Reduction of Arterial Stiffness After Kidney Transplantation: A Systematic Review and Meta‐Analysis

Study Design

We conducted a systematic review based on methodological recommendations from the Cochrane Handbook for Systematic Reviews of Interventions21, on a registered protocol in Prospero (CRD42016045383) and reported according to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta‐Analyses) statement.22

Eligibility Criteria

We included in our systematic review before‐after design studies conducted in adult (age ≥18) end‐stage kidney disease patients who underwent KTx with functioning graft and with repeated measurements of arterial stiffness parameters. The outcomes of interest were arterial stiffness parameters including PWV (the clinical gold standard) and other indirect parameters such as central PP and increased wave reflection as measured by AIx11 (see Table 1 for details on population, intervention, comparators, and outcome).

Information Sources and Search Strategy

We performed a search using electronic databases (Medline, Cochrane Library, Embase, and the Web of Science) from their inception until January 2016. The search strategy is based on keywords related to the intervention (renal transplantation and synonyms) and the outcomes (arterial stiffness, its parameters, and their synonyms). A search strategy was first set for PubMed/Medline and Embase and then adapted to Cochrane Library and the Web of Science with no restriction. The search strategy established for PubMed/Medline is shown in Table 2. We then hand searched additional references from the reference lists of relevant articles and in gray literature (Google Scholar, thesis repositories including Thesis portal Canada, EtHOS, DART‐Europe E‐Thesis Portal, the National Library of Australia's Trove, and ProQuest Dissertations & Theses Global).

Study Selection and Data Management

After removing duplicates of identified records from our search strategies using EndNote (version x7.2.1, Thomson Reuters, New York, 1988‐2014), 2 independent reviewers (A.S. and C.F.) screened each study by title and abstract using standardized and pilot‐tested screening forms. Full text was also screened when title and abstract were not enough for inclusion of a study in the review under the eligibility criteria described above.

Data Extraction and Risk of Bias Assessment

Data of included studies were then independently extracted using a standardized and pilot‐tested data extraction form. In each step discrepancies between the 2 reviewers were resolved through consensus or with the involvement of a third reviewer (M.P.D.), as required. Extracted data included information on the study, characteristics of the study population and intervention, pre‐KTx therapies, and arterial stiffness parameters from eligible studies.

Two authors (A.S. and M.P.D.) assessed the quality (risk of bias) in included studies with a tool developed by the National Health Institute to assess the quality in before‐after design studies.23 Likewise, we assessed conflicts of interest using information on the source of funding for each study.

Data Synthesis

We analyzed and summarized the study data with Review Manager Software (Revman, Computer program, Version 5.3,The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark, 2014). To assess the association between KTx and arterial stiffness, we considered the mean difference between arterial stiffness outcomes (PWV [overall, central, peripheral, and brachial‐ankle], central PP, and AIx) assessed after KTx and before KTx as effect size. PWV was first analyzed globally regardless of the arterial territory or the period of assessment (overall PWV). Second, we considered a posteriori each arterial territory reported as an outcome: central PWV (composed of carotid‐femoral PWV and aorta‐femoral PWV), peripheral PWV (composed of carotid‐radial PWV and femoral‐distal PWV), and brachial‐ankle PWV. Two global analyses were performed for each outcome for which more than 2 measurements were performed within a study. Indeed, we included either the earliest measure after the first week (first global analysis) or the latest measure within 1 year (second global analysis). Because we anticipated a potential heterogeneity between the included studies, we estimated the pooled mean difference and its 95% confidence interval using a random‐effects model with inverse variance method.24

Statistical heterogeneity was assessed using the Higgins I² and interpreted as low between 0% and 30%, moderate between 30% and 60%, considerable between 60% and 80%, and substantial from 80% and above.21, 25

For each outcome of interest, we explored the publication bias using the Begg rank correlation test and the Egger regression as appropriate given the known limitations of these methods26 using an available macro for SAS.27 All tests of statistical inference were 2‐sided with an α of 5%.

Subgroup Analyses, Meta‐Regression, and Sensitivity Analyses

Subgroup analyses were performed for each outcome according to the period of assessment, which was decided in post hoc analysis using the most used period (≤3 months, at 6 months, and at 12 months post‐KTx).

A meta‐regression was performed in order to quantify the amplitude of influence of some potential factors modifying the effect size estimated in meta‐analysis. We considered period of assessment, arterial territory of PWV, and changes in mean arterial pressure (ΔMAP) as independent variables. Each of these independent variables was analyzed alone in a univariate meta‐regression. Then, ΔMAP was analyzed in a multivariate meta‐regression with period of assessment and/or type of PWV as covariates. All meta‐regressions were performed in SAS Version 9.4 (Copyright © 2015, SAS Institute Inc, Cary, NC) using general linear mixed model with random effects.28 MAP was estimated by the Lian formula (MAP=2/3 DBP+1/3 SBP) in studies that did not report it but for which systolic and diastolic brachial blood pressures were available. Changes in MAP were estimated by computing the absolute difference between the MAP post‐KTx and MAP pre‐KTx. Sensitivity analyses in meta‐analysis were performed by removing studies 1 by 1 to see the impact on the estimated pooled effect size and its statistical heterogeneity. Likewise, sensitivity analyses were performed in meta‐regression by removing studies with poor quality (high risk of bias).

Grading of the Evidence

Two reviewers assessed the overall quality of evidence for each outcome using the GRADE tool.29 Five domains were considered to assess the quality of evidence: quality (risk of bias), inconsistency, indirectness and imprecision, which can be considered serious or very serious, and publication bias, which can be undetected, likely, or very likely. The quality of evidence of observational studies begins low. Although, grading upward may be warranted if the magnitude of the treatment effect is very large, if there is evidence of a dose‐response relation, or if all plausible biases would decrease the magnitude of an apparent treatment effect.29

Study Selection and Description of Included Studies

From the 868 citations identified through electronic databases, we included 12 unique studies in meta‐analysis (see details in Figure 1), with sample sizes ranging from 18 to 168 observations (9‐84 pairs). All studies were quasiexperimental with before‐after design study and published in English between 2003 and 2016. Seven studies8, 9, 20, 30, 31, 32, 33 assessed aortic PWV (central), 3 studies8, 9, 34 assessed peripheral PWV (carotid‐radial or femoral‐distal PWV), and 4 studies35, 36, 37, 38 assessed brachial‐ankle PWV. Two studies8, 9 assessed both central and peripheral PWV, and 3 studies20, 33, 35 used multiple periods of assessment. Central PP was reported in 4 studies8, 9, 20, 32 and AIx in 7 studies.8, 9, 20, 31, 32, 33, 34 All studies were considered to have a good quality (low risk of bias), except 1 considered as fair quality (moderate risk of bias) and 230, 34 adjudicated as poor quality (high risk of bias). Tables 3 and 4 show the detailed characteristics of the included studies.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Flow diagram. The figure shows the selection process for the studies included in the meta‐analysis.

Pulse Wave Velocity

Overall PWV.

Analyses of overall PWV regardless of the vascular territory, taking into account either the earliest measure or the latest measure within a study, showed globally a significant reduction by 1.17 m/s (0.64‐1.71) and by 1.20 m/s (0.67‐1.73), respectively, with considerable heterogeneity (I²=72; Figures 2 and 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

First global analysis of pulse wave velocity. Pulse wave velocity (PWV) is analyzed overall and according to the arterial territory. In the first global analysis, we included the earliest measure reported in studies using multiple measurements (studies from Kaur at 3 months and Ro at 6 months). CI indicates confidence interval; mo, months.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Second global analysis of pulse wave velocity. Pulse wave velocity (PWV) is analyzed overall and according to the arterial territory. In the second global analysis, we included the latest measure reported in studies using multiple measurements (studies from Kaur at 6 months and Ro at 12 months). CI indicates confidence interval; mo, months.

Almost all studies were adjudicated to have a good or fair quality, except the study from Keven30 and that from Covic34 in which the numbers of participants before and after KTx were different (poor quality). When those studies were excluded in sensitivity analyses, overall PWV remained significantly decreased globally by 1.20 m/s (0.57‐1.83, I²=75%) and 1.23 m/s (0.61‐1.85, I²=75%) for the earliest measure and latest measure within a study, respectively (Table 5). While remaining significantly decreased, the statistical heterogeneity of overall PWV was reduced from considerable to moderate when the study from Kovacs33 was excluded and to low when the 2 studies from Zoungas9 and Kovacs33 were excluded (Table 5).

In subgroup analysis according to the period of assessment post‐KTx, overall PWV decreased at ≤3, 6, and 12 months post‐KTx, with substantial, low, and considerable heterogeneity, respectively. Although it remained significant in sensitivity analyses, the statistical heterogeneity of overall PWV effect size was reduced to low at 3 months or less and at 12 months post‐KTx (Table 6). Publication bias was likely according to Egger regression but unlikely with Begg rank correlation. The overall strength of evidence was considered to be low for overall PWV (Tables 7 and 8).

View this table:

• View inline
• View popup

Table 5. Main and Sensitivity Global Analyses of Arterial Stiffness Outcomes

View this table:

• View inline
• View popup

Table 6. Subgroup and Sensitivity Analyses of Arterial Stiffness Outcomes According to the Period of Assessment

View this table:

• View inline
• View popup

Table 7. Test for Publication Bias in the Estimate of Pulse Wave Velocity, Central Pulse Pressure, and Augmentation Index

View this table:

• View inline
• View popup

Table 8. Summary of Grading the Evidence of Findings in PWV, Central PP, and AIx: Kidney Transplantation Compared to Pretransplantation for Reduction of Arterial Stiffness

Central Pulse Pressure

Analyses of central PP at any period of assessment, considering either the earliest measure or the latest measure within a study, was not significantly reduced (Table 5). In sensitivity analysis, after exclusion of the study from Zoungas9 in which heart rate had significantly declined after KTx, central PP significantly decreased by 6.0 mm Hg (2.2‐9.9, I²=0%) and 7.0 mm Hg (2.8‐11.2, I²=18%), respectively, when considering the earliest or the latest measure within a study. The statistical heterogeneity was also considerably reduced to low (Table 5).

Subgroup analyses of central PP, although limited by the small number of studies, were performed with regard to the period of assessment post‐KTx (Table 6). Publication bias was unlikely according to both tests. All studies were adjudicated to have a good quality for central PP. The overall strength of evidence was considered as very low for central PP (Tables 7 and 8).

Augmentation Index

The AIx at any period of assessment post‐KTx, considering either the earliest measure or the latest measure within a study, declined by 10.0% (6.7‐13.4, I²=61%) and 10.5% (6.9‐14.1, I²=64%), respectively (Table 5). All studies were adjudicated to have a good quality for AIx except the study from Covic,34 which was considered to have poor quality. When the latter was excluded, AIx remained significantly decreased globally (Table 5). Although it remained significantly reduced, the statistical heterogeneity of both global effect sizes of AIx was considerably reduced in sensitivity analysis. Although limited, subgroup analyses were performed according to the period of assessment (Table 6). Publication bias was unlikely according to both tests. The overall strength of evidence was considered as low for AIx (Tables 7 and 8).

Meta‐Regression

Pulse Wave Velocity.

The between‐study variance in PWV was mostly explained by the period of assessment (68% when considering all studies except Covic34 and 80% when studies from both Covic34 and Keven30 were excluded in univariate analysis). The percentage of variance increased to 78% and 87%, respectively, when period of assessment and type of PWV were considered in the same model (Table 9). In univariate meta‐regression, only ΔMAP was associated with the effect size of PWV observed. In the model including all studies except the 1 from Covic,34 in which MAP was not available, this association was no longer significant with adjustments for either type of PWV or period of assessment. In the model excluding both studies from Covic34 and Keven,30 the association between ΔMAP and the effect size of PWV remained significant when adjusting for the type of PWV but was no longer significant when adjusting for the period of assessment (Figure 4).

View this table:

• View inline
• View popup

Table 9. Estimation of Study Residual Variance According to the Arterial Stiffness Parameters

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Association between change in mean arterial pressure (ΔMAP) and the effect size of pulse wave velocity (PWV). The figure shows the association between ΔMAP and the effect size of pulse wave velocity analyzed by meta‐regression. MAP indicates mean arterial pressure.

Sources of Heterogeneity

Statistical heterogeneity was present in the pooled data, coming from different sources. First, different vascular territories were used to assess PWV. Indeed, 7 studies8, 9, 20, 30, 31, 32, 33 looked at measures of aortic stiffness: the aorta is an elastic capacitive vessel and is more heavily affected with age and pathological conditions such as hypertension, CKD, and diabetes mellitus.4, 10, 39, 40, 41, 42 Three studies8, 9, 34 evaluated medium‐sized muscular vascular beds, which are physiologically stiffer and generally less affected by age and pathological conditions.4, 10, 39, 40, 41, 42 Finally, 4 studies35, 36, 37, 38 looked at brachial‐ankle PWV, which gives a general index of overall stiffness. Three studies20, 33, 35 performed measurement at different periods post‐KTx, and 2 studies8, 9 used multiple sites.

We first performed meta‐analyses regardless of the arterial territory and then by considering each arterial territory of PWV as an outcome using a random‐effect method. To deal with multiple measurements within a study, we considered in our global analysis the first measure after 1 week (first global analysis) or the latest measure until 1 year (second global analysis). In sensitivity analyses the statistical heterogeneity of the different effect sizes was considerably reduced in overall PWV and central PWV, from considerable to moderate and from substantial to moderate, respectively. But central PWV remained significantly decreased only in the latest period of assessment within a study. Likewise, central PP and AIx significantly decreased in sensitivity analyses, and the statistical heterogeneity of the effect sizes was considerably reduced.

Mean Arterial Pressure, Arterial Stiffness, and Period of Assessment

Vascular remodeling in CKD is complex, and it involves remodeling that is mediated not only through the effects of hypertension but also through mineral disorder–induced vascular calcification,40 endothelial dysfunction,43 and alterations of extracellular matrix by uremic toxins such as advanced glycation end products.44 A successful KTx improves fluid overload and improves the metabolic pathways that are potentially involved in CKD‐related vascular stiffness.7, 8 An improvement in arterial stiffness post‐KTx does suggest that arterial stiffness is potentially reversible. However, vascular stiffness is also affected by the operating MAP without any changes in the structure of vascular wall (ie, functional stiffness). In this systematic review only 2 studies reported a mean difference of PWV adjusted for MAP.8, 9 It is for this reason that we performed a meta‐regression to evaluate the association between changes in MAP (estimated by the Lian formula in 3 studies30, 33, 36) and the effect size of arterial stiffness parameters. A positive association between the effect size of arterial stiffness parameters and changes in MAP suggests that the improvement of arterial stiffness parameters may be related to either functional or structural improvement of vascular wall properties. Although this association was significant in univariate analysis, it was no longer significant after adjustment for the combined effect of the period of measurements and/or type of PWV. The association between changes in MAP and the effect size of central PP could not be evaluated due to a small number of studies, and we did not observe an association between changes in MAP and the effect size of AIx. In fact, the period of assessment seemed to explain the most important part of the residual variance in PWV, central PP, and AIx. These findings suggest that vascular wall remodeling post‐KTx may not follow a homogeneous pattern. Accordingly, it is possible that other factors, such as post‐KTx medication or activation of immune system (by HLA mismatching or infection), outweigh the initial improvement of uremia.

Risk of Bias

Risk of bias is a major concern for the internal validity of the results. Almost all studies were adjudicated to have a good (low risk of bias) or fair quality (moderate risk of bias). Indeed, the risk of selection bias was adjudicated as low in all except 2 studies, where it was adjudicated as high due to an important difference between the number of patients in baseline and after KTx.30, 34

Publication bias was likely for overall PWV, central PWV, brachial‐ankle PWV, and AIx. It was unlikely for peripheral PWV and central PP. Finally, the strength of evidence was classified as low for overall PWV, brachial‐ankle PWV, and AIx but very low for central PWV, peripheral PWV, and central PP.

Strengths and Limitations

The strengths of this review include the completeness of the search including hand searching of the gray literature and references of relevant citations, the statistical approach including random effect, and the confirmation of the findings in most of the subgroups, the strength of evidence classified as low for overall PWV, brachial‐ankle PWV, and AIx.

There are, however, some limitations in the interpretation of the results of this meta‐analysis. First, the use of various devices for the assessment of arterial stiffness parameters, kidney source (living or cadaveric), dialysis modality and duration before KTx, the timing of assessment (before and after transplantation), as well as immunosuppressive regimen could all constitute potential sources of heterogeneity. The impact of immunosuppressive drugs, considered as cointervention, could not be explored because little information was provided by the studies. Likewise, the impact of dialysis modality and duration before KTx could not be assessed. Blood pressure was not available for 1 study.34 Furthermore, the use of the Lian formula to calculate MAP also has its limits when applied to the average of a group's systolic and diastolic blood pressures. The number of included studies could not provide a statistical power necessary to detect all the differences, so some results of meta‐regression are at risk of type 2 error. Finally, we could not take into account, in the same model, many sources of heterogeneity in meta‐regression by using a random effect for each.

In summary, this study supports the conclusion that CKD‐related arterial stiffness may be reversible after a successful KTx. This is an important finding because it gives hope for a potential therapeutic intervention, targeting arterial stiffness in CKD. However, an individual data meta‐analysis may be required to perform additional analysis using a more sophisticated approach in determining the blood pressure–independent effect of a regression in arterial stiffness after KTx.45