, by anorexia (leading to reduced energy intake resulting in reduced nutrient absorption) and also by intestinal malabsorption. An increased resting energy expenditure, as a result of deteriorating pulmonary function and chronic sepsis, also contributes to weight loss (Elborn 1996). This results in a recurring cycle of weight loss and malnutrition, contributing to reduced lung function, a lower quality of life (QoL) and increased morbidity and mortality (Hardin 2002; Sharma 2001; Sinaasappel 2002). Furthermore, pulmonary exacerbations have important adverse effects on body protein metabolism (Shepherd 1998). Symptoms of anorexia, weight loss and tissue wasting, combined with a decrease in muscle mass and adipose tissue, are together known as anorexia-cachexia syndrome (Lopez 2004).
The importance of maintaining optimal nutrition in people with CF is well-recognised. However, the exact mechanism of anorexia in CF remains uncertain (Berenstein 2005) and there is as yet no objective method of assessing appetite in CF. Inadequate appetite tends to be diagnosed through elimination of all other contributory factors (Nasr 2008). The aetiology of anorexia is likely to be multifactorial; it may be caused in part by chronic infection due to factors such as increased mucus production and the anorectic effects of elevated serum inflammatory cytokines (Elborn 1996). Tumour necrosis factor (TNF)-α in particular may be implicated (Suter 1989). In addition, anorexia may be related to the presence of severe sinusitis, gastroesophageal reflux, and protein or energy malnutrition or both (Eubanks 2002).
Description of the intervention
Whilst appetite stimulants are prescribed, they are not currently licensed for use in CF in either adults or children. They are used as part of an array of treatment for anorexia and weight loss, but their use is controversial because of doubts about efficacy and concerns about toxicity. Research to date has consisted of small, sometimes poorly controlled studies. Multiple agents have been studied in the CF population which may have a secondary effect on appetite stimulation. These agents have a range of primary characteristics and include hormones (ghrelin, growth hormone, insulin), antihistamines (cyproheptadine (CH), pizotifen, steroids (megesterol acetate (MA), oxandrolone, prednisone), cannabinoids (dronabinol), antidepressants (mirtazapine) and antipsychotics (olanzapine). In addition to the agents which have already been studied in CF, there are potential interventions which may have implications for patients with CF since similar effects could well occur. Given this wide range of agents, a definition of an appetite stimulant for consideration in this review will be: "an agent with a biologically plausible mechanism by which it may stimulate appetite and where it is prescribed specifically for that indication". Thus we will restrict the interventions considered to be appetite stimulants for the purpose of this review to: CH; MA; oxandrolone; dronabinol; mirtazapine; pizotifen; risperidone; and olanzapine. All agents are administered orally.
This agent is used as an antihistamine and clinical recommended dosages differ in children and adults (Auden Mckenzie 2014; BNF 2014). It has been investigated for use as an appetite stimulant in CF (Homnick 2004).
Capsules of synthetic tetrahydrocannabinol (THC) (dronabinol) have been available for restricted medical use in the USA since 1985. Nabilone, a synthetic THC analogue taken orally, is the only cannabinoid licensed for prescription in the UK for the treatment of nausea and vomiting caused by chemotherapy; its use in other indications is only possible on a ‘named patient’ basis if the drug is supplied by a hospital pharmacy (EMC 2014a). Dronabinol has been shown to be effective as an oral appetite stimulant in HIV and cancer patients using doses of 2.5 mg to a maximum of 5 mg twice daily (Anstead 2003).
The progestogen steroid MA (also known as Megace
It is not very water-soluble and thus its bio-availability is low; but bio-availability is improved if MA is taken with food. Several formulations have been developed in an attempt to improve bio-availability, e.g. a micronized tablet form and a concentrated oral suspension. The most recent is an oral suspension form using nanocrystal technology, which is licensed for anorexia-cachexia in patients with AIDS.
Olanzipine is an atypical anti-psychotic drug (EMC 2014d). Whilst limited data is available, olanzapine has been used to stimulate appetite, improve BMI and improve other disease-related symptoms (e.g., eating attitudes, anxiety) in anorexia patients aged nine years and older. The dose used to stimulate appetite has been higher than in psychiatric practice (Nasr 2008).
Therapy with the anabolic steroid oxandrolone (Oxandrin
Pizotifen is an antihistamine and serotonin antagonist used to treat migraine at differing age-dependent doses (EMC 2014e).
Risperidone is indicated in the treatment of acute and chronic psychoses, and in the management of aggression in moderate to severe Alzheimer’s dementia. Recommended doses vary depending on the condition being treated and the age and weight of the patient (EMC 2014f). It is not recommended for use in children, except for conduct disorder and then only for children over five years of age.
How the intervention might work
Given that the agents are of the varying types, the mechanisms of action are varied and mostly unclear in CF patients.
Cyproheptadine hydrochloride is a serotonin and histamine antagonist approved by the Food and Drug Administration (FDA) in the USA for use in children for allergic rhinitis, allergic conjunctivitis, urticaria, dermatographism and mild angioedema. Unexplained weight gain has been observed in patients with CF who have taken CH.
It has been elucidated that MA may cause appetite stimulation and weight gain in individuals with anorexia or cachexia or both (Loprinzi 1993). The mechanism by which it does this has not been established, but, it has been shown to have the secondary effect of appetite stimulation (Homnick 2004). Increased levels of cytokines are known to be associated with anorexia and cachexia in people with cancer (Eubanks 2002). It has been reported that MA inhibits cytokines and so may be a treatment option for cachexia (Taylor 2007). Additionally, it has been hypothesized that cytokines inhibit the action of TNF on fatty tissue and its products (Marchand 2000); and that cytokines released during inflammation and malignancy act on the central nervous system to alter the release and function of a number of neurotransmitters, thereby altering both appetite and metabolic rate (Grossberg 2010). However, this has not been elucidated in CF.
Mirtazapine has noradrenergic and serotonergic enhancing properties as well antihistamine effects and a common side effect observed is appetite stimulation (Young 2000).
Olanzipine, an antipsychotic, is associated with clearly documented weight gain and adverse metabolic effects. Although increased appetite or caloric intake and various receptors, hormones and peptides, have been implicated, the biological mechanisms contributing to the increase in weight and glucose and lipid abnormalities with antipsychotics are largely unknown (Nasr 2008; Nasrallah 2003).
Oxandralone is weak oral androgen which has anabolic properties with minimal androgenic effects (Varness 2009).
Pizotifen is a sedating antihistamine which is reported to have an orexigenic effect in patients with pulmonary tuberculosis (Ohnhaus 1974). However, like CH its mechanism of action as an appetite enhancer is unclear.
Risperidone may cause weight gain due to the blockade of certain receptors, for example, 5-HT2c, that modulate appetite and body weight and associated with modest weight changes that are not dose-related. However, the mechanisms involved in drug-related weight gain for both risperidone and olanzapine are as yet uncertain (Nasr 2008; Nasrallah 2003).
Why it is important to do this review
Nutrition and weight are corner-stones of CF management. A lack of appetite has not only been reported as a common indicator of pulmonary exacerbation in CF patients (Abbott 2011), but also plays a key role in weight loss. The use of appetite stimulants in CF is controversial because of doubts concerning efficacy and also because of possible side effects. Hence the aim of the review is to establish whether appetite stimulants should be recommended in people with CF.
The aim of this review is to systematically search for and evaluate evidence on the beneficial effects of appetite stimulants in the management of CF-related anorexia and synthesize reports of any side-effects.
Criteria for considering studies for this review
Types of studies
Randomised controlled trials (RCTs) and quasi-RCTs (with no language restraints).
Types of participants
People with CF (diagnosed clinically and confirmed with sweat test or genetic testing or both) of any age, irrespective of pancreatic insufficiency or sufficiency and of any disease severity.
Types of interventions
We considered an appetite stimulant to be an agent with a biologically plausible mechanism by which it may stimulate appetite and where it is prescribed specifically for that indication (addition of this definition is a post hoc change). In light of this definition, we considered trials eligible for inclusion if they compared appetite stimulants or any agent used as an appetite stimulant to placebo, control or no treatment; different appetite stimulants and the same appetite stimulants at different doses or regimens of at least one month duration.
Types of outcome measures
- Change in body weight (kg)
- Change in body composition
- lean body mass (LBM)
- fat mass
- body mass index (BMI)
- Change in pulmonary function
- forced expiratory volume in one second (FEV
1) (absolute values)
- Subjective report of anorexia or change in appetite or both
- QoL (subjective report or measured by a validated questionnaire)
- Dietary intake
- energy intake (measured in kcal per day)
- protein intake (measured in grams of protein per day)
- Any adverse events directly related to the intervention
- Change in the number of pulmonary exacerbations
Search methods for identification of studies
Relevant trials were identified from the Group's Cystic Fibrosis Trials Register using the terms: appetite stimulant OR treatment of growth failure OR depression OR psychosis OR insulin OR anabolic steroid OR headache.
The Cystic Fibrosis Trials Register was 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 1980 and the prospective handsearching of two journals - Pediatric Pulmonology and the Journal of Cystic Fibrosis. Unpublished work was 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 last search of the Cystic Fibrosis Trials Register: 08 April 2014.
We used Healthcare Databases Advanced Search (HDAS) via www.library.nhs.uk to search MEDLINE (1950 to 01 April 2014), CINAHL (1981 to 01 May 2012) and Embase (1980 to 01 April 2014). Details of the search strategies can be found in the appendices (Appendix 1; Appendix 2; Appendix 3).
These searches were completed: 01 April 2014.
Searching other resources
We contacted the authors of conference abstracts to determine if further publications were in press, hand searched and reviewed citations and references located in articles we identified.
We also requested additional material such as unpublished further trials and negative trials from personal contacts with experts and the suppliers of appetite stimulants (Bristol-Myers Squibb Company and Actavis Mid Atlantic LLC).
Data collection and analysis
Selection of studies
Authors (RC and JD) independently screened titles and abstracts of trials identified through the searches and selected those that met selection criteria. They extracted and entered trial details in the generic study selection and data extraction form developed by the Cochrane Cystic Fibrosis and Genetic Disorders Review Group. This process encouraged adherence to the criteria for trials to be included in the review in order to avoid including those trials which were not exclusively researching agents prescribed for appetite or which may not work on appetite stimulation.
In the event of uncertainty or disagreement on study selection, the authors resolved this through discussion and consultation with a third author (DRB).
Data extraction and management
Authors independently extracted, cross-checked data reported for the outcomes listed above, discussed any differences and reached a consensus on the extracted data. If authors were unable to extract data, they reported the outcome results narratively.
The authors planned to assess outcome measures at the time-points over one and up to six months and at six-monthly intervals thereafter. In the review, the authors actually present data at three and six months, leading to a difference between the protocol and the review. However, it was not considered clinically relevant to combine the time-points at three and six months.
Assessment of risk of bias in included studies
Authors independently assessed the risk of bias for each included trial (without blinding to authorship or journal publication) following the domain-based assessment tool described in Chapter 8 of the Cochrane Handbook for Systematic Reviews of Interventions 5.1.0 (Higgins 2011). This comprised of a description and a judgement for each entry in a ‘Risk of bias’ table, where each entry addresses a specific feature of the trial. The judgement for each entry involved answering a question, with answers of low risk of bias, high risk of bias, and unclear risk of bias (indicating either lack of information or uncertainty over the potential for bias). The authors assessed the following risk of bias items: randomisation procedure; allocation concealment; blinding of investigators, participants or outcome assessors; intention-to-treat analysis, completeness of follow-up and incomplete outcome data; and selective reporting. The authors also noted details of statistical assessment such as differences in means, overall treatment effects, heterogeneity, subgroup and sensitivity analyses. The information was presented in the 'Risk of Bias' tables (Characteristics of included studies). They resolved any discrepancies by consensus and when necessary by discussion with a third author (DRB). Further details regarding the risk of bias tool are set out in the additional tables ( Table 1).
Measures of treatment effect
The authors conducted the primary analysis using the Cochrane Collaboration's software (RevMan 2011). They measured any treatment effects for dichotomous data using the odds ratio (OR) and 95% confidence intervals (CIs). They measured any treatment effects for continuous data by analysing the mean changes from baseline measures and their standard deviations (SDs) to calculate the difference in means (MD) and their 95% CIs.
Originally, if studies measured data longitudinally, the authors planned to base the analysis on the final time-point results, since methods are not yet available to carry out a meta-analysis of aggregate longitudinal data, unless individual patient data (IPD) (Jones 2005). However, when completing the data analysis, the authors decided to present all available data at selected time-points separately. Trial investigators measured data longitudinally in one included trial of MA (Eubanks 2002); the review authors reported data at each time-point independently and did not combine these.
Unit of analysis issues
When conducting the meta-analysis combining results from cross-over trials, the authors used the methods recommended by Elbourne (Elbourne 2002). Where the individual data are available, the within-participant changes and variation can be calculated directly and authors were able to include data from both arms of the trial. If the authors needed to combine data from cross-over trials with data from parallel trials in a meta-analysis, they used the weighted mean difference method discussed by Curtin (Curtin 2002a; Curtin 2002b; Curtin 2002c), where the SDs entered into the meta-analysis are adjusted to allow for within-person correlations and produce the correct standard errors.
Dealing with missing data
The review authors described the drop outs and reasons given for these in the primary papers. They contacted the original investigators if there were any missing data.
Assessment of heterogeneity
Authors considered the extent to which results of trials were consistent by considering the chi
- 0% to 40%: might not be important;
- 30% to 60%: may represent evidence of moderate heterogeneity;
- 50% to 90%: may represent evidence of substantial heterogeneity;
- 75% to 100%:may represent evidence of considerable heterogeneity (Higgins 2003)
Assessment of reporting biases
The review authors assessed publication bias by contacting authors for the trials listed as 'Awaiting classification' to seek clarification for the details of these (Epifanio 2012; Kissner 2000). Although, the authors originally planned to assess the existence of publication bias from the meta-analyses by a funnel plot, there were not sufficient numbers of trials combined (minimum of 10) to allow this. Furthermore, the authors also acknowledge that the reasons for funnel plot asymmetry extend beyond reporting bias alone, for example methodological differences or pure chance.
The review authors assessed outcome reporting bias by obtaining data from the clinical trial registry, or by comparing the 'Methods' section to the 'Results' section of the full publications and using knowledge of the clinical area. If the authors suspected outcome reporting bias, they contacted the trial investigators to firstly ascertain if they had measured and analysed the outcome and secondly to obtain the data.
The authors used a fixed-effect model in the analyses. They had planned to use a random-effects model if they had identified at least moderate heterogeneity (e.g. I
Subgroup analysis and investigation of heterogeneity
If the authors had identified clinical and statistical heterogeneity and they had included a sufficient number of trials (at least 10 trials) with enough published or reported details to allow the authors to extract data about separate participant types, they planned to undertake subgroup analyses to investigate the following further:
- different appetite stimulants;
- dosage of appetite stimulants
The authors were able to include several trials in the review; however, no single meta-analysis combined a sufficient number of data sets to allow the authors to undertake a subgroup analysis to investigate heterogeneity.
There were insufficient trials combined within any single meta-analysis to justify the use of a sensitivity analysis. If there are sufficient trials combined in future updates of this review, the authors plan to analyse data combining and splitting cross-over and parallel trials in order to test if the current findings are robust.
Summary of findings table
In a post hoc change to the protocol, a summary of findings (GRADE) table was generated to rate the quality of evidence for change in body weight (Balshem 2011).
Description of studies
Results of the search
The combined searches identified 108 titles and abstracts in total. Of the 108 references, 99 references (to 68 individual trials) were excluded from the review. There are two references (to two individual trials) listed as 'Awaiting classification' until we are able to obtain further information to allow us to categorically include or exclude these trials (we have contacted the authors of these trials for clarification); no trials are currently in progress. Therefore, nine references to three trials are included in the review.
All trials were conducted in the USA (Eubanks 2002; Homnick 2004; Marchand 2000). One was of cross-over design (Marchand 2000) and the remaining two were of parallel design (Eubanks 2002; Homnick 2004). Data from parallel and cross-over trials were combined, so adjusted SDs were used (Curtin 2002a; Curtin 2002b; Curtin 2002c). The duration of trials varied in duration from three months (Homnick 2004) to six months (Eubanks 2002).
One trial recruited only children, including pre-pubertal children (Marchand 2000), with the remaining two trials including both adults and children (Eubanks 2002; Homnick 2004). All three trials reported the gender split between females and males (Eubanks 2002; Homnick 2004; Marchand 2000). Male to female ratios were different across all trials; the male to female ratio being eight males to nine females in the Eubanks trial (Eubanks 2002), six males to 10 females in the Homnick trial (Homnick 2004) and three males to nine females in the Marchand trial (Marchand 2000).
Two of the trials looked at MA (Eubanks 2002; Marchand 2000) and one studied CH (Homnick 2004). The dosage of oral appetite stimulants varied from trial to trial; in two trials, MA was administered at a dose of 10 mg/kg/day (Eubanks 2002; Marchand 2000); and in the third study CH was administered at a dose of 4 mg four times daily (Homnick 2004). All trials used placebo treatments as comparators (Eubanks 2002; Homnick 2004; Marchand 2000).
All three trials reported on the change in body weight, change in pulmonary function and adverse events (Eubanks 2002; Homnick 2004; Marchand 2000). Two trials reported on change in body composition (Eubanks 2002; Homnick 2004); two trials reported change in appetite (Homnick 2004; Marchand 2000); and one trial reported on the change of dietary intake (Marchand 2000). One trial reported on the change in the number of pulmonary exacerbations (Eubanks 2002); and two trials reported on change in QoL (Homnick 2004; Marchand 2000) - one of these reported on QoL in the Results section even though this was not stated as planned in the Methods section of the paper (Marchand 2000).
In total 99 references to 68 individual trials were excluded from the review. Of these, in three trials participants were not diagnosed with CF or were not human (Loprinzi 1993; Rogan 2010; Weisberg 2002). A total of 26 trials did not use the researching agent primarily as an appetite stimulant, instead the agent was given for the following reasons: growth hormone therapy (Bucuvalas 2001; Darmaun 2004; Hardin 2001; Hardin 2005a; Hardin 2005b; Hardin 2006; Hutler 2002; Schibler 2003; Schnabel 2007; Stalvey 2011); insulin therapy (Ballmann 2013;Grover 2008; Minicucci 2012; Moran 2009; Moran 2001; Teeter 2004); zinc supplementation (Safai 1991); prednisone therapy (Auerbach 1985; Cohen-Cymberknoh 2008; Dovey 2007; Greally 1992; Linnane 2001; Nyamugunduru 1998; Pantin 1986; Rosenstein 1991); and treatment with ghrelin (Cohen 2008; Cohen 2010). A total of 19 trials were not RCTs or quasi-RCTs (Alemzadeh 1998; Anstead 2003; Canfield 1998; Dowsett 1999; Eubanks 2000; Guillot 2011; Nasrallah 2003; Newkirk 2000; Ohnhaus 1974; Parsons 2009; Paterson 2010; Phung 2010; Ross 2005; Sackey 1995; Stylianou 2007; Taylor 1997; Tongudai 1971; von Haehling 2009; Young 2000). Six papers were review articles (Berenstein 2005; Chinuck 2007; Chung 2006; Hardin 2007; Lopez 2004; Nasr 2008), seven were case studies or case series (Crawley 2003; Durant 1998; Hardin 1997; Hardin 2004; Nasr 1999; Stalvey 2008; Varness 2009), three were retrospective analyses of medical records (Hardin 2005c; Switzer 2009; Sykes 2006), one was a consensus document (CF Trust 2002) and one was an overview of appetite regulation (Dhillo 2007).
Risk of bias in included studies
The risk of bias in included trials is summarised in the risk of bias summary (Figure 1).
|Figure 1. Risk of bias summary: review authors' judgements about each risk of bias item for each included study.|
All the included trials referred to random allocation, ranging from a brief comment of fact that the patients were randomised to a detailed description of the sequence generation. We judged the single trial that was described as randomised, but which gave no details, as having an unclear risk of bias (Marchand 2000). The remaining two trials stated the use of a computer-generated randomisation procedure (Eubanks 2002) or more specifically SAS small block randomisation (Homnick 2004) and we judged them both as having a low risk of bias.
Two of the trials were referred to as double-blind (Eubanks 2002; Marchand 2000) and one stated that only the pharmacist investigator and trial coordinator remained unblinded (Homnick 2004). We judged all these trials as having a low risk of bias.
Incomplete outcome data
One trial had minimal drop-outs or drop-outs which were unrelated to the intervention and we classified this as having a low risk of bias (Homnick 2004). The remaining trials were judged to have a high risk of bias (Eubanks 2002; Marchand 2000). One trial had a drop-out explicitly linked to the intervention (no effect in the placebo group) and no evidence of a treatment of the missing data to reduce the bias (Eubanks 2002) and the second trial had a 50% drop-out rate (6 from 12 participants) with the missing data being excluded and no clear evidence that bias was not introduced (Marchand 2000).
The risk of reporting bias in the three included trials was judged by comparing the published methods with the reported results (Eubanks 2002; Homnick 2004; Marchand 2000). None of these trials appeared to be free from selective reporting as the outcomes stated in the Methods sections were not reported in the Results sections: dietary intake (Eubanks 2002); dietary intake and pulmonary function (Homnick 2004; Marchand 2000). In contrast, outcomes not stated in the Methods section were subsequently reported in the Results section as follows: dietary energy intake and spirometry (Homnick 2004); QoL (Marchand 2000). In addition, outcomes stated in the Methods section of the Eubanks paper, were reported using unexpected measures i.e. weight for age z-score only, instead of being additional to weight as a mean (SD) (Eubanks 2002). Furthermore, Eubanks reported LBM and fat mass for the MA group but not for the placebo group (Eubanks 2002).
Other potential sources of bias
Homnick reported significant differences in FEV
Effects of interventions
1. Change in body weight (kg)
The effect of MA compared with placebo on change in body weight was reported in one paediatric trial (Marchand 2000) and one trial of children and adults (Eubanks 2002). The effect of CH compared with placebo on change in body weight was reported in one trial of both children and adults (Homnick 2004).
Eubanks reported a significant change in weight in favour of the treatment group at both three and six months (Eubanks 2002), MD 3.00 kg (95% CI 0.92 to 5.08) and MD 3.80 kg (95% CI 1.27 to 6.33) respectively ( Analysis 1.1).
All three trials reported change in weight z score (WAZ) at three months and results were statistically significant when combined, MD 0.61 (95% CI 0.29 to 0.93) (P <0.001) ( Analysis 1.2). The quality was assessed as moderate using the GRADE approach and the heterogeneity was low (I² statistic = 0%) ( Analysis 1.2). In one cross-over trial the individual patient data for WAZ score were available from a graph (Marchand 2000). Therefore the within-subject variation could be calculated and the meta-analysis carried out using the methodology of Elbourne with the paired subject intervention and control periods used as the unit of analysis (Elbourne 2002). Only Eubanks reported WAZ at six months, MD 0.74 (95% CI 0.26 to 1.22), which is presented separately from the three-month data ( Analysis 1.2). This trial also presented results for one and two months on a graph, but not in the text or tables, with the WAZ effect size and standard error similar to the three-month value (Eubanks 2002).
The subgroup analysis of three-month data for WAZ by appetite stimulant type did not give any evidence for different effects of MA and CH on weight gain (test for subgroup difference Chi² = 0.22, df = 1, P = 0.64; I² = 0%) with both subgroups showing a significant weight gain; MA 0.68 (95% CI 0.24 to 1.13) (P = 0.003) and CH 0.53 (95% CI 0.07 to 0.99) (P = 0.02) ( Analysis 1.3).
In the nine-month paediatric trial, Marchand reported the "average" change in weight with ranges, so data could not be entered into the analysis (Marchand 2000). Marchand reported an "average" weight gain of 3.05 kg in the MA group (range 0.1 kg to 7.0 kg) versus 0.3 kg (range -0.3 kg to 0.8 kg) in the placebo group; this was significant (P = 0.04).
2. Change in body composition
None of the trials reported the change in LBM in sufficient detail to be included into a meta-analysis; and no further data were obtained when the trialists were contacted.
Eubanks reported significant increases in triceps skin-fold measurements and mid-arm circumference in the treatment group at three and six months (P < 0.01) (Eubanks 2002). Marchand also reported an improvement in LBM in the group receiving MA (Marchand 2000).
b. fat mass
Likewise, none of the trials reported the change in fat mass in sufficient detail to be included into a meta-analysis.
Eubanks reported an increase in fat mass and fat-free mass in the MA group, assessed by DEXA (P < 0.02, at three and six months) (Eubanks 2002). Marchand also reported an improvement in body fat in the MA group (Marchand 2000). In the 2004 trial, Homnick showed a significant increase fat and fat-free mass in the CH group over 12 weeks (Homnick 2004).
None of the trials reported the change in BMI in sufficient detail to be included into a meta-analysis.
In the 2004 study, Homnick showed a significant increase in BMI in the group receiving CH and no significant change in BMI for the placebo group (Homnick 2004); however, variation of the change in BMI was not given and no direct comparison between groups was carried out.
3. Change in pulmonary function
Whilst change in lung function is a primary outcome measure of appetite stimulant use, it is important to highlight that the included trials were not performed to directly affect pulmonary function; the use of appetite stimulants may take longer than any of the trials reviewed to show improvement in respiratory muscle function.
1 (absolute values)
None of the trials reported any change in absolute values of FEV
1 (% predicted)
Eubanks reported an improvement in FEV
In the 2004 paper, Homnick did not report means or SDs for the difference between baseline and follow up. The trial reported there were no significant differences in spirometric measures, but no FEV
Marchand reported FEV
1. Subjective report of anorexia or loss of appetite or both
Homnick reported that at three months five participants in the CH treatment group demonstrated increased appetite versus two participants in the placebo group; this was after assessment using part of a brief five-question questionnaire (Homnick 2004). After assessment on interview, Marchand reported that all patients demonstrated an increase in appetite whilst receiving MA (Marchand 2000).The meta-analysis for the two trials shows a statistically significant result, OR 45.25 (95% CI 3.57 to 573.33) (P = 0.003) ( Analysis 1.5).
2. QoL (subjective report or measured by a validated questionnaire)
The effect of MA compared with placebo on change in QoL was either not measured or reported (Eubanks 2002; Marchand 2000). Only the Homnick paper reported on the effect of CH versus placebo on the change in QoL. In this paper, Homnick reported less fatigue in two patients in the treatment group (Homnick 2004).
3. Dietary intake
a. energy intake (measured in kcal/day)
The effect of MA versus placebo on change in dietary energy and protein intake was reported in one paediatric trial (Marchand 2000). Marchand reported calorific intake was calculated from three-day food records, but did not differ between the treatment and placebo groups (Marchand 2000). The effect of CH versus placebo on energy intake was reported in one trial where mean caloric intake was determined by three-day food records prior to visits at 4 weeks and 12 weeks; no significant differences between groups was demonstrated (Homnick 2004).
There was insufficient detail for the data to be included in a meta-analysis.
b. protein intake (measured in grams of protein/day)
None of the trials reported protein intake results.
4. Any adverse events directly related to the intervention
Both objective and subjective adverse events were reported for both MA and CH. However, trials reported adverse effects subjectively without specifying which groups (treatment or placebo) the participants belonged to. Reported adverse effects included effects on glucose tolerance, decreased cortisol levels, increased insulin levels, insomnia, pulmonary exacerbations, blocked port-a-cath, constipation, haemoptysis and mild transient sedation; these are shown in the additional tables ( Table 2).
Eubanks reported MA to significantly decrease morning cortisol levels compared to placebo. Furthermore, bone mineral density was stated to remain stable in the MA-treated participants over the entire six-month treatment period; however, the data were not reported (Eubanks 2002). Homnick reported no significant side effects except for increased general fatigue in the CH group (Homnick 2004).
5. Change in the number of pulmonary exacerbations
Two trials reported on this outcome (Eubanks 2002; Homnick 2004). Eubanks reported that in the MA group six patients required intravenous (IV) antibiotics and four required aerosolised antibiotics (Eubanks 2002). Homnick did not report any significant differences from baseline to week 12 in oral or IV antibiotic use with CH (Homnick 2004).
Whilst intuitively an increase in appetite should result in an increase in weight, the precise relationship between the two has not been studied, and is difficult to ascertain for the following reasons. Firstly, there are no good measures of appetite per se; also, the relationship between appetite and weight gain is likely to differ significantly between individuals, and any impact of appetite on weight will be modulated by each individuals eating habits, exercise levels, and metabolic demands. That given, it seems sensible to assume that an increase in appetite should result in weight gain, but further trials are required to delineate this relationship more carefully.
We know that lung function is closely associated with nutritional status in CF and this is an independent predictor of survival (Bell 2008; Borowitz 1996; Corey 1998). However, due to the condition, most CF patients are on a high-calorie diet to help achieve normal growth and development and maintain good lung function (Nasr 2008). Achieving this energy intake from food can be difficult, and is usually not successful (Poustie 2006), but the consequences of inadequate calorie intake, i.e. anorexia, can lead to malnutrition (Nasr 2008).
Appetite stimulation and increasing food intake may be one way to try and address the anorexia. The aim of the review was to ascertain the side effects and effect of appetite stimulants on CF- anorexia, and hence explore their clinical usefulness.
Summary of main results
Response to treatment
Both MA and CH are effective in improving weight and WAZ in children and adults with CF; statistically significant increases in both weight and WAZ were seen in the appetite stimulant group compared to the placebo group at three and six months ( Analysis 1.1; Analysis 1.2). While data showed significant increases in WAZ for both MA and CH separately, there was no significant difference between the two stimulants, so it cannot be concluded that one stimulant is more effective than the other ( Analysis 1.3). However, what constitutes a clinically significant weight gain for children and adults with CF has yet to be directly investigated. Recommendations from the USA aim for a BMI of 22 and 23 in females and males respectively (Stallings 2008). In the UK, it is recommended that weight loss of more than 5% body weight for more than two months duration should be prevented and an adult BMI of less than 19 should be avoided (CF Trust 2002). In children, USA recommendations have reported weight for length should be at least the 50th percentile from birth up to two years and in children from 2 to 20 years BMI should be at least equal to the 50th percentile (Stallings 2008). Thus, the definition of significant weight gain is not clearly agreed.
The quality of weight gain is also clinically significant as correlations have been found between FEV
It is important to highlight that owing to the lack of objective markers for appetite change, the assumption has been made that improved dietary intake and body composition are indicative of an improved appetite. No RCTs have been published assessing appetite using a validated tool. Published trial data did provide some subjective support that MA and CH may improve appetite in children and adults, but the evidence was scant (Homnick 2004). In one trial, during the three months of MA treatment, there was evidence that all children reported increase in appetite whilst receiving MA; however, there was no evidence that MA acted to increase calorific intake (Marchand 2000). In a trial of CH, there was limited evidence in children and adults to support an increase in appetite (Homnick 2004). The pooled data from the two trials, which reported an increase in appetite, showed a larger proportion of patients with increased appetite in the appetite stimulant group compared to the control group, OR 45.25 (95% CI 3.57 to 573.33) (P = 0.003) ( Analysis 1.5). Whilst this was statistically significant, the change also appeared to represent a clinically significant improvement in appetite which was sustained over six months. Although the large CIs and subjective reporting of appetite change do indicate a low level of precision, they also endorse the need for further rigorous trials, particularly since the change in appetite is clinically meaningful. For clinicians to judge which appetite stimulant to use with individual patients and make an informed decision, they require information on: significant gains in appetite and weight; type; age of use; starting and maximum dosages; and the effect of ceasing stimulants.
There was no evidence that CH improved calorific intake (Homnick 2004), and so although MA and CH were both reported to improve appetite, this outcome was not reported consistently.
Adverse effects of stimulants
The adverse effects of both MA and CH were not fully determined; the only significant effect reported in the papers was transient mild sedation (Eubanks 2002). However unlike MA, which seems to display a propensity to induce glucose dysregulation, CH does not appear to affect glucose tolerance ( Table 2).
Dosage, duration and timing of appetite stimulants
Although, the data suggested weight gain may be optimised by treating patients with 10 mg/kg/day of MA (Eubanks 2002; Marchand 2000) or 4 mg four times daily of CH for three months (Homnick 2004); the available data from RCTs do not present conclusive evidence for the dose and duration of anorexia in adults and children with appetite stimulants. Long duration of treatment or the time-dependency could not be formally assessed due to insufficient data points from the RCTs and hence also the meta-analyses.
Overall completeness and applicability of evidence
All of the trials included in the analysis directly investigated the impact of the appetite stimulants MA and CH on relevant outcome measures in patients with CF. All relevant types of participants, interventions and outcomes have been investigated. However, currently there are no validated measures of appetite per se available for research purposes in CF, and therefore, any effects of these appetite stimulants seen in the reported outcome measures are at best surrogate markers of appetite. More research is required to delineate valid measures of appetite in CF, which could then be applied to the outcome measures included in this review, and to the study of appetite stimulants.
Hence, within the constraints of the current literature and research tools available, the objectives of this review have been satisfied. The effects of appetite stimulants in CF-related anorexia and any side effects reported have been rigorously evaluated.
Three other appetite stimulants, oxandrolone, dronabilol and mirtazapine, have been studied in CF. However, these trials were not included within the review because they did not meet eligibility criteria. The role of these agents in appetite stimulation in CF deserves further stringent study.
Whilst more research is required to delineate the role of appetite stimulants in CF and so inform clinical practice, at present the evidence presented within this review suggests that there is a rationale for the short-term use of MA and CH (six months). Clinicians need to be aware of potential side effects of these agents and monitor patients accordingly. The clinical benefits need to be balanced, not only against the risks of potential adverse events, but on a case-by case basis. However, there is insufficient evidence at present to recommend the use of these agents on a long-term basis (over six months).
Quality of the evidence
Trial quality was frequently sub-optimal, which may bias any observed treatment effects, on average, in the direction of over-estimating the true treatment effect.
There were areas of the reporting within the included trials that would have been greatly improved if the authors had followed the CONSORT reporting guidelines for RCTs, specifically in reporting the details of the random sequence generation, allocation concealment and blinding (Moher 2001; Moher 2003; Moher 2004). We suspect that many of the trials were correctly randomised, but the evidence was not presented in the published papers and so the risk of bias was marked as unclear.
The risk of bias of the included trials was graded as moderate ( Summary of findings for the main comparison). All three trials reported sufficient detail for a selection of outcomes to be included in the meta-analyses. All the trials used oral appetite stimulants compared to a placebo. All of the meta-analyses were of moderate quality, due to attrition bias caused by either a proportionately large drop-out rate or drop outs due to the intervention which was not corrected for in the analysis (Figure 1). Thus, whilst the data may suggest MA and CH have a potential use as appetite stimulants in children and adults due to the improvements in weight, WAZ score and appetite; this is based upon moderate quality data from a small number of trials and so cannot be conclusively recommended based upon the findings in the review.
Potential biases in the review process
The number of trials included in the review did not allow for an analysis of publication bias and so we are not able to comment on this aspect of potential bias.
In several trials it was not possible to obtain the effect size and SDs. The outcomes and SDs at baseline and follow up were reported accurately, but often the SD of the difference in outcome measure between baseline and follow up was not reported and could not be calculated from reported data. The difference can be calculated from baseline and follow-up measurements, but the SD of the difference cannot be simply calculated due to within-subject correlations. It is strongly recommended that outcome differences and SDs of differences are reported in future trials.
Trial quality was frequently sub-optimal, which may bias any observed treatment effects on average in the direction of over-estimating the true effect.
Strengths of the review were the methods used for searching, trial selection and analysis, which in our opinion did not introduce any bias.
Agreements and disagreements with other studies or reviews
The review concurs with the findings of two other reviews of appetite stimulants (Chinuck 2007; Nasr 2008); although the evidence was only reviewed systematically by Chinuck, whose review served to illuminate the potential role of appetite stimulants in the management of anorexia associated with CF (Chinuck 2007). This review concluded that larger RCTs were necessary to confirm the safety and validate the efficacy of their use in CF and also highlighted the impossibility of drawing firm long-term conclusions for the other agents or stimulants given the low numbers of patients in the trials (Chinuck 2007).
The side effects of adrenal insufficiency, testicular failure (McKone 2002) and bone metabolism (Wermers 2004) have been elucidated in the literature and we would have reported these outcomes under adverse effects in the review if the included trials had measured and documented these accordingly.
Implications for practice
The data included in this review suggest that MA and CH may be useful for short-term (i.e. six months) treatment of anorexia in adults and children with CF. However, it cannot be concluded that any one stimulant is more effective than another. The included RCTs and the meta-analyses were unable to suggest an optimal dosage, duration or timing of appetite stimulant therapy. Furthermore, if cost and availability of the agents are compared, MA prescribed at 160 mg once per day costs GBP17.83 for 30 tablets, but the most cost competitive agent is CH, 4 mg tablets prescribed once per day at a cost of GBP0.77 for 30 tablets (Chinuck 2013 [pers comm]). Although all trials of appetite stimulants reported adverse events, data were difficult to interpret because trials were underpowered to detect clinically important differences. Furthermore, the reporting of adverse events was not consistent and there were no reports on the frequency in adverse events per patient years. Despite these important limitations, the results suggested a positive effect of appetite stimulants on both weight gain and appetite.
Hence, at present, the dose of appetite stimulants and duration of therapy should fall within the short-term range (e.g. six months) as used in the trials reviewed. Clinicians need to be aware of the potential adverse effects of these medications and actively monitor patients accordingly. The clinical benefits of appetite stimulants in CF need to be balanced, not only against the risks of potential adverse events, but also considered on a case-by-case basis.
Implications for research
In order to further our understanding of the role of appetite stimulants in CF, it is first necessary to determine meaningful surrogate measures for appetite and also define what constitutes quality weight gain. This will then allow much more precise and meaningful research to be conducted into appetite stimulants. Trials should evaluate the effectiveness of MA and CH on poor appetite in CF. Further research must define the best direct method of documenting the presence of poor appetite amongst patients. Trials should use a validated measure of symptoms, and should include a disease-specific instrument for measuring poor appetite. There are, as yet, no validated scoring systems for grading appetite and the best objective measure of evaluating appetite is yet to be defined.
Since there is insufficient evidence based on RCTs to recommend any more than short-term use of appetite stimulants in CF, this review highlights the need for multicentred, adequately powered and well-designed trials to prove or disprove their potential to increase appetite safely in CF, and establish the optimal mode of treatment.
Research is further complicated by the fact that the aetiology of poor appetite may be multifactorial and not fully understood. There are also uncertainties not only about the effective duration and appropriate dose, but also the side effects of appetite stimulants in both adults and children with CF.
Gaps in the current knowledge and issues for future trials are:
These questions remain unanswered because currently available trials lack important clinical outcomes and are underpowered to detect differences in treatment effects across subgroups of patients, resulting in overall poor quality data. Future trials need to improve the quality of conducting RCTs and reporting clinically significant outcomes so they are unbiased and clear by following the CONSORT statement (Ioannidis 2004; Moher 2001; Moher 2003; Moher 2004).
We recommend that future trials:
The authors would like to acknowledge the input from the statistician Dr Graham Warren for his help in drafting the protocol and his great assistance in appropriately analysing the data presented within this review.
The authors would also like to thank Nikki Jahnke and Professor Alan Smyth for their support and also the patients who have inspired this review. We hope it will lead to more meaningful trials and make a small difference in the lives of those who suffer with CF-related anorexia.
Data and analyses
- Top of page
- Summary of findings [Explanations]
- Authors' conclusions
- Data and analyses
- Contributions of authors
- Declarations of interest
- Sources of support
- Differences between protocol and review
- Index terms
Appendix 1. MEDLINE (HDAS) search strategy (1950 to 01 April 2014)
Appendix 2. CINAHL (HDAS) search strategy (1981 to 01 May 2012)
Appendix 3. EMBASE (HDAS) search strategy (1980 to 01 April 2014)
Appendix 4. Glossary
Contributions of authors
Ruth Chinuck, Dr David Baldwin and Dr Jane Dewar assessed all trials for inclusion and completed the final version of the review. Elizabeth Hendron completed the literature search.
Ruth Chinuck will act as the guarantor for the review.
Declarations of interest
Ruth Chinuck declares no potential conflict of interest.
Jane Dewar declares no potential conflict of interest.
David Baldwin declares no potential conflict of interest.
Elizabeth Hendron declares no potential conflict of interest.
Sources of support
- Nottingham University Hospitals, City Campus, UK.
- Nottingham University, UK.
Differences between protocol and review
There were four post hoc changes to the Methods section of the review regarding data analysis.
1. We introduced a definition of an appetite stimulant to make the eligibility criteria clearer.
2. Originally, if trials had measured data longitudinally, the authors planned to base the analysis on the final time-point results (Jones 2005). However, when completing the data analysis, the authors decided to present all available data at selected time-points separately.
3. The authors originally planned to present data at over one and up to six months and at six-monthly intervals thereafter; however, it was not considered clinically relevant to combine the time-points at three and six months, hence data are presented at three, six and twelve months.
4. Although GRADE Tables were not planned as per protocol, they were generated for change in body weight by a statistician.
Medical Subject Headings (MeSH)
Anorexia [*drug therapy; etiology]; Appetite [drug effects]; Appetite Stimulants [*administration & dosage]; Body Mass Index; Cyproheptadine [*administration & dosage]; Cystic Fibrosis [*complications]; Megestrol Acetate [*administration & dosage]; Randomized Controlled Trials as Topic
MeSH check words
* Indicates the major publication for the study