Steroid avoidance or withdrawal for kidney transplant recipients

Summary of findings for the main comparison. Steroid withdrawal versus steroid maintenance for kidney transplant recipients

Steroid withdrawal versus steroid maintenance for kidney transplant recipients
Patient or population: kidney transplant recipients Intervention: steroid withdrawal Comparison: steroid maintenance
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	Steroid maintenance	Steroid withdrawal
Mortality Follow‐up: 1 year	22 per 1000	15 per 1000 (8 to 29)	RR 0.68 (0.36 to 1.3)	1913 (10)	⊕⊕⊝⊝ low^1,2
Graft loss (excluding death) Follow‐up: 1 year	32 per 1000	38 per 1000 (23 to 62)	RR 1.17 (0.72 to 1.92)	1817 (8)	⊕⊕⊝⊝ low^2,3
Acute rejection Follow‐up: 1 year	152 per 1000	268 per 1000 (182 to 396)	RR 1.77 (1.2 to 2.61)	1913 (10)	⊕⊕⊕⊝ moderate¹
NODAT Follow‐up: 5 years	57 per 1000	44 per 1000 (28 to 69)	RR 0.77 (0.49 to 1.21)	1439 (6)	⊕⊕⊝⊝ low^2,4
CMV infection Follow‐up: 5 years	100 per 1000	104 per 1000 (80 to 137)	RR 1.04 (0.8 to 1.36)	1758 (5)	⊕⊕⊝⊝ low^2,5
The assumed risk* is the baseline risk in the control group treated with steroid maintenance. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; NODAT: new‐onset diabetes after transplantation; CMV ‐ cytomegalovirus
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
¹ Most studies were unblinded (9 studies) and did not report details about random sequence generation or allocation concealment or both (8 studies). One study had inappropriate random sequence generation. Four studies were industry sponsored. ITT analysis was unclear in four. ² Total number of events were fewer than 300. ³ Most studies were unblinded (7 studies) and did not report details about random sequence generation or allocation concealment or both (6 studies). One study had inappropriate random sequence generation. Four studies were industry sponsored. ITT analysis was unclear in two. ⁴ Most studies were unblinded (5 studies) and did not report details about random sequence generation or allocation concealment or both (5 studies). Three studies were industry sponsored. ITT analysis was unclear in three studies. One study had selective outcome reporting. ⁵ Most studies were unblinded (4 studies) and did not report details about random sequence generation or allocation concealment or both (4 studies). Three studies were industry sponsored. ITT analysis was unclear in two studies. One study had selective outcome reporting.

See: Summary of findings for the main comparison Steroid withdrawal versus steroid maintenance for kidney transplant recipients; Summary of findings 2 Steroid avoidance versus steroid maintenance for kidney transplant recipients

Summary of findings 2. Steroid avoidance versus steroid maintenance for kidney transplant recipients

Steroid avoidance versus steroid maintenance for kidney transplant recipients
Patient or population: kidney transplant recipients Intervention: steroid avoidance Comparison: steroid maintenance
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Steroid avoidance versus steroid maintenance
Mortality Follow‐up: 1 year	31 per 1000	30 per 1000 (16 to 56)	RR 0.96 (0.52 to 1.8)	1462 (10)	⊕⊕⊝⊝ low^1,2
Graft loss (excluding death) Follow‐up: 1 year	42 per 1000	46 per 1000 (27 to 79)	RR 1.09 (0.64 to 1.86)	1211 (7)	⊕⊕⊝⊝ low^2,3
Acute rejection Follow‐up: 1 year	204 per 1000	323 per 1000 (221 to 470)	RR 1.58 (1.08 to 2.3)	835 (7)	⊕⊕⊕⊝ moderate⁴
NODAT Follow‐up: 5 years	107 per 1000	80 per 1000 (54 to 117)	RR 0.75 (0.51 to 1.1)	1618 (9)	⊕⊕⊝⊝ low^2,5
CMV Infection Follow‐up: 5 years	106 per 1000	101 per 1000 (74 to 138)	RR 0.96 (0.7 to 1.31)	1454 (6)	⊕⊕⊝⊝ low^2,6
The assumed risk* is the baseline risk in the control group treated with steroid maintenance. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; NODAT: new‐onset diabetes after transplantation; CMV ‐ cytomegalovirus
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate
¹ All studies were unblinded. Six studies were industry sponsored. In six studies random sequence generation or allocation concealment or both was unclear. In two studies ITT was either not performed or unclear. One study had selective outcome reporting ² Total number of events was fewer than 300 ³ All studies were unblinded. Five studies were industry sponsored. In four studies random sequence generation or allocation concealment or both was unclear. ITT was unclear in one study. One study had selective outcome reporting ⁴ All studies were unblinded. Five studies were industry sponsored. In four studies random sequence generation or allocation concealment or both was unclear. In three studies ITT was either not performed or unclear. One study had selective outcome reporting ⁵ Most studies were unblinded (8 studies). Five studies were industry sponsored. In four studies random sequence generation or allocation concealment or both was unclear. One study had selective outcome reporting ⁶ Most studies were unblinded (5 studies). Four studies were industry sponsored. One study had unclear ITT

Background

Description of the condition

Patients with end‐stage kidney disease (ESKD) have to undergo renal replacement therapy which is available either as dialysis or kidney transplantation. Kidney transplantation is the preferred treatment for eligible patients with ESKD, because it offers a nearly normal life and is associated with better survival and quality of life compared to dialysis treatment. More than 16,000 kidney transplants are currently performed annually in the USA (OPTN/SRTR 2014) and more than 12,000 in Europe (ERA‐EDTA 2013). Despite kidney transplants from live donors, organ demand exceeds organ availability worldwide and the number of patients wait listed for kidney transplantation continues to rise (ANZDATA 2012; ERA‐EDTA 2013; OPTN/SRTR 2014).

Although short‐term outcomes of kidney transplantation have continuously improved since the 1980s, long‐term results have only marginally improved until today. Death with a functioning graft and chronic allograft nephropathy are the most important causes of graft loss (Pascual 2002). Thus, strategies that prolong patient survival and graft patency have become a priority in kidney transplantation.

One of the key factors that influence transplant outcomes is immunosuppression which prohibits progressive immune mediated injury of the allograft. Standard immunosuppressive protocols nowadays consist of an initial induction treatment followed by a maintenance regimen. Immunosuppression is induced by an intensive treatment for the initial days after transplantation either with higher dosages of the immunosuppressive drugs or by adding an additional immunosuppressive agent, such as anti‐T‐cell antibodies or interleukin 2 receptor antibodies. Maintenance immunosuppression usually comprises a combination of three drug groups: calcineurin inhibitors, such as cyclosporin (CsA) or tacrolimus (TAC), anti‐proliferative agents, such as azathioprine (AZA) or mycophenolate mofetil (MMF) and corticosteroids, such as prednisolone.

Corticosteroids are long known for their anti‐inflammatory and immunosuppressive properties and have been used to prevent rejection since the early days of kidney transplantation. Although steroids are effective in preventing acute rejection, chronic steroid use may be an important cause of morbidity and mortality (Opelz 2005). Steroids exhibit a wide range of adverse effects, such as skin fragility, bodyweight gain, osteoporosis and cataracts, can adversely affect important cardiovascular and metabolic risk factors including hypertension, hyperglycaemia and dyslipidaemia and may contribute to an increased risk of infection (Coutinho 2011; Czock 2005; Matas 2005; Patel 2001). A literature review on the safety of low dose glucocorticoid treatment in rheumatoid arthritis suggested that the toxicity of steroids is overestimated, because adverse effects of chronic low dose steroid treatment (≤ 10 mg/d prednisolone equivalent) were found to be modest and rarely statistically significantly different from placebo (Da Silva 2006).

Description of the intervention

With the aim to reduce the adverse effects of long‐term corticosteroid therapy, there has been much effort to limit the exposure of kidney transplant recipients to steroids. Lessening exposure to steroids can be achieved by either steroid avoidance or steroid withdrawal. In steroid avoidance, steroids are either avoided completely or withdrawn within the first days after kidney transplantation and steroid withdrawal refers to discontinuation of steroids at a certain time point in the later post‐transplant phase. This review evaluated all steroid avoidance or withdrawal strategies in kidney transplant recipients.

How the intervention might work

Steroids show adverse cardiovascular and metabolic effects and therefore discontinuing steroid treatment may take effect by a decrease in this accelerated cardiovascular risk. However, while steroid avoidance and withdrawal potentially reduces post‐transplant atherosclerosis, ischaemic heart disease, post‐transplant diabetes and death, it may significantly increase the risk of acute rejection. Acute rejection is associated with late graft loss, especially if rejection episodes are severe, followed by impaired kidney function, occur late and affect arteries (Basadonna 1993; Massy 1996). The new immunosuppressants TAC and MMF have led to important declines in the incidence of acute rejection and may provide a more potent substrate to attempt safe steroid‐free immunosuppression or steroid withdrawal.

Why it is important to do this review

It is important to reduce the cardiovascular risk in kidney transplant recipients, who area population at increased cardiovascular risk, but at the same time it is important to avoid rejection and graft loss. Steroids have been associated with increased cardiovascular risk in kidney transplant recipients, but long‐term benefits and harms of steroid discontinuation have not yet been established with controlled long‐term data (Knight 2010). Prednisone was perceived as the least effective and least favoured immunosuppressive drug compared to calcineurin inhibitors, MMF and AZA in a survey among Canadian kidney transplant recipients and the majority of US transplant physicians and surgeons stated that steroid‐free immunosuppression was a goal for future organ transplant recipients (Hricik 2002; Prasad 2003). Steroid use varies largely in clinical practice around the globe. While steroids are discontinued in many centres worldwide, they are at the same time frequently used for long‐term treatment in kidney transplant recipients to protect the allograft. There is no consensus whether discontinuation of steroids is safe, what type of patients benefit from steroid discontinuation and at what time point after transplantation steroids are best stopped. A number of RCTs evaluating steroid avoidance or withdrawal at various time‐points after kidney transplantation with different immunosuppressive regimes have been performed during the last decades and were first systematically reviewed in 2009 (Pascual 2009). Steroid avoidance and steroid withdrawal strategies in kidney transplantation were not associated with increased patient mortality or graft loss, despite an overall higher incidence of acute rejection for steroid withdrawal strategies compared with steroid maintenance. The aim of this review was to update the benefits and harms of steroid withdrawal and avoidance in kidney transplant recipients with new evidence from RCTs.

Objectives

To evaluate the benefits and harms of steroid withdrawal or avoidance for kidney transplant recipients.

Methods

Criteria for considering studies for this review

Types of studies

All RCTs or quasi‐RCTs (in which allocation to treatment was obtained by alternation, use of alternate medical records, date of birth or other predictable methods), whether published or unpublished, in which steroids were avoided or withdrawn at any time point after kidney transplantation were eligible for inclusion. RCTs evaluating any other steroid‐sparing strategy (i.e. dose reduction) or attempting other interventions in addition to steroid withdrawal (i.e. switch from AZA to MMF, induction treatment in addition to steroid withdrawal) were excluded in this review.

Types of participants

Adult and paediatric recipients of a first or subsequent kidney transplant from a cadaveric or living donor. Recipients of multiorgan transplants (kidney‐pancreas, kidney‐liver, kidney‐heart) were excluded.

Types of interventions

Steroid avoidance, defined as steroid use during less than 14 days after kidney transplantation versus steroid maintenance
Steroid withdrawal, defined as steroid use for more than 14 days after transplantation versus steroid maintenance
Steroid avoidance versus steroid withdrawal.

Types of outcome measures

Outcome measures used by transplant registries to report patient and graft survival were selected for this review. Outcome events were assessed within the first year and up to five years after kidney transplantation. A secondary outcome looking at infection has been amended for this update to specify cytomegalovirus (CMV) infection.

Primary outcomes

All‐cause mortality
Graft loss or death with a functioning graft; and graft loss censored for death with a functioning graft (loss of graft function resulting in either return to dialysis or retransplantation)
Acute rejection (clinically suspected and treated) and biopsy‐proven acute rejection.

Secondary outcomes

Cardiovascular events
New‐onset diabetes after transplantation (NODAT)
Malignancy
Infection and CMV infection
Kidney function measures (serum creatinine (mg/dL); creatinine clearance (mL/min)).

Search methods for identification of studies

Electronic searches

We searched the Cochrane Kidney and Transplant Specialised Register up to 15 February 2016 through contact with the Information Specialist using search terms relevant to this review. The Specialised Register contains studies identified from several sources.

Monthly searches of the Cochrane Central Register of Controlled Trials (CENTRAL)
Weekly searches of MEDLINE OVID SP
Handsearching of kidney‐related journals and the proceedings of major kidney conferences
Searching of the current year of EMBASE OVID SP
Weekly current awareness alerts for selected kidney journals
Searches of the International Clinical Trials Register (ICTRP) Search Portal and ClinicalTrials.gov.

Studies contained in the Specialised Register are identified through search strategies for CENTRAL, MEDLINE, and EMBASE based on the scope of Cochrane Kidney and Transplant. Details of these strategies, as well as a list of handsearched journals, conference proceedings and current awareness alerts, are available in the Specialised Register section of information about Cochrane Kidney and Transplant.

See Appendix 1 for search terms used in strategies for this review.

Searching other resources

Reference lists of review articles, relevant studies and clinical practice guidelines.
Letters seeking information about unpublished or incomplete studies to investigators known to be involved in previous studies.

Data collection and analysis

Selection of studies

The search strategies described was used to obtain title and abstracts of studies relevant to this review. Three authors independently screened titles and abstracts, and discarded reports that were not applicable. Studies and reviews that might include relevant data or information on studies were retained initially and two authors independently assessed retrieved abstracts and, if necessary the full text, of these studies to determine which studies satisfied the inclusion criteria. Disagreement about inclusion was resolved by discussion with a third author.

Data extraction and management

Two authors independently carried out data extraction using standard data extraction forms. Studies reported in non‐English language journals will be translated before assessment. Where more than one report of a study existed, reports were grouped together and the publication with the most complete data was used in the analyses. We examined any prior or subsequent report for supplementary outcomes or data to ensure the inclusion of all relevant information. If data were unclear, ambiguous or missing, authors were contacted for further information and any provided additional data was included in the review. Whenever necessary, disagreements were resolved by discussion.

Assessment of risk of bias in included studies

Two authors independently assessed the following items using the risk of bias assessment tool (Higgins 2011) (see Appendix 2).

Was there adequate sequence generation (selection bias)?
Was allocation adequately concealed (selection bias)?
Was knowledge of the allocated interventions adequately prevented during the study?
- Participants and personnel (performance bias)
- Outcome assessors (detection bias)
Were incomplete outcome data adequately addressed (attrition bias)?
Are reports of the study free of suggestion of selective outcome reporting (reporting bias)?
Was the study apparently free of other problems that could put it at a risk of bias?

Measures of treatment effect

For dichotomous outcomes results were expressed as risk ratio (RR) with 95% confidence intervals (CI). Where continuous scales of measurement were used to assess the effects of treatment, the mean difference (MD) was used, or the standardised mean difference (SMD) if different scales had been used.

Unit of analysis issues

The unit of analysis was the study participant and not the events; that is the number of study participants with an acute rejection rather than the number of episodes of acute rejection.

Dealing with missing data

Any further information required from the original author was requested by written correspondence (e.g. emailing corresponding author) and any relevant information obtained in this manner was to be included in the review. Evaluation of important numerical data such as screened, randomised patients as well as intention‐to‐treat, as‐treated and per‐protocol population will be carefully performed. Attrition rates, for example drop‐outs, losses to follow‐up and withdrawals were investigated. Issues of missing data and imputation methods (for example, last‐observation‐carried‐forward) were critically appraised (Higgins 2011). If standard deviation was not available, it was estimated using standard error (if provided) (Higgins 2011).

Assessment of heterogeneity

Heterogeneity was analysed using a Chi² (on N‐1 degrees of freedom, with an alpha of 0.05 used for statistical significance) and with the I² statistic, calculated to measure the proportion of total variation in the estimates of treatment effect that was due to heterogeneity beyond chance (Higgins 2003). I² values of 25%, 50% and 75% correspond to low, medium and high levels of heterogeneity.

Assessment of reporting biases

We assessed publication bias by constructing funnel plots for primary outcomes if there was sufficient data available to enable this analysis (at least 10 included studies in the meta‐analysis).

Data synthesis

Data were pooled for summary estimates using the random‐effects model but the fixed‐effect model was also to be used to ensure robustness of the model chosen and susceptibility to outliers. Results reported used the random‐effects model because this is more conservative in the presence of known or unknown heterogeneity (Deeks 2001).

Subgroup analysis and investigation of heterogeneity

Subgroup analyses were used to explore possible sources of heterogeneity and potential effect modifiers were defined a priori. The main source of heterogeneity among participants could be related to age, therefore adults and children who were kidney transplant recipients were analysed separately. Heterogeneity in treatments could be related to duration of steroid therapy and concomitant immunosuppressants. Therefore subgroup analysis was undertaken using stratified meta‐analysis for type of calcineurin inhibitor, type of antimetabolite and whether an induction treatment was administered.

Sensitivity analysis

Sensitivity analysis was performed to demonstrate that final results did not vary where low quality studies were included or excluded. Low quality studies were defined based on publication type (conference abstract or peer reviewed journal) and methodological conduct (whether intention‐to‐treat analysis was assessed as adequate or inadequate/unclear).

Results

Description of studies

Results of the search

A search in 15 February 2016 identified 151 reports. Additionally three previously excluded studies were re‐evaluated and included; these had been incorrectly excluded for reasons of insufficient data (Aswad 1998; Kacar 2004; Pisani 2001). All three are published as abstract only. Pisani 2001 contributed data for the meta‐analysis. We also re‐evaluated three previously included studies and excluded them because they had been incorrectly included despite a wrong co‐intervention (CARMEN Study 2005; Tarantino 1991; ter Meulen 2002). In CARMEN Study 2005 and ter Meulen 2002 induction treatment with daclizumab was only given to patients in the steroid withdrawal group and in Tarantino 1991 AZA was given solely to patients in the steroid maintenance group. We included 21 new studies (59 reports) that involved 1854 participants, two of these new studies (seven reports) concerned children. We found that 88 new reports were additional reports of previously included studies. This update includes 48 studies (224 reports) that involved 7803 participants, including three studies (11 reports) that involved 346 children. See Figure 1.

Figure 1

Study flow diagram.

Included studies

See Characteristics of included studies.

The 48 included studies were published in 22 different journals and seven had preliminary abstract data only available (Aswad 1998; Burke 2000; del Castillo 2005; INFINITY Study 2013; Kacar 2004; Kim 2002; Pisani 2001).The effect of steroid withdrawal compared versus steroid maintenance was investigated in 26 studies (4022 participants) and the effect of steroid avoidance compared versus steroid maintenance was investigated in 19 studies (3401 participants). We identified three studies (380 participants) that evaluated the effect of steroid avoidance compared versus steroid withdrawal. Numbers of participants per study varied from 21 (Aswad 1998) to 560 patients (THOMAS Study 2002). It is noteworthy that 25 studies randomised fewer than 100 participants, 15 studies included between 100 and 300 participants, and eight studies randomised more than 300 participants.

Trials in adult kidney transplant recipients

This update included 45 studies (208 reports, 7457 participants) of steroid withdrawal or avoidance in adult kidney transplant recipients.

Participants

Trials recruited participants who were older than 18 years of age, except two studies which recruited participants older than 12 years (Stiller 1983) or between five and 62 years (Nagib 2015). In 14 studies the age range was not further specified (Albert 1985; Aswad 1998; Gulanikar 1991; INFINITY Study 2013; Isoniemi 1990; Johnson 1989a; Kacar 2004; Kim 2002; Ratcliffe 1993; Schulak 1989; Smak Gregoor 1999; Sola 2002; THOMAS Study 2002; Zhu 2008a).The majority of studies included cadaveric and living kidney transplant recipients (25 studies: Ahsan 1999; ATLAS Study 2005; Boletis 2001; Boots 2002; Burke 2000; DOMINOS Study 2012; EVIDENCE Study 2014; Farmer 2006; FREEDOM Study 2008; Gulanikar 1991; Jankowska‐Gan 2009; Kim 2002; Kumar 2005; Laftavi 2005; Lebranchu 1999; Matl 2000; Montagnino 2005; Nott 1985; Pelletier 2006; Schulak 1989; Stiller 1983; Smak Gregoor 1999; THOMAS Study 2002; Vincenti 2003a; Woodle 2005). Kidney transplantation was limited to cadaveric donor sources in 11 studies (Bouma 1996; De Vecchi 1986; FRANCIA Study 2007; Isoniemi 1990; Johnson 1989a; Maiorca 1988; Ponticelli 1997; Ratcliffe 1993; Sandrini 2009; Sola 2002; Zhu 2008a) and to living donors in four studies (Aswad 1998; Nagib 2015; Nematalla 2007; Park 1994) In 17 studies first or subsequent kidney transplant recipients were eligible (Boots 2002; Bouma 1996; DOMINOS Study 2012; EVIDENCE Study 2014; Farmer 2006; Gulanikar 1991; Johnson 1989a; Lebranchu 1999; Montagnino 2005; Nott 1985; Pisani 2001; Ponticelli 1997; Ratcliffe 1993; Schulak 1989; Stiller 1983; THOMAS Study 2002; Woodle 2005), while in 19 studies limited participants to recipients of first kidney transplants (Ahsan 1999; ATLAS Study 2005; Boletis 2001; Burke 2000; del Castillo 2005; FRANCIA Study 2007; FREEDOM Study 2008; INFINITY Study 2013; Isoniemi 1990; Kumar 2005; Laftavi 2005; Maiorca 1988; Matl 2000; Nagib 2015; Nematalla 2007; Park 1994; Pelletier 2006; Sandrini 2009; Vincenti 2003a).

Study comparisons

The 45 included studies evaluated three different comparisons in adults.

Steroid withdrawal compared versus steroid maintenance was investigated in 24/45 studies in adult patients (Ahsan 1999; Albert 1985; Aswad 1998; Boletis 2001; Bouma 1996; Burke 2000; del Castillo 2005; EVIDENCE Study 2014; Farmer 2006; Gulanikar 1991; Isoniemi 1990; Jankowska‐Gan 2009; Kacar 2004; Lebranchu 1999; Maiorca 1988; Matl 2000; Park 1994; Pelletier 2006; Pisani 2001; Ratcliffe 1993; Smak Gregoor 1999; Sola 2002; THOMAS Study 2002; Zhu 2008a). Steroids were withdrawn three months after transplantation in eight studies (Ahsan 1999; EVIDENCE Study 2014; Gulanikar 1991; Isoniemi 1990; Lebranchu 1999; Park 1994; Sola 2002; THOMAS Study 2002); six months after transplantation in eight studies (Albert 1985; Aswad 1998; Boletis 2001; Burke 2000; del Castillo 2005; Pisani 2001; Smak Gregoor 1999; Zhu 2008a); one year after transplantation in one study (Matl 2000), and beyond one year after transplantation in six studies (Bouma 1996; Farmer 2006; Jankowska‐Gan 2009; Kacar 2004; Maiorca 1988; Ratcliffe 1993). In one study, steroids were withdrawn at different time points after transplantation and the time point of withdrawal was not reported, but all patients had steroids for more than 14 days (Pelletier 2006).
Steroid avoidance compared versus steroid maintenance was investigated in 18/45 studies in adult kidney transplant recipients (ATLAS Study 2005; De Vecchi 1986; FRANCIA Study 2007; FREEDOM Study 2008; Nott 1985; INFINITY Study 2013; Johnson 1989a; Kim 2002; Kumar 2005; Laftavi 2005; Stiller 1983; Montagnino 2005; Nagib 2015; Nematalla 2007; Ponticelli 1997; Schulak 1989; Vincenti 2003a; Woodle 2005). In two studies steroids were not given at any time point before, during or after transplantation (FREEDOM Study 2008; Stiller 1983). Steroids were withdrawn until day seven after transplantation in 12 studies (ATLAS Study 2005; De Vecchi 1986; FRANCIA Study 2007; Nott 1985; Johnson 1989a; Kim 2002; Kumar 2005; Laftavi 2005; Montagnino 2005; Nematalla 2007; Ponticelli 1997; Vincenti 2003a) and between day 8 and day 14 in two studies (Schulak 1989; Woodle 2005).
Steroid avoidance was compared versus steroid withdrawal in 3/45 studies with adults (Boots 2002; DOMINOS Study 2012; Sandrini 2009). In all of these three studies, steroids were withdrawn until day seven after transplantation in the avoidance group and between three to six months after transplantation in the withdrawal group.

Immunosuppression

CsA was used in 34 studies evaluating steroid withdrawal or steroid avoidance (Ahsan 1999; Albert 1985; Boletis 2001; Bouma 1996; Burke 2000; del Castillo 2005; De Vecchi 1986; DOMINOS Study 2012; EVIDENCE Study 2014; Farmer 2006; FRANCIA Study 2007; FREEDOM Study 2008; Gulanikar 1991; INFINITY Study 2013; Isoniemi 1990; Jankowska‐Gan 2009; Johnson 1989a; Kim 2002; Kumar 2005; Lebranchu 1999; Maiorca 1988; Matl 2000; Montagnino 2005; Nott 1985; Park 1994; Pelletier 2006; Pisani 2001; Ponticelli 1997; Ratcliffe 1993; Sandrini 2009; Schulak 1989; Smak Gregoor 1999; Vincenti 2003a). TAC was used in 10 studies investigating steroid withdrawal or steroid avoidance (Aswad 1998; ATLAS Study 2005; Boots 2002; Laftavi 2005; Nagib 2015; Nematalla 2007; Sola 2002; THOMAS Study 2002; Woodle 2005; Zhu 2008a). One study provided no information about the baseline immunosuppression used (Kacar 2004). Of the three studies comparing steroid avoidance with steroid withdrawal, two used a CsA‐based immunosuppression (DOMINOS Study 2012; Sandrini 2009) and one used a TAC‐based immunosuppression (Boots 2002).

Five studies investigated steroid withdrawal compared versus steroid maintenance in patients without an additional antiproliferative immunosuppressant (either MMF or enteric‐coated mycophenolate sodium or AZA or mTOR‐inhibitor) (Albert 1985; Bouma 1996; Gulanikar 1991; Maiorca 1988; Park 1994) and five studies investigated steroid avoidance compared versus steroid maintenance without an additional antiproliferative (De Vecchi 1986; Johnson 1989a; Nott 1985; Stiller 1983; Ponticelli 1997). Steroid avoidance compared versus steroid withdrawal in patients without an antiproliferative was investigated in Boots 2002. An immunosuppressive regimen including an additional antiproliferative agent was used in 18 studies that investigated steroid withdrawal compared versus steroid maintenance (Ahsan 1999; Aswad 1998; Boletis 2001; Burke 2000; del Castillo 2005; EVIDENCE Study 2014; Farmer 2006; Isoniemi 1990; Jankowska‐Gan 2009; Lebranchu 1999; Matl 2000; Pelletier 2006; Pisani 2001; Ratcliffe 1993; Smak Gregoor 1999; Sola 2002; THOMAS Study 2002; Zhu 2008a). Of these 18 studies, 12 used MMF (Ahsan 1999; Boletis 2001; del Castillo 2005; Burke 2000; Jankowska‐Gan 2009; Pelletier 2006; Pisani 2001; Smak Gregoor 1999; Sola 2002; THOMAS Study 2002; Lebranchu 1999; Zhu 2008a), five used AZA (Aswad 1998; Farmer 2006; Isoniemi 1990; Matl 2000; Ratcliffe 1993), and one used Everolimus (EVIDENCE Study 2014). Steroid avoidance compared versus steroid maintenance using an additional antiproliferative immunosuppressant was used in 13 studies (ATLAS Study 2005; FRANCIA Study 2007; FREEDOM Study 2008; INFINITY Study 2013; Kim 2002; Kumar 2005; Laftavi 2005; Montagnino 2005; Nagib 2015; Nematalla 2007; Schulak 1989; Vincenti 2003a; Woodle 2005). Of these, nine used MMF (ATLAS Study 2005; FRANCIA Study 2007; Kim 2002; Kumar 2005; Laftavi 2005; Nagib 2015 Nematalla 2007; Vincenti 2003a; Woodle 2005), two used enteric‐coated mycophenolate sodium (FREEDOM Study 2008; INFINITY Study 2013), one used AZA (Schulak 1989), and one used everolimus (Montagnino 2005). Steroid avoidance compared versus steroid withdrawal in patients treated with an additional antiproliferative was investigated in two studies (DOMINOS Study 2012; Sandrini 2009). One study used enteric‐coated mycophenolate sodium (DOMINOS Study 2012) and one used sirolimus (Sandrini 2009) as the third immunosuppressant.

Induction treatment was administered in 17 studies with adult kidney transplant recipients in three studies comparing steroid withdrawal with steroid maintenance (EVIDENCE Study 2014; Pelletier 2006; Pisani 2001), in 12 studies comparing steroid avoidance with steroid maintenance (FRANCIA Study 2007; FREEDOM Study 2008; INFINITY Study 2013; Kim 2002; Kumar 2005; Laftavi 2005; Montagnino 2005; Nagib 2015; Nematalla 2007; Schulak 1989; Vincenti 2003a; Woodle 2005), and in two studies comparing steroid avoidance with steroid withdrawal (DOMINOS Study 2012; Sandrini 2009). In 12 studies an IL‐2 receptor antagonist was used for induction treatment (DOMINOS Study 2012; EVIDENCE Study 2014; FREEDOM Study 2008; INFINITY Study 2013; Kim 2002; Kumar 2005; Montagnino 2005; Nagib 2015; Nematalla 2007; Pisani 2001; Sandrini 2009; Vincenti 2003a), in three studies an anti‐lymphocytic depleting antibodies was used (FRANCIA Study 2007; Laftavi 2005; Schulak 1989) and two studies allowed the type of induction treatment to be chosen by the investigator (Pelletier 2006; Woodle 2005).

Studies in child kidney transplant recipients

This update included three studies (11 reports, 346 participants) of steroid withdrawal or avoidance in child kidney transplant recipients (Benfield 2005; Höcker 2009; Mericq 2013).

Participants

Studies recruited participants who were younger than 20 years of age. All three studies included cadaveric and living kidney transplant recipients. In Benfield 2005 and Mericq 2013 only first kidney transplant recipients were eligible; in Höcker 2009 first or subsequent kidney transplantation was included.

Study comparisons

The three studies evaluated two different comparisons in children. Benfield 2005 and Höcker 2009 investigated steroid withdrawal versus steroid maintenance; Mericq 2013 investigated steroid avoidance versus steroid withdrawal.

Immunosuppression

All three studies used a calcineurin inhibitor‐based immunosuppressive regimen including an additional antiproliferative agent. Höcker 2009 used CsA and MMF, Benfield 2005 allowed either CsA or TAC to be used with sirolimus and Mericq 2013 used TAC in combination with MMF. Benfield 2005 and Mericq 2013 also used basiliximab for induction treatment, but Benfield 2005 was terminated early when the Data Safety Monitoring Board noted an excess risk of post‐transplant lymphoproliferative disease in both treatment groups.

Reported outcome measures

The reporting of outcome measures varied across studies. Of the 45 included studies, 34 reported patient mortality and 23 reported acute rejection (see Figure 1). Reporting of harms was more limited and inconsistent among studies (six studies reported cardiovascular events with varying definitions of cardiovascular events or definitions not reported). Frequently, studies reported incomplete data for harm outcomes or expressed their results as 'episodes', which complicated meaningful use of such data in the meta‐analysis.

Excluded studies

We excluded a total of 48 studies because studies: were not randomised (12), concerned ineligible populations (3), involved ineligible interventions ( 11) or ineligible co‐interventions (22).

Risk of bias in included studies

Reporting of details of study methodology regarding design and conduct of the study was incomplete in most studies. The assessment of risk of bias is shown in Figure 2 and Figure 3. Figure 2 shows the risk of bias indicators for individual studies. Figure 3 shows the proportion of studies assessed as low, high or unclear risk of bias for each risk of bias indicator.

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study

Figure 3

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies

Allocation

Random sequence generation was judged to be at low risk of bias in 19 studies (Ahsan 1999; ATLAS Study 2005; Benfield 2005; DOMINOS Study 2012; EVIDENCE Study 2014; FRANCIA Study 2007; FREEDOM Study 2008; Gulanikar 1991; Höcker 2009; Johnson 1989a; Kumar 2005; Laftavi 2005; Mericq 2013; Montagnino 2005; Nematalla 2007; Ponticelli 1997; Schulak 1989; Stiller 1983; Woodle 2005) and considered at high risk in two studies (Aswad 1998; Matl 2000). Randomisation methods were not reported in 27 studies (Albert 1985; Boletis 2001; Boots 2002; Bouma 1996; Burke 2000; del Castillo 2005; De Vecchi 1986; Farmer 2006; INFINITY Study 2013; Isoniemi 1990; Jankowska‐Gan 2009; Kacar 2004; Kim 2002; Lebranchu 1999; Maiorca 1988; Nagib 2015; Nott 1985; Park 1994; Pelletier 2006; Pisani 2001; Ratcliffe 1993; Sandrini 2009; Smak Gregoor 1999; Sola 2002; THOMAS Study 2002; Vincenti 2003a; Zhu 2008a).

Allocation concealment was assessed to be at low risk of bias in 14 studies (ATLAS Study 2005; Boots 2002; De Vecchi 1986; DOMINOS Study 2012; Farmer 2006; FRANCIA Study 2007; Gulanikar 1991; Isoniemi 1990; Mericq 2013; Montagnino 2005; Nematalla 2007; Smak Gregoor 1999; Stiller 1983; Woodle 2005); no study was judged to be at high risk of bias. Methods used for allocation concealment were unclear in the remaining 34 studies (Ahsan 1999; Albert 1985; Aswad 1998; Benfield 2005; Boletis 2001; Bouma 1996; Burke 2000; del Castillo 2005; EVIDENCE Study 2014; FREEDOM Study 2008; Höcker 2009; INFINITY Study 2013; Jankowska‐Gan 2009; Johnson 1989a; Kacar 2004; Kim 2002; Kumar 2005; Laftavi 2005; Lebranchu 1999; Maiorca 1988; Matl 2000; Nagib 2015; Nott 1985; Park 1994; Pelletier 2006; Pisani 2001; Ponticelli 1997; Ratcliffe 1993; Sandrini 2009; Schulak 1989; Sola 2002; THOMAS Study 2002; Vincenti 2003a; Zhu 2008a).

Blinding

Participants and investigators were blinded in only five studies (Ahsan 1999; Benfield 2005; Burke 2000; Gulanikar 1991; Woodle 2005). The absence of blinding was judged as high risk of bias because clinical management could be influenced by knowledge of treatment group. Blinding of outcome assessment was considered as low risk of bias because outcomes were objective and therefore more robust against influence by knowledge of treatment group (e.g. death, graft loss, serum creatinine).

Incomplete outcome data

Incomplete outcome data was judged to be at low risk of bias in 22 studies (Ahsan 1999; ATLAS Study 2005; Benfield 2005; Boots 2002; Bouma 1996; del Castillo 2005; DOMINOS Study 2012; FRANCIA Study 2007; FREEDOM Study 2008; Gulanikar 1991; Höcker 2009; Isoniemi 1990; Kumar 2005; Matl 2000; Montagnino 2005; Ponticelli 1997; Ratcliffe 1993; Schulak 1989; Smak Gregoor 1999; THOMAS Study 2002; Vincenti 2003a; Woodle 2005). Exclusion of participants after randomisation and attrition were considered at high risk in four studies (Boletis 2001; Burke 2000; De Vecchi 1986; Nagib 2015). Methods for addressing incomplete outcome data remained unclear in 22 studies (Albert 1985; Aswad 1998; EVIDENCE Study 2014; Farmer 2006; INFINITY Study 2013; Jankowska‐Gan 2009; Johnson 1989a; Kacar 2004; Kim 2002; Laftavi 2005; Lebranchu 1999; Maiorca 1988; Mericq 2013; Nematalla 2007; Nott 1985; Park 1994; Pelletier 2006; Pisani 2001; Sandrini 2009; Sola 2002; Stiller 1983; Zhu 2008a).

Selective reporting

Selective outcome reporting was judged as low risk in 37 studies (Ahsan 1999; Aswad 1998; ATLAS Study 2005; Benfield 2005; Boots 2002; Bouma 1996; del Castillo 2005; De Vecchi 1986; DOMINOS Study 2012; EVIDENCE Study 2014; FRANCIA Study 2007; FREEDOM Study 2008; Höcker 2009; INFINITY Study 2013; Isoniemi 1990; Jankowska‐Gan 2009; Kacar 2004; Kumar 2005; Lebranchu 1999; Maiorca 1988; Matl 2000; Mericq 2013; Montagnino 2005; Nagib 2015; Nematalla 2007; Park 1994; Pelletier 2006; Pisani 2001; Ponticelli 1997; Sandrini 2009; Schulak 1989; Smak Gregoor 1999; Sola 2002; Stiller 1983; THOMAS Study 2002; Vincenti 2003a; Woodle 2005). Eleven studies did not report all hard clinical outcomes that were considered primary outcomes for this review and were assessed as high risk of bias for selective outcome reporting (Albert 1985; Boletis 2001; Burke 2000; Farmer 2006; Gulanikar 1991; Nott 1985; Johnson 1989a; Kim 2002; Laftavi 2005; Ratcliffe 1993; Zhu 2008a).

Other potential sources of bias

Funding from academic independent sources was considered as low risk of bias in four studies (De Vecchi 1986; Isoniemi 1990; Matl 2000; Mericq 2013). In 16 studies a pharmaceutical company was reported as funding source, which was judged as high risk of bias (Ahsan 1999; ATLAS Study 2005; Benfield 2005; Bouma 1996; DOMINOS Study 2012; FRANCIA Study 2007; FREEDOM Study 2008; Kumar 2005; Montagnino 2005; Gulanikar 1991; Smak Gregoor 1999; Stiller 1983; THOMAS Study 2002; Vincenti 2003a). In 27 studies funding sources were not disclosed (Albert 1985; Aswad 1998; Boletis 2001; Boots 2002; Burke 2000; del Castillo 2005; EVIDENCE Study 2014; Farmer 2006; Höcker 2009; INFINITY Study 2013; Jankowska‐Gan 2009; Johnson 1989a; Kacar 2004; Kim 2002; Laftavi 2005; Lebranchu 1999; Maiorca 1988; Nematalla 2007; Nott 1985; Park 1994; Pelletier 2006; Pisani 2001; Ponticelli 1997; Ratcliffe 1993; Sandrini 2009; Schulak 1989; Sola 2002; Woodle 2005; Zhu 2008a). Publication bias was assessed by constructing funnel plots for three comparisons that included at least 10 studies in the meta‐analysis (death and acute rejection for steroid withdrawal versus steroid maintenance and acute rejection for steroid avoidance versus steroid maintenance). All funnel plots are symmetric and do not indicate publication bias (see Figure 4).

Figure 4

Funnel plot of comparisons that included at least 10 studies in the meta‐analysis

Effects of interventions

Studies in adults with kidney transplant recipients

Steroid withdrawal versus steroid maintenance

Steroid withdrawal may lead to little of no difference in patient mortality at either one year (Analysis 1.1.1 (10 studies, 1913 participants): RR 0.68, 95% CI 0.36 to 1.30; I² = 0%) or one to five years post transplantation (Analysis 1.1.2 (7 studies, 1118 participants): RR 1.26, 95% CI 0.73 to 2.17; I² = 0%). Likewise steroid withdrawal may lead to little or no difference in graft loss excluding death at either one year (Analysis 1.1.5 (8 studies, 1817 participants): RR 1.17, 95% CI 0.72 to 1.92; I² = 0%) or one to five years post transplantation (Analysis 1.1.6 (7 studies, 1092 participants): RR 1.61, 95% CI 0.98 to 2.64; I² = 0%).

The risk of acute rejection significantly increased by 77% in patients withdrawn from steroids compared versus patients maintained on steroids within the first year after transplantation (Analysis 1.2.1 (10 studies, 1913 participants): RR 1.77, 95% CI 1.20 to 2.61; I² = 54%), but there was no difference in the incidence of biopsy‐proven acute rejection (Analysis 1.2.2 (5 studies, 1292 participants): RR 1.32, 95% CI 0.78 to 2.22; I² = 65%).

The incidence of NODAT (Analysis 1.3.1 (6 studies, 1439 participants): RR 0.77, 95% CI 0.49 to 1.21; I² = 0%) as well as the incidence of cardiovascular events (Analysis 1.3.2 (2 studies, 607 participants): RR 0.98, 95% CI 0.42 to 2.33; I² = 0%) up to five years after transplantation were not significantly different between groups, mainly because of the low number of studies reporting these rarely occurring outcomes. Likewise data was sparse for harmful events, such as infection (Analysis 1.4.1 (5 studies, 1819 participants): RR 1.02, 95% CI 0.84 to 1.22; I² = 30%), CMV infection (Analysis 1.4.2 (RR 1.04, 95% CI 0.80 to 1.36; participants = 1758; studies = 5; I² = 0%), and malignancy (Analysis 1.4.3 (3 studies, 756 participants): RR 0.77, 95% CI 0.41 to 1.46; I² = 0%) and a difference in these outcomes could not be demonstrated up to five years after transplantation. There was also no evidence of difference in kidney function as determined by measurement of serum creatinine and creatinine clearance up to one as well as up to five years after transplantation (Analysis 1.5) (See also summary of findings Table for the main comparison).

Sensitivity and subgroup analyses for steroid withdrawal versus steroid maintenance studies

Results of the sensitivity and subgroup analyses are summarised in Table 1.

Table 1. Steroid withdrawal versus steroid maintenance ‐ stratified subgroup and sensitivity analysis for death, graft loss and acute rejection up to one year after transplantation

	Death			Graft loss			Acute rejection			Biopsy‐proven acute rejection
	Studies	RR	95% CI	Studies	RR	95% CI	Studies	RR	95% CI	Studies	RR	95% CI
*Publication status*
Peer reviewed journal	8	0.60	0.31 to 1.17	7	1.25	0.73 to 2.13	8	2.02	1.23 to 3.23	4	1.32	0.66 to 2.66
Abstract only	2	3.04	0.33 to 28.29	1	0.82	0.23 to 2.94	2	1.25	0.67 to 2.32	1	1.37	0.70 to 2.69
*ITT analysis*
ITT analysis used	6	0.69	0.30 to 1.61	6	1.31	0.69 to 2.46	6	2.07	1.10 to 3.91	3	1.37	0.64 to 2.94
ITT analysis not used/unclear	4	0.67	0.25 to 1.81	2	1.00	0.46 to 2.17	4	1.65	0.81 to 3.36	2	1.04	0.24 to 4.59
*Calcineurin inhibitor*
CsA	9	0.75	0.36 to 1.54	7	0.90	0.50 to 5.14	9	2.08	1.29 to 3.35	4	1.60	0.87 to 2.92
TAC	1	0.50	0.13 to 1.97	1	2.13	0.88 to 5.14	1	1.11	0.82 to 1.51	1	0.89	0.61 to 1.30
*Antimetabolite*
MMF or EC‐MPS	6	0.67	0.31 to 1.47	5	1.25	0.75 to 2.08	6	1.41	1.02 to 1.94	3	1.27	0.81 to 2.00
AZA	2	0.93	0.26 to 3.40	2	0.25	0.03 to 2.18	2	2.61	0.62 to 10.91	1	0.33	0.04 to 2.56
MMF or EC‐MPS or AZA	8	0.73	0.38 to 1.43	7	1.15	0.70 to 1.89	8	1.46	1.07 to 1.98	4	1.19	0.75 to 1.90
none	2	0.31	0.03 to 2.95	1	3.00	0.13 to 71.61	2	5.80	2.16 to 15.57	1	9.00	1.19 to 67.93
*Induction treatment*
Induction (yes)	2	3.00	0.32 to 27.87	1	0.33	0.01 to 8.02	2	0.80	0.22 to 2.91	NA	‐‐	‐‐
Induction (no)	8	0.60	0.31 to 1.17	7	1.21	0.74 to 1.99	8	1.93	1.26 to 2.94	NA	‐‐	‐‐

AZA ‐ azathioprine; CI ‐ confidence interval; CsA ‐ cyclosporin A; EC‐MPS ‐ enteric‐coated mycophenolate sodium; ITT ‐ intention to treat; MMF ‐ mycophenolate mofetil; NA ‐ not available; RR ‐ risk ratio; TAC ‐ tacrolimus

We have performed sensitivity analysis to assess the impact of publication status and use of intention‐to‐treat‐analysis on primary endpoints (mortality, death censored graft loss, acute rejection and biopsy‐proven acute rejection) using data from studies reporting these outcomes at any time point within the first year after transplantation. There was no evidence to suggest a difference in effect estimates of mortality, graft loss and biopsy‐proven acute rejection for studies depending on whether they have performed intention‐to‐treat analysis or whether the study was published in a peer‐reviewed journal. The significant increase in risk for acute rejection in patients withdrawn from steroids compared versus those maintained on steroids was further increased in studies published in a peer‐reviewed journal (8 studies, 1741 participants: RR 2.02, 95% CI 1.26 to 3.23) and in studies that applied intention‐to‐treat analysis (6 studies, 1199 participants: RR 2.07, 95% CI 1.10 to 3.91), but was lost in studies published as abstract‐only and in studies where intention‐to‐treat analysis was either not used or unclear.

We performed subgroup analysis stratified by calcineurin‐inhibitor type, type of antimetabolite and induction treatment on primary endpoints (mortality, death censored graft loss, acute rejection and biopsy‐proven acute rejection) using data from studies reporting these outcomes at any time point within the first year after transplantation. There was no difference in mortality and graft loss in any of the subgroups. The risk of acute rejection after steroid withdrawal was further increased in patients treated with CsA (9 studies, 1357 participants: RR 2.08, 95% 1.29 to 3.35), especially among those who did not receive an additional antimetabolite (2 studies, 150 participants: RR 5.80, 95% CI 2.16 to 15.57) and in patients who did not receive induction treatment (8 studies, 1765 participants: RR 1.93, 95% CI 1.26 to 2.94), but was decreased in patients who received either MMF or enteric‐coated mycophenolate sodium (6 studies, 1612 participants: RR 1.41, 95% CI 1.02 to 1.94) or any type of antimetabolite (8 studies, 1763 participants: RR 1.46, 95% CI 1.07 to 1.98).

Steroid avoidance versus steroid maintenance

Results are summarised in summary of findings Table 2.

Steroid avoidance did not show a significant effect on patient mortality at either one year (Analysis 2.1.1 (10 studies, 1462 participants): RR 0.96, 95% CI 0.52 to 1.80; I² = 0%) or one to five years post transplantation (Analysis 2.1.2 (7 studies, 1201 participants): RR 0.57, 95% CI 0.32 to 1.01; I² = 0%). Likewise steroid avoidance did not show any significant effects on graft loss excluding death at either one year (Analysis 2.1.5 (7 studies, 1211 participants): RR 1.09, 95% CI 0.64 to 1.86; I² = 0%) or one to five years post transplantation (Analysis 2.1.6 (7 studies, 1245 participants): RR 0.98, 95% CI 0.66 to 1.45; I² = 0%).

Steroid avoidance significantly increased the risk of acute rejection within the first year after transplantation by 58% compared versus patients maintained on steroids (Analysis 2.2.1 (7 studies, 835 participants): RR 1.58, 95% CI 1.08 to 2.30; I² = 63%). This effect of steroid avoidance was also demonstrated for biopsy‐proven acute rejection with a risk increase of 94% within the first year after transplantation (Analysis 2.2.2 (6 studies, 1073 participants): RR 1.94, 95% CI 1.26 to 2.98; I² = 45%).

There was no evidence of difference in the occurrence of NODAT, cardiovascular events, infection, CMV infection and malignancy between groups up to five years after transplantation (Analysis 2.3; Analysis 2.4). Kidney function determined as serum creatinine and creatinine clearance up to one year as well as up to five years after transplantation was not different for patients treated with steroids for less than 14 days compared versus patients maintained on steroids (Analysis 2.5).

Sensitivity and subgroup analysis for steroid avoidance versus steroid maintenance ‐ studies

We performed sensitivity analysis to assess the impact of use of intention‐to‐treat‐analysis on primary endpoints (mortality, death censored graft loss, acute rejection and biopsy‐proven acute rejection) using data from studies reporting these outcomes at any time point within the first year after transplantation. There was no study investigating steroid avoidance compared versus steroid maintenance that was published as abstract only, consequently the influence of publication status on the effect estimates could not be tested. There was no evidence to suggest a difference in effect estimates of mortality and graft loss for studies depending on whether they have performed intention‐to‐treat analysis. The increased risk for acute rejection and biopsy‐proven acute rejection in patients treated with steroids for less than 14 days after kidney transplantation compared versus those maintained on steroids was further increased in studies that applied intention‐to‐treat analysis (acute rejection: 4 studies, 655 participants: RR 1.92, 95% CI 1.18 to 3.14; biopsy‐proven acute rejection: 4 studies, 918 participants: RR 2.31, 95% CI 1.47 to 3.63), but lost significance in studies where intention‐to‐treat analysis was either not used or unclear.

We have performed subgroup analysis stratified by type of calcineurin inhibitor, type of antimetabolite and induction treatment on primary endpoints (mortality, death censored graft loss, acute rejection and biopsy‐proven acute rejection) using data from studies reporting these outcomes at any time point within the first year after transplantation. Stratified analysis did not reveal any difference in patient mortality and graft loss. The significant increase in risk for biopsy‐proven acute rejection persisted in patients treated with CsA (3 studies, 615 participants: RR 1.89, 95% CI 1.29 to 2.79), while patients treated with TAC did not have an increased risk for biopsy‐proven acute rejection (See Table 2).

Table 2. Steroid avoidance versus steroid maintenance ‐ stratified subgroup and sensitivity analysis for death, graft loss and acute rejection up to one year after transplantation

	Death			Graft loss			Acute rejection			Biopsy‐proven acute rejection
	Studies	RR	95% CI	Studies	RR	95% CI	Studies	RR	95% CI	Studies	RR	95% CI
*ITT analysis*
ITT analysis used	7	1.16	0.48 to 2.83	5	1.09	0.56 to 2.11	4	1.92	1.18 to 3.14	4	2.31	1.47 to 3.63
ITT analysis not used/unclear	3	0.51	0.07 to 3.83	2	1.11	0.46 to 2.67	3	1.24	0.97 to 1.59	2	1.05	0.49 to 2.23
*Calcineurin inhibitor*
CsA	8	0.88	0.47 to 1.66	5	1.08	0.59 to 1.99	5	1.31	1.05 to 1.63	3	1.89	1.29 to 2.79
TAC	2	6.82	0.36 to 130.81	2	1.14	0.39 to 3.3	2	2.40	1.05 to 5.49	3	1.81	0.66 to 4.99
*Antimetabolite*
MMF or EC‐MPS	6	1.15	0.36 to 3.69	6	1.09	0.56 to 2.11	5	1.87	1.20 to 2.91	6	1.94	1.26 to 2.98
AZA	1	1.64	0.29 to 9.2	NA	‐‐	‐‐	NA	‐‐	‐‐	NA	‐‐	‐‐
MMF or EC‐MPS or AZA	8	1.16	0.48 to 2.83	NA	‐‐	‐‐	NA	‐‐	‐‐	NA	‐‐	‐‐
None	2	0.51	0.07 to 3.83	1	1.11	0.46 to 2.67	2	1.26	0.95 to 1.65	NA	‐‐	‐‐
*Induction treatment*
Induction (yes)	7	0.97	0.38 to 2.48	5	1.06	0.45 to 2.46	4	1.50	0.97 to 2.32	5	1.67	1.19 to 2.36
Induction (no)	3	0.92	0.17 to 5.01	2	1.12	0.57 to 2.2	3	1.72	0.89 to 3.32	1	3.89	2.16 to 7.03

Steroid avoidance versus steroid withdrawal

Only three studies investigating the effect of steroid avoidance compared versus steroid withdrawal were identified, wherefore data is specifically sparse for this comparison. There is no evidence to suggest a difference in any outcome (death: Analysis 3.1; rejection: Analysis 3.2; NODAT, infection, malignancy: Analysis 3.3; kidney function: Analysis 3.4). Sensitivity and subgroup analysis could not be performed due to the small number of studies identified.

Studies in children with kidney transplant recipients

Steroid withdrawal versus steroid maintenance

We identified only two studies that investigated the effect of steroid withdrawal compared versus steroid maintenance in children (Benfield 2005; Höcker 2009). Death and graft loss at five years were significantly lower for children withdrawn from steroids, but these results were drawn from Benfield 2005 only, since neither death nor graft loss were observed in Höcker 2009 (Analysis 4.1.2: RR 0.16, 95% CI 0.02 to 1.35). The effect of steroid withdrawal on acute rejection is unclear due to the small number of studies and wide confidence intervals (Analysis 4.2). Kidney function was reported in Höcker 2009 only and was not significantly different between groups (Analysis 4.3).

Benfield 2005 was terminated early due to an unanticipated high incidence of post‐transplant lymphoproliferative disease. Of the 274 enrolled participants, 19 developed post‐transplant lymphoproliferative disease, 10 before randomisation. Sensitivity and subgroup analysis could not be performed due to the small number of studies identified.

Steroid avoidance versus steroid maintenance

Only Mericq 2013 investigated the effect of steroid avoidance compared versus steroid maintenance in children. Neither death nor graft loss was observed in this study, and due to sparse data, a difference in biopsy‐proven acute rejection could not be demonstrated. Kidney function was not reported. Sensitivity and subgroup analysis could not be performed on a single study.

Discussion

Summary of main results

The aim of this review was to provide updated evidence addressing the benefits and harms of steroid avoidance and withdrawal in kidney transplant recipients. All identified studies concerned one of the three comparisons defined for this review. The majority of studies compared steroid withdrawal versus steroid maintenance (24 adult studies, 2 studies in children). Steroid avoidance was compared versus steroid maintenance in 19 studies, one of which was conducted in child kidney transplantation. Of the three studies that compared steroid avoidance versus steroid withdrawal, none involved children. In adult kidney transplantation meta‐analysis could be carried out for all three comparisons, but data was particularly scarce for the comparison of steroid avoidance with steroid withdrawal. The low number of studies with child kidney transplant recipients did not enable data synthesis through meta‐analysis.

We were unable to demonstrate clear beneficial effects, such as a reduction in mortality or NODAT within five years after transplantation for steroids withdrawal or avoidance in adult kidney transplant recipients. Both steroid withdrawal and steroid avoidance showed little or no effect on mortality, graft loss, and CMV infection. The risk of acute rejection did significantly increase by 77% after steroid withdrawal and by 58% after steroid avoidance compared to steroid maintenance (see summary of findings Table for the main comparison, summary of findings Table 2).

The effect of steroid withdrawal in children is uncertain. The available data allowed only one meta‐analysis for acute rejection in children, which found no significant difference. Death and graft loss had not been observed in one of the two studies in children and outcomes such as biopsy‐proven acute rejection and malignancy were only reported in one of the two studies which further reduced the quantity of the available data. Only one study investigated the effect of steroid avoidance compared versus steroid maintenance in children, thus a meta‐analysis was not possible.

Overall completeness and applicability of evidence

An extensive literature review was performed to identify studies that assessed the benefits and harms of steroid withdrawal or avoidance in kidney transplant recipients. In general, two parameters are particularly relevant for assessing benefits and harms of steroid withdrawal in kidney transplant recipients: firstly, the time‐point of steroid withdrawal after kidney transplantation and secondly, the duration of follow‐up to observe outcome events in kidney transplant patients.

Steroids are withdrawn at various time points after kidney transplantation in clinical practice and this fact was reflected by the variety of time points used to investigate the effects of steroid withdrawal in clinical studies. We used a cut‐off of 14 days after transplantation to discriminate between steroid withdrawal and steroid avoidance. With this approach we were able to combine different time points for steroid withdrawal within these clinically relevant time frames. The majority of steroid avoidance studies used steroids for seven days or less, and the majority of the steroid withdrawal studies withdrew steroids between three to six months after transplantation. Thus, our findings may not be applicable for patients who are withdrawn from steroids at other time‐points after transplantation.

Most studies had between one and three years of follow‐up after either steroid avoidance or withdrawal which constitutes a major limitation for conclusions regarding long‐term consequences for patient and graft survival. Acute rejection is a major risk factor for reduced long‐term graft survival and typically occurs within the first year after transplantation. The impact of acute rejection on long‐term graft outcomes depends on the severity, recurrence and treatment of the acute rejection. While particularly severe and recurrent rejections increase the risk of graft loss, a single early acute rejection with complete functional recovery after treatment appears to be less harmful for long‐term graft outcomes. Most of the acute rejections reported in the included studies occurred early after transplantation and were mild and easily controlled with steroids which could be an argument to conclude that an increased risk of long‐term graft loss after steroid withdrawal is unlikely. However, recognizing that potential harms arising from steroid withdrawal may remain hidden for up to five years after steroid withdrawal (Gulanikar 1991); follow‐up periods of the included studies were too short to determine long‐term graft survival. Furthermore, it is important to stress that only half of the studies reported acute rejection. Consequently, potential harmful effects of steroid withdrawal on long‐term graft survival cannot be ruled out with this review with sufficient confidence.

Reporting of harmful events was especially limited and inconsistent. More than half of the studies did not report adverse events such as infection and CMV infection and less than a third of the studies reported malignancy and cardiovascular events. Even though we did not find evidence to suggest a difference in harmful events, it is important to point out that the absence of evidence does not mean there is evidence for absence of effect. It is unclear which outcomes occurred in the studies that provided no data. Although we believe this is the most comprehensive evidence summary on this topic, interpretation of our findings must consider the limitations of available data from this cohort. The value of increasing available evidence of potential harms associated with interventions has been widely recognised and is also not a problem peculiar to this review, but is common to many randomised studies and systematic reviews (Cuervo 2003; Tunis 2003).

Only one study investigating steroid avoidance included an mTOR‐inhibitor as baseline immunosuppression. Consequently, we cannot extrapolate the safety of steroid avoidance or withdrawal to protocols including mTOR‐inhibitors.

The inclusion and exclusion criteria for participation in the included studies may mean that our findings are not generalizable to all kidney transplant recipients. Eight studies did not specify any exclusion criteria, of which four did not specify any inclusion criteria. In three studies only recipients of a living kidney transplant were included and 11 studies included solely recipients of a cadaveric kidney transplant. Seventeen studies limited participation for patients who received their first kidney transplant and 16 studies excluded kidney transplant recipients who had experienced previous acute rejection. Kidney transplant recipients with a PRA > 50% were excluded in 13 studies. It is unclear whether the findings of this review apply to kidney transplant recipients with a higher immunologic transplant risk.

Although almost all studies included participants of a wide range of adult ages, none of the studies reported results for different age groups. Therefore we were unable to determine whether there is any difference in results depending on age. Due to the low number of studies in child kidney transplantation, our findings need to be interpreted with great caution in the light of a clear lack of evidence in children.

Quality of the evidence

The quality of the included studies was rather variable. The main limitations in the quality of the studies were allocation concealment, incomplete outcome data, blinding of participants and personnel and disclosure of funding. Of the 48 included studies only five studies blinded participants and personnel. This was considered a high risk of bias because clinical decision making could be influenced by knowledge of the treatment, such as for example that patients withdrawn from steroids were more closely monitored for signs of acute rejection. Adequate allocation concealment was reported in 14 studies and 19 studies demonstrated adequate sequence generation. The lack of adequate sequence generation and allocation concealment can lead to biased estimates of treatment effects in the original study and thus in a systematic review (Hollis 1999; Juni 1999; Moher 1998; Schulz 1995). All hard clinical outcomes (mortality, graft loss, acute rejection) were reported in 37 studies, but incomplete reporting of relevant data for a meta‐analysis in many studies hampered use of the provided data in our analysis. Comparison of kidney function was only possible in a limited number of studies because frequently either the number of participants in whom kidney function was measured or a measure of variability of the effect estimate were not provided. It might be more informative to compare the number of patients at risk of graft loss with a low creatinine clearance rather than assessing mean data. However, these data were not provided in any of the studies. Similarly dichotomous outcomes, especially infection and acute rejection were frequently reported as rates or episodes which complicated the use of such data for meta‐analysis. For disclosure of funding sources, 16 studies reported receiving of funding from pharmaceutical companies and 28 studies did not disclose their sponsor. We found that blinding of outcome assessors was adequate in 43 studies where the primary outcome were hard‐clinical endpoints (mortality, graft loss, acute rejection) and considered unlikely to be influenced by lack of blinding.

Potential biases in the review process

We searched multiple databases without language restriction in attempt to reduce publication bias. The Cochrane Kidney and Transplant's Specialised Register contains handsearched reports of studies presented at conferences and meetings, but there is a possibility that we missed unpublished data presented at smaller conferences or studies published in foreign language journals and low impact journals. Studies may have been added since our last search of the register. Not all included studies reported all outcomes which may have affected the results of the meta‐analysis.

Agreements and disagreements with other studies or reviews

Several previous systematic reviews have addressed steroid avoidance and withdrawal after kidney transplantation. The first review included three steroid withdrawal and four steroid avoidance studies in patients on CsA with or without AZA and showed a significant increase in acute rejection with an incidence of acute rejection of 48% in those withdrawn from steroids versus 30% in those maintained on steroids (P = 0.012) (Hricik 1993). The review published seven years later (Kasiske 2000) included 10 studies and showed an increased proportion of patients with acute rejection by 0.14 (95% CI 0.10 to 0.17; P < 0.001) and an increase in graft failure after steroid withdrawal by 40% (RR 1.40, 95% CI 1.09 to 1.70; P = 0.012). Most studies included in this meta‐analysis used CsA‐based immunosuppression with either no anti‐metabolite added or in combination with AZA. Only two studies with MMF were included and subgroup analysis showed similar results for these studies compared versus those that did not include MMF. A review of six studies of steroid withdrawal in kidney transplant recipients on triple therapy with calcineurin inhibitors and MMF showed an increase in acute rejection and no difference in graft failure (Pascual 2004). Due to the relative short follow‐up in these six studies long‐term consequences for graft survival given the observed increase in acute rejection after steroid withdrawal is unclear. A meta‐analysis published in 2012 by Knight 2010 found an increased risk of acute rejection and a reduced cardiovascular risk after steroid withdrawal or avoidance, but these findings resulted from a combined analysis of all steroid withdrawal or avoidance time points and were based on surrogate outcomes such as hypercholesterolaemia, hypertension and NODAT. Another review (Pascual 2012) with nine studies comparing steroid avoidance to steroid maintenance in kidney transplant recipients who received an immunosuppressive regimen consisting of antibody induction, either CsA or TAC and MMF reported that the increased risk of acute rejection in steroid avoidance was lost when patients received TAC‐based immunosuppression.

Figure 1

Study flow diagram.

Figure 2

Risk of bias summary: review authors' judgements about each risk of bias item for each included study

Figure 3

Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies

Figure 4

Funnel plot of comparisons that included at least 10 studies in the meta‐analysis

Analysis 1.1

Comparison 1 Steroid withdrawal versus steroid maintenance, Outcome 1 Death and graft loss.

Analysis 1.2

Comparison 1 Steroid withdrawal versus steroid maintenance, Outcome 2 Rejection.

Analysis 1.3

Comparison 1 Steroid withdrawal versus steroid maintenance, Outcome 3 New‐onset diabetes after transplantation and cardiovascular events.

Analysis 1.4

Comparison 1 Steroid withdrawal versus steroid maintenance, Outcome 4 Infection and malignancy.

Analysis 1.5

Comparison 1 Steroid withdrawal versus steroid maintenance, Outcome 5 Kidney function.

Analysis 2.1

Comparison 2 Steroid avoidance versus steroid maintenance, Outcome 1 Death and graft loss.

Analysis 2.2

Comparison 2 Steroid avoidance versus steroid maintenance, Outcome 2 Rejection.

Analysis 2.3

Comparison 2 Steroid avoidance versus steroid maintenance, Outcome 3 New‐onset diabetes after transplantation and cardiovascular events.

Analysis 2.4

Comparison 2 Steroid avoidance versus steroid maintenance, Outcome 4 Infection and malignancy.

Analysis 2.5

Comparison 2 Steroid avoidance versus steroid maintenance, Outcome 5 Kidney function.

Analysis 3.1

Comparison 3 Steroid avoidance versus steroid withdrawal, Outcome 1 Death and graft loss.

Analysis 3.2

Comparison 3 Steroid avoidance versus steroid withdrawal, Outcome 2 Rejection.

Analysis 3.3

Comparison 3 Steroid avoidance versus steroid withdrawal, Outcome 3 New‐onset diabetes after transplantation, infection, malignancy.

Analysis 3.4

Comparison 3 Steroid avoidance versus steroid withdrawal, Outcome 4 Kidney function.

Analysis 4.1

Comparison 4 Steroid withdrawal versus maintenance in children, Outcome 1 Death and graft loss.

Analysis 4.2

Comparison 4 Steroid withdrawal versus maintenance in children, Outcome 2 Rejection, malignancy.

Analysis 4.3

Comparison 4 Steroid withdrawal versus maintenance in children, Outcome 3 Kidney function.

Analysis 5.1

Comparison 5 Publication bias, Outcome 1 Funnel plots.

Summary of findings for the main comparison. Steroid withdrawal versus steroid maintenance for kidney transplant recipients

Steroid withdrawal versus steroid maintenance for kidney transplant recipients
Patient or population: kidney transplant recipients Intervention: steroid withdrawal Comparison: steroid maintenance
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	Steroid maintenance	Steroid withdrawal
Mortality Follow‐up: 1 year	22 per 1000	15 per 1000 (8 to 29)	RR 0.68 (0.36 to 1.3)	1913 (10)	⊕⊕⊝⊝ low^1,2
Graft loss (excluding death) Follow‐up: 1 year	32 per 1000	38 per 1000 (23 to 62)	RR 1.17 (0.72 to 1.92)	1817 (8)	⊕⊕⊝⊝ low^2,3
Acute rejection Follow‐up: 1 year	152 per 1000	268 per 1000 (182 to 396)	RR 1.77 (1.2 to 2.61)	1913 (10)	⊕⊕⊕⊝ moderate¹
NODAT Follow‐up: 5 years	57 per 1000	44 per 1000 (28 to 69)	RR 0.77 (0.49 to 1.21)	1439 (6)	⊕⊕⊝⊝ low^2,4
CMV infection Follow‐up: 5 years	100 per 1000	104 per 1000 (80 to 137)	RR 1.04 (0.8 to 1.36)	1758 (5)	⊕⊕⊝⊝ low^2,5
The assumed risk* is the baseline risk in the control group treated with steroid maintenance. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; NODAT: new‐onset diabetes after transplantation; CMV ‐ cytomegalovirus
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.
¹ Most studies were unblinded (9 studies) and did not report details about random sequence generation or allocation concealment or both (8 studies). One study had inappropriate random sequence generation. Four studies were industry sponsored. ITT analysis was unclear in four. ² Total number of events were fewer than 300. ³ Most studies were unblinded (7 studies) and did not report details about random sequence generation or allocation concealment or both (6 studies). One study had inappropriate random sequence generation. Four studies were industry sponsored. ITT analysis was unclear in two. ⁴ Most studies were unblinded (5 studies) and did not report details about random sequence generation or allocation concealment or both (5 studies). Three studies were industry sponsored. ITT analysis was unclear in three studies. One study had selective outcome reporting. ⁵ Most studies were unblinded (4 studies) and did not report details about random sequence generation or allocation concealment or both (4 studies). Three studies were industry sponsored. ITT analysis was unclear in two studies. One study had selective outcome reporting.

Summary of findings for the main comparison. Steroid withdrawal versus steroid maintenance for kidney transplant recipients

Summary of findings 2. Steroid avoidance versus steroid maintenance for kidney transplant recipients

Steroid avoidance versus steroid maintenance for kidney transplant recipients
Patient or population: kidney transplant recipients Intervention: steroid avoidance Comparison: steroid maintenance
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	No of participants (studies)	Quality of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Control	Steroid avoidance versus steroid maintenance
Mortality Follow‐up: 1 year	31 per 1000	30 per 1000 (16 to 56)	RR 0.96 (0.52 to 1.8)	1462 (10)	⊕⊕⊝⊝ low^1,2
Graft loss (excluding death) Follow‐up: 1 year	42 per 1000	46 per 1000 (27 to 79)	RR 1.09 (0.64 to 1.86)	1211 (7)	⊕⊕⊝⊝ low^2,3
Acute rejection Follow‐up: 1 year	204 per 1000	323 per 1000 (221 to 470)	RR 1.58 (1.08 to 2.3)	835 (7)	⊕⊕⊕⊝ moderate⁴
NODAT Follow‐up: 5 years	107 per 1000	80 per 1000 (54 to 117)	RR 0.75 (0.51 to 1.1)	1618 (9)	⊕⊕⊝⊝ low^2,5
CMV Infection Follow‐up: 5 years	106 per 1000	101 per 1000 (74 to 138)	RR 0.96 (0.7 to 1.31)	1454 (6)	⊕⊕⊝⊝ low^2,6
The assumed risk* is the baseline risk in the control group treated with steroid maintenance. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio; NODAT: new‐onset diabetes after transplantation; CMV ‐ cytomegalovirus
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate
¹ All studies were unblinded. Six studies were industry sponsored. In six studies random sequence generation or allocation concealment or both was unclear. In two studies ITT was either not performed or unclear. One study had selective outcome reporting ² Total number of events was fewer than 300 ³ All studies were unblinded. Five studies were industry sponsored. In four studies random sequence generation or allocation concealment or both was unclear. ITT was unclear in one study. One study had selective outcome reporting ⁴ All studies were unblinded. Five studies were industry sponsored. In four studies random sequence generation or allocation concealment or both was unclear. In three studies ITT was either not performed or unclear. One study had selective outcome reporting ⁵ Most studies were unblinded (8 studies). Five studies were industry sponsored. In four studies random sequence generation or allocation concealment or both was unclear. One study had selective outcome reporting ⁶ Most studies were unblinded (5 studies). Four studies were industry sponsored. One study had unclear ITT

Summary of findings 2. Steroid avoidance versus steroid maintenance for kidney transplant recipients

Table 1. Steroid withdrawal versus steroid maintenance ‐ stratified subgroup and sensitivity analysis for death, graft loss and acute rejection up to one year after transplantation

	Death			Graft loss			Acute rejection			Biopsy‐proven acute rejection
	Studies	RR	95% CI	Studies	RR	95% CI	Studies	RR	95% CI	Studies	RR	95% CI
*Publication status*
Peer reviewed journal	8	0.60	0.31 to 1.17	7	1.25	0.73 to 2.13	8	2.02	1.23 to 3.23	4	1.32	0.66 to 2.66
Abstract only	2	3.04	0.33 to 28.29	1	0.82	0.23 to 2.94	2	1.25	0.67 to 2.32	1	1.37	0.70 to 2.69
*ITT analysis*
ITT analysis used	6	0.69	0.30 to 1.61	6	1.31	0.69 to 2.46	6	2.07	1.10 to 3.91	3	1.37	0.64 to 2.94
ITT analysis not used/unclear	4	0.67	0.25 to 1.81	2	1.00	0.46 to 2.17	4	1.65	0.81 to 3.36	2	1.04	0.24 to 4.59
*Calcineurin inhibitor*
CsA	9	0.75	0.36 to 1.54	7	0.90	0.50 to 5.14	9	2.08	1.29 to 3.35	4	1.60	0.87 to 2.92
TAC	1	0.50	0.13 to 1.97	1	2.13	0.88 to 5.14	1	1.11	0.82 to 1.51	1	0.89	0.61 to 1.30
*Antimetabolite*
MMF or EC‐MPS	6	0.67	0.31 to 1.47	5	1.25	0.75 to 2.08	6	1.41	1.02 to 1.94	3	1.27	0.81 to 2.00
AZA	2	0.93	0.26 to 3.40	2	0.25	0.03 to 2.18	2	2.61	0.62 to 10.91	1	0.33	0.04 to 2.56
MMF or EC‐MPS or AZA	8	0.73	0.38 to 1.43	7	1.15	0.70 to 1.89	8	1.46	1.07 to 1.98	4	1.19	0.75 to 1.90
none	2	0.31	0.03 to 2.95	1	3.00	0.13 to 71.61	2	5.80	2.16 to 15.57	1	9.00	1.19 to 67.93
*Induction treatment*
Induction (yes)	2	3.00	0.32 to 27.87	1	0.33	0.01 to 8.02	2	0.80	0.22 to 2.91	NA	‐‐	‐‐
Induction (no)	8	0.60	0.31 to 1.17	7	1.21	0.74 to 1.99	8	1.93	1.26 to 2.94	NA	‐‐	‐‐
AZA ‐ azathioprine; CI ‐ confidence interval; CsA ‐ cyclosporin A; EC‐MPS ‐ enteric‐coated mycophenolate sodium; ITT ‐ intention to treat; MMF ‐ mycophenolate mofetil; NA ‐ not available; RR ‐ risk ratio; TAC ‐ tacrolimus

Table 1. Steroid withdrawal versus steroid maintenance ‐ stratified subgroup and sensitivity analysis for death, graft loss and acute rejection up to one year after transplantation

Table 2. Steroid avoidance versus steroid maintenance ‐ stratified subgroup and sensitivity analysis for death, graft loss and acute rejection up to one year after transplantation

	Death			Graft loss			Acute rejection			Biopsy‐proven acute rejection
	Studies	RR	95% CI	Studies	RR	95% CI	Studies	RR	95% CI	Studies	RR	95% CI
*ITT analysis*
ITT analysis used	7	1.16	0.48 to 2.83	5	1.09	0.56 to 2.11	4	1.92	1.18 to 3.14	4	2.31	1.47 to 3.63
ITT analysis not used/unclear	3	0.51	0.07 to 3.83	2	1.11	0.46 to 2.67	3	1.24	0.97 to 1.59	2	1.05	0.49 to 2.23
*Calcineurin inhibitor*
CsA	8	0.88	0.47 to 1.66	5	1.08	0.59 to 1.99	5	1.31	1.05 to 1.63	3	1.89	1.29 to 2.79
TAC	2	6.82	0.36 to 130.81	2	1.14	0.39 to 3.3	2	2.40	1.05 to 5.49	3	1.81	0.66 to 4.99
*Antimetabolite*
MMF or EC‐MPS	6	1.15	0.36 to 3.69	6	1.09	0.56 to 2.11	5	1.87	1.20 to 2.91	6	1.94	1.26 to 2.98
AZA	1	1.64	0.29 to 9.2	NA	‐‐	‐‐	NA	‐‐	‐‐	NA	‐‐	‐‐
MMF or EC‐MPS or AZA	8	1.16	0.48 to 2.83	NA	‐‐	‐‐	NA	‐‐	‐‐	NA	‐‐	‐‐
None	2	0.51	0.07 to 3.83	1	1.11	0.46 to 2.67	2	1.26	0.95 to 1.65	NA	‐‐	‐‐
*Induction treatment*
Induction (yes)	7	0.97	0.38 to 2.48	5	1.06	0.45 to 2.46	4	1.50	0.97 to 2.32	5	1.67	1.19 to 2.36
Induction (no)	3	0.92	0.17 to 5.01	2	1.12	0.57 to 2.2	3	1.72	0.89 to 3.32	1	3.89	2.16 to 7.03
AZA ‐ azathioprine; CI ‐ confidence interval; CsA ‐ cyclosporin A; EC‐MPS ‐ enteric‐coated mycophenolate sodium; ITT ‐ intention to treat; MMF ‐ mycophenolate mofetil; NA ‐ not available; RR ‐ risk ratio; TAC ‐ tacrolimus

Table 2. Steroid avoidance versus steroid maintenance ‐ stratified subgroup and sensitivity analysis for death, graft loss and acute rejection up to one year after transplantation

Comparison 1. Steroid withdrawal versus steroid maintenance

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Death and graft loss Show forest plot	15		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

1.1 Death up to one year	10	1913	Risk Ratio (M‐H, Random, 95% CI)	0.68 [0.36, 1.30]
1.2 Death one to five years	7	1118	Risk Ratio (M‐H, Random, 95% CI)	1.26 [0.73, 2.17]
1.3 Graft loss including death up to one year	8	1817	Risk Ratio (M‐H, Random, 95% CI)	0.97 [0.64, 1.49]
1.4 Graft loss including death one to five years	7	1092	Risk Ratio (M‐H, Random, 95% CI)	1.41 [1.00, 2.01]
1.5 Graft loss excluding death up to one year	8	1817	Risk Ratio (M‐H, Random, 95% CI)	1.17 [0.72, 1.92]
1.6 Graft loss excluding death one to five years	7	1092	Risk Ratio (M‐H, Random, 95% CI)	1.61 [0.98, 2.64]
2 Rejection Show forest plot	11		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

2.1 Acute rejection up to one year	10	1913	Risk Ratio (M‐H, Random, 95% CI)	1.77 [1.20, 2.61]
2.2 Biopsy‐proven acute rejection up to one year	5	1292	Risk Ratio (M‐H, Random, 95% CI)	1.32 [0.78, 2.22]
3 New‐onset diabetes after transplantation and cardiovascular events Show forest plot	6		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

3.1 New onset diabetes after transplantation up to five years	6	1439	Risk Ratio (M‐H, Random, 95% CI)	0.77 [0.49, 1.21]
3.2 Cardiovascular events up to five years	2	607	Risk Ratio (M‐H, Random, 95% CI)	0.98 [0.42, 2.33]
4 Infection and malignancy Show forest plot	7		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

4.1 Infection (all) up to five years	5	1819	Risk Ratio (M‐H, Random, 95% CI)	1.02 [0.84, 1.22]
4.2 CMV infection up to five years	5	1758	Risk Ratio (M‐H, Random, 95% CI)	1.04 [0.80, 1.36]
4.3 Malignancy up to five years	3	756	Risk Ratio (M‐H, Random, 95% CI)	0.77 [0.41, 1.46]
5 Kidney function Show forest plot	8		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only

5.1 Serum creatinine (mg/dL) up to one year	4	644	Std. Mean Difference (IV, Random, 95% CI)	‐0.04 [‐0.21, 0.13]
5.2 Serum creatinine (mg/dL) one to five years	5	762	Std. Mean Difference (IV, Random, 95% CI)	0.08 [‐0.06, 0.23]
5.3 Creatinine clearance (mL/min) up to one year	2	215	Std. Mean Difference (IV, Random, 95% CI)	‐0.07 [‐0.35, 0.21]
5.4 Creatinine clearance (mL/min) one to five years	3	669	Std. Mean Difference (IV, Random, 95% CI)	‐0.21 [‐0.56, 0.13]

Comparison 1. Steroid withdrawal versus steroid maintenance

Comparison 2. Steroid avoidance versus steroid maintenance

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Death and graft loss Show forest plot	13		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

1.1 Death up to one year	10	1462	Risk Ratio (M‐H, Random, 95% CI)	0.96 [0.52, 1.80]
1.2 Death one to five years	7	1201	Risk Ratio (M‐H, Random, 95% CI)	0.57 [0.32, 1.01]
1.3 Graft loss including death up to one year	7	1211	Risk Ratio (M‐H, Random, 95% CI)	1.08 [0.72, 1.62]
1.4 Graft loss including death one to five years	7	1245	Risk Ratio (M‐H, Random, 95% CI)	0.79 [0.53, 1.18]
1.5 Graft loss excluding death up to one year	7	1211	Risk Ratio (M‐H, Random, 95% CI)	1.09 [0.64, 1.86]
1.6 Graft loss excluding death one to five years	7	1245	Risk Ratio (M‐H, Random, 95% CI)	0.98 [0.66, 1.45]
2 Rejection Show forest plot	9		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

2.1 Acute rejection up to one year	7	835	Risk Ratio (M‐H, Random, 95% CI)	1.58 [1.08, 2.30]
2.2 Biopsy‐proven acute rejection up to one year	6	1073	Risk Ratio (M‐H, Random, 95% CI)	1.94 [1.26, 2.98]
3 New‐onset diabetes after transplantation and cardiovascular events Show forest plot	9		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

3.1 New onset diabetes after transplantation up to five years	9	1618	Risk Ratio (M‐H, Random, 95% CI)	0.75 [0.51, 1.10]
3.2 Cardiovascular events up to five years	4	1013	Risk Ratio (M‐H, Random, 95% CI)	0.56 [0.30, 1.05]
4 Infection and malignancy Show forest plot	11		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

4.1 Infection (all) up to five years	9	1833	Risk Ratio (M‐H, Random, 95% CI)	0.93 [0.84, 1.03]
4.2 CMV Infection up to five years	6	1454	Risk Ratio (M‐H, Random, 95% CI)	0.96 [0.70, 1.31]
4.3 Malignancy up to five years	7	1635	Risk Ratio (M‐H, Random, 95% CI)	0.97 [0.61, 1.52]
5 Kidney function Show forest plot	10		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only

5.1 Serum creatinine (mg/dL) up to one year	5	735	Std. Mean Difference (IV, Random, 95% CI)	0.02 [‐0.12, 0.17]
5.2 Serum creatinine (mg/dL) one to five years	3	688	Std. Mean Difference (IV, Random, 95% CI)	‐0.01 [‐0.16, 0.14]
5.3 Creatinine clearance (mL/min) up to one year	6	1104	Std. Mean Difference (IV, Random, 95% CI)	‐0.08 [‐0.23, 0.08]
5.4 Creatinine clearance (mL/min) one to five years	3	563	Std. Mean Difference (IV, Random, 95% CI)	‐0.08 [‐0.25, 0.08]

Comparison 2. Steroid avoidance versus steroid maintenance

Comparison 3. Steroid avoidance versus steroid withdrawal

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Death and graft loss Show forest plot	3		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

1.1 Death up to one year	1	222	Risk Ratio (M‐H, Random, 95% CI)	0.39 [0.08, 1.98]
1.2 Death one to five years	2	152	Risk Ratio (M‐H, Random, 95% CI)	2.67 [0.63, 11.32]
1.3 Graft loss including death up to one year	1	222	Risk Ratio (M‐H, Random, 95% CI)	0.86 [0.32, 2.29]
1.4 Graft loss including death one to five years	2	152	Risk Ratio (M‐H, Random, 95% CI)	2.44 [0.89, 6.70]
1.5 Graft loss excluding death up to one year	1	222	Risk Ratio (M‐H, Random, 95% CI)	1.64 [0.40, 6.68]
1.6 Graft loss excluding death one to five years	2	152	Risk Ratio (M‐H, Random, 95% CI)	1.91 [0.48, 7.67]
2 Rejection Show forest plot	2		Risk Ratio (M‐H, Random, 95% CI)	Totals not selected

2.1 Acute rejection up to one year	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
2.2 Biopsy‐proven acute rejection up to one year	1		Risk Ratio (M‐H, Random, 95% CI)	0.0 [0.0, 0.0]
3 New‐onset diabetes after transplantation, infection, malignancy Show forest plot	3		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

3.1 New onset diabetes after transplantation up to five years	3	351	Risk Ratio (M‐H, Random, 95% CI)	0.63 [0.36, 1.09]
3.2 Infection (all) up to five years	3	374	Risk Ratio (M‐H, Random, 95% CI)	1.07 [0.76, 1.50]
3.3 CMV Infection up to five years	2	284	Risk Ratio (M‐H, Random, 95% CI)	0.53 [0.30, 0.92]
3.4 Malignancy up to five years	1	90	Risk Ratio (M‐H, Random, 95% CI)	1.57 [0.28, 8.94]
4 Kidney function Show forest plot	2		Std. Mean Difference (IV, Random, 95% CI)	Subtotals only

4.1 Serum creatinine (mg/dL) up to one year	2	88	Std. Mean Difference (IV, Random, 95% CI)	‐0.05 [‐0.47, 0.37]
4.2 Creatinine clearance (mL/min) up to one year	2	206	Std. Mean Difference (IV, Random, 95% CI)	‐0.13 [‐0.41, 0.14]

Comparison 3. Steroid avoidance versus steroid withdrawal

Comparison 4. Steroid withdrawal versus maintenance in children

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Death and graft loss Show forest plot	2		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

1.1 Death up to five years	2	174	Risk Ratio (M‐H, Random, 95% CI)	0.16 [0.02, 1.35]
1.2 Graft loss including death up to five years	2	174	Risk Ratio (M‐H, Random, 95% CI)	0.09 [0.01, 0.69]
1.3 Graft loss excluding death up to five years	2	174	Risk Ratio (M‐H, Random, 95% CI)	0.09 [0.00, 1.64]
2 Rejection, malignancy Show forest plot	2		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only

2.1 Acute rejection up to one year	2	174	Risk Ratio (M‐H, Random, 95% CI)	0.37 [0.13, 1.02]
2.2 Biopsy‐proven acute rejection up to one year	1	42	Risk Ratio (M‐H, Random, 95% CI)	0.17 [0.01, 3.27]
2.3 Malignancy (PTLD) up to five years	1	132	Risk Ratio (M‐H, Random, 95% CI)	1.89 [0.51, 6.98]
3 Kidney function Show forest plot	1		Mean Difference (IV, Random, 95% CI)	Totals not selected

3.1 Creatinine clearance (mL/min) up to five years	1		Mean Difference (IV, Random, 95% CI)	0.0 [0.0, 0.0]

Comparison 4. Steroid withdrawal versus maintenance in children

Comparison 5. Publication bias

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Funnel plots Show forest plot	20	5288	Risk Ratio (M‐H, Fixed, 95% CI)	1.36 [1.15, 1.62]

1.1 Death, steroid withdrawal versus maintenance	10	1913	Risk Ratio (M‐H, Fixed, 95% CI)	0.70 [0.39, 1.29]
1.2 Acute rejection steroid withdrawal versus maintenance	10	1913	Risk Ratio (M‐H, Fixed, 95% CI)	1.55 [1.28, 1.87]
1.3 Death, steroid avoidance versus maintenance	10	1462	Risk Ratio (M‐H, Fixed, 95% CI)	0.92 [0.52, 1.63]

Comparison 5. Publication bias