DESCRIPTION OF SUPPLEMENT (METHODS):

Transcription

1 DESCRIPTION OF SUPPLEMENT (METHODS): Included in this supplement is additional detail for the: I. Search strategy including databases, search dates, and the specific terms and combinations used II. Statistical approach to the quantitative analyses. III. Statistical approach to analysis of treatment era

2 I. SEARCH STRATEGY On October 21, 2015, the research librarian (L.K.) conducted searches in the following databases using a combination of controlled vocabulary when possible and keywords. The search strategy for each database is described below as follows: MEDLINE (A), Embase (B and C), Cochrane Library (D), Web of Science (E), and Scopus (F). At the time of the initial search, the research team utilized a subscription to Embase accessed via OvidSP with duplicative MEDLINE records removed from the results using the included Ovid filter. To remain current in the context of the time to completion for the complex data extraction and statistical analyses, the search strategy was re-ran within each database on November 23, 2016 and again on June 16, For these two latter searches, the research team had full access to Embase through Elsevier (C). In an effort to include grey literature in the search, three additional search methods were conducted: (1) cited references of included studies were screened, (2) the conference proceedings for the American Society of Hematology (ASH) annual meetings (December 2015 and 2016) were hand-searched, and lastly, (3) the National Institutes of Health's (NIH) clinical registry website ClinicalTrials.gov (G) was searched for relevant information. All abstracts from manual review of the ASH conference proceedings and review of cited reference were found via search as well. Search Terms by Database (dates): A) MEDLINE/Pubmed (1946 June 19, 2017) 1. Leukemia, Biphenotypic, Acute[Mesh] 2. MPAL 3. 1 OR 2 4. mixed phenotype 5. mixed-phenotype 6. mixed lineage 7. mixed-lineage 8. mixed cell 9. mixed-cell 10. ambiguous lineage 11. ambiguous origin 12. bilineal 13. bi-lineal 14. undifferentiated 15. World Health 16. EGIL 17. biphenotype 18. bi-phenotype 19. biphenotypic 20. bi-phenotypic 21. biclonal 22. bi-clonal 23. hybrid 24. (B AND T) cell / OR 26. acute 27. Leukemia[Mesh] 28. leukemia 29. leukemias 30. leukaemia 31. leukaemias 32. leucocythaemia 33. leucocythaemias 34. leucocythemia 35. leucocythemias /OR AND AND OR Therapeutics[Mesh] 41. Treatment Outcome[Mesh] 42. Remission, Spontaneous[Mesh] 43. Remission Induction[Mesh] 44. Survival Analysis[Mesh] 45. Survival Rate[Mesh] 46. Survival[Mesh] 47. therapeutic 48. treatment 49. treatments 50. therapy 51. therapies 52. clinical effectiveness 53. clinical effectivenesses 54. clinical efficacy 55. clinical efficacies 56. patient relevant 57. patient relevant s 58. patient-relevant 59. patient-relevant s 60. remission 61. remissions 62. relapse 63. relapses 64. survival 65. recurrence 66. recurrences 67. recrudescence 68. recrudescences

3 69. survivorship /OR 71. Neoadjuvant Therapy[Mesh] 72. Drug Therapy[Mesh] 73. Antineoplastic Agents[Mesh] 74. Antineoplastic Agents[Pharmcological Action] 75. chemotherapy 76. chemotherapies 77. chemotherapeutic 78. pharmacotherapy 79. pharmacotherapies 80. antineoplastic 81. antineoplastics 82. antitumor 83. anti-tumor 84. anticancer 85. anti-cancer /OR 87. Stem Cell Transplantation[Mesh] 88. Stem Cells[Mesh] 89. stem cell 90. stem cells 91. progenitor cell 92. progenitor cells 93. cord blood 94. placental blood 95. hematopoietic 96. mesenchymal 97. peripheral blood 98. bone marrow 99. haploidentical 100. autologous 101. allogeneic /OR 103. Transplantation[Mesh] 104. transplant 105. transplants 106. transplantation 107. transplantations /OR AND OR OR 86 OR AND 111 B) EMBASE/OvidSP ( October 9, 2015) 1. Mixed Phenotype Leukemia/exp 2. MPAL 3. 1 OR 2 4. mixed phenotype 5. mixed lineage 6. mixed cell 7. ambiguous lineage 8. ambiguous origin 9. bilineal 10. bi-lineal 11. undifferentiated 12. World Health Organization Classification 13. EGIL 14. biphenotype 15. bi-phenotype 16. biphenotypic 17. bi-phenotypic 18. biclonal 19. bi-clonal 20. hybrid 21. (B AND T) cell /OR 23. acute 24. Leukemia/exp 25. leukemia 26. leukemias 27. leukaemia 28. leukaemias 29. leucocythaemia 30. leucocythemia 31. leucocythemias /OR AND AND OR Therapy/exp 36. Treatment Outcome/exp 37. Remission/exp 38. Survival/exp 39. therapeutic 40. therapeutics 41. treatment 42. treatments 43. therapy 44. therapies 45. clinical effectiveness 46. clinical efficacy 47. clinical efficacies 48. patient relevant 49. patient relevant s 50. patient-relevant 51. patient-relevant s 52. remission 53. remissions 54. relapse 55. relapses 56. survival 57. recurrence 58. recurrences 59. recrudescence 60. recrudescences 61. survivorship /OR 63. Cancer Adjuvant Therapy/[+NT] 64. Adjuvant Chemotherapy/exp 65. Drug Therapy/exp 66. Antineoplastic Agents/exp 67. chemotherapy 68. chemotherapies 69. chemotherapeutic 70. pharmacotherapy 71. pharmacotherapies 72. antineoplastic 73. antineoplastics 74. antitumor 75. anti-tumor

4 76. anticancer 77. anti-cancer /OR 79. Stem Cell Transplantation/exp 80. Stem Cells/exp 81. stem cell 82. stem cells 83. progenitor cell 84. progenitor cells 85. cord blood 86. placental blood 87. hematopoietic 88. mesenchymal 89. peripheral blood 90. bone marrow 91. haploidentical 92. autologous 93. allogeneic /OR 95. Transplantation/exp 96. transplant 97. transplants 98. transplantation 99. transplantations /OR AND OR OR 78 OR AND REMOVED MEDLINE RECORDS C) EMBASE/Elsevier ( June 19, 2017) 1. Acute Biphenotypic Leukemia/exp 2. MPAL 3. 1 OR 2 4. mixed phenotype 5. mixed lineage 6. mixed cell 7. ambiguous lineage 8. ambiguous origin 9. bilineal 10. bi-lineal 11. undifferentiated 12. World Health 13. EGIL 14. biphenotype 15. bi-phenotype 16. biphenotypic 17. bi-phenotypic 18. biclonal 19. bi-clonal 20. hybrid /OR 22. acute 23. 'Leukemia'/exp 24. leukemia 25. leukemias 26. leukaemia 27. leukaemias 28. leucocythaemia 29. leucocythaemias 30. leucocythemia 31. leucocythemias /OR AND AND OR 'Therapy'/exp 37. 'Treatment Outcome'/exp 38. 'Remission'/exp 39. 'Survival'/exp 40. therapeutic 41. therapeutics 42. treatment 43. treatments 44. therapy 45. therapies 46. 'clinical effectiveness' 47. 'clinical efficacy' 48. 'clinical efficacies' 49. 'patient relevant ' 50. 'patient relevant s' 51. 'patient-relevant ' 52. 'patient-relevant s' 53. remission 54. remissions 55. relapse 56. relapses 57. survival 58. recurrence 59. recurrences 60. recrudescence 61. recrudescences 62. survivorship /OR 64. 'Cancer Adjuvant Therapy'/exp 65. 'Drug Therapy'/exp 66. 'Anti-neoplastic Agent'/exp 67. chemotherapy 68. chemotherapies 69. chemotherapeutic 70. pharmacotherapy 71. pharmacotherapies 72. antineoplastic 73. antineoplastics 74. antitumor 75. anti-tumor 76. anticancer 77. anti-cancer /OR 79. 'Stem Cell Transplantation'/exp 80. 'Stem Cell'/exp 81. 'stem cell' 82. 'stem cells' 83. 'progenitor cell' 84. 'progenitor cells' 85. 'cord blood' 86. 'placental blood' 87. hematopoietic 88. mesenchymal 89. 'peripheral blood' 90. 'bone marrow' 91. haploidentical

5 92. autologous 93. allogeneic /OR 95. 'Transplantation'/exp 96. transplant 97. transplants 98. transplantation 99. transplantations /OR AND OR OR 78 OR AND 103 D) COCHRANE LIBRARY/Wiley (June 2017, Issue 6 of 12 Search June 19, 2017) 1. MeSH descriptor: [Leukemia, Biphenotypic, Acute] explode all trees 2. MPAL 3. 1 OR 2 4. mixed phenotype 5. mixed lineage 6. mixed-lineage 7. mixed cell 8. mixed-cell 9. ambiguous lineage 10. ambiguous origin 11. bilineal 12. undifferentiated 13. World Health 14. EGIL 15. biphenotype 16. biphenotypic 17. bi-phenotypic 18. hybrid 19. (B and T) cell /OR 21. acute 22. MeSH descriptor: [Leukemia] explode all tree 23. leukemia 24. leukemias 25. leukaemia 26. leukaemias /OR AND AND OR 29 E) WEB OF SCIENCE/ Thompson Reuters (1900 June 19, 2017) 1. MPAL 2. mixed phenotype 3. mixed-phenotype 4. mixed lineage 5. mixed-lineage 6. mixed cell 7. mixed-cell 8. ambiguous lineage 9. ambiguous origin 10. bilineal 11. bi-lineal 12. undifferentiated 13. World Health 14. EGIL 15. biphenotype 16. bi-phenotype 17. biphenotypic 18. bi-phenotypic 19. biclonal 20. bi-clonal 21. hybrid 22. (B AND T) cell /OR 24. acute 25. leukemia 26. leukaemia 27. leucocythaemia 28. leucocythemia /OR AND OR therapeutic 33. treatment 34. therapy 35. clinical effectiveness 36. clinical efficacy 37. patient relevant 45. patient-relevant 46. remission 47. relapse 48. survival 49. recurrence 50. recrudescence 51. survivorship 52. mortality 53. disease-free /OR 55. drug therapy 56. neoadjuvant 57. chemotherapy 58. chemotherapies 59. chemotherapeutic 60. pharmacotherapy 61. pharmacotherapies 62. antineoplastic 63. anti-neoplastic 64. antineoplastics 65. antitumor 66. anti-tumor 67. anticancer 68. anti-cancer /OR 70. stem cell 71. progenitor cell 72. cord blood 73. placental blood 74. hematopoietic 75. mesenchymal 76. peripheral blood 77. bone marrow 78. haploidentical 79. bone marrow 80. autologous 81. allogeneic /OR 83. transplant 84. transplantation

6 OR AND OR 69 OR AND 87 F) SCOPUS/Elsevier ( June 19, 2017) 1. MPAL 2. "mixed phenotype" 3. "mixed lineage" 4. "mixed cell" 5. "ambiguous lineage" 6. bilineal 7. bi-lineal 8. undifferentiated 9. "World Health " 10. EGIL 11. biphenotype 12. bi-phenotype 13. biphenotypic 14. bi-phenotypic 15. biclonal 16. bi-clonal 17. hybrid /OR 19. acute 20. leukemia 21. leucocythemia OR AND AND OR therapeutic 27. treatment 28. therapy 29. "clinical effectiveness" 30. "clinical effectivenesses" 31. "clinical efficacy" 32. "clinical efficacies" 33. "patient relevant " 34. "patient relevant s" 35. remission 36. relapse 37. survival 38. recurrence 39. recrudescence 40. survivorship /OR 42. chemotherapy 43. chemotherapies 44. chemotherapeutic 45. pharmacotherapy 46. pharmacotherapies 47. antineoplastic 48. antineoplastics 49. anti-neoplastic 50. anti-neoplastics 51. antitumor 52. anti-tumor 53. anticancer 54. anti-cancer /OR 56. stem cell 57. progenitor cell 58. cord blood 59. placental blood 60. hematopoietic 61. mesenchymal 62. peripheral blood 63. bone marrow 64. haploidentical 65. autologous 66. allgeneic /OR 68. transplant* AND OR 55 OR AND 70 G) CLINICALTRIALS.GO V/ US NIH (2000 June 19, 2017) 1. mixed phenotype 2. mixed lineage 3. mixed cell 4. ambiguous 5. undifferentiated 6. EGIL 7. World Health 8. biphenotypic 9. bilineage /OR 11. acute leukemia AND MPAL OR 12

7 II. DETAILED STATISTICAL APPROACH: The primary endpoints of interest were tumor response (CR or not), event-free survival (EFS), and overall survival (OS). The data comprise both summary data and individual case data that were abstracted from the literature. The independent variables of primary interest were type of induction therapy (ALL, AML, hybrid), age group (adult, pediatric), tumor phenotype (B/My, B/T, B/T/My, T/My), tumor lineage (Bi-L, Bi-P), and transplant (yes, no). Transplant was not relevant to analyses of response. The available data include both aggregate data and detailed case-level data. A. Meta-Analyses Aggregate data included overall patient counts, as well as patient counts in individual subgroups defined by the independent variables above, along with data relevant to one or more of the primary endpoints above, from each of a number of publications. These data were from publications that reported only summarized statistics, as well as from publications that reported case-level data. For the latter, summarized statistics were computed for each publication and included as part of the summary data. These case-level data were also used in the case-level analyses described below. For tumor response, the numbers of patients and numbers of responses overall and within subgroup were available. For overall survival, the number of patients as well as the actuarial estimates of OS percent at a reference time was provided. These estimates were often but not always accompanied by an estimate of the precision of the estimate, either as a stated standard error (SE) or as a 95% confidence interval, from which a SE could be computed. A number of estimates were provided only as median OS times. These estimates could not be used because they lacked direct estimates of precision and including them in the analysis would require strict and unverifiable parametric assumptions. Meta-Analysis of complete response (CR) Aggregate response data included patients counts (N) and numbers of complete responses (ncr), and thus were amenable to analyses using generalized linear models (GLM). Specifically, a univariable GLM regression, under the binomial family with logistic link, of number of responses on each of the factor independent variables of interest was performed for each publication. In order to avoid convergence failures when the number of responses was either 0 or the number of patients, the number of responses and denominator for analysis were derived from the posterior mean and effective sample size from a Bayesian analysis of a binomial distribution with a minimally informative Beta prior with α = β =.01; i.e., ncr + α. and N + α + β were used as number of responses and sample size, respectively. From this analysis, a log odds ratio estimate for response comparing pairs of categories for each independent variable was obtained, along with corresponding standard error estimates. For variables with more than two categories all pairwise comparisons were made based on a single regression model. Note that not all publications included data relevant to each pairwise comparison. Also, multivariable regression analyses were not possible, as in general only marginal totals of patients and responses were provided even if more than one independent variable of interest was addressed in the publication. Meta-Analysis of event-free survival (EFS) and overall survival (OS) Summary EFS and OS data included patient counts (N) and an estimate if the EFS and OS probability at a fixed reference time, which differed depending on the publication. We made a

8 simplifying assumption that differences among subgroups of independent variables were consistent with the proportional hazard assumption, so that information from EFS and OS evaluated at different times could nevertheless be used to estimate the relative failure rate. EFS or OS estimates with value p with corresponding standard error estimate s were transformed to the complementary log-log scale, L = ln( ln(p )), with corresponding standard error SE = s p ln(p ). The estimated log hazard ratio between two groups was therefore L i L j, with estimated standard error SE i 2 + SE j 2 for groups i and j, respectively. In cases for which a standard error estimate was not available, the standard error was approximated by s = p (1 p )/N, i.e., treating p as a binomial proportion. This is anti-conservative since it ignores censoring and therefore implicitly underestimates the standard error of the estimate. Log odds ratio and log relative failure rate estimates from the response, EFS, and OS analyses were pooled using the random effect method described by DerSimonian and Laird 1 to obtain an overall estimate of these quantities, corresponding standard errors, and Q and I 2 statistics. B. Multivariable analyses of patient-level compiled case series A number of publications provided detailed case-level data for response and for overall survival. These were combined in a separate analysis that allowed multivariable logistic regression analysis for response and multivariable Cox regression analysis for overall survival. The independent variables considered for analyses of CR, EFS, and OS were type of induction therapy, age group, tumor phenotype, and tumor lineage. The transplant variable was not included in these analyses, as it was not relevant (CR) or as a time-dependent variable, could not be properly included in a time-dependent analysis (EFS, OS) because the timing of transplant was not known. Independent variables which were unknown were included in an unknown category in these analyses. Since the majority of patients were classified as satisfying the EGIL definition of MPAL or satisfied both the WHO and EGIL definitions of MPAL, with only a small number classified as WHO MPAL alone, all cases were included in a model with an additional independent variable denoting whether the case was known to satisfy the EGIL, WHO, or EGIL and WHO criteria. Note that these are not mutually exclusive sets. Note also that patients classified as EGIL alone or WHO alone may or may not also satisfy the alternate MPAL criteria as this information is not routinely reported. Both the logistic regression and Cox regression analyses were performed as backward stepwise main effects analyses that initially included all of the variables described above. At each step, variables were removed from the model in order of least significance of the global Wald test of each variables contribution to the model. All independent variables with p < 0.10 were retained in the model. As final steps, individual variables that had been removed were retested to see if they could significantly improve the final model, and interactions between the MPAL definition variable and main effects in the final model were tested for evidence that their effect might differ depending on the definition. Note that while no interaction effects for definition of MPAL were observed, the relatively small number of WHO cases would possibly insufficient to discover plausible interaction effects between WHO and EGIL classification and other independent variables in the model. 1 DerSimonian R, Laird N: Meta-analysis in clinical trials. Control Clin Trials 7: , 1986

9 III. STATISTICAL APPROACH TO ANALYSIS OF TREATMENT ERA As individual treatment year for each case was not available, but rather only the range of years during which each case was treated, an exact analysis of the effect of treatment year on CR rate or on OS was not possible. We could, however, address this indirectly using a simulation approach, wherein we assigned to every case a (uniform) random treatment year within the range of years that applied to the case and performed an analysis, repeated this multiple times (x10,000), and examined the range of results. For the CR analysis, a continuous term for year of treatment was added to the logistic model in manuscript Table 2. For the OS analysis, a continuous term for year of treatment was added to the second Cox regression model in manuscript Table 3. Results of this simulation test are described in Supplemental Results (Table 2).