Journal Logo

Original Clinical Science

Corticosteroid-free Kidney Transplantation Improves Growth

2-Year Follow-up of the TWIST Randomized Controlled Trial

Webb, Nicholas J. A.1,2; Douglas, Sarah E.1; Rajai, Azita3; Roberts, Stephen A.4; Grenda, Ryszard5; Marks, Stephen D.6; Watson, Alan R.7; Fitzpatrick, Maggie8; Vondrak, Karel9; Maxwell, Heather10; Jaray, Jeno11; Van Damme-Lombaerts, Rita12; Milford, David V.13; Godefroid, Nathalie14; Cochat, Pierre15; Ognjanovic, Milos16; Murer, Luisa17; McCulloch, Mignon18; Tönshoff, Burkhard19

Author Information
doi: 10.1097/TP.0000000000000498
  • Free

Whilst the height of children with chronic kidney disease (CKD) undergoing renal transplantation has increased over recent years, this population remains on average 1.8 standard deviations shorter than their peers, this deficit being greater in boys and younger children.1,2 Some, particularly the under-sixes may experience posttransplant catch-up growth; however, for many, attaining an adequate final adult height remains a significant challenge.3

Corticosteroid administration after transplantation is known to suppress growth,4-6 and a number of recent studies have investigated corticosteroid avoidance (CA) and withdrawal protocols to assess their impact on growth and patient safety.6-18 We have previously reported the results of the Tacrolimus and Withdrawal of Steroids (TWIST) study, a large, multicenter, industry-sponsored randomized controlled trial (RCT) which evaluated the impact of early corticosteroid withdrawal (CW) on posttransplantation growth.19 At 6 months follow-up, CW subjects demonstrated a significant improvement in height standard deviation score (SDS) as well as improved lipid profiles and less hyperglycaemia. There was more anemia and infection in CW subjects, though no difference in acute rejection (AR) incidence. One major limitation of TWIST was the short duration of follow-up; for this reason, the industry-sponsored trial was followed by an investigator-driven 2-year extension study. The aims of this were to ascertain whether the improved growth observed at 6 months was sustained at 1 and 2 years, and whether early CW was associated with more AR or other adverse events (AEs).



Figure 1 summarizes study participant flow. Two hundred subjects were randomized in the initial 6-month study: 196 were included in the intention to treat (ITT) analysis. Demographics and baseline characteristics have been previously reported.19 One hundred sixty-nine (86% of ITT population) subjects completed the initial 6-month study. Complete 1 year follow-up data were available for 127; however, 14 exclusions (CW = 9, corticosteroid continuation [CC]=5) were made (visits outside 3-month window or incompletely recorded visit dates), leaving eligible data for 113 (66.7%, CW = 54, CC = 59). Complete 2-year follow-up data were returned for 118, with 12 (CW = 6, CC = 6) exclusions (same reasons as at year 1), leaving 106 (62.7%) subjects with data available for analysis (CW = 53, CC = 53). Baseline characteristics of subjects in the ITT data set and completing 1-year and 2-year follow-up did not differ between the two groups other than a significantly higher number of non-white CW subjects (Table 1). Analysis of baseline characteristics for those subjects where 1-year and 2-year data were not collected showed that loss to follow-up was not associated with any particular characteristic.

Flow of patients through study.
Baseline demographics for patients in the full ITT dataset and completing 1-year and 2-year follow-up

Linear Growth–Primary Endpoint

There was significantly improved growth in CW subjects at 1 year follow-up. The mean (SE) change in height SDS (adjusted for pubertal status and baseline value) was 0.50 (0.09) compared with 0.22 (0.09) in the CC group. The treatment group difference in adjusted mean change using the mixed model repeated measures (MMRM) model, 0.25 (95% confidence interval [95% CI], 0.10, 0.40) was statistically significant (P = 0.001). Subgroup analysis showed that the observed improvement in growth was more pronounced in prepubertal subjects, but failed to reach statistical significance in pubertal subjects (Table 2a). At the 2-year follow-up, there remained a trend toward improved growth in the overall CW group, though the treatment group difference in adjusted mean change, 0.20 (95% CI, −0.01, 0.41) failed to reach statistical significance (P = 0.06). However, on subgroup analysis, there remained statistically significant better growth in prepubertal subjects (Table 2b).

Change in height SDS from baseline to 1 year follow-up
Change in height SDS from baseline to 2 year follow-up

Expressing these observed differences in change in height SDS as absolute changes in height, between baseline and 1 year, CW subjects grew on average 1.8 cm more than CC subjects, the corresponding figures for the prepubertal and pubertal subgroups being 2.4 and 0.6 cm, respectively. Between baseline and 2 years, CW subjects grew on average 1.5 cm more than CW subjects (prepubertal, 3.2 cm and pubertal, −0.5 cm).

Further per-protocol analysis was performed, including only outcomes for subjects who remained on the randomized corticosteroid regimen to which they were randomized. This per protocol population excluded 9 CW subjects who were receiving corticosteroids by the time of their 1-year visit and 15 CW subjects who were receiving corticosteroids by the time of their 2-year visit. Also excluded were 1 CC subject in whom corticosteroids had been discontinued by the time of their 1-year visit and 8 CC subjects who had discontinued corticosteroids by the time of their 2-year visit. A total of 45 and 38 CW and 58 and 34 CC subjects therefore remained on the corticosteroid regimen to which they were randomized at 1 and 2 years, respectively. The precise timing and reasons for either recommencing or discontinuing corticosteroids were not documented. At 1-year follow-up, the treatment group difference in adjusted mean change was significant in the overall per-protocol population as well as both the prepubertal and pubertal subgroups (Table 2c). At 2 years, this difference remained significant in the overall study population as well as for the prepubertal subgroup (Table 2d).

Change in height SDS from baseline to 1 year follow up in patients who remained on the treatment to which they were randomized
Change in height SDS from baseline to 2 year follow up in patients who remained on the treatment to which they were randomized

Body Mass Index

There were no major significant differences in change in body mass index (BMI). Using the MMRM model, the difference in change in BMI SDS was 0.26 higher (95% CI −0.06, 0.57; P = 0.11) in CC subjects at 1 year and 0.12 higher (95% CI, −0.26, 0.50; P = 0.53) in CW at 2 years. Prepubertal CC subjects showed a significantly greater difference at 1 year only (0.64 [95% CI, 0.19, 1.09; P = 0.006]).

Patient and Graft Survivals

During the initial 6-month study, three grafts were lost in each group, including one CW subject who died from sepsis with a functioning graft 1 month after study withdrawal.19 Two CW subjects lost grafts after refractory AR and died shortly after their first year follow-up from cardiorespiratory arrest complicating mucomycosis, and posttransplant lymphoproliferative disease, respectively. There were no deaths or graft losses in the CC group during year 1. During year 2, two CC subjects lost grafts (acute refractory rejection and chronic rejection).

Rejection Episodes

Acute rejection occurred with comparable frequency in both treatment groups. During year 1, five reported episodes (four biopsy-proven) occurred in CW subjects (one was receiving corticosteroids) and two (both biopsy-proven) in CC subjects. A further two episodes (both biopsy-proven) occurred during year 2 in CW subjects (one was receiving corticosteroids). Two CC subjects (one not receiving corticosteroids) also experienced biopsy-proven AR; both had previously reported AR during year 1. Kaplan-Meier estimates of freedom from biopsy-proven AR at 2 years were 82.3% and 88.1% in the CW and CC groups, respectively (P = 0.3) (Fig. 2).

Kaplan-Meier analysis of freedom from biopsy proven acute rejection.

Kidney Function

The median (interquartile range [IQR]) estimated glomerular filtration rate was comparable at both 1 year (CW 57.2 [52.2, 70.4] mL per min per 1.73 m2 vs. CC 62.3 [52.4, 77.1] mL per min per 1.73 m2, P = 0.27) and 2 years (CW 57.3 [50.5, 71.8] mL per min per 1.73 m2 vs. CC 64.3 [46.1, 78.5] mL per min per 1.73 m2, P = 0.35).

Immunosuppressive Therapy Exposure

There was no difference between the two groups in the median (IQR) total daily tacrolimus dose administered at 1 year (CW 0.11 [0.07, 0.16] mg/kg vs. CC 0.09 [0.07, 0.16] mg/kg; P = 1.0) or 2 years (CW 0.07 [0.06, 0.14] mg/kg vs. CC 0.08 [0.07, 0.15] mg/kg; P = 0.29). Mean (SD) tacrolimus trough levels were also no different at 1 year (CW 5.3 [4.1, 7.3] ng/mL vs. CC 6.5 [5.1, 7.6] ng/mL; P = 0.14) and 2 years (CW 5.0 (3.7, 7.5) ng/mL vs. CC 5.7 (3.4, 6.4) ng/mL; P = 0.81). There was similarly no difference in the median (IQR) daily mycophenolate mofetil (MMF) dose at both timepoints (1 year CW 547 [443, 642] mg/m2 vs. CC 566 [504, 654] mg/m2; P = 0.29 and 2 years CW 535 (404, 673) mg/m2 vs. 537 (434, 633) mg/m2; P = 0.81).

Blood Pressure

There was no difference in blood pressure (BP) SDS or the number of subjects taking antihypertensive therapies between the two treatment groups at both 1 and 2 years (Table 3).

Blood pressure SDS and medications

Lipid Profiles

At one year, median (IQR) plasma cholesterol was 3.6 (1.9, 4.7) mmol/L in CW subjects vs. 3.4 (1.9, 4.1) mmol/L in CC subjects (P = 0.8). At 2 years, there remained no difference between the two groups (3.3 (1.5, 4.1) mmol/L vs. 3.2 (1.6, 4.0) mmol/L, respectively; P = 0.8). Two CW and three CC subjects were receiving lipid lowering drugs at both year 1 and year 2 follow-up.

Adverse Events

Adverse event data are shown in Table 4. Other than infections in year 1 only, no significant differences between CW and CC groups were observed. Of the nine reported cases of disturbed glucose metabolism, seven occurred in CC subjects (four at year 1: two cases on-going from 6-month study; three at year 2: one case ongoing from 6-month study, one case resolved at year 1 then recurred at year 2) and two in CW subjects, both of whom had recommenced corticosteroids. Though 1-year and 2-year ITT analyses were insignificant (P = 0.12 and 1.00, respectively), the observation that no cases of disturbed glucose metabolism were observed in patients who remained off corticosteroid therapy is of interest. Only one case of recurrent primary renal disease was observed in the CW group. In addition to the two cases of malignancy reported in the initial 6-month study, there was one reported case of posttransplant lymphoproliferative disease in the CW group; this subject died in the second year after transplantation. We are also aware of a further CW subject who developed multivisceral Kaposi sarcoma at around 12 months after transplantation, though no follow-up data were submitted on this individual. No cases of malignancy were reported in the CC group (P = 0.12).

Adverse events by randomized group (intention to treat)


This investigator-led follow-up study has demonstrated that early CW improves growth up to 2 years after transplantation, most prominently in prepubertal subjects. The overall study population as well as the prepubertal subgroup demonstrated better growth at 1 year; by 2 years, significant benefit was lost in the overall study population, though persisted in prepubertal patients. A subgroup analysis of the per-protocol population, who remained on their randomized treatment, demonstrated an even greater benefit of CW, significantly improved growth being detected at 2 years in the overall study population as well as in prepubertal subjects.

As in the initial 6-month study, early CW was associated with a small but significant increase in infection in year 1 only. This is difficult to explain; systematic review has not shown an increased risk of infection with the use of anti-CD25 monoclonal antibodies,20 and patients in the CW arm did not receive higher doses of either tacrolimus or MMF. The incidence of other adverse events did not differ between the two groups; differences in metabolic profiles observed at 6 months were lost with increased follow-up. Although the incidence of disturbed glucose metabolism did not differ between the two groups, it is noteworthy that on per-protocol analysis, no CW subject who remained off corticosteroids developed abnormal glucose metabolism.

A number of other pediatric kidney transplant series have reported good outcomes after early CW or CA. The Stanford group reported an extended six-month course of daclizumab, tacrolimus and MMF CA protocol. Linear growth was significantly improved compared with corticosteroid-treated historic controls out to two years with no excess AR and less hypertension and hyperlipidemia.9,21 Other similar small series have also shown promising outcomes.7,8 However, when the Stanford protocol was investigated in the 130-subject SNS01 RCT, no overall difference was observed in change in height Z score between CA and CC subjects, though significantly better growth was observed in the under-fives.10 Similar to our findings, graft survival and AR incidence were similar, and no significant difference was found in the incidence of adverse effects, including new-onset diabetes after transplantation, hyperlipidemia and infection.

One possible explanation for the difference in outcomes for the overall study population in this and the SNS01 study is the inclusion of 63 adolescent subjects in the latter; many of these may have had little potential for further growth. Differences in the corticosteroid dose in CC patients appear unlikely to be responsible; in the SNS01 study, all subjects received a prednisone dose of less than 0.1 mg/kg daily by 6 months; in our study, the mean daily prednisolone dose in the CC group was similar at 4.01 mg (equivalent to 0.11 mg/kg) at 1 year and 3.52 mg (0.09 mg/kg) at 2 years. The lack of uniform positive impact of CW demonstrated in both studies indicates that factors other than corticosteroids, including age at transplantation, may limit posttransplantation growth.

Although the TWIST and SNS01 studies were the first to use growth as a primary end point in newly transplanted subjects, previous later stage CW studies using this endpoint have similarly demonstrated improved growth in CW subjects. A German prospective RCT randomized children with stable function 12 to 24 months after transplantation to CW or CC, demonstrating superior longitudinal growth and improved BP and lipid control in CW subjects at 2 years.22,23

Limitations of These Data

There are a number of limitations of this study which require both acknowledgement and further discussion. We were only able to obtain complete one-year follow up data on 113 of the 198 subjects (57.1%) who were included in the initial study analysis; complete 2-year follow-up data were available on 106 (53.5%). This attrition may have introduced bias as subjects who were lost to follow up may have differed demographically or in their clinical outcomes to those who were successfully followed up. Analysis of baseline characteristics of those lost to follow-up found no substantive differences other than a centre effect; certain centres returned complete data on all or the large majority of their subjects whilst others returned no data. We used a standard maximum-likelihood analysis methodology which explicitly allows for missing data, although it does assume that the missingness mechanism is encapsulated in the available data. Thus, the possibility remains that subjects who were lost to follow up differed in their clinical characteristics and outcomes in ways that we have not measured. The fact that a naive complete case analysis gave very similar results and that the main driver of missing data seems to be center-related and service-related rather than patient-related gives some reassurance that the data and analysis are representative of the whole trial cohort.

A further limitation is the lack of information regarding subjects’ pubertal status at follow-up. At baseline, subjects underwent formal Tanner staging and randomization was stratified according to pubertal status (prepubertal or pubertal). Data reporting ongoing Tanner staging was rarely completed, rendering more complex growth analysis incorporating ongoing pubertal development impossible.

Subjects recruited into this study were of generally low immunologic risk, potentially restricting the generalizability of our observations to a wider transplant population. The overall lack of research regarding transplantation outcomes in high immunologic risk subjects has been identified; with one notable exception,24 the large majority of reported studies have included first transplant subjects with low PRA values and substantial use of living donors.17 We have not in any way redressed this imbalance; further studies are necessary to confirm the growth-related efficacy and safety of early CW in higher immunologic risk subjects.

In conclusion, these data demonstrate that early CW effectively and safely improves growth up to 2 years after transplantation, particularly in prepubertal children.


Trial Design

The initial 6-month study protocol (EudraCT 2005-001348022) has been previously reported.19 In short, 200 pediatric kidney recipients were randomized 1:1, stratified by pubertal status, to early CW (tacrolimus, MMF, daclizumab (×2) and corticosteroids until day 4, n = 100) or CC (tacrolimus, MMF and on-going corticosteroids, n = 100) and were followed for 6 months. The primary end point for the 6-month study was change in height SDS from baseline to 6 months after transplantation.

Investigators submitted data regarding growth, patient and graft outcomes, laboratory values, on-going medications and AEs including infection, malignancy and disordered glucose metabolism (as interpreted by investigators) at 1 and 2 years after transplantation on a standardised case record form and sent it to the co-ordinating center.


All subjects in the initial 6-month study were eligible for inclusion in this follow-up study. Full inclusion and exclusion criteria have been previously reported.18 The study received approval at the relevant ethics committees for each of the participating centers in accordance with local requirements (UK North West 5 Research Ethics Committee reference 10/H1010/2).

Immunosuppressive and Other Therapy

After completion of the initial 6-month study there was no set immunosuppressive therapy protocol; this and other aspects of clinical care were performed in accordance with local standards of care. There was, however, an expectation that subjects would continue to receive their randomized treatment protocol unless clinical reasons mandated change, for example, the development of corticosteroid-related adverse effects or AR. Specific recommendations were not made regarding commencement or cessation of corticosteroids.

Outcome Measures

The primary outcome was change in height SDS from baseline assessed at 1 and 2 years after transplantation with a nominal 3-month window for data inclusion. Height SDS values were generated from Freeman et al.25 Secondary outcomes included the incidence of AR, patient and graft survival, change in BMI SDS, kidney function and incidence of adverse effects, including hypertension, dyslipidemia, obesity, abnormal glucose metabolism, malignancy and infection.

Estimated glomerular filtration rate was calculated using the Schwartz formula.26 Systolic and diastolic BP values were converted to SDS using QuesGen software.27

Renal biopsy was performed as clinically indicated. Protocol biopsies were not performed, and there was no routine monitoring for donor specific antibodies.

Statistical Analysis

The primary endpoint was analyzed using a MMRM model, which assumes that the missing outcomes are “Missing at Random” rather than a complete case analysis which assumes missing data are “Missing Completely at Random”.28–30 Thus, if the missingness is dependent on baseline or outcome variables that are included in the model then unbiased estimates will be obtained despite the missing data. Additionally the analysis utilises all the available data on all patients. A complete case analysis was performed for comparison and in this case gave similar results, suggesting that the missing data were largely random. The growth outcomes at all time points (6 months, 1 year, and 2 years) were modeled, with treatment group, baseline height SDS and pubertal status included as covariates, and an unstructured covariance matrix was used to represent the correlations between the outcomes at the different time points avoiding any assumptions as to the structure. The treatment effects at 1 and 2 years were estimated from this model along with associated 95% CI. Additionally, treatment by pubertal interaction terms was added to allow direct estimation of the treatment effects in the two subgroups and a Wald test of the relevant interaction term used to ascertain whether the treatment had a differential effect according to baseline pubertal status. Similar analyses were performed in the subset of observations where the participants remained on the allocated treatment (the per-protocol populations).

Complete case analyses of renal function, BP and lipid profiles were performed where visits occurred within the 3-month visit window. Data on death, AR, and AEs incorporated all subjects in whom follow-up data were available, regardless of timing.

Kaplan-Meier curves were computed for the time to biopsy-proven AR and groups compared with a log-rank test. Blood pressure SDS were compared using unpaired t tests, laboratory values, and immunosuppression doses using the nonparametric Mann-Whitney U test and AE rates using Fisher exact tests.

Stata 13 software (StataCorp. 2013, College Station, TX) was used to perform the MMRM analyses.


The authors thank Astellas Pharma for unrestricted educational funding of this project, in particular Dr. Malcolm Brown and Mr. Boris Wotroba. Drs. Clare Hamer and Raja Padidela provided invaluable assistance with data collection and analysis. The authors are most grateful to the children and families who participated and also to those investigators who participated in the initial 6-month study.


1. Fine RN, Martz K, Stablein D. What have 20 years of data from the North American Pediatric Renal Transplant Cooperative Study taught us about growth following renal transplantation in infants, children, and adolescents with end-stage renal disease? Pediatr Nephrol. 2010; 25: 739.
2. NAPRTCS 2010 Annual Transplant Report. Available at:
3. Harambat J, Cochat P. Growth after renal transplantation. Pediatr Nephrol. 2009; 24: 1297.
4. López-Espinosa JA, Yeste-Fernández D, Iglesias-Berengue J, et al. Factors affecting catch-up growth after liver transplantation. J Pediatr Endocrinol Metab. 2004; 17: 1097.
5. Peterson RE, Perens GS, Alejos JC, et al. Growth and weight gain of prepubertal children after cardiac transplantation. Pediatr Transpl. 2008; 12: 436.
6. Sarwal MM, Yorgin PD, Alexander S, et al. Promising early outcomes with a novel, complete steroid avoidance immunosuppression protocol in pediatric renal transplantation. Transplantation. 2001; 72: 13.
7. Bhakta N, Marik J, Malekzadeh M, et al. Can pediatric steroid-free renal transplantation improve growth and metabolic complications? Pediatr Transplant. 2008; 12: 854.
8. Silverstein DM, Aviles DH, LeBlanc PM, et al. Results of one-year follow-up of steroid-free immunosuppression in pediatric renal transplant patients. Pediatr Transplant. 2005; 9: 589.
9. Sarwal MM, Vidhun JR, Alexander SR, et al. Continued superior outcomes with modification and lengthened follow-up of a steroid-avoidance pilot with extended daclizumab induction in pediatric renal transplantation. Transplantation. 2003; 76: 1331.
10. Sarwal MM, Ettenger RB, Dharnidharka V, et al. Complete corticosteroid avoidance is effective and safe in children with renal transplants: a multicentre randomized trial with three year follow-up. Am J Transpl. 2012; 12: 2719.
11. Pape L, Lehner F, Blume C, et al. Pediatric kidney transplantation followed by de novo therapy with everolimus, low-dose cyclosporine A, and steroid elimination: 3-year data. Transplantation. 2011; 92: 658.
12. Höcker B, John U, Plank C, et al. Successful withdrawal of steroids in pediatric renal transplant recipients receiving cyclosporine A and mycophenolate mofetil treatment: results after four years. Transplantation. 2004; 78: 228.
13. Birkeland SA, Larsen KE, Rohr N. Pediatric renal transplantation without steroids. Pediatr Nephrol. 1998; 12: 87.
14. Delucchi A, Valenzuela M, Ferrario M, et al. Early steroid withdrawal in pediatric renal transplant on newer immunosuppressive drugs. Pediatr Transplant. 2007; 11: 743.
15. Oberholzer J, John E, Lumpaopong A, et al. Early discontinuation of steroids is safe and effective in pediatric kidney transplant recipients. Pediatr Transplant. 2005; 9: 456.
16. Delucchi A, Valenzuela M, Lillo AM, et al. Early steroid withdrawal in pediatric renal transplant: five years of follow-up. Pediatr Nephrol. 2011; 26: 2235.
17. Grenda R, Webb NJA. Steroid minimization in pediatric renal transplantation: early withdrawal or avoidance? Pediatr Transplant. 2010; 14: 961.
18. Nehus E, Goebel J, Abraham E. Outcomes of steroid-avoidance protocols in pediatric kidney transplant recipients. Am J Transplant. 2012; 12: 3441.
19. Grenda R, Watson A, Trompeter R, et al. A randomized trial to assess the impact of early steroid withdrawal on growth in pediatric renal transplantation: the TWIST study. Am J Transplant. 2010; 10: 828.
20. Webster AC, Ruster LP, McGee RG, et al. Interleukin 2 receptor antagonists for kidney transplant recipients. Cochrane Database of Systematic Reviews. 2010, Issue 1. Art. No.: CD003897. DOI 10.1002/14651858.CD003897.pub3.
21. Li L, Chang A, Naesens M, et al. Steroid-free immunosuppression since 1999: 129 pediatric renal transplants with sustained graft and patient benefits. Am J Transplant. 2009; 9: 1362.
22. Höcker B, Weber LT, Feneberg R, et al. Prospective, randomized trial on late steroid withdrawal in pediatric renal transplant recipients under cyclosporine microemulsion and mycophenolate mofetil. Transplantation. 2009; 87: 934.
23. Höcker B, Weber LT, Feneberg R, et al. Improved growth and cardiovascular risk after late steroid withdrawal; 2-year results of a prospective, randomised trial in paediatric renal transplantation. Nephrol Dial Transplant. 2010; 25: 617.
24. Li L, Chaudhury A, Chen A, et al. Efficacy and safety of thymoglobulin induction as an alternative approach for steroid-free maintenance immunosuppression in pediatric renal transplantation. Transplantation. 2010; 90: 1516.
25. Freeman JV, Cole TJ, Chinn S, et al. Cross sectional stature and weight reference curves for the UK 1990. Arch Dis Child. 1995; 73: 17.
26. Schwartz GJ, Muñoz A, Schneider M, et al. New equations to estimate GFR in children with CKD. J Am Soc Nephrol. 2009; 20: 629.
28. Ohidul Siddiqui HM, Hung J, O’Neill R. MMRM vs. LOCF: a comprehensive comparison based on simulation study and 25 NDA datasets. J Biopharm Stat. 2009; 19: 227.
29. Mallinckrodt CH, Clark WS, David SR. Accounting for dropout bias using mixed-effects models. J Biopharm Stat. 2001; 11: 9.
30. Mallinckrodt CH, Sanger TM, Dube S, et al. Assessing and interpreting treatment effects in longitudinal clinical trials with missing data. Biol Psychiatry. 2003; 53: 754.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.