The effect of intravenous iron supplementation on exercise capacity in iron-deficient but not anaemic patients with chronic kidney disease: study design and baseline data for a multicentre prospective double-blind randomised controlled trial

Participants in this group received a total iron dose of 1000 mg Ferinject® as a one-off infusion in 100 ml normal saline administered over a minimum of 15 min.

Participants in this group received 100 ml normal saline administered over a minimum of 15 min.

Participants were assessed at 4-weeks and offered three 1-h sessions of exercise training each week for 8 consecutive weeks, concentrating on large muscle groups of the lower limbs. Following a 6-min warm-up on the cycle ergometer (3 min of warm up at 1/3 of training workload at constant and rhythmical pace and 3 min of warm up at 1/2 of training workload), training involved a combination of cycle ergometry and resistance training, the latter performed on the isokinetic dynamometer machine Biodex, Mirion Technologies (Biodex), Inc. USA. To conclude, patients spent 6 min cooling down on the cycle ergometer (3 min of cool down at 1/3 of training workload and 3 min of cool down with very little or no resistance at slower but constant pace). Both the intensity and the duration of the exercise programme were progressively increased, as individually tolerated. Cycle ergometry was progressed from 65 to 80% of peak watts achieved during testing, determined during a graded incremental exercise test performed to peak endurance capacity, and monitored with heart rate monitors. Resistance training involved patients performing 3 sets of 10 repetitions with a 30 s rest between sets. Isokinetic knee extension at 90 degrees/sec was completed at 65–85% of the maximum torque achieved at 90 degrees flexion, which was determined using the isokinetic dynamometer. Each participant’s training programme was re-assessed and progressed at 4 weeks into the programme. The intervention was delivered by trained physiotherapy assistants, on an individual basis, in a hospital gym.

Due to Covid-19 related restrictions on study visits, and related patient concerns about travelling and attending hospital, patients were offered the option of completing these sessions of exercise training at home via an online kidney-specific exercise platform, called Kidney Beam.

https://​beamfeelgood.​com/​onDemand/​list/​kidney-disease that was available on any technological device. All participants were offered support to sign-up to the platform and engage with the resource. Those participants who did not have access to any technological device, were offered a loan of a tablet. Each participant was provided with a heart rate monitor for use during the 8-week exercise training programme, written guidance on how to access the platform, and exercise recording sheets.

Following screening, willing, and eligible participants were invited to attend a baseline visit, at which baseline investigations were performed (Table 2). The baseline/randomisation visit(s) occurred no more than four weeks after the patient had successfully passed a screening visit. The assessments were split over two visits if the participant preferred. Demographic data (date of birth, sex, ethnicity, smoking status, alcohol intake, body mass index (BMI), and primary aetiology of CKD were collected and recorded. A detailed disease history including cardiovascular co-morbidity and list of concomitant medications was taken. Vital signs were recorded including blood pressure. A physical examination was not required.

Baseline investigations included estimated glomerular filtration rate (eGFR) and biochemical profile (BCP), full blood count (FBC), serum ferritin (SF), transferrin saturation (TSAT) and C-reactive protein (CRP). Quantification of proteinuria was carried out by measurement of urinary protein:creatinine ratio (uPCR) or if diabetic, urinary albumin:creatinine ratio (uACR) levels in a spot urine sample using standard laboratory techniques.

Muscle biopsies for participants and controls were taken on a separate visit (baseline visit 2) as they needed to be collected following 1–2 days of physical inactivity. Patients from the other London sites participating in the muscle biopsy/MRI sub-study attended for the baseline visit 2 assessments at King’s College Hospital.

Qualitative interviews at entry to the trial were preferably completed on a further separate visit but could be combined with the main baseline or muscle biopsy visit if the participant preferred and was completed prior to administration of iron therapy or placebo. Interviews were conducted either face-to-face, in a quiet room separate from where the visit commenced, or via telephone depending upon the preference of the participant.

Individual semi-structured interviews were conducted to understand the impact of CKD on the lived experience, expectations of the iron and exercise interventions before the trial, and their experiences afterwards. A sample of 20 participants were interviewed at entry to and exit from the study. Participants were selected using purposive sampling to ensure varied experiences and viewpoints are represented, for example, including people in both study arms, and including people with different participation rates and response to the intervention. Interviews were conducted face-to-face or via telephone, according to participant preference, and lasted approximately one hour. Face-to-face interviews were conducted in a private room, on a separate visit or combined with the baseline visit depending on participant preference. The interviews were digitally recorded and transcribed verbatim by a professional transcription service. Thematic analysis (Braun and Clarke, 2006) was used to interpret the data and identify themes. The data was stored and managed in QSR NVivo software.

The choice of endpoint measures to be studied were chosen to provide clinical relevance. Studies that have evaluated the clinically meaningful change in the distance walked on the 6MWT in patients with heart failure [25, 26] indicate that a clinically meaningful improvement in 6MWT distance is between 32 and 45 m. Calculations from the FAIR-HF study [27] and a study in patients with CKD [28] suggested a clinically meaningful improvement in 6MWD of up to 40 m. We have therefore taken this to be the mean clinically important difference (MCID) at the 4-week primary outcome endpoint. An estimated sample size to detect a clinically meaningful difference of 40 m with a SD of 56 m (derived from the higher SD reported by Tang et al. [28]) at 80% power and 5% alpha revealed a requirement of 62 participants. We aimed to recruit at least 70 patients to enable dropouts and ensure that 4-week primary outcome data were collected for at least 62 patients.

The results will be constructed according to CONSORT. To assess the comparability of randomised groups at baseline, descriptive summaries of clinical and socio-demographic variables will be presented for each study group. Means and SD will be presented for quantitative variables, and numbers and proportions for categorical variables. No significance tests will be undertaken. If any quantitative variables are highly skewed, medians and interquartile ranges will also be presented.

Descriptive summaries will similarly be provided for each outcome variable, at both 4 weeks and 12 weeks, by study group. The proportions of participants with missing data for each outcome variable will be presented by study group. Baseline characteristics of those with missing follow-up data for key outcomes will be compared descriptively to those with complete data.

Adverse events and serious adverse events (SAE), including serious adverse reactions (SAR) and suspected unexpected serious adverse reactions (SUSAR) will be tabulated. The reasons for any withdrawals from the trial will be summarised.

Data recorded outside of these time windows will be included in the primary ITT analysis. However, a sensitivity analysis excluding these data points will be performed to assess if this makes any considerable difference to the model results.

As the stratification variable was necessarily dichotomised in the design stage, the ANCOVA will also use the dichotomised form. However, sensitivity analysis using the continuous form of the stratification variable will also be performed to ensure substantive conclusions are not highly dependent on the dichotomisation [32].

It is unlikely that there will be large variation between centres on the primary outcome. Therefore, as this is a relatively small trial with a low number of centres, and some centres with a small number of patients, it is unlikely that any small variation between centres can be accurately modelled, and so centre will not be included in the analysis as a fixed effect. However, variation in centre can be included as a random intercept as a sensitivity analysis without sacrificing many additional parameters, an approach that often has advantages over fixed effect models [33].

Questionnaire-based variables to be analysed are the Chalder Fatigue Questionnaire, KDQOL-36 [16], and the Work and Social Adjustment Scale (WSAS). The number (%) with complete data will be reported. Where questionnaire-specific guidance is available for the treatment of missing items in the construction of scales, this will be used. Where such guidance is not available, the King’s Clinical Trials Unit guideline will be followed, i.e. scales will be pro-rated for an individual if 20% or fewer items are missing. For example, in a scale with 10 items, pro-rating will be applied to individuals with 1 or 2 items missing. The average value for the 8 or 9 complete items will be calculated for that individual and used to replace the missing values.

We expect baseline 6MWD data to be nearly complete, if not fully complete. It is unlikely that excluding any patients with missing baseline data will bias treatment effect estimates or lead to much loss of precision, if any [34], and so a complete case analysis will be undertaken in such a scenario.

If there are missing data for the primary outcome of 6MWT at 4 weeks, several strategies will be considered. If missingness is below 5%, and missingness is not thought to be isolated to a particular clinically relevant subgroup, then a complete-case analysis (CCA) will be performed [35]. If there is more than 5% missing data, then alternative approaches will be used.

If such post-randomisation variables are not present, patterns of missingness in baseline variables will be assessed. If the data is thought to be MAR conditional on observed baseline variables, then a complete-case analysis (CCA) adjusted for variables predictive of missingness will be performed. Under the MAR assumption, this gives unbiased and efficient treatment effect estimates, and generally performs better than alternative methods like MI [34].

If there is evidence of an interaction being predictive of both the outcome and missingness of the outcome (e.g., see “Method for handling the impact of COVID-19” section below), then MI will be used instead of the LMM and adjusted-CCA approaches outlined above. MI has the flexibility to handle both the interaction term and the other variables predictive of missingness and gives an unbiased estimate of the average treatment effect (ATE) in the analysis model [34].

The impact of the MAR assumption in all the above approaches will be assessed via a sensitivity analysis, in which a range of plausible values for the missing outcome data will be assumed, and corresponding treatment effect ranges reported.

For the primary outcome variable (6MWT), there may be cases where the participant has recorded a value for the first test but does not have a value for the second test (which is what will be used in the analyses). As it is likely such values will be generated under a missing-not-at-random mechanism, these values will not be included in the primary analysis.

The IV iron or placebo administration comprises a single intravenous bolus, given by the unblinded research nurse at the baseline visit (or shortly after) as per the prescription. Non-adherence with treatment is therefore likely to be low, and so no per protocol/CACE analyses will be undertaken.

Residual plots in all ANCOVA models described above will be checked for normality and homoscedasticity and inspected for outliers and influential points. If any outcome measures have a highly skewed distribution, a transformation of the outcome (e.g., logarithm) may also be considered. The primary outcome (6MWT) is considered likely to be normally distributed.

If there are high degrees of heteroscedasticity then robust standard errors will be considered. If there are any strong outliers then nonparametric (e.g. re-sampling) methods will be considered as an alternative.

The characteristics of patients recruited before and after the COVID-19-related study pause will be compared, as suggested by Meyer et al., 2020 [36].

The same procedure above will also be performed for the 12-week 6MWT and other outcome variables too. The 12-week 6MWT outcome may be more likely to have an interaction between treatment arm and before/after break status, due to the change in delivery format of the exercise training taking place between weeks 4 and 12. Some of the questionnaire-based outcomes may also be associated with before/after break status.

With regards to missing data, any patients with missing outcome data due to a COVID-19-related site closure will be treated as missing completely at random (MCAR) and excluded from the final primary analysis.

We anticipate that the number of patients becoming infected with COVID-19 during the 4-week window between baseline and the primary outcome timepoint will likely be very low, and so any patients with a missing primary outcome due to COVID-19 infection will be excluded [35]. A similar rationale will also be assumed for all other outcome variables in this case.