www.nature.com/scientificreports




                        OPEN       Cholesterol-Lowering Treatment in
                                   Chronic Kidney Disease: Multistage
                                   Pairwise and Network Meta-
Received: 25 October 2018
Accepted: 5 June 2019              Analyses
Published: xx xx xxxx
                                   Francisco Herrera-Gómez 1,2, M. Montserrat Chimeno3, Débora Martín-García4,
                                   Frank Lizaraso-Soto5, Álvaro Maurtua-Briseño-Meiggs6, Jesús Grande-Villoria2,
                                   Juan Bustamante-Munguira7, Eric Alamartine8, Miquel Vilardell9, Carlos Ochoa-Sangrador                             10

                                   & F. Javier Álvarez 1,11
                                   Pairwise and network meta-analyses on the relationship between the efficacy of the use of statins
                                   with or without ezetimibe and reductions in low-density lipoprotein cholesterol (LDLc) and C-reactive
                                   protein (CRP) in patients with chronic kidney disease (CKD) are presented. In the pairwise meta-
                                   analysis, statins with or without ezetimibe were shown to be efficacious in reducing major adverse
                                   cardiovascular events (MACE) in patients with CKD and an estimated glomerular filtration rate (eGFR) of
                                   less than 60 ml/min/1.73 m2, in the context of both primary prevention [odds ratio (OR)/95% confidence
                                   interval (95% CI)/I2/number of studies (n): 0.50/0.40–0.64/0%/6] and primary/secondary prevention
                                   (0.66/0.57–0.76/57%/18). However, in the Bayesian network meta-analysis, compared to the placebo,
                                   only atorvastatin 80 mg daily and atorvastatin and rosuvastatin at doses equivalent to simvastatin
                                   20 mg daily reduced the odds of MACEs in this patient population. The network meta-analysis for LDLc
                                   and CRP treatment objectives also showed that, regardless of eGFR and excluding dialysis patients,
                                   the number of MACEs decreased in patients with CKD, with reductions in both LDLc and CRP of less
                                   than 50% (surface under the cumulative ranking (SUCRA)/heterogeneity (vague)/n: 0.77/0.14/3). The
                                   evaluation of the benefits of drugs may lead to individualized therapy for CKD patients: Cholesterol-
                                   lowering treatment for CKD patients with high levels of both LDLc and CRP is suggested.


                                   Cardiovascular disease (CVD) occurs early in individuals with chronic kidney disease (CKD)1: Inflammation
                                   is central to the development of the first and subsequent CVD events in CKD patients (as well as in non-CKD
                                   patients)2 and plays an essential role in linking CVD and kidney function decline1. However, classic predic-
                                   tors of coronary and cerebrovascular events seem to perform better in the early stages of the CKD spectrum3.
                                   Importantly, in CKD patients, potential new markers of CVD events that are derived from analyses of the com-
                                   plex pathophysiology of CKD act as independent prognostic factors4,5. Despite advances in the qualification of
                                   biomarkers for the use of medicines6, the translation of real treatment effect modifiers for CVD in the CKD pop-
                                   ulation is needed to evaluate the effects of the available therapies7.



                                   1
                                     Pharmacological Big Data Laboratory, Pharmacology and Therapeutics, Faculty of Medicine, University of
                                   Valladolid, Valladolid, Spain. 2Nephrology, Hospital Virgen de la Concha – Sanidad de Castilla y León, Zamora,
                                   Spain. 3Internal Medicine, Hospital Virgen de la Concha – Sanidad de Castilla y León, Zamora, Spain. 4Cardiovascular
                                   risk unit, Hospital Clínico Universitario de Valladolid – Sanidad de Castilla y León, Valladolid, Spain. 5Instituto de
                                   Investigación de la Facultad de Medicina Humana, Universidad de San Martín de Porres, Lima, Peru. 6Woodland
                                   Medical Practice – NHS, Lincolnshire, United Kingdom. 7Cardiac Surgery, Hospital Clínico Universitario de Valladolid
                                   – Sanidad de Castilla y León, Valladolid, Spain. 8Nephrology, Dialysis and Transplantation, Centre Hospitalier
                                   Universitaire de Saint-Etienne, Saint-Priest-en-Jarez, France. 9Medicine, Faculty of Medicine, Universidad Autónoma
                                   de Barcelona, Barcelona, Spain. 10Research unit, Hospital Virgen de la Concha – Sanidad de Castilla y León,
                                   Zamora, Spain. 11CEIm Área de Salud Valladolid Este, Hospital Clínico Universitario de Valladolid, Valladolid, Spain.
                                   Correspondence and requests for materials should be addressed to F.H.-G. (email: fherrera@med.uva.es)


        Scientific Reports |   (2019) 9:8951 | https://doi.org/10.1038/s41598-019-45431-5                                                              1
www.nature.com/scientificreports/                                                              www.nature.com/scientificreports


                              The risk of coronary events in patients with CKD is high8, and the related evidence is in favor of reducing
                           CVD risk using cholesterol-lowering treatment strategies9. However, the relationship between serum levels of
                           low-density lipoprotein cholesterol (LDLc) and CVD is more difficult to understand in the context of CKD;
                           malnutrition and chronic inflammation are additional factors to consider as CKD progresses, and these factors
                           increase the risk of death in CKD patients9.

                           Current context and study objective. The Kidney Disease–Improving Global Outcomes (KDIGO)
                           guidelines recommend the use of statins with or without ezetimibe for adults 50 years of age or older with CKD
                           and an estimated glomerular filtration rate (eGFR) of lower than 60 ml/min/1.73 m2 (KDIGO GFR categories
                           G3a to G5) who are not being treated with chronic dialysis or kidney transplantation and not aiming to achieve
                           specific LDLc targets10. Statins are also recommended for individuals older than 50 years when the eGFR is 60 ml/
                           min/1.73 m2 or higher (KDIGO GFR categories G1 to G2).
                               Notably, in the evaluation of the impact of cholesterol-lowering treatment along the wide spectrum of CKD,
                           an unanswered question remains: are statins with or without ezetimibe similarly efficacious in patients with an
                           eGFR lower than 60 ml/min/1.73 m2 and in patients with an eGFR of 60 ml/min/1.73 m2 or higher? If not, which
                           cholesterol-lowering treatment strategies are the best for lowering CVD risk in patients with an eGFR lower than
                           60 ml/min/1.73 m2? Furthermore, is the cardiovascular efficacy of statins with or without ezetimibe related to
                           reductions in LDLc and C-reactive protein (CRP) in CKD patients as is the case for LDLc in non-CKD patients?
                           Conventional pairwise meta-analysis may be insufficient to answer these questions, as the efficacy between multi-
                           ple interventions that compare cholesterol-lowering treatments and placebo/no treatment cannot be studied sep-
                           arately. Only one possibility exists based on this technique: to compare cholesterol-lowering treatments as a whole
                           with placebo/no treatment. The same concept applies for the reduction in LDLc and CRP with respect to the
                           sole reduction in LDLc when meta-analysis on exposures is performed. Network meta-analysis provides a viable
                           solution even though some interventions/exposures have never been compared head-to-head in randomized con-
                           trolled trials that evaluate the cardiovascular efficacy of statins with or without ezetimibe in the context of CKD.
                               These multistage pairwise and network meta-analyses presents (1) a summary of the efficacy of statins with or
                           without ezetimibe along the wide spectrum of CKD and (2) the relation of this treatment to reductions in LDLc
                           and CRP with the aim of evaluating the real impact of cholesterol lowering-treatment on CVD in CKD patients.

                           Results
                           Literature search. Data from 106 050 participants from 22 randomized controlled trials (RCT) are pre-
                           sented here11–35. Figure 1 presents the Preferred Reporting Items for Systematic reviews and Meta-Analyses
                           (PRISMA) flowchart corresponding to the two systematic reviews36. Irrelevant citations were mostly interven-
                           tional or observational studies of patients with early CKD (KDIGO GFR categories G1 to G2) that did not assess
                           CVD-related outcomes or studies involving patients without CKD but affected by other pathologies. All of the
                           following RCTs that were eligible at the treatment objectives evaluation phase had already been included at the
                           cholesterol-lowering treatment efficacy evaluation phase: The Study of Heart and Renal Protection (SHARP)17,
                           the Air Force/Texas Coronary Atherosclerosis Prevention study (AFCAPS/TexCAPS)23,37, the Collaborative
                           Atorvastatin Diabetes Study (CARDS)26,38, and the Justification for the Use of Statins in Primary Prevention–an
                           Intervention Trial Evaluating Rosuvastatin (JUPITER) study28,39.
                               All these pharmaceutical-industry-sponsored trials were published in peer-reviewed journals and corre-
                           sponded to the phase 3 or 4 evaluation of various statins with or without ezetimibe, with both administered per
                           os. Nineteen of these trials separately assessed persons with normal kidney function (NKF)/CKD KDIGO GFR
                           categories G1–G2 and patients with CKD KDIGO GFR categories G3a–G5 but not patients treated with chronic
                           dialysis16–35; 12 of the abovementioned studies were post hoc analyses of other RCTs conceived to evaluate par-
                           ticipants from the general population or patients with other pathologies22–33. Five RCTs studied dialysis patients
                           (hemodialysis and peritoneal dialysis modalities)13–18, and 1 RCT evaluated patients who underwent kidney
                           transplantation11,12. The eligible RCTs are presented in Table S1.

                           Cholesterol-lowering treatment efficacy evaluation.                    Moderate- to high-quality trials (Table S2)
                           indicate that atorvastatin 80 mg daily, simvastatin 20 mg daily combined with ezetimibe 10 mg daily, simvastatin
                           20 mg daily and milligram-equivalent doses of fluvastatin, atorvastatin, rosuvastatin, pravastatin and lovastatin
                           all reduced the incidence of major adverse cardiovascular events (MACEs) in patients with CKD KDIGO GFR
                           categories G3a–G5 (excluding dialysis patients)16–35. Figure 2a,b shows the pooled odds ratios (OR) and 95%
                           confidence intervals (95% CI) corresponding to pairwise meta-analysis of all trials conducted in this population
                           (primary/secondary prevention; 0.66, 0.57 to 0.76) and those that enrolled only patients without known CVD
                           (primary prevention; 0.50, 0.40 to 0.64). The heterogeneity (I2) was 57% and 0%, respectively. However, Bayesian
                           network meta-analysis confirmed that compared to placebo, only atorvastatin 80 mg daily and atorvastatin and
                           rosuvastatin at doses equivalent to simvastatin 20 mg daily reduced the odds of incident MACEs when the eGFR
                           was lower than 60 ml/min/1.73 m2 (excluding dialysis patients). The ORs and 95% CIs for the abovementioned
                           treatments were 0.35, 0.16 to 0.74; 0.50, 0.26 to 0.92; and 0.67, 0.40 to 0.92, respectively. Figure 3 shows a league
                           table ranking all statins and the simvastatin/ezetimibe combination on the basis of the surface under the cumu-
                           lative ranking (SUCRA)40. In the generated network diagram (Fig. 4), the heterogeneity (vague) was 0.08 (Fig. 5),
                           and there was no relevant inconsistency (Fig. 6).
                               In persons with NKF/CKD KDIGO GFR categories G1–G217,22–35 and in dialysis patients13–18, the ORs and
                           95% CIs for MACEs were 0.78, 0.71 to 0.85 and 0.95, 0.81 to 1.07, respectively (Figs S1 and S2).

                           Treatment objectives evaluation. Serum levels of LDLc and/or CRP decreased in response to treatment
                           with simvastatin 20 mg daily combined with ezetimibe 10 mg daily17, lovastatin 40 mg daily37, atorvastatin 10 mg


Scientific Reports |   (2019) 9:8951 | https://doi.org/10.1038/s41598-019-45431-5                                                             2
www.nature.com/scientificreports/                                                              www.nature.com/scientificreports




                           Figure 1. The PRISMA flowcharts presenting the selection processes used in the two systematic reviews.
                           PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.


                           daily38, and rosuvastatin 20 mg daily39, regardless of eGFR and history of diabetes or CVD. Most patients with
                           early CKD (eGFR ≥ 60 ml/min/1.73 m2) had high LDLc levels before treatment17,37–39. A reduction in CRP of
                           more than 50% was not observed. Reductions in LDLc and/or CRP of less than 50% were noted in patients with
                           CKD in any KDIGO GFR category37–39.
                              Pairwise meta-analysis showed that, regardless of eGFR and excluding dialysis patients, reductions in both
                           LDLc and CRP of less than 50%, and reductions in LDLc (</≥50%) and CRP (<50%) separately were associated
                           with a low frequency of MACEs (Figs S3–S6)17,37–39. However, in the network meta-analysis, only the combined
                           reduction in LDLc and CRP appeared to have the most pronounced impact on this outcome37–39. The SUCRA was
                           0.77 (Table 1)40. In the network diagram built by treatment objectives (Fig. S7), the heterogeneity (vague) was 0.14
                           (Fig. S8) and there was no inconsistency (Fig. S9). Meta-regression with the combined reduction in LDLc and
                           CRP and cholesterol-lowering treatments was not possible given the number of studies included41.
                              Although clear details were not obtained, post hoc analyses evaluating the combined reduction in LDLc and
                           CRP as predictors of future MACEs in CKD patients treated with a statin or statin/ezetimibe combination were
                           conducted following a retrospective biomarker-stratified design that allowed for the evaluation of codependent
                           health technologies (Table S3)37–39.

                           Discussion
                           Key messages. Statins at doses equivalent to simvastatin 20 mg daily (even considering the addition of eze-
                           timibe) and statins at higher doses showed similar efficacy in reducing MACEs when the eGFR was lower than
                           60 ml/min/1.73 m2: Only atorvastatin and rosuvastatin were associated with a clear benefit in terms of MACE
                           reduction. Irrespective of eGFR, MACEs were less frequent in patients with a moderate reduction in LDLc and
                           CRP than in patients with reductions in either of these two parameters separately.
                               LDLc and/or CRP decrease as a result of treatment with statins alone or in combination with ezetimibe in
                           CKD patients17,37–39. However, it should not be forgotten that malnutrition and chronic inflammation appear
                           quickly once the eGFR is lower than 60 ml/min/1.73 m2 9. As a consequence, the levels of LDLc and other lipopro-
                           teins decrease, resulting in an increased risk of all-cause and CVD mortality42. The evaluation of the benefits of
                           statins with or without ezetimibe along the wide spectrum of CKD is thus a real necessity.
                               This study found that most patients with early-stage CKD (eGFR ≥ 60 ml/min/1.73 m2) presented with high
                           LDLc levels before treatment17,37–39 and that the reductions in LDLc and CRP were moderate regardless of the


Scientific Reports |   (2019) 9:8951 | https://doi.org/10.1038/s41598-019-45431-5                                                            3
www.nature.com/scientificreports/                                                              www.nature.com/scientificreports




                           Figure 2. The effect of cholesterol-lowering treatment on MACEs when the eGFR is lower than 60 ml/
                           min/1.73 m2 (excluding dialysis patients). (a) All trials. (b) Primary prevention trials. CI, confidence interval;
                           eGFR, estimated glomerular filtration rate; EZE, ezetimibe; M-H, Mantel–Haenszel test; PBO, placebo; SE,
                           standard error; STA, statins.




                           Figure 3. SUCRA-based ranking of all statins and the combination of simvastatin plus ezetimibe when the
                           eGFR is lower than 60 ml/min/1.73 m2 (excluding dialysis patients). eGFR, estimated glomerular filtration rate;
                           SUCRA, surface under the cumulative ranking.



                           eGFR of the patients37–39; it is likely that in the early stages of CKD, patients develop high LDLc levels, which
                           then decrease as a result of the treatment (similar to non-CKD patients). Importantly, CRP is more difficult to
                           decrease in CKD patients, and this becomes even more challenging as the need for dialysis approaches43: Our
                           meta-analytic calculations revealed that reductions in LDLc and CRP, compared to the sole reduction in LDLc,
                           were associated with a low frequency of MACEs. In our opinion, both LDLc and CRP should be considered for
                           CKD patients initiating cholesterol-lowering treatment, as is currently the case for LDLc for non-CKD patients.
                               This study explored which cholesterol-lowering treatment strategies are the best for treating CVD risk in
                           this particular population. CKD constitutes the common final manifestation of a constellation of pathologies
                           that affect kidneys in a chronic and irreversible way, so therapies should be adapted to the complex and intricate
                           pathophysiology resulting from the incremental disease burden; evidence-based treatment strategies should thus
                           be conceived as a pragmatic option to improve decision-making for the daily management of patients3.
                               Statins, in addition to lowering LDLc levels, exert anti-inflammatory, antioxidant, anti-proliferative and
                           immunomodulatory effects, such as influencing plaque stability, normalizing sympathetic outflow, and prevent-
                           ing platelet aggregation, among other pleiotropic beneficial effects44–46.
                               The effects of statins on CVD risk in CKD patients have been discussed in the most recent systematic reviews
                           and meta-analyses47–50 and other previous evidence summaries51–53. Our results are in line with the findings
                           of these studies. In particular, our findings in the Bayesian network meta-analysis are consistent with those
                           obtained in The Cholesterol Treatment Trialists (CTT) Collaboration meta-analysis of individual participant


Scientific Reports |   (2019) 9:8951 | https://doi.org/10.1038/s41598-019-45431-5                                                               4
www.nature.com/scientificreports/                                                               www.nature.com/scientificreports




                           Figure 4. Network diagram for the competing cholesterol-lowering treatment strategies (Markov chain Monte
                           Carlo simulation).



                           data published in 2016, which concluded that the efficacy of statins is modest in CKD47. However, our study
                           cannot confirm the effect on cerebrovascular events that was presented in a pairwise meta-analysis conducted
                           in 201548, probably because this outcome was included as a composite of MACEs in the trials that were eligible
                           for our meta-analysis. Our study also found that no effect of ezetimibe alone can be expected in individuals with
                           CKD, and this finding is also consistent with the available evidence54.

                           Strengths and limitations.             To our knowledge, this study is the first to present a summary of the efficacy
                           of statins with or without ezetimibe in relation to reductions in LDLc and CRP levels along the wide spectrum of
                           CKD. Importantly, monitoring LDLc levels is no longer recommended by the latest guidelines from the KDIGO
                           lipid management work group, and it should be reserved for instances in which the results would alter the man-
                           agement of patient treatment10. However, the KDIGO lipid management work group recommends a full lipid
                           profile upon first presentation (a statement made on the basis of low-quality evidence)10; according to the findings
                           of this meta-analysis, CRP should be measured in addition to LDLc before initiating cholesterol-lowering treat-
                           ment in CKD patients.
                               Our approach guarantees the provision of evidence that pertinently answers the formulated review questions.
                           The evidence was obtained from general sources to focused sources according to a method that resembles the
                           “mixed-criteria” quality appraisal method of Wortman55. In addition, a multidisciplinary supervision mechanism
                           was planned for contextualizing the search findings56. Nevertheless, some limitations must be mentioned. First,
                           only 3 post hoc analyses of RCTs including patients with and without CKD provided evidence for the combined
                           reduction in LDLc and CRP in CKD patients37–39. The criticism against attributing more value to the outcome
                           of an unplanned analysis is difficult to ignore57. In addition, considering that there was no possibility to perform
                           meta-regression41, our results should be considered prudently, particularly considering the current paucity of
                           clinical research on CVD in CKD3, which should highlight the need for more research58. Second, publication
                           bias is likely to affect our results. No unpublished studies were found, but asymmetry in the funnel plots may be a
                           discouraging finding. Reporting bias occurs because significant results suggesting a beneficial effect of interven-
                           tions are more likely to be published than nonsignificant results59. Caution is thus necessary when interpreting
                           the manuscript, and common sense is required when applying our results into everyday clinical practice. Finally,
                           the power of the trials evaluating lovastatin and fluvastatin was probably insufficient, resulting in an inappropriate
                           ranking of all statins in the network meta-analysis23,29. Once more, caution is needed. For inflated associations,
                           being fair with the interpretation of the results is an important consideration60.

                           Conclusion and study contribution.            In conclusion, statins with and without ezetimibe have a modest
                           efficacy for reducing CVD risk in patients with CKD and eGFRs lower than 60 ml/min/1.73 m2. Irrespective of


Scientific Reports |   (2019) 9:8951 | https://doi.org/10.1038/s41598-019-45431-5                                                              5
www.nature.com/scientificreports/                                                            www.nature.com/scientificreports




                           Figure 5. Network forest plot of the fixed and random effects of the competing cholesterol-lowering treatment
                           strategies. CrI, credible intervals; OR, odds ratio.



                           eGFR, if LDLc and CRP levels are high before treatment, reductions in such parameters are appropriate to obtain
                           a benefit from cholesterol-lowering treatment. In other words, the use of statins should be initiated when levels
                           of LDLc and CRP are high to reduce the CVD risk of patients. Importantly, patients with advanced CKD may
                           present with low levels of LDLc. Such patients have a higher risk of CVD events, much like the risk of patients
                           presenting with high levels of LDLc42. The benefit of cholesterol-lowering treatment in CKD patients is evident,
                           but the impact on those with advanced CKD remains unclear. Finally, in accordance with the current guidelines,
                           after an initial measurement, monitoring levels of LDLc is probably not required for CKD patients10. However,



Scientific Reports |   (2019) 9:8951 | https://doi.org/10.1038/s41598-019-45431-5                                                          6
www.nature.com/scientificreports/                                                             www.nature.com/scientificreports




                           Figure 6. Inconsistency plot of the random effects for the competing cholesterol-lowering treatment strategies.



                            Treatment objectives†                     SUCRA‡
                            LDLc reduction < 50% plus CRP
                                                                      0.7719
                            reduction < 50%
                            LDLc reduction ≥ 50%                      0.6261
                            LDLc reduction < 50%                      0.5933
                            CRP reduction < 50%                       0.4995
                            None                                      0.0092

                           Table 1. The ranking of the LDLc and CRP treatment objectives based on SUCRA. †In a Bayesian context
                           (Markov chain Monte Carlo simulation), the LDLc and/or CRP treatment objectives were the parameters used
                           for ranking according to probabilities for being the best, the second best, the third best, and so on P(o = b),
                           b = 1, … , a. ‡SUCRA for each treatment objective o out of the a competing treatment objectives requires
                           calculation of the a vector of the cumulative probabilities cumo , b to be among the b best treatment objectives,
                           b = 1, … , a. Abbreviations: CRP, C-reactive protein; LDL-c; low-density lipoprotein cholesterol; SUCRA,
                           surface under the cumulative ranking.


                           CRP should be assessed in addition to LDLc before initiating cholesterol-lowering treatment. Evidence-based
                           treatment strategies may lead to individualized therapies for CKD patients.

                           Methods
                           Study design, eligibility criteria and systematic searches.               These multistage pairwise and network
                           meta-analyses were carried out and reported according to the PRISMA guidelines56. Two protocols were regis-
                           tered in the International Prospective Register of Systematic Reviews (PROSPERO) under the following registra-
                           tion IDs: CRD42017075166 and CRD42017055787. A two-stage systematic review design constitutes the basis
                           of our multistage approach58,61. A systematic mapping (cholesterol-lowering treatment efficacy evaluation phase)
                           followed by an in-depth systematic review (treatment objectives evaluation phase) was planned, and the former
                           was complemented by an independent parallel one-stage systematic review (Material S1). An expert advisory
                           group was formed to contextualize the search findings (E.A., F.J.A., J.B.-M. and M.V.).
                               The review questions corresponding to each of the 2 systematic reviews and the participants/population,
                           intervention(s), and comparators are presented in Table 2. The primary outcome was incident MACEs, which
                           included all fatal and nonfatal coronary events, including revascularization procedures, and all cerebrovascular
                           events, including transient ischemic attacks (TIAs). The secondary outcome was material codependency when
                           combining technologies related to treatment and the potential treatment effect modifiers for a low frequency of
                           MACEs. The eligible studies were all randomized controlled trials (RCTs) with extended follow-up periods and
                           post hoc analysis that were carried out on or that included CKD patients.
                               In parallel, published and unpublished literature sources were searched at Pharmacological Big Data
                           Laboratory, Pharmacology and Therapeutics, Faculty of Medicine, University of Valladolid, Valladolid, Spain
                           and Centro de Investigación en Salud Pública, Instituto de Investigación de la Facultad de Medicina Humana,
                           Universidad de San Martín de Porres, Lima, Perú through September 2018 according to the homogenous PICO
                           element-based search strategies that are available online from the systematic review protocols registered in
                           PROSPERO (Material S2).

                           Data collection and analyses.              Anonymized datasets describing characteristics of the studies and their
                           participants, interventions, comparators, outcomes, and follow-up periods were constructed. Before analysis, the
                           risk of bias in the eligible studies was assessed using the standard tool produced by the Cochrane Collaboration62.
                           A two-stage pairwise and network meta-analysis of aggregate-level data was planned (C.O.-S. and F.H.-G.).
                           However, if there were not enough data or if there was a doubt about the comparability of the data, a two-staged
                           systematic narrative synthesis was envisaged63.


Scientific Reports |   (2019) 9:8951 | https://doi.org/10.1038/s41598-019-45431-5                                                              7
www.nature.com/scientificreports/                                                                            www.nature.com/scientificreports


                                                           Systematic mapping (stage 1)/ systematic review support      In-depth meta-analysis (stage 2)
                                                                                                                         Are serum levels of LDLc and CRP
                                                           Are statins with or without ezetimibe efficacious in reducing related to CVD events in patients with
                            Review question
                                                           CVD risk in patients with CKD?   §
                                                                                                                         CKD receiving treatment with statins
                                                                                                                         alone or in combination with ezetimibe?
                                                                                                                        Adult individuals with NKF/CKD
                                                           Adult individuals with NKF/CKD KDIGO GFR categories
                                                                                                                        KDIGO GFR categories G1–G2, patients
                                                           G1–G2, patients with CKD KDIGO GFR categories G3a–
                            Participants/population                                                                     with CKD KDIGO GFR categories G3a–
                                                           G5,# and patients treated with chronic dialysis or kidney
                                                                                                                        G5, and patients treated with chronic
                                                           transplantation.§
                                                                                                                        dialysis or kidney transplantation.
                                                                                                                        Serum levels of LDLc and/or CRP in
                            Intervention(s)/exposures(s)   Statins with or without ezetimibe.§                          patients treated with statins with or
                                                                                                                        without ezetimibe.
                                                                                                                        Serum levels of LDLc and/or CRP under
                            Comparators                    Placebo/usual care.@
                                                                                                                        placebo/usual care.@

                           Table 2. Review questions and study eligibility. §Only statins were studied, and patients treated with chronic
                           dialysis or kidney transplantation were included in the systematic review to support systematic mapping.
                           #
                            Individuals with CKD were divided according to eGFR (in ml/min/1.73 m2): ≥60 (KDIGO GFR categories
                           G1–G2) or < 60 (G3a–G5). @Usual care may include a statin if a statin at a higher dose or a statin/ezetimibe
                           combination was already the intervention. Abbreviations: CKD, chronic kidney disease; CRP, C-reactive
                           protein; CVD, cardiovascular disease; eGFR, estimated glomerular filtration rate; KDIGO, the Kidney Disease:
                           Improving Global Outcomes; LDLc, low-density lipoprotein cholesterol; NKF, normal kidney function.


                               The overall odds ratios and their 95% confidence intervals for the outcome of incident MACEs correspond-
                           ing to cholesterol-lowering treatment (stage 1) and the treatment objectives of LDLc and/or CRP (stage 2) were
                           obtained (Mantel-Haenszel random-effect method) through conventional pairwise meta-analysis. Heterogeneity
                           was examined by computing the I² statistic (inconsistency) and the χ² statistic, and the presence of reporting bias
                           was assessed by visual inspection of funnel plots of the estimates against their standard errors. A calculation of the
                           regression coefficient corresponding to cholesterol-lowering treatment and the treatment objectives of LDLc and/
                           or CRP (stage 2) was planned (random-effects meta-regression). Review Manager software (RevMan) version 5.3
                           (Cochrane Collaboration) and the ‘metareg’ macro from Stata version 12.1 (StataCorp) were used to conduct the
                           pairwise meta-analytic calculations and meta-regression, respectively. Pooled ORs and the corresponding 95%
                           credible intervals for incident MACEs (considered a ‘bad’ outcome) were calculated (random-effects model using
                           vague priors with correction for zero values) via Bayesian network meta-analysis (Markov chain Monte Carlo sim-
                           ulation); the competing strategies (stage 1) and treatment objectives of LDLc and/or CRP (stage 2) were ranked
                           based on SUCRA (surface under the cumulative ranking) after verifying convergence (Brooks-Gelman-Rubin
                           method) and inconsistency. Network meta-analysis was performed using NetMetaXL software (Canadian Agency
                           for Drugs and Technologies in Health and Cornerstone Research Group)64. Analysis of subgroups was planned
                           at stage 1 (NKF/CKD KDIGO GFR categories G1–G2, CKD KDIGO GFR categories G3a–G5, chronic dialysis,
                           transplantation) and stage 2 (LDLc reduction < 50% combined with CRP reduction < 50%, LDLc reduction <
                           50%, LDLc reduction ≥ 50%, CRP reduction < 50%).
                               Codependency when combining technologies related to the treatment and the potential treatment effect mod-
                           ifiers for a low frequency of MACEs (LDLc combined with CRP) was assessed using an adaptation of Merlin’s
                           tool included in the guidelines for preparing a submission to the Pharmaceutical Benefits Advisory Committee
                           (PBAC) from the Department of Health of Australia (F.H.-G. and A.M.-B.-M.)65. The tool sections of the eco-
                           nomic evaluation and use of the medicine in practice were not considered.

                           Data Availability
                           All data generated and analyzed during this study are included in this published article and the supplementary
                           materials.

                           References
                            1. Stenvinkel, P. Chronic kidney disease: a public health priority and harbinger of premature cardiovascular disease. J. Intern. Med. 268,
                               456–467, https://doi.org/10.1111/j.1365-2796.2010.02269.x (2010).
                            2. Yousuf, O. et al. High-sensitivity C-reactive protein and cardiovascular disease: a resolute belief or an elusive link? J. Am. Coll.
                               Cardiol. 62, 397–408, https://doi.org/10.1016/j.jacc.2013.05.016 (2013).
                            3. Formentini, I. et al. Current drug development challenges in chronic kidney disease (CKD)–identification of individualized
                               determinants of renal progression and premature cardiovascular disease (CVD). Nephrol. Dial. Transplant. 27, iii81–iii88, https://
                               doi.org/10.1093/ndt/gfs270 (2012).
                            4. Joshi, S. & Viljoen, A. Renal biomarkers for the prediction of cardiovascular disease. Curr. Opin. Cardiol. 30, 454–460, https://doi.
                               org/10.1097/HCO.0000000000000177 (2015).
                            5. Zoccali, C. Traditional and emerging cardiovascular and renal risk factors: an epidemiologic perspective. Kidney Int. 70, 26–33,
                               https://doi.org/10.1038/sj.ki.5000417 (2006).
                            6. Bai, J. P. et al. Translational biomarkers: from preclinical to clinical a report of 2009 AAPS/ACCP Biomarker Workshop. AAPS J. 13,
                               274–283, https://doi.org/10.1208/s12248-011-9265-x (2011).
                            7. Merlin, T. et al. Assessing personalized medicines in Australia: a national framework for reviewing codependent technologies. Med.
                               Decis. Making. 33, 333–342, https://doi.org/10.1177/0272989X12452341 (2013).
                            8. Tonelli, M. et al. Risk of coronary events in people with chronic kidney disease compared with those with diabetes: a population-
                               level cohort study. Lancet 380, 807–814, https://doi.org/10.1016/S0140-6736(12)60572-8 (2012).
                            9. Massy, Z. A. & de Zeeuw, D. LDL cholesterol in CKD–to treat or not to treat? Kidney Int. 84, 451–456, https://doi.org/10.1038/
                               ki.2013.181 (2013).


Scientific Reports |   (2019) 9:8951 | https://doi.org/10.1038/s41598-019-45431-5                                                                                   8
www.nature.com/scientificreports/                                                                            www.nature.com/scientificreports


                           10. KDIGO Clinical Practice Guideline for Lipid Management in CKD: summary of recommendation statements and clinical approach
                               to the patient. Kidney Int. 85, 1303–1309, https://doi.org/10.1038/ki.2014.31 (2014).
                           11. Holdaas, H. et al. Effect of fluvastatin on cardiac outcomes in renal transplant recipients: a multicentre, randomised, placebo-
                               controlled trial. Lancet 361, 2024–2031, https://doi.org/10.1016/S0140-6736(03)13638-0 (2003).
                           12. Holdaas, H. et al. Long-term cardiac outcomes in renal transplant recipients receiving fluvastatin: the ALERT extension study. Am.
                               J. Transplant. 5, 2929–2936, https://doi.org/10.1111/j.1600-6143.2005.01105.x (2005).
                           13. Wanner, C. et al. Atorvastatin in patients with type 2 diabetes mellitus undergoing hemodialysis. N. Engl. J. Med. 353, 238–248,
                               https://doi.org/10.1056/NEJMoa043545 (2005).
                           14. Krane, V. et al. Long-term effects following 4 years of randomized treatment with atorvastatin in patients with type 2 diabetes
                               mellitus on hemodialysis. Kidney Int. 89, 1380–1387, https://doi.org/10.1016/j.kint.2015.12.033 (2016).
                           15. Fellström, B. C. et al. Rosuvastatin and cardiovascular events in patients undergoing hemodialysis. N. Engl. J. Med. 360, 1395–1407,
                               https://doi.org/10.1056/NEJMoa0810177 (2009).
                           16. Stegmayr, B. G. et al. Low-dose atorvastatin in severe chronic kidney disease patients: a randomized, controlled endpoint study.
                               Scand. J. Urol. Nephrol. 39, 489–497, https://doi.org/10.1080/00365590500329304 (2005).
                           17. Baigent, C. et al. The effects of lowering LDL cholesterol with simvastatin plus ezetimibe in patients with chronic kidney disease
                               (Study of Heart and Renal Protection): a randomised placebo-controlled trial. Lancet 377, 2181–2192, https://doi.org/10.1016/
                               S0140-6736(11)60739-3 (2011).
                           18. Landray, M. et al. The second United Kingdom Heart and Renal Protection (UK-HARP-II) Study: a randomized controlled study of
                               the biochemical safety and efficacy of adding ezetimibe to simvastatin as initial therapy among patients with CKD. Am. J. Kidney Dis.
                               47, 385–395, https://doi.org/10.1053/j.ajkd.2005.11.018 (2006).
                           19. Ueshima, K. et al. Effects of atorvastatin on renal function in patients with dyslipidemia and chronic kidney disease: rationale and
                               design of the ASsessment of clinical Usefulness in CKD patients with Atorvastatin (ASUCA) trial. Clin. Exp. Nephrol. 17, 211–217,
                               https://doi.org/10.1007/s10157-012-0676-5 (2013).
                           20. Nanayakkara, P. W. et al. Effect of a treatment strategy consisting of pravastatin, vitamin E, and homocysteine lowering on carotid
                               intima-media thickness, endothelial function, and renal function in patients with mild to moderate chronic kidney disease: results
                               from the Anti-Oxidant Therapy in Chronic Renal Insufficiency (ATIC) Study. Arch. Intern. Med. 167, 1262–1270, https://doi.
                               org/10.1001/archinte.167.12.1262 (2007).
                           21. Fassett, R. G., Robertson, I. K., Ball, M. J., Geraghty, D. P. & Coombes, J. S. Effect of atorvastatin on kidney function in chronic
                               kidney disease: a randomised double-blind placebo-controlled trial. Atherosclerosis 213, 218–224, https://doi.org/10.1016/j.
                               atherosclerosis.2010.07.053 (2010).
                           22. Chonchol, M., Cook, T., Kjekshus, J., Pedersen, T. R. & Lindenfeld, J. Simvastatin for secondary prevention of all-cause mortality and
                               major coronary events in patients with mild chronic renal insufficiency. Am. J. Kidney Dis. 49, 373–382, https://doi.org/10.1053/j.
                               ajkd.2006.11.043 (2007).
                           23. Kendrick, J. et al. Effect of lovastatin on primary prevention of cardiovascular events in mild CKD and kidney function loss: a post
                               hoc analysis of the Air Force/Texas Coronary Atherosclerosis Prevention Study. Am. J. Kidney Dis. 55, 42–49, https://doi.
                               org/10.1053/j.ajkd.2009.09.020 (2010).
                           24. Rahman, M. et al. Pravastatin and cardiovascular outcomes stratified by baseline eGFR in the lipid- lowering component of
                               ALLHAT. Clin. Nephrol. 80, 235–248, https://doi.org/10.5414/CN107922 (2013).
                           25. Koren, M. J. et al. Focused atorvastatin therapy in managed-care patients with coronary heart disease and CKD. Am. J. Kidney Dis.
                               53, 741–750, https://doi.org/10.1053/j.ajkd.2008.11.025 (2009).
                           26. Colhoun, H. M. et al. Effects of atorvastatin on kidney outcomes and cardiovascular disease in patients with diabetes: an analysis
                               from the Collaborative Atorvastatin Diabetes Study (CARDS). Am. J. Kidney Dis. 54, 810–819, https://doi.org/10.1053/j.
                               ajkd.2009.03.022 (2009).
                           27. Holme, I. et al. Cardiovascular outcomes and their relationships to lipoprotein components in patients with and without chronic
                               kidney disease: results from the IDEAL trial. J. Intern. Med. 267, 567–575, https://doi.org/10.1111/j.1365-2796.2009.02176.x (2010).
                           28. Ridker, P. M., MacFadyen, J., Cressman, M. & Glynn, R. J. Efficacy of rosuvastatin among men and women with moderate chronic
                               kidney disease and elevated high-sensitivity C-reactive protein: a secondary analysis from the JUPITER (Justification for the Use of
                               Statins in Primary Prevention–an Intervention Trial Evaluating Rosuvastatin) trial. J. Am. Coll. Cardiol. 55, 1266–1273, https://doi.
                               org/10.1016/j.jacc.2010.01.020 (2010).
                           29. Lemos, P. A. et al. Long-term fluvastatin reduces the hazardous effect of renal impairment on four-year atherosclerotic outcomes (a
                               LIPS substudy). Am. J. Cardiol. 95, 445–451, https://doi.org/10.1016/j.amjcard.2004.10.008 (2005).
                           30. Nakamura, H. et al. Pravastatin and cardiovascular risk in moderate chronic kidney disease. Atherosclerosis 206, 512–517, https://
                               doi.org/10.1016/j.atherosclerosis.2009.03.031 (2009).
                           31. Deedwania, P. C., Stone, P. H., Fayyad, R. S., Laskey, R. E. & Wilson, D. J. Improvement in Renal Function and Reduction in Serum
                               Uric Acid with Intensive Statin Therapy in Older Patients: A Post Hoc Analysis of the SAGE Trial. Drugs Aging. 32, 1055–1065,
                               https://doi.org/10.1007/s40266-015-0328-z (2015).
                           32. Shepherd, J. et al. Intensive lipid lowering with atorvastatin in patients with coronary heart disease and chronic kidney disease: the
                               TNT (Treating to New Targets) study. J. Am. Coll. Cardiol. 51, 1448–1454, https://doi.org/10.1016/j.jacc.2007.11.072 (2008).
                           33. Tonelli, M. et al. Effect of pravastatin on cardiovascular events in people with chronic kidney disease. Circulation 110, 1557–1563,
                               https://doi.org/10.1161/01.CIR.0000143892.84582.60 (2004).
                           34. Asselbergs, F. W. et al. Effects of fosinopril and pravastatin on cardiovascular events in subjects with microalbuminuria. Circulation
                               110, 2809–2816, https://doi.org/10.1161/01.CIR.0000146378.65439.7A (2004).
                           35. Brouwers, F. P. et al. Long-term effects of fosinopril and pravastatin on cardiovascular events in subjects with microalbuminuria: Ten
                               years of follow-up of Prevention of Renal and Vascular End-stage Disease Intervention Trial (PREVEND IT). Am. Heart J. 161,
                               1171–1178, https://doi.org/10.1016/j.ahj.2011.03.028 (2011).
                           36. Moher, D. et al. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 6, e1000097,
                               https://doi.org/10.1371/journal.pmed.1000097 (2009).
                           37. Ridker, P. M. et al. Measurement of C-reactive protein for the targeting of statin therapy in the primary prevention of acute coronary
                               events. N. Engl. J. Med. 344, 1959–1965, https://doi.org/10.1056/NEJM200106283442601 (2001).
                           38. Soedamah-Muthu, S. S. et al. Effect of atorvastatin on C-reactive protein and benefits for cardiovascular disease in patients with type
                               2 diabetes: analyses from the Collaborative Atorvastatin Diabetes Trial. Diabetologia 58, 1494–1502, https://doi.org/10.1007/s00125-
                               015-3586-8 (2015).
                           39. Ridker, P. M. et al. Reduction in C-reactive protein and LDL cholesterol and cardiovascular event rates after initiation of rosuvastatin:
                               a prospective study of the JUPITER trial. Lancet 373, 1175–1182, https://doi.org/10.1016/S0140-6736(09)60447-5 (2009).
                           40. Salanti, G., Ades, A. E. & Ioannidis, J. P. Graphical methods and numerical summaries for presenting results from multiple-
                               treatment meta-analysis: an overview and tutorial. J. Clin. Epidemiol. 64, 163–171, https://doi.org/10.1016/j.jclinepi.2010.03.016
                               (2010).
                           41. Baker, W. L. et al. Understanding heterogeneity in meta-analysis: the role of meta-regression. Int. J. Clin. Pract. 63, 1426–1434,
                               https://doi.org/10.1111/j.1742-1241.2009.02168.x (2009).




Scientific Reports |   (2019) 9:8951 | https://doi.org/10.1038/s41598-019-45431-5                                                                                     9
www.nature.com/scientificreports/                                                                          www.nature.com/scientificreports


                           42. Kovesdy, C. P., Anderson, J. E. & Kalantar-Zadeh, K. Inverse association between lipid levels and mortality in men with chronic
                               kidney disease who are not yet on dialysis: effects of case mix and the malnutrition-inflammation-cachexia syndrome. J. Am. Soc.
                               Nephrol. 18, 304–311, https://doi.org/10.1681/ASN.2006060674 (2007).
                           43. Abraham, G. et al. C-Reactive protein, a valuable predictive marker in chronic kidney disease. Saudi J. Kidney Dis. Transpl. 20,
                               811–815, http://www.sjkdt.org/text.asp?2009/20/5/811/55367 (2009).
                           44. Kavalipati, N., Shah, J., Ramakrishan, A. & Vasnawala, H. Pleiotropic effects of statins. Indian J. Endocrinol. Metab. 19, 554–562,
                               https://doi.org/10.4103/2230-8210.163106 (2015).
                           45. Zhou, Q. & Liao, J. K. Statins and cardiovascular diseases: from cholesterol lowering to pleiotropy. Curr. Pharm. Des. 15, 467–478,
                               https://doi.org/10.2174/138161209787315684 (2009).
                           46. Liao, J. K. & Laufs, U. Pleiotropic effects of statins. Annu. Rev. Pharmacol. Toxicol. 45, 89–118, https://doi.org/10.1146/annurev.
                               pharmtox.45.120403.095748 (2005).
                           47. Cholesterol Treatment Trialists’ (CTT) Collaboration. Impact of renal function on the effects of LDL cholesterol lowering with
                               statin-based regimens: a meta-analysis of individual participant data from 28 randomised trials. Lancet Diabetes Endocrinol. 4,
                               829–839, https://doi.org/10.1016/S2213-8587(16)30156-5 (2016).
                           48. Yan, Y. L. et al. High-intensity statin therapy in patients with chronic kidney disease: a systematic review and meta-analysis. BMJ
                               Open 5, e006886, https://doi.org/10.1136/bmjopen-2014-006886 (2015).
                           49. Major, R. W., Cheung, C. K., Gray, L. J. & Brunskill, N. J. Statins and Cardiovascular Primary Prevention in CKD: A Meta-Analysis.
                               Clin. J. Am. Soc. Nephrol. 10, 732–739, https://doi.org/10.2215/CJN.07460714 (2015).
                           50. Zhang, X. et al. Effect of statins on cardiovascular events in patients with mild to moderate chronic kidney disease: a systematic
                               review and meta-analysis of randomized clinical trials. BMC Cardiovasc. Disord. 14, 19, https://doi.org/10.1186/1471-2261-14-19
                               (2014).
                           51. Palmer, S. C. et al. Benefits and harms of statin therapy for persons with chronic kidney disease: a systematic review and meta-
                               analysis. Ann Intern Med. 157, 263–275, https://doi.org/10.7326/0003-4819-157-4-201208210-00007 (2012).
                           52. Upadhyay, A. et al. Lipid-lowering therapy in persons with chronic kidney disease: a systematic review and meta-analysis. Ann
                               Intern Med. 157, 251–262, https://doi.org/10.7326/0003-4819-157-4-201208210-00005.
                           53. Strippoli, G. F. et al. Effects of statins in patients with chronic kidney disease: meta-analysis and meta-regression of randomised
                               controlled trials. BMJ 336, 645–651, https://doi.org/10.1136/bmj.39472.580984.AE (2008).
                           54. Battaggia, A., Donzelli, A., Font, M., Molteni, D. & Galvano, A. Clinical efficacy and safety of Ezetimibe on major cardiovascular
                               endpoints: systematic review and meta-analysis of randomized controlled trials. PLoS One 10, e0124587, https://doi.org/10.1371/
                               journal.pone.0124587 (2015).
                           55. Wortman, P. M. In Cooper, H. & Hedges, L. V. (eds) The Handbook of Research Synthesis, pp 97–110 (Russell Sage Foundation, New
                               York, 1994).
                           56. Trivedi, D. et al. A protocol for a systematic review of research on managing behavioural and psychological symptoms in dementia
                               for community-dwelling older people: evidence mapping and syntheses. Syst. Rev. 2, 70, https://doi.org/10.1186/2046-4053-2-70
                               (2013).
                           57. Schühlen, H. Pre-specified vs. post-hoc subgroup analyses: are we wiser before or after a trial has been performed? Eur. Heart J. 35,
                               2055–2057, https://doi.org/10.1093/eurheartj/ehu202 (2014).
                           58. Herrera-Gómez, F. et al. The number of FoxP3 regulatory T cells in the circulation may be a predictive biomarker for kidney
                               transplant recipients: A multistage systematic review. Int. Immunopharmacol. 65, 483–492, https://doi.org/10.1016/j.
                               intimp.2018.10.028 (2018).
                           59. Sterne, J. A. et al. Recommendations for examining and interpreting funnel plot asymmetry in meta-analyses of randomised
                               controlled trials. BMJ 343, d4002, https://doi.org/10.1136/bmj.d4002 (2011).
                           60. Ioannidis, J. P. Why most discovered true associations are inflated. Epidemiology 19, 640–648, https://doi.org/10.1097/
                               EDE.0b013e31818131e7 (2008).
                           61. Herrera-Gómez, F. et al. Correction of hyponatremia may be a treatment stratification biomarker: a two-stage systematic review and
                               meta-analysis. J. Clin. Med. 7, 262, https://doi.org/10.3390/jcm7090262 (2018).
                           62. Higgins, J. P. T., Altman, D. G. & Sterne, J. A. C. In Higgins, J. P. T. & Green, S. (eds) Cochrane Handbook for systematic reviews of
                               interventions – Version 5.1.0., (The Cochrane Collaboration), http://handbook-5-1.cochrane.org/.
                           63. Systematic reviews: Centre for Reviews and Dissemination’s (CRD) guidance for undertaking reviews in health care, www.york.
                               ac.uk/media/crd/Systematic_Reviews.pdf (University of York, York, 2008).
                           64. Brown, S. et al. A Microsoft-Excel-based tool for running and critically appraising network meta-analyses–an overview and
                               application of NetMetaXL. Syst. Rev. 3, 110, https://doi.org/10.1186/2046-4053-3-110 (2014).
                           65. Merlin, T. et al. Assessing personalized medicines in Australia: a national framework for reviewing codependent technologies. Med.
                               Decis. Making 33, 333–342, https://doi.org/10.1177/0272989X12452341 (2013).

                           Acknowledgements
                           The authors thank B. Muñoz (Research unit, Hospital Virgen de la Concha – Sanidad de Castilla y León, Zamora,
                           Spain) who helped coordinate the data retrieval. This research did not receive any specific grant from funding
                           agencies in the public, commercial or not-for-profit sectors. However, the edition and publication costs were
                           covered by Consejería de Educación, Junta de Castilla y León, Spain (reference: VA161G18).

                           Author Contributions
                           A.M.-B.-M., C.O.-S. and F.H.-G. developed the hypothesis and study design. A.M.-B.-M., D.M.G., F.H.-G., F.J.A.,
                           F.L.-S., J.G.V. and M.M.C. performed the literature searches and/or screened papers. E.A., F.J.A., J.B.-M. and M.V.
                           formed the expert advisory group. C.O.-S. and F.H.-G. performed the data analyses. All authors contributed to
                           the drafting and critical revision of all manuscript versions. F.H.-G. and M.M.C. are co-first authors. C.O.-S. and
                           F.J.A. are co-senior authors.

                           Additional Information
                           Supplementary information accompanies this paper at https://doi.org/10.1038/s41598-019-45431-5.
                           Competing Interests: The authors declare no competing interests.
                           Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and
                           institutional affiliations.




Scientific Reports |   (2019) 9:8951 | https://doi.org/10.1038/s41598-019-45431-5                                                                                10
www.nature.com/scientificreports/                                                             www.nature.com/scientificreports


                                         Open Access This article is licensed under a Creative Commons Attribution 4.0 International
                                         License, which permits use, sharing, adaptation, distribution and reproduction in any medium or
                           format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre-
                           ative Commons license, and indicate if changes were made. The images or other third party material in this
                           article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the
                           material. If material is not included in the article’s Creative Commons license and your intended use is not per-
                           mitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the
                           copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

                           © The Author(s) 2019




Scientific Reports |   (2019) 9:8951 | https://doi.org/10.1038/s41598-019-45431-5                                                          11