The PREDNOS study has not shown any clinical benefit for a 16-week EC of prednisolone compared with the 8-week SC as described by the ISKDC in UK children presenting with SSNS. There was no significant difference between the two treatment groups in time to first relapse of nephrotic syndrome or in any other of the clinically important secondary end points, including the number of relapses experienced, the proportion of participants who went on to develop FRNS or SDNS or the requirement for alternative non-corticosteroid immunosuppressive therapies. However, despite showing no clinical benefit, the cost-effectiveness analysis suggested that EC therapy may be cheaper, with the possibility of a small QALY benefit.

These findings differ from the six studies published prior to the commencement of the PREDNOS study, which had compared the ISKDC regimen with prednisolone regimens of > 3 months’ duration. A Cochrane review²⁹ of these studies performed in 2005 showed a benefit of longer-course prednisolone therapy, with a lower rate of relapse at 12–24 months, and a significant reduction in the number of children with FRNS. Based upon this, a recommendation was made that children presenting with SSNS should be treated with a minimum of 3 months of prednisolone therapy. This did not, however, lead to international consensus, and significant clinical equipoise and variation in practice persisted.

More recent studies have reported results that are consistent with PREDNOS. A Japanese study of 255 participants that compared the ISKDC regimen with a 6-month course of prednisolone found no benefit associated with longer-duration prednisolone therapy.³⁶ Yoshikawa et al.³⁶ chose a primary end point of time to the development of FRNS, a clinically important end point identifying those children who have developed a complicated disease course and who are likely to develop disease- and treatment-related morbidity and, therefore, require alternative, more potent, immunosuppressive therapies. There was no difference in FRNS, and the time to first relapse and the incidence of adverse effects were also similar in the two groups. An Indian study comparing 3 and 6 months of prednisolone did not show any benefit associated with increased duration of prednisolone,³⁵ and neither did the Dutch study by Teeninga et al.,³³ which also compared 3 and 6 months of therapy. The inclusion of three well-designed studies in a 2015 update of the Cochrane review¹³ resulted in a change in the overall conclusions. It was noted that these studies of longer versus shorter duration of corticosteroids had heterogeneous treatment effects, with the older, higher risk of bias studies tending to overestimate the effect of longer-course therapy compared with the more recently published, lower risk of bias studies. Among studies at low risk of bias, there was no significant difference in the risk of FRNS between participants given prednisone for 2 or 3 months and those receiving therapy of longer duration or higher total dose, indicating that there is no benefit of increasing the duration of prednisone beyond 2 or 3 months in the initial episode of SSNS. However, when the meta-analysis was restricted to those addressing the same question as the PREDNOS study, comparing the 8-week ISKDC regimen with regimens of ≥ 3 months (i.e. adding the study of Yoshikawa et al.³⁶ to the six studies reported in the original Cochrane review), there remained a benefit for the longer (≥ 3 months) treatment regimen, although this only just reached statistical significance. The risk of FRNS was significantly lower (RR 0.68, 95% CI 0.47 to 1.00), as was the number of participants relapsing by 12–24 months (RR 0.8, 95% CI 0.64 to 1.00). It is likely that once the results of the PREDNOS study are added to the Cochrane review that the overall result will show no difference in outcome between the ISKDC and longer treatment regimens.

The data reported in our study are similar to those reported in previous studies. The proportion of participants experiencing a relapse was 80.3% (179/223) over a median follow-up of 37 months, which is comparable to the rate of 60–90% reported in the literature.^13,87 Teeninga et al.³³ reported a relapse rate of 78.6% (99/126) in a European population with a median follow-up of 47 months, with a median time to first relapse of 6 months for the 3-month prednisolone group and 8 months for the 6-month group. Sinha et al.³⁵ reported a lower relapse rate of 57.8% (104/180); however, participants were followed up for only 12 months. In the Japanese study, the overall relapse rate was not stated; however, the median time to relapse was 242 days and 243 days in the 2-month and 6-month groups, respectively, significantly longer than the 87 days for the SC group and 139 days for EC group observed in the PREDNOS study. It is noteworthy that Yoshikawa et al. used urine dipstick values of ++ or higher as their definition of relapse, although, if anything, this would have overdiagnosed relapse and reduced the time to first relapse.³⁶

We also found rates of FRNS similar (50% and 53%) to those previously reported. Using the same ISKDC definition as used in PREDNOS, Sinha et al.³⁵ reported a rate of FRNS at last follow-up of 50.4% in the 6-month group and 60.4% in the 3-month group. The time to FRNS was 23.0 months and 17.6 months, respectively. Teeninga et al.³³ reported FRNS in 45% of the 3-month prednisolone group and in 50% of the 6-month prednisolone group, commenting that this was higher than expected. In previous studies, FRNS has been reported in 32–78% of participants who received a 2-month course of prednisolone^{15,19–21,47,88} and in 18–44% of those who received prednisolone for 3 months.^20,26,47 It has been proposed that this variation may, in part, be explained by regional differences or variations in definitions of FRNS, length of observation and relapse treatments.

Interestingly, although we found no clinical benefit for the EC prednisolone treatment regimen, we did find that this regimen was cheaper and more effective in QALY terms. The cost analysis showed that, over a 24-month period, the EC treatment regimen cost less because of a lower rate of hospital admission, a shorter duration of hospital stay, fewer hospital emergency visits and fewer outpatient and primary care visits; therefore, on average, it was cheaper than the SC treatment regimen by £1673 per patient. Furthermore, the EC treatment regimen produced more QALYs than the SC treatment regimen. Using commonly applied threshold values for how much society is willing to pay for a QALY gain, the EC treatment regimen is cost-effective. At first glance, this result may seem surprising, as the clinical outcomes have shown little or no benefit of extending prednisolone treatment, yet the heath economics reveals possible evidence of cost-effectiveness. These differences, in part, relate to the differences in cost and QALYs, but also to the different methods of analysis adopted in the health economic and clinical evaluations.

Unlike the objectives of the clinical evaluation, which are about testing whether or not extending prednisolone therapy leads to an improved patient outcome relative to a control group, the objectives of the economic evaluation are to provide an estimation of the value of the extended therapy reflecting both efficiency and equity, and, thus, an estimate of whether or not the difference in cost between the treatment groups is worth the difference in effect, taking into account the opportunity cost of that investment and the fact that the resources could have been invested elsewhere across all parts of the NHS.

The key thing to note is that small insignificant clinical benefits can be cost-effective. The health economics focus is about comparing two things: costs and effects. For the health economic analysis, the effects are measured using QALYs, which reflect societal values and incorporate preferences for domains of QoL. These values are measured using preference-based QoL instruments, for example the CHU-9D questionnaire that was used in this study. Measuring cost differences between different treatment therapies has no meaning until these are offset against differences in effects: it is the simultaneous consideration of costs and effects and, therefore, the joint density of cost and effect differences⁷⁶ that is the focus of a health economic evaluation.

Within the PREDNOS study, the clinical analysis quantified the difference in time to first relapse, at the individual participant level, using statistical inference. The economic analysis compared the per-participant cost of EC versus SC therapy, and found that the cost was, on average, £1673 lower in the EC group, and the QALYs gained were, on average, 0.0162 higher in the EC group. When the costs and QALY differences are assessed separately, these differences are not statistically significant; however, when assessed simultaneously, the ICER (the ratio of the mean cost and the mean QALY difference) produces a cost-effective result, as the EC regimen is cheaper and produces more QALYs, on average. Therefore, it is dominant, as it is not only more effective in QALY terms but also saves health-care resources, relative to the SC group.

Furthermore, there are different methods within the health economic analysis for representing the uncertainty in the cost and QALY differences. QALYs and cost data tend to have unusual distributional properties and are often skewed, exhibiting ceiling effects or having a bimodal distribution; for this reason, the stochastic bootstrapping method was applied. Bootstrapping generates multiple samples of joint cost and effect estimates from the same trial data, and these cost and effect pairs are then represented on a scatterplot on an incremental cost-effectiveness plane. Figure 15 in Chapter 6 presents the bootstrapped cost and QALY pairs from the PREDNOS study. It shows that most of the pairs lie in the south-east quadrant, indicating that there are cost-savings and QALY gains from the extended therapy versus the standard therapy; however, there are some points spread within the north-east quadrant (indicating that the extended therapy is more costly) and the south-west quadrant (indicating that the extended therapy leads to a QALY loss). This reflects some uncertainty regarding the cost savings and QALY gains to be achieved from extended therapy compared with standard therapy, which is consistent with the finding of a non-significant difference for both costs and effects, when considered independently, between the two treatment groups.

To account for the uncertainty in the cost and effect pairs, the proportions of points falling above and below a willingness-to-pay threshold line are simply counted and then the threshold line is varied to produce a CEAC. CEACs are regarded as an alternative method of calculating CIs and indicate the probability that the extended therapy is cost-effective, compared with standard therapy, for different threshold values of willingness to pay for a QALY gain. Using the PREDNOS trial data, the probability that the extended therapy is cost-effective, at the commonly applied threshold value of £20,000 per QALY, is 0.988. Therefore, despite there being no statistically significant differences in costs and effects for extended therapy compared with standard therapy, the CEAC shows that there is very little uncertainty, from a cost-effectiveness perspective, about the choice to treat patients with EC therapy compared with SC therapy. It is also worth noting that, regardless of benefit measured in QALYs, parents and children value avoidance of hospital admission. This is also valued by clinicians and reduces demand pressures on the NHS.

Previous studies have been somewhat inconsistent in their reporting of the adverse effects associated with using corticosteroids; however, the most recent (2015) Cochrane review¹³ found no significant differences in the risk of adverse events between extended duration and 2 or 3 months of prednisolone. We found no differences in the adverse effect profiles between the two treatment groups, with the exception of parentally reported poor behaviour, which was significantly more common in the SC group. At 24 months, the cumulative incidence of poor behaviour was 93% in the SC group, compared with 82% in the EC group (RR 0.90, 95% CI 0.82 to 0.98). There was no difference in the incidence of any other adverse effects including Cushingoid facies, striae, hypertrichosis, acne, increased appetite, glycosuria, cataract and abdominal pain. These findings are broadly comparable with those of multiple other larger-³⁶ and smaller-scale^20,21,23,25 trials addressing this same clinical question. The large majority of these have found no significant difference in the incidence of adverse effects; however, there was significant heterogeneity in the extent to which these were monitored. Most adverse events were transient and occurred relatively early on during the course of treatment, when the prednisolone dose was at its highest.

We were particularly interested in the impact that the two prednisolone regimens had on behaviour, as expert clinical opinion and advice from our patient and public involvement group indicated that this was the adverse effect of greatest prevalence and significance to families. When the PREDNOS study was designed, no previous study had objectively and systematically investigated this using a quantitative measure. In PREDNOS, we collected quantitative data on behaviour using the ACBC. Although parentally reported poor behaviour was significantly more common in the SC group, when behaviour was assessed objectively through the ACBC questionnaire completed by the parents, there was no significant difference in either the total behaviour score or t-score. The proportion of participants assessed as having abnormal behaviour by the ACBC was also not different between the two groups and varied between 21% and 31% at different time points throughout the study. The proportion of participants whose parents reported poor behaviour was higher than the proportion whose scores were outside the normal ACBC range. This provides some reassurance to parents that perceived poor behaviour is generally within normal bounds and is not greatly impacted by corticosteroid treatment, a finding of relevance in other paediatric conditions treated with corticosteroids. Teeninga et al.³³ assessed behaviour using visual analogue scales. Compared with baseline, participants scored significantly higher on eating, overactive behaviour and aggressive behaviour at 3 months’ follow-up (p < 0.01); however, these scores returned to baseline within 1 year in both groups. Scores for happiness temporarily dropped in the first 6 months, while scores for sleeping remained relatively stable over the entire observation period.

Subgroup analyses showed that there was no clear evidence to suggest that the treatment effect differed according to ethnicity, age or gender, although we were not powered to detect differences in subgroups. For age, there may be some suggestion that the time to first relapse was extended in those in the EC group in participants aged ≤ 5 years, with no difference between the two groups in participants aged ≥ 6 years. This remains a topic of some debate, as a number of studies have reported young age at diagnosis to be associated with an increased risk of FRNS and/or corticosteroid dependence,^8,16,47,89 whereas others have not reported this association.^12,90,91 In a post hoc analysis of Sinha et al.’s study,⁴⁰ Cox regression suggested that participants aged ≤ 3 years benefited from prolonged therapy, with a reduction in the risk of a first relapse, but not of frequent relapses, and Poisson regression confirmed a higher relative relapse rate in younger participants. Other reports have strongly argued that age may be a predictor of disease severity, including FRNS, corticosteroid dependence and response to cyclophosphamide therapy.^92,93 A few non-randomised studies have investigated the role of gender in the disease course and have reported males to be at a disadvantage.^16,47 Cox regression analysis in the study of Teeninga et al.³³ did not identify boys as being at significantly greater risk of developing FRNS. We found also no evidence of a difference in treatment effect according to gender.

Systolic and diastolic blood pressure z-scores were similar in both treatment groups throughout the course of the study. z-scores were relatively high at the time of the week 4 visit, presumably as a result of the high dose of prednisolone being administered at this point. However, the z-scores decreased progressively during study follow-up. These observations are entirely consistent with the findings of other similar studies.

Interestingly, over the course of the study, following an initial slight fall during the first 16 weeks, the height z-scores increased in both treatment groups. This is an interesting observation, given the fact that these participants received multiple courses of prednisolone for treatment of relapses, a treatment that is known to have a negative impact on linear growth. Previous studies have also reported a fall in height velocity during the first few months of high-dose prednisolone treatment, with a subsequent return to baseline by 12 months.³³ Others have described a dose-dependent effect of corticosteroids on growth in children with SSNS.^94–96 A small number of studies have noted the baseline height SDS to be relatively low in children presenting with SSNS, although no satisfactory explanation has been found for this observation.^33,97 We did not observe this in PREDNOS study. Weight z-scores were relatively constant throughout the study, and BMI z-scores decreased over time.

The main strength of the PREDNOS study is its randomised, double-blind, placebo-controlled design. This ensured a low risk of selection, performance, detection and selective reporting bias. Inclusion criteria were defined to ensure that the study population was representative of the population of children presenting with SSNS in the UK. Outcomes were assessed using internationally accepted ISKDC definitions. Our primary outcome measure of time to first relapse was felt by UK clinicians to be of clinical importance, and previous studies⁸⁸ have shown a link between timing of first relapse and subsequent clinical course. Baseline features were well balanced and there was a low rate of attrition. Regular safety assessment was ensured through regular clinical review. Prior to the commencement of PREDNOS, previous studies had been small (largest 184 participants) and no previous double-blind placebo controlled RCTs had ever been conducted in children with SSNS. PREDNOS successfully recruited 237 participants from 124 sites throughout all regions of the UK into a double-blind, placebo-controlled RCT. Since then, the studies of both Teeninga et al.³³ and Sinha et al.³⁵ were double-blinded, although one further study comparing the ISKDC regimen with longer duration therapy was not blinded.³⁶ The authors acknowledged that this may have introduced preconception bias; however, they proposed that because their design was a non-inferiority trial with regular visits and with relapses being measured objectively, they could not assume positive placebo effects. It must be remembered that this would not have been the case with the reporting of AEs, for which there is considerable scope for bias.

The sample size calculations for the study were based on detecting an absolute difference of 20% in relapse rate at 1 year from 60% in the SC group to 40% in the EC group using a log-rank test (80% power, α = 0.05). The 1-year relapse rate observed in the SC group was 77%. This means that the study has > 85% power to detect an absolute difference of 20% between groups (i.e. from 77% to 57%) using a log-rank test. This makes the likelihood of our results being the result of a type 2 statistical error small.

An additional strength of the PREDNOS study is the generalisability of its findings. We recruited participants with a first presentation of INS from across the UK with broad inclusion criteria. We selected an age range of 1–14 years as this is the range in which the large majority of patients present and are treated empirically with a course of corticosteroids without recourse to renal biopsy. One of the key purposes of the PREDNOS study was to ascertain the optimal prednisolone treatment regimen for UK children by comparing the 8-week SC ISKDC regimen with a longer 16-week EC regimen. We chose the ISKDC SC regimen for one group because this was the regimen in use in the very large majority of UK centres at the time of the planning of the PREDNOS study and chose a 16-week EC regimen as a comparator because longer duration treatment regimens have previously been shown to result in lower rates of relapse and FRNS. The ethnic mix of the study population broadly represented that of the wider population of children with nephrotic syndrome, including significant representation from the South Asian community. We recruited 44 participants (20% of the study population) from families of South Asian origin and 31 (14%) from families recorded as other non-white ethnic origin. This is a very similar figure to that reported in the study of Teeninga et al.,³³ which included 35% of participants of non-Western European descent. This is an important achievement, as the UK South Asian community in particular is significantly over-represented in the SSNS population, the incidence being around six times greater than in the UK white population. Furthermore, the UK South Asian population is generally under-represented in clinical trials and recruitment poses a number of particular challenges.⁹⁸ Finally, although formal screening logs were requested, in keeping with other studies these were not kept well; however, based on known epidemiological data, we estimate that we have managed to include 34% of newly presenting patients over a recruitment period of 3 years and 2 months. This indicates a high level of acceptability of the trial among both families and clinicians.

One of the greatest challenges in setting up the PREDNOS study was facilitating the recruitment of participants in district general hospitals. Children with INS have traditionally been, and continue to be, investigated and managed within district general hospitals rather than tertiary paediatric nephrology centres. Referral to tertiary centres generally takes place only if the presenting features are atypical or when investigation or management proves problematic. For this reason, any study that recruited solely from tertiary centres would not reach the large majority of potential participants and would risk sampling a preselected, somewhat atypical group of participants. In the early 2000s, when the PREDNOS study was being planned, there was little paediatric experience in the conduct of RCTs involving an investigational medicinal product and little funded infrastructure to support this work. Our Kidney Research UK and Kids Kidney Research jointly funded pilot study confirmed that there was great willingness among principal investigators to participate in the study and similar interest in the study from participants from both the white and South Asian communities. The pilot study allowed the trial design and infrastructure to be tested, including aspects such as the provision of study medication from a single national clinical trials pharmacy with delivery direct to the participant, attendance rates for study visits and the completion of questionnaires and the study case report forms. The success of the pilot study was significantly enhanced by the development of the NIHR Clinical Research Network: Children (formerly the NIHR MCRN), which commenced operating in 2005.⁹⁹ The PREDNOS pilot study was one of the first studies to be adopted onto the MCRN study portfolio and the infrastructure put in place facilitated study set-up and participant recruitment in many sites. The main PREDNOS study was also adopted onto the study portfolio and similarly benefited.

Possible weaknesses of the study include the possibility that the choice and preparation of study drug might have influenced the age profile of the population under study, with a potential trend towards the relative overinclusion of older participants. Because the PREDNOS study drug was supplied in crushable tablet form rather than in suspension or in a soluble or dispersible form, this may have resulted in younger children not participating in the study because of inability or perceived inability to swallow the crushed tablets. The initial ISKDC studies reported that the median age at presentation of MCD nephrotic syndrome was 3 years,² and a UK series from the county of Yorkshire reported the incidence to be greatest in the 1–4 years age group.¹ The mean age of participants in our study was 4.9 years, with 65% of the study participants being < 6 years of age. This rather suggests that there was a trend towards the recruitment of slightly older participants, perhaps as a result of this study medication formulation issue. However, our study participant age profile is broadly comparable to those reported in the three most recent RCTs of corticosteroid therapy in SSNS by Teeninga et al.³³ (median age 4.2 years, IQR 3.2 to 6.2 years), Yoshikawa et al.³⁶ (mean age 6.7 years, SD 4.1 years, in the SC group and mean age 6.3 years, SD 4.1 years, in the EC group) and Sinha et al.³⁵ (median 42.4 months in 3-month treatment group and median 44.2 months in the 6-month treatment group).

The parents of two participants recruited at the chief investigator’s site commented that they felt that they knew which group their child had been randomised to because their child had noticed slight differences in taste between the active prednisolone and placebo tablets. Prednisolone and other oral corticosteroids tend to have a somewhat bitter taste and it may be that other study participants noticed this same phenomenon. This was not, however, reported by other principal investigators and, therefore, it seems unlikely that this would have had a significant impact on the study results.

One further potential minor limitation is the fact that study participants were, in the majority of cases, observed and treated at their local hospital, where the study visits took place. As such, observation and scoring of adverse effects in the study was performed by multiple observers. To avoid this issue would have meant that all study participants would have had to travel to a single or small number of centres, which would have proved a very significant barrier to participation. Randomisation was not minimised by centre, as individual centre contributions were difficult to predict.

In designing the study, we wanted to ensure that we objectively and comprehensively collected prednisolone-related AEs to adequately compare the two treatment regimens. Earlier studies lacked consistency in the level of information that was recorded and none reported quantitative data regarding behavioural change. Our adverse event reporting was, however, somewhat less comprehensive than that of some more recent studies and this warrants some further discussion. Yoshikawa et al.³⁶ performed formal ophthalmological assessment, including measurement of intraocular pressure, which was found to be elevated in 32 out of 246 participants (13%).³⁶ In the study of Teeninga et al.,³³ participants underwent formal ophthalmological assessment at diagnosis and after 6 months, specifically looking for evidence of cataract and glaucoma.³³ No cases of glaucoma were detected and only one single case of posterior subcapsular cataract was detected in the 3-month prednisolone group. In the PREDNOS study, we did not ask for participants to have a formal ophthalmological review, although principal investigators screened participants for cataract on an annual basis. Only one case of cataract was detected in each group, a similar frequency to the single case in the study of Sinha et al.³⁵ On the basis of their observations, Teeninga et al.³³ commented that cataract and glaucoma have been reported with much greater frequency in cohorts of Japanese children than those from other races and that their findings indicate that routine ophthalmological screening was not indicated at an early stage in Dutch children.

Yoshikawa et al. additionally performed regular blood tests and detected minor abnormalities of liver function tests and plasma amylase in up to 21% of participants.³⁶ We elected not to perform regular blood tests as part of the study protocol, principally because this is not routine clinical practice in the UK; children with SSNS generally have very few blood tests performed unless they are commenced on alternative immunosuppressive therapies mandating monitoring of drug levels or adverse effects. Furthermore, we were of the opinion that the introduction of regular blood tests into the study protocol would have a negative impact on study participation. We did include a single episode of blood sampling for the purpose of collecting deoxyribonucleic (DNA) samples for a separate study aiming to identify potential genetic changes associated with SSNS. A recommendation was made that this be performed at the time of venepuncture for other clinical reasons if this occurred, although our Research Ethics Committee approval permitted us to perform a standalone venepuncture solely for this sample. Sampling was successfully performed in 173 study participants.

In the study by Teeninga et al.³³ of Dutch children, lumbar spine bone mineral density was measured using dual-energy X-ray absorptiometry (DEXA) at baseline and after 6 months. We did not perform this investigation, principally because there is little in the published literature that indicates that significant abnormality of bone mineral density is likely to occur within the first 24 months of treatment, particularly in an unselected cohort of newly presenting SSNS patients. Although our own work has reported a minor reduction in trabecular bone mineral density in adult ‘survivors’ of childhood relapsing nephrotic syndrome, such changes required the use of peripheral quantitative computerised tomography, which is not widely available and would not have been detected using DEXA alone.¹⁰⁰ Many of the district general hospitals participating in PREDNOS would have also experienced difficulties in providing DEXA services for paediatric study participants. Furthermore, a high-quality prospective study examined 60 children with INS and 195 control children and found no deficits in spine or whole body bone mineral content.¹⁰¹ In the study of Teeninga et al., no difference was detected in bone mineral density from baseline to 6 months in either group, and they were not able to achieve bone assessment in all participants.³³