Supplemental information
Risk of bias across studies:
We used the Cochrane risk of bias tool36 to evaluate the quality of RCTs in six domains: (i) selection bias (how the random sequence of numbers was generated and how allocations were concealed), (ii) detection bias (blinding of participants and personnel), (iii) performance bias (blinding of outcome assessment), (iv) attrition bias (how incomplete outcome data was handled), (v) reporting bias (selective outcome reporting), and (vi) other bias (e.g., baseline imbalance, carryover, funding, etc.). Studies were described as having low, high, or unknown risk. We computed an overall risk of bias score for any given study by assigning scores to each of the domains assessed. For each domain there were three response options offered, low risk of bias (score 2), high risk of bias (score 0), or unclear risk of bias (score 1). We used the New Castle-Ottawa Scale (NOS) to evaluate the quality of the non-randomized studies (NRS) in 3 domains: (i) selection and detection bias, (ii) bias due to confounding (comparability of the interventions), (iii) performance and attrition bias (blinding of outcome assessment and missing data –more than 20% drop-outs).37 The NOS uses a ‘star system’ to assess the quality of the included studies across eight items categorized into the above mentioned 3 domains. A maximum of one star for each item is awarded to studies of highest quality (exception - item related to comparability allows the assignment of two stars). The NOS ranges from 0 to 9 stars. A summary of the number of stars was used to provide the overall risk of bias for each study presented in table 1.
Patient characteristics:
The donor type was from un-manipulated marrow from HLA-identical siblings in 367 (57.1%) participants,14, 15, 17-23, 26-29, 31 matched unrelated donor in 81 (12.6%),17-23, 26, 27, 31 cord donor in 24 (3.7%),21-23 and unknown in 171 (26.6%).16, 24, 25, 30, 32 The source of the stem cells were: from bone marrow in 49 participants (7.6%),17-20 from peripheral blood in 120 (18.7%),17-20 and unknown in 474 (73.8%).14-16, 21-32Race was reported in only one study (3 publications) where 59 (97%) were Caucasians and 2 (3%) Asian/American Indian.21-23 One study mentioned smoking status (2 publications) in which 12 participants (17%) were active smokers, 27 ex-smokers (37%), and 33 (46%) non-smokers.26, 27Primary diseases included acute myeloid leukemia 19.0% (n=122), acute lymphoblastic leukemia(ALL) 5.6% (n=36), ALL in second complete remission 0.5% (n=3), chronic myeloid leukemia (CML) 12.7% (n=81), CML in chronic phase 1.9% (n=12), myelodysplastic syndrome 1.9% (12), Hodgkin’s lymphoma 1.6% (n=10), Burkitt’s lymphoma 0.1% (n=1), aplastic anemia 0.8% (n=5), severe aplastic anemia 0.3% (n=2), chronic lymphocytic leukemia/small lymphocytic leukemia 1.2% (n=8), non-Hodgkins lymphoma 2.0% (n=13), chronic myelomonocytic leukemia 0.1% (n=1), lymphoma but type not specified 1.9% (n=12), myelofibrosis 0.6% (n=4), myelodysplastic syndrome 0.9% (n=6), paroxysmal nocturnal hemoglobinuria 0.1% (n=1) and other malignancies/hematological diseases not specified 48.8% (n=314).
Subgroup analysis:
Zoledronic acid combined with Ca/VitD versus Ca/VitD alone:
Four studies reported on this comparison but only 3 studies were used in the meta-analysis, 15, 21-23, 28, 29 since one study had insufficient information.16 There wasa statistically significant lower mean BMD percentage change of the lumbar spine and femoral neck in the zoledronic acid group compared to the control group at 12 months (MD 11.5, 95% CI 9.3, 13.8 and 9.4, 95% CI 8.2, 10.7 respectively). Similar differences were observed with T-score and Z-score between the groups (Table 4).
A greater reduction in bone alkaline phosphatase levels (at 12 months) and urinary hydroxyproline excretion (at 6 and 12 months) were reported in the zoledronic acid group compared to the control group (Supplemental table 1).
Pamidronate combined with Ca/VitD and HRT versus Ca/VitD and HRT:
Two studies reported on this comparison .18-20, 26, 27 Patients in the pamidronate group had a statistically significant lower mean BMD percentage change of the lumbar spine at 6 and 12 months; femoral neck at 3 and 6 months; trochanter at 6 and 12 months, and total hip at 3-24 months compared to the control group (Table 4).
When analyzing BMD by quantity of steroids received, patients in the pamidronate group under medium dose steroids had a lower mean BMD percentage change compared to the control group in the lumbar spine, femoral neck and total hip at 12 months (MD 6.4, 95%CI 1.5, 11.3; 10.5, 95%CI 6.1, 14.8 and 7.4, 95%CI 3.8, 11.0) respectively. Furthermore patients in the pamidronate group under high dose steroids had a lower mean BMD percentage change compared to the control group in the lumbar spine, femoral neck and total hip at 12 months (MD 10.9, 95%CI 3.8, 18.1; 10.2, 95%CI 3.9, 16.5 and 13.0, 95%CI 7.7, 18.2) respectively. Similarly, patients in the pamidronate group under long duration of cyclosporine had a lesser mean percentage change in BMD compared to the control group in the lumbar spine, femoral neck and total hip at 12 months (MD 8.7, 95%CI 4.6, 12.9; 10.1, 95%CI 6.5, 13.8 and 6.8, 95%CI 3.4, 10.1) respectively.
A greater reduction in various bone turnover and resorption marker levels was observed in the pamidronate group compared to control group. Assessment of adequacy of treatment was measured using markers such as estradiol, sex hormone-binding globulin , serum osteocalcin and serum total alkaline phosphatase .The overall survival was greater in patients in the pamidronate group compared to patients in the control group at 1 year (MD 19.2, 95%CI 1.6, 36.8) (Supplemental Table 1).
Risedronate combined with Ca/VitD versus Ca/VitD alone:
Three studies reported on this comparison.14, 15, 24, 25 Patients in the risedronate group had a statistically significant lower mean BMD percentage change of the lumbar spine, total hip, and femoral neck from baseline compared to the control group at 6 and 12 months (Table 4). Similar differences were observed with T-score between the groups (Table 4). Patients in the risedronate group had a lower change in urinary hydroxyproline excretion compared to the control group at 12 months (MD -18.9, 95%CI -33.6, -4.3) (Supplemental Table 1).
Zoledronic acid combined with Ca versus Ca alone:
One non randomized study reported on this comparison.17 Patients in the zoledronic acid combined with Ca group had a statistically significant lower mean Z-score percentage change of the lumbar spine and femoral neck when compared to the control group at 6, 12 and 24 months (Table 4).
Ca/VitD combined with HRT versus Ca/VitD alone; 15Ca plus calcitonin versus no intervention;31Ca alone versus no intervention:31No statistically significant differences were observed in BMD and fracture occurrence between all 3 of the comparison groups.