Article Text
Abstract
Background The relationship between delayed gastric emptying and upper GI symptoms (UGI Sx) is controversial.
Objective To assess association between gastric emptying and UGI Sx, independent of treatment.
Design We performed a systematic review and meta-analysis of the literature from 2007 to 2017, review of references and additional papers identified by content expert. We included studies evaluating the association between gastric emptying and nausea, vomiting, early satiety/postprandial fullness, abdominal pain and bloating. Covariate analyses included optimal gastric emptying test method, gastric emptying type (breath test or scintigraphy) and patient category. Meta-regression compared the differences based on type of gastric emptying tests.
Results Systematic review included 92 gastric emptying studies (26 breath test, 62 scintigraphy, 1 ultrasound and 3 wireless motility capsule); 25 of these studies provided quantitative data for meta-analysis (15 scintigraphy studies enrolling 4056 participants and 10 breath test studies enrolling 2231 participants). Meta-regression demonstrated a significant difference between optimal and suboptimal gastric emptying test methods when comparing delayed gastric emptying with nausea and vomiting. On evaluating studies using optimal gastric emptying test methodology, there were significant associations between gastric emptying and nausea (OR 1.6, 95% CI 1.4 to 1.8), vomiting (OR 2.0, 95% CI 1.6 to 2.7), abdominal pain (OR 1.5, 95% CI 1.0 to 2.2) and early satiety/fullness (OR 1.8, 95% CI 1.2 to 2.6) for patients with UGI Sx; gastric emptying and early satiety/fullness in patients with diabetes; gastric emptying and nausea in patients with gastroparesis.
Conclusions The systematic review and meta-analysis supports an association between optimally measured delayed gastric emptying and UGI Sx.
- gastric emptying
- upper gastrointestinal symptoms
- gastric scintigraphy
- breath test
- association
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
The relationship between delayed gastric emptying (GE) and upper GI symptoms (UGI Sx) is controversial.
Prior meta-regression study did not discriminate quality of GE measurements and included both therapeutic and non-therapeutic studies, including those with ineffective prokinetic action.
We assessed the association between GE UGI Sx, independent of treatment, and differentiated results between studies with optimal, in contrast to suboptimal, GE tests.
What are the new findings?
Twenty five of 92 eligible studies provided quantitative data for meta-analysis (15 by scintigraphy with 4056 participants and 10 by breath test with 2231 participants).
Meta-regression demonstrated a significant difference between optimal and suboptimal GE test methods when comparing the association of delayed GE with nausea and vomiting.
For studies using optimal GE test methods, there were significant associations between GE and nausea, vomiting, abdominal pain and early satiety/fullness for patients with UGI Sx.
How might it impact on clinical practice in the foreseeable future?
These results emphasise the importance of an optimal method to measure GE in assessing patients with UGI Sx.
This analysis restores the rationale to seek and treat delayed GE in patients with UGI Sx.
Introduction
Upper gastrointestinal symptoms (UGI Sx) constitute a significant burden on the healthcare system annually.1 After excluding mucosal disease, UGI Sx are usually due to abnormalities in gastric emptying, accommodation or visceral hypersensitivity.2–5 Identifying pathophysiology may lead to individualised or precise treatments.
Although the community prevalence of delayed gastric emptying was estimated to be 37.8 women and 9.6 men per 100 000 persons,6 a telephone survey of >21 000 US adults showed that 44.9% reported one or more UGI Sx during the past 3 months, the most common being early satiety, heartburn and postprandial fullness.7
Gastric emptying of liquids and solids are influenced by the meal’s calorie and macronutrient content.8 Multiple methods9–15 are available for measurement of gastric emptying: scintigraphy, breath test, ultrasound, wireless motility capsule and MRI. The association of gastric emptying with UGI Sx is controversial, based on numerous positive9–12 or negative13 14 studies, and the overall negative meta-regression of diverse treatment effects on the association between gastric emptying and UGI Sx.15
Our aim was to systematically review the existing literature and conduct a meta-analysis to appraise the association between gastric emptying and the following UGI Sx: composite symptoms, nausea, vomiting, postprandial fullness/early satiety, abdominal pain and bloating.
Methods
This systematic review was performed according to the Cochrane Handbook for Systematic Reviews of Interventions16 and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.17 We followed an a priori established protocol. Access to the protocol is available by contacting the authors.
The study objective was to identify studies for a systematic review and meta-analysis using PRISMA guidance: patients with UGI Sx, gastroparesis and/or diabetes. Intervention was assessment of gastric emptying and UGI Sx, and the outcome was calculation of the association between individual UGI Sx and gastric emptying. The studies were conducted in the outpatient setting. There was no comparison, since the studies were all observational.
Literature search
A database search was performed in April 2017, using Ovid MEDLINE (2007 to present) including Epub ahead of print, in-process and other non-indexed citations and EMBASE (2007 to 2017 week 16). The start date was determined by the publication of consensus approaches of gastric emptying measurements by scintigraphy and breath tests (2008–2011). An expert librarian (PJE) conducted the electronic literature search with input from the investigator (online supplemental figure 1A, B). Studies were identified using a combination of subject headings and free text terms including: ‘stomach emptying’, ‘gastric emptying’, ‘gastroparesis’, ‘dyspepsia’, ‘symptom’, ‘bloating’, ‘satiety’, ‘fullness’, ‘nausea’, ‘discomfort’, ‘distress’, ‘pain’, and ‘vomiting’. The search was restricted to papers and abstracts in the English language. Further studies were added after review of the reference lists in the publications, manual searching of PubMed and discussion with an expert in this field of study (MC) for relevant articles missed by our search criteria. The search strategy is available in the online supplemental materials. Abstracts and full text review were evaluated independently by two investigators (SJ-O and PV) in order to evaluate for inclusion and exclusion criteria. Disagreements were resolved by a third reviewer (MC) who made the final judgement.
Supplementary file 1
Inclusion criteria
We included studies which directly evaluated the association between quantitative gastric emptying parameters based on any method of gastric emptying measurement, composite or individual UGI Sx or symptom scores, and any patient categories. We excluded studies involving children because of small study samples, and heterogeneous population compared with studies involving adult patients with UGI Sx.18–21 In studies involving adults, there were no exclusions based on age, sex, patient population or sample size. Therapeutic trials were also included, but the only data used were baseline characteristics, symptoms and gastric emptying. These studies included both participants with UGI Sx and normal and delayed gastric emptying.
Outcome assessment
The main outcome was to quantify the association between delayed gastric emptying and UGI Sx. We focused on six specific UGI symptoms: composite symptoms, nausea, vomiting, postprandial fullness/early satiety, abdominal pain and bloating. Composite symptom was defined by the papers. We did not combine UGI Sx to create the composite symptoms. Both outcome (UGI Sx) and predictor (gastric emptying) were divided into binary outcomes based on the study definition of positive symptoms and delayed gastric emptying. Covariates included optimal gastric emptying test method, gastric emptying type and participant population.
Major covariate: optimal measurement of gastric emptying
Previous literature has documented that the type of meal and length of gastric emptying measurement make a significant difference in the quality and accuracy of the results.22 Additionally, a consensus guideline was created to help standardise measurement of gastric emptying,23 but current literature and gastroenterology practices continue to use suboptimal gastric emptying test methods. Therefore, to assess whether there is a significant relationship between gastric emptying and UGI Sx, it is imperative to determine the relationship independently in studies with optimal versus suboptimal gastric emptying test methodology.
An optimal gastric emptying test method was defined as breath test or scintigraphy appraising the emptying of a solid meal,22 monitored for at least 3 hours,22 well-documented end points of interest, validated calculations, realistic results and absence of confounders in the interpretation of the results or association. Suboptimal gastric emptying test method was defined as gastric emptying studies that used incorrect mathematical equations to calculate gastric emptying, liquid meals, <3 hours of gastric emptying monitoring and unrealistic results, for example, Tlag >140 min.
The following criteria were applied to disqualify gastric emptying studies from the systematic review: absence of original gastric emptying measurements (T1/2, Tlag or gastric emptying rate replaced by normal vs delayed gastric emptying); and confounding by use of chronic medications that potentially alter both UGI Sx assessment and gastric emptying such as narcotics and tricyclic antidepressants.
In addition, we removed from the meta-analysis studies based on ultrasound techniques (usually only liquid meal and predominantly children) and wireless motility capsule, which typically exits the stomach after the meal has passed into the small intestine and may not be reported in similar terms (t1/2) as scintigraphy and breath tests.24 Thus, gastric emptying type was either breath test or scintigraphy.
Patient categories
When multiple publications appeared to derive from the same database with cumulated patients over time, we selected the most recent or largest patient cohort in those databases.
We combined patients with UGI Sx and functional dyspepsia because of similar symptom profile. Idiopathic, diabetic and postsurgical gastroparesis were included as gastroparesis, and type 1, type 2 and insulin-dependent type 2 diabetes were included as diabetes mellitus.
Data extraction
Data extraction was completed independently by two reviewers (PV and SJ-O). The number of patients with delayed versus normal gastric emptying with composite or individual UGI Sx was extracted. We catalogued in tables the first author’s name, year of publication, number and diagnoses of subjects studied, method of symptom assessment, type of study, method of gastric emptying measurement and delayed gastric emptying criteria.
Assessment of risk of bias
Two reviewers (PV and VC) independently determined the risk of bias based on the Newcastle-Ottawa Scale.25 UGI Sx was considered the exposure, and delayed gastric emptying was considered the outcome. Disagreements were resolved after discussion. Publication bias was assessed using Funnel plots.
Data synthesis
We narratively summarised data about the studied association from all studies in tables and presented pictorially using waterfall plots, which were subdivided based on gastric emptying type (scintigraphy or breath test) and size of study sample. In general, the left sides of the waterfall plots show the number of participants in studies showing positive association between gastric emptying and UGI Sx shown as a positive deflection and the right sides show the number of participants in studies with no positive association.26
For the meta-analysis, we performed comparisons with a DerSimonian-Laird random effects model using the software package Open Meta.27 OR and 95% CIs were either extracted or calculated from the binary data from each study. Subgroup analysis was conducted when at least two studies were available. Meta-regression analysis was completed with two objectives: first by comparison of results between gastric emptying type (breath test and scintigraphy) in order to determine if there was any difference between these two gastric emptying tests to assess association with UGI Sx, and second, by comparison of results based of optimal versus suboptimal gastric emptying test methodology. A statistically significant p value was based on <0.05 (two-tailed). We assessed heterogeneity between the studies using the I2 statistic. I2 <25%, >25% and <50% and >50% represented low, moderate and high heterogeneity, respectively that is not attributable to chance and rather reflects true differences across the studies population or setting.
We also conducted separate meta-analyses based on studies that used optimal gastric emptying test method, to determine whether there were differences in the associations detected based on the type of gastric emptying test (breath test or scintigraphy) and the patient population included in the studies (those who presented with UGI Sx, those with diabetes or studies exclusively involving gastroparesis).
Results
Literature search
The search strategy identified 393 citations (figure 1). Based on abstract and full-text review, 335 studies were removed because the study did not directly evaluate the association between UGI Sx and gastric emptying. After discussion with the content expert and reviewing article references, 34 additional studies were included for review, for a total of 92 studies included in the systematic review. The studies included predominantly prospective cohort studies. Details regarding each study are available in online supplemental tables 1–3. Waterfall plots in the supplemental material depict that the majority of the studies, 19/24 of breath test (online supplemental figure 2) and 28/49 of scintigraphy studies (online supplemental figure 3) demonstrated a positive association between UGI Sx and delayed gastric emptying. In studies with documented gastroparesis, the majority (13/17) (online supplemental figure 4) demonstrated no significant association between UGI Sx and gastric emptying.
Although 92% of the breath test studies were performed using optimal gastric emptying test methods, 10 studies4 10 11 28–34 were published from the same group with additional patients accrued over time. Three studies did not provide gastric emptying times within the manuscript, and nine papers used an inaccurate equation, artificially prolonging gastric emptying times.
Of the gastric scintigraphy studies, 19/49 used a suboptimal gastric emptying test method, including most commonly solid meal emptying measurements <3 hours and exclusive use of liquid meals. Fifteen papers were duplicates from research groups with varying participant numbers over time, nine had no gastric emptying data within the paper itself and two papers had confounding medication effect, that is the medication may have altered both sensation (and therefore symptoms) and gastric emptying times (figure 1).
Of the 17 studies in the gastroparesis group, 14 studies had optimal gastric emptying test methodology, but there were 7 duplicate studies from the same research groups over time, 3 studies without gastric emptying data available within the manuscript and 1 study with medication confounder (figure 1). Therefore, only three studies were deemed informative for the meta-analysis.
Meta-analysis
A total of 10 breath test studies with 2231 participants and 15 scintigraphy studies with 4056 participants were included (figure 1). Additional details regarding these studies are bolded in online supplemental tables 1–3. Table 1 details the number of studies for each symptom and is subdivided based on the covariates studied. Gastric emptying method, the first subdivision in the table, divides all gastric emptying studies into optimal or suboptimal test methodology, based on criteria detailed in the ’Methods' section. In the subsequent covariates, optimal gastric emptying refers to analysing that covariate in studies that performed gastric emptying test with the optimal gastric emptying test methodology. Among breath test studies, at least 50% of the studies had suboptimal gastric emptying test methodology, whereas scintigraphy studies demonstrated higher proportions of optimal gastric emptying test methodology. The specific OR and related measures of heterogeneity appear in table 2A and B. The detailed information is also listed in the online supplemental material. There was a significant association between gastric emptying and the following:
Nausea: for the full complement of 16 studies, as well as for studies using optimal gastric emptying methodology (both breath test and scintigraphy), and patients with UGI Sx and gastroparesis, but not in diabetics (figures 2 and 3).
Vomiting: for the full complement of 13 studies, as well as for studies using optimal gastric emptying methodology (both breath test and scintigraphy), and patients with UGI Sx and gastroparesis, but not in diabetics (figure 2).
Abdominal pain: for the full complement of 16 studies, as well as for studies using optimal gastric emptying methodology (for breath test only) and patients with UGI Sx when appraised using optimum gastric emptying methodology, but not in gastroparesis or diabetics (figure 3).
Bloating: for the full complement of 15 studies, as well as for studies using optimal gastric emptying methodology (for breath test only), and patients with UGI Sx (when including all gastric emptying studies) and no association with gastroparesis or diabetics (figures 2 and 3).
Early satiety/fullness: for the full complement of 17 studies, as well as for studies using optimal or suboptimal gastric emptying methodology (both breath test and scintigraphy), and patients with UGI Sx and diabetics, but not in patients with gastroparesis.
Composite symptoms: for the full complement of nine studies; since only three studies used optimal gastric emptying methodology, the associations with method and patient groups are incomplete. Meta-regression demonstrated a significant difference between studies with optimal versus suboptimal gastric emptying test methodology in studies that evaluated nausea and vomiting.
There was moderate heterogeneity with all studies among the individual UGI Sx. However, there was a significant improvement to low heterogeneity when studies were isolated to optimal gastric emptying test methodology in studies that evaluated nausea and vomiting. There was an improvement in the level of heterogeneity when analysing breath test versus scintigraphy for abdominal pain and early satiety/fullness. There was no improvement in I2 for studies evaluating bloating and composite symptoms, but there were only three studies with optimal gastric emptying test methodology for composite symptoms.
Risk of bias assessment
The Newcastle-Ottawa form used in this study is available in the online supplemental materials. Table 3 contains each study assessment of the individual components of the Newcastle-Ottawa form. Because of the types of studies included in the meta-analysis, three questions in the Newcastle-Ottawa form do not appear to closely appraise risk of bias in the studies evaluated. First, delayed gastric emptying (the predictor of interest) was likely present at the start of all studies. Second, follow-up in these studies was not necessary and sufficient in all studies. Lastly, since the exposure was UGI Sx, every study obtained this information via a validated questionnaire; however, symptoms cannot be assessed by an objective measure. Overall, the risk of bias in the included studies was moderate. Funnel plots (not shown) demonstrated no evidence of publication bias when evaluating the individual symptoms: nausea (p=0.33), vomiting (p=0.85), bloating (p=0.38) and early satiety/fullness (p=0.18).
Discussion
This systematic review and meta-analysis demonstrates that delayed gastric emptying is associated with UGI Sx (nausea, vomiting, early satiety/fullness, abdominal pain, bloating and composite symptoms). This was demonstrated quantitatively using meta-analysis (showing significant association between gastric emptying and UGI Sx in 10 breath test and 15 scintigraphy studies with a total of 6287 patients), as well as in the systematic appraisal which evaluated a larger group of studies. This association had a large magnitude (OR 2.0) except perhaps for the association with abdominal pain. In patients with documented gastroparesis, the association of delayed gastric emptying and symptoms is less clear. There may be an association with nausea and vomiting in this group of patients. Gastric emptying using optimal test methodology is an important evaluation for patients who present with UGI Sx. Data with optimal gastric emptying test methodology are limited for patients with diabetes and gastroparesis and require further evaluation before utility can be determined.
Unfortunately, about half of the studies evaluating the six UGI Sx were performed with suboptimal gastric emptying test methods. There was no significant difference in appraising the associations between gastric emptying and UGI Sx based on breath test or scintigraphy.
The observation of an association of early satiety/fullness (prominent symptoms of postprandial distress syndrome based on the nomenclature used in Rome III and IV criteria) in patients with diabetes with upper GI symptoms is consistent with recent observations of early satiety/fullness reported in a cohort of 198 patients with gastroparesis (134 idiopathic, 64 diabetic) in the NIH gastroparesis consortium study.35 Nausea and vomiting were reported to be significant symptoms of gastroparesis in a single-centre study of 157 patients with idiopathic or diabetic gastroparesis,36 and were more likely to be the symptoms prompting evaluation in diabetic than idiopathic gastroparesis in the multicentre database of the NIH Consortium.37
This meta-analysis demonstrated the importance of using optimal gastric emptying test methods in seeking the association between gastric emptying and symptoms. Except for composite symptoms and early satiety/fullness, the remaining UGI Sx had no significant associations in studies that used suboptimal gastric emptying test methodology. For example, the largest study using breath test evaluated 798 patients with functional dyspepsia or with type 1 diabetes and UGI Sx and reported that no individual symptom or severity of symptoms was associated with gastric emptying based on T1/2 estimated by 13C-octanoate breath test and a calculation that inflates the estimated gastric emptying T1/2.38 The latter calculation requires a steady state of 13CO2 breath excretion (to calculate parameter m in the formula) at the end of the 4 hour measurements of 13CO2.39 However, the reported gastric emptying values question the accuracy of the gastric emptying T1/2 data used to explore the relationship with symptoms. Thus, the mean lag time (typically defined as the time for 10% of the meal to empty) values in all groups were >100 min, and the T1/2 values were extremely high, especially in female patients with diabetes (top quartile >600 min, with mean T1/2 of 231 min).40 In patients with diabetes, we have shown that, if the 13CO2 breath excretion did not reach a steady state, the gastric emptying T1/2 is overestimated when using the same mathematical formula.41 Newer calculations have enhanced the accuracy of gastric emptying breath test measurements, and those gastric emptying T1/2 estimates have correlated well with simultaneously acquired gastric scintigraphy data.42 43
From the scintigraphy cohort, a number of studies were removed from the final analysis because of concern that medication use may have altered both gastric emptying and the ability to assess UGI Sx. For example, two scintigraphy studies from the same research group included 425 and 262 participants. After subdividing UGI Sx based on severity, the authors reported there were no significant differences in symptoms between those with normal and delayed gastric emptying. A careful appraisal of this study shows that ~45% of patients in the gastroparesis group had >30% gastric retention at 4 hours of a 2% fat, 200 kcal egg-substitute meal. Such retardation of gastric emptying is typical of patients with the most severe gastroparesis, or secondary to a drug effect. Indeed, medication use in the gastroparesis group in the study included 16% on anxiolytics, 39.2% antidepressants (15.1% tricyclic antidepressants) and 42.3% narcotics.13 These medications could certainly alter gastric emptying, and can influence the perception of UGI Sx. These dual effects of medications in such a sizeable proportion of the gastroparesis cohort could conceivably confound the ability to assess the relationship between gastric emptying and UGI Sx.
Among the participants with UGI Sx, there was a significant correlation between gastric emptying and nausea, vomiting, abdominal pain and early satiety/fullness, particularly in the studies with optimal gastric emptying methodology. This indicates that gastric emptying is a useful tool in the assessment of these symptoms. However, based on a recent review of the literature and subsequent meta-regression analysis, some may argue that identifying delayed gastric emptying, and then treating patients with delayed gastric emptying with promotility agents may not be beneficial in improving UGI Sx. After careful evaluation of the gastric emptying methodology in the meta-regression analysis,15 there were 20 scintigraphy studies and 1 breath test study that measured gastric emptying over <2 hours, 3 studies used ultrasound, 1 radiopaque markers and 4 scintigraphy and 4 breath tests measured over 3–4 hours period. Therefore, only 24% (8/33) of these studies were performed with optimal gastric emptying methodology. In fact, Janssen et al acknowledged that the results of the meta-regression analysis should be ‘interpreted with caution’.15
In the diabetes group, we only found strong association of gastric emptying with early satiety/fullness. However, this limited association of gastric emptying with symptoms may be secondary to the wide range of gastric emptying results in diabetics with upper GI symptoms. In one study, 42% of patients with diabetes had normal gastric emptying, 36% delayed gastric emptying and 22% rapid gastric emptying.44 Delayed gastric emptying is more likely than accelerated gastric emptying to be associated with nausea, vomiting, bloating, abdominal discomfort and weight loss >10 kg.5 Additionally, neuropathy may result in both alterations of gastric emptying and decreased UGI Sx sensations secondary to the sensory neuropathy causing reduced sensation as has been reported in the literature.45
Lastly, abnormal gastric accommodation may play a role in symptom generation. Indeed, in the recent report on 1287 patients evaluated in clinical gastroenterology practice at Mayo Clinic,5 21.9% of patients with UGI Sx had abnormal gastric accommodation (GA) and 21.1% had both abnormal GA and gastric emptying. Moreover, one study showed significant association of accelerated gastric emptying and UGI Sx.46
Limitations
Several studies did not provide quantitative data for meta-analysis and were narratively summarised. The exclusion of these studies would have led to a form of reporting bias with unclear direction. However, 9 of the 12 excluded scintigraphy studies were small (~70 participants or less in each study), and 2 larger scintigraphy studies with >250 participants did actually demonstrate a significant association of gastric emptying with symptoms, and yet we excluded them because of the absence of primary gastric emptying data. Therefore, we believe that this meta-analysis has provided a conservative estimate of the association of delayed gastric emptying and UGI Sx.
A second limitation is that each study labelled symptoms differently based on regional or investigator preferences, which can introduce heterogeneity. Third, the risk of bias assessment via the Newcastle-Ottawa Scale had a number of questions that did not pertain to our question and included studies. The studies overall had moderate risk of bias, inherent to observational studies. Fourth, there was significant heterogeneity that persisted for abdominal pain, bloating and composite symptoms, despite covariate analysis. This indicates that the heterogeneity may be present because of differences in symptom definition, test meal composition and lack of sufficient studies to accurately assess the association, providing continued evidence that standardisation of gastric emptying and symptom assessment is vital for future analysis. Lastly, different gastric emptying meals will have different normal ranges that need to be rigorously validated in the healthy population. Although we were thorough to ensure similar definition of UGI Sx, we accepted the author’s definition of abnormal gastric emptying in each paper without definite documentation of the validation of the cut-offs for normality being available in each study.
Conclusions
This systematic review and meta-analysis demonstrates a significant association between delayed gastric emptying and UGI Sx, particularly in patients who present with UGI Sx and functional dyspepsia. Future studies are required to assess patients with diabetes and various manifestations of gastric emptying alterations and the association of UGI Sx. Additionally, more studies with optimal gastric emptying testing methodology are required to assess if promotility agents help improve gastric emptying and UGI Sx in patients with gastroparesis. Earlier clinical trials (outside the period of the current review) with the effective prokinetic, cisapride, had demonstrated that, in patients with idiopathic gastroparesis or dyspepsia, relative changes in symptom scores correlated significantly with the relative changes in gastric emptying rates.47 Future studies using optimal gastric emptying test methods, diverse cohorts including healthy and disease groups and truly effective prokinetics are required to conclusively appraise the associations of gastric emptying response to medications and effects on UGI Sx.
Acknowledgments
The authors would like to thank Mrs Cindy Stanislav for excellent secretarial assistance.
References
Footnotes
Contributors MC: staff supervisor and senior authorship. PV: data collation, analysis and coauthorship. SJ-O: data collation and coauthorship. VC: duplicate assessment of study bias and coauthorship. PJE: library search for literature review. MHM: methodology expert and coauthorship.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent Not required.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement MC accepts full responsibility for the conduct of the study and has had access to the data and control of the decision to publish.