Results
Search results
We identified 48 studies eligible for this systematic review.22-69 After data extraction and further screening, we found 18 studies containing sufficient non-overlapping outcome data suitable for metaanalysis.24 27 30-34 36 42 48 50 52 61 62 64 66 67 68 Figure 1 shows a PRISMA flow diagram detailing the process of study selection.
HOW LONG DOES IT TAKE FOR CISTANCHE TO WORK FOR KIDNEY PATIENTS?
Study characteristics
All 48 studies were observational cohort in design: 42 studies were retrospective (of which 37 were based on registries with prospectively collected data), five studies were prospective,27 48 51 52 68, and one study lacked clarification concerning the timing of data collection.22 Thirty-eight studies were based on national or regional registry data, three were multisite studies, and seven were single-site studies. Concerning the geographical representation, studies were conducted in Europe (n=24), North America (n=20), South America (n=3), and Oceania (n=1).
Patient characteristics
Study sample sizes had large variability, ranging from 81 to 449937 (median 2040); enrolment of study participants ranged from 1971 to 2016, and the maximum duration of follow-up ranged from three to 25 years. In total, 452119 patients underwent transplantation and 793731 waitlisted patients remained on dialysis. In studies reporting sex, the percentage of male patientswas 61% (n=144748/235789) in the transplantation group and 59% (n=146991/248394) in the waitlisted dialysis group. The mean age of patients was also comparable between the two groups at 51.1 (range 34-73) years and 51.3 (37-73) years for transplantation and dialysis patients, respectively.
Fig 1 | Study selection for systematic review and meta-analysis. Tx=transplantation; WL=waiting list
Quality of studies
The mean quality score of the eligible studies according to the NOS scale was 8.5 (range 7-9) out of 9, representing overall highly quality observational data (supplementary table B). Supplementary table C summarises the baseline study characteristics, patient characteristics, and total quality score. A detailed summary of the statistical procedures used, and covariates adjusted for, across the studies is provided in supplementary table D.
Although 44 studies showed an overall benefit favoring transplantation, 11 of those studies identified a stratum in which transplantation offered no statistically significant benefit over remaining on dialysis. No difference was reported in patients with kidney failure caused by glomerulonephritis25; patients aged 65-70 years28; patients with kidney failure caused by hypertension or a hereditary cause 30; patients with body mass index ≥4135; patients aged ≥70 years with kidney failure caused by glomerulonephritis45; patients receiving standard criteria donor transplantation compared with those on nocturnal haemodialysis49; patients aged ≥70 years on dialysis between 1990 and 199950; patients classified as being at low risk by the American Society of Transplantation52; patients with peripheral arterial disease receiving deceased donor allografts59; patients aged ≥70 years67; and patients with chronic obstructive pulmonary disease, aged ≥70 years, or with kidney failure caused by diabetes.58 No study described an overall lower mortality risk associated with dialysis, with 8% (n=4/48) of studies showing no difference in long-term survival between treatment modalities. Two of those studies were single-site studies published after 1975 but represented data from the early 1970s and were likely to be non-representative of contemporary clinical practice.22 23 One study was based across two Dutch sites representing data from only a small cohort of patients (n=156) between 1990 and 1997 with no regression analysis.32 Finally, one study presented data on patients aged ≥70 years from a French registry between 2002 and 2013.63 Although this study found that risk in the transplantation group had halved by nine months compared with the dialysis group, it was not enough to offset the high perioperative risk associated with transplantation by the end of the study period (month 36).
Systematic review of evidence
92% (n=44/48) of studies reported a significant overall long-term (one year or more) survival benefit from transplantation compared with dialysis. Thirty-one of those 44 studies reported adjusted hazard ratios for all-cause mortality (hazard ratio range 0.18-0.73; 95% confidence range 0.08 to 0.96). Eight studies reported the adjusted risk in discrete periods,25 26 30 34 38 53 59 63, and all found that mortality risk in transplant recipients was significantly higher than in waitlisted dialysis patients immediately after transplantation (hazard ratio range 1.3-17.7 at 0-30 days; 1.5-4.8 at 0-3 months). However, in all eight studies, the increased mortality risk attributable to perioperative and postoperative risk factors was eventually offset, and by one year the mortality risk was significantly lower in transplant recipients (hazard ratio range 0.19-0.49). Fourteen of the 44 studies reported adjusted relative hazards stratified by donor type. In all cases, where respective comparisons were available, studies reported living donor transplantation as conferring a greater survival benefit over deceased donor transplantation,31 35 45 49 55 59 60 61 63 standard criteria donors over extended criteria donors,41 47 55 60 69 and finally hepatitis C virus seronegative over hepatitis C virus seropositive donor transplantation.37 65 Detailed
outcome data with adjusted hazard ratios for each study and stratum are provided in supplementary table D. Benefit was found in all adults aged ≥18, older adults (≥60,24 28 29 38 47 66 67 ≥65,28 55 58 61 and ≥7045 50 62), and population groups with obesity (body mass index ≥30),35 diabetes as comorbidity,43 44 58 systemic sclerosis,39 peripheral arterial diseases,43 60 and patients seropositive for hepatitis C virus.65
Meta-analysis
Of the 18 studies included in the meta-analysis,11 studies presented adjusted regression models reporting both risk (hazard ratio or relative risk) and precision (standard error, P-value, or confidence intervals).30 31 34 36 42 50 61 62 64 66 67 Hazard ratios were not directly reported in seven studies and were extrapolated from the reported Kaplan-Meier survival curves. Overall, the pooled estimate showed that transplantation was associated with a 55% lower long-term mortality risk in patients with kidney failure compared with dialysis (hazard ratio 0.45,
95% confidence interval 0.39 to 0.54; P<0.001), as shown in fig 2. Heterogeneity was significant by the Q statistic (405, df=19; P<0.01) and by I2 (95.3%; P<0.01).
Subgroup analyses
To investigate the significant heterogeneity, we stratified studies by geographical region (continent; fig 2), donor type (living versus deceased; fig 3), and population type (general versus aged ≥60 years; fig 4). The pooled hazard ratio for long-term, all-cause mortality for the transplantation group, compared with the dialysis group, was lowest in Oceania (n=1; hazard ratio 0.19, 0.17 to 0.22), followed by North America (n=2; 0.34, 0.22 to 0.53; Cochran Q statistic 2.81, P=0.09; I2=64.4%, P<0.01) and Europe (n=13; 0.49, 0.40 to 0.60; Cochran Q statistic 131.87,
P<0.01; I2=90.1%, P<0.01), which were comparable.
However, the result between the two modalities was not statistically different in South America (n=2; 0.67, 0.33 to 1.35; Cochran Q statistic=29.2, P<0.01; I2=93.2%, P<0.01). Although subgroup analyses by continent helped to explain some of the heterogeneity, it remained largely significant. Stratification by donor type showed living donor transplantation to have a
lower long-term mortality risk compared with deceased donor transplantation, with hazard ratios of 0.30 (0.23 to 0.39) and 0.45 (0.38 to 0.55), respectively. However, the result was not significant owing to overlapping confidence intervals and only a single study contributing data for living donor transplantation. Furthermore, this analysis was also unable to account
for the significant heterogeneity present. Finally, stratification by population type showed a similar trend to the previous subgroup analyses, with both the general (hazard ratio 0.47, 0.38 to 0.59) and ≥ 60-year group (0.42, 0.34 to 0.53) showing transplantation as conferring a lower mortality risk compared with dialysis; however, stratification failed to elucidate the cause of the significant heterogeneity.
Fig 2 | Forest plot of hazard ratios (with 95% confidence interval (CIs)) for transplantation versus waitlisted dialysis, stratified by geographical region. lnHR=log hazard ratio; seHR=standard error hazard ratio
