7/23/2019 1-s2.0-S0109564115000810-main epr14.pdf

1/12

d e nt a l m a te r ia l s 3 1 ( 2 0 1 5 ) 657668

Available

online

at

www.sciencedirect.com

ScienceDirect

j o u rn a l h om epag e : ww w. in t l . e l s ev i e rhea l t h . com/ jou rna l s / dema

Inuence of surface treatment on theresin-bonding of zirconia

Seda Sanl a , Mine Dndar Cmleko glu b, Erhan C mleko glu b,Mehmet Sonugelen b, Tijen Pamir c, B.W. Darvell d,

a Ulukent Dental Clinic, Menemen, Izmir, Turkeyb Department of Prosthodontics, School of Dentistry, Ege University, Izmir, Turkeyc Department of Restorative Dentistry, School of Dentistry, Ege University, Izmir, Turkeyd

Dental Materials Science, Faculty of Dentistry, Kuwait University, Kuwait

a r t i c l e i n f o

Article history:Received 29 September 2014Received in revised form8 January 2015Accepted 13 March 2015

Keywords:

Zirconia surface treatmentAdhesive cementationFour-point bending test

a b s t r a c t

Objective. To compare the effect of various surface treatments on the bonding of luting resincements to zirconia under four-point bending.Methods. Barspecimens( n =200)(2mm 5mm 25mm) were preparedfromzirconiablocks(VITA In-Ceram YZ, Vita Zahnfabrik) with the cementation surface (2mm 5mm) of groups of 40 treated in one of ve ways: airborne particle abrasion with 50 m Al2O3(GB), zirconia primer (Z-Prime Plus, Bisco) (Z), glaze ceramic (Crystall.Glaze spray, IvoclarVivadent)+ hydrouoricacid (GHF), fusion glass-ceramic (Crystall.Connect, IvoclarVivadent)(CC), or left untreated as control (C). Within each treatment, bars were cleaned ultrasoni-

cally

for 15min in ethanol and then deionized water before bonding in pairs with one of two luting resins: Panavia F 2.0, (Kuraray) (P); RelyX U-200 (3M/Espe) (R), to form 10 testspecimens for each treatment and lute combination. Mechanical tests were performedand bond strengths (MPa) were subject, after log transformation, to analysis of variance,ShapiroWilk and HolmSidak tests; also log-linear contingency analysis of failure modedistribution; all with = 0.05. Fracture surfaces were examined under light and scanning electron microscopy.Results. While the effect of surface treatment was signicant ( p =1.27 10 9), there was nodetected effect due to resin ( p = 0.829). All treatments except CC (30.1MPa / 1.44)* weresignicantlybetter than the untreated control (24.8MPa / 1.35) ( p =3.28 10 9). While theeffect of GB which gave the highest mean strength (50.5MPa / 1.29) was not distin-guishable from that of GHF (39.9MPa / 1.29) ( p = 0.082), it was signicantly better thantreatment with either CC or Z (33.1MPa / 1.48) ( p < 0.05). (* After log transformation for

analysis and back; asymmetric error bounds as s.d. in log values.)Signicance. Thenoveltest methoddesign, which hasgood discriminatory power, conrmedthe valueof gritblastingas a simpleand effective treatmentwith low operatorhazard. Itgavethe highest bond strengths regardless of the cement type. Glaze layer application followed

Corresponding author . Tel.: +44 1225 81 06 71.E-mail address: b.w.darvell@hku.hk (B.W. Darvell).

http://dx.doi.org/10.1016/j.dental.2015.03.004

0109-5641/ 2015 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.

http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004http://www.sciencedirect.com/science/journal/01095641http://www.intl.elsevierhealth.com/journals/demamailto:b.w.darvell@hku.hkhttp://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004mailto:b.w.darvell@hku.hkhttp://crossmark.crossref.org/dialog/?doi=10.1016/j.dental.2015.03.004&domain=pdfhttp://www.intl.elsevierhealth.com/journals/demahttp://www.sciencedirect.com/science/journal/01095641http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004

7/23/2019 1-s2.0-S0109564115000810-main epr14.pdf

2/12

658 d en ta l m at er ia ls 3 1 ( 2 0 1 5 ) 657668

by hydrouoric acid-etching on zirconia before cementation might be viable for adhesivezirconia cementation, but represents a much greater hazard as well as having problemswith thickness control.

2015 Academy of Dental Materials. Published by Elsevier Ltd. All rights reserved.

1. Introduction

Zirconia, or rather yttria-partially stabilized tetragonal zirco-nia (YTZP), is increasingly used in dentistry in view of itsremarkable strength [1,2]. However, a major impediment toits effective use is its lack of reactivity since it is a non-polarandmore or less inert material whichbecomes very dense andhomogeneous on sintering [3]. This means that all the usualmeans of bonding (in the broad chemical sense of establisheda covalentbond betweenadhesivematerialand thesubstrate),effective enough for other ceramic systems, do not work [3],

and although phosphate-based systems have shown someimprovement [3], nothing is yet considered reliable enough.This also includes the mechanical retention afforded by grit-blasting, where particle size has to be limited to minimizesubsurface damage and hence the risk of cracking [4]. Thus, itis of great interest to develop an effective means of bonding to zirconia to enable long-term retention. For this purpose, agreat number of surface treatment methods have been tried,for example to roughen the surface with rotary systems, laserirradiation, or selective inltration etching, or to modify thesurface through silica coating, silane application, hydrouoricacid etching after fused glass-ceramic application, or nano-structured alumina coating [3,57].

The nature of dentistry, in terms of economy, processing time, and the fact that all restorative devices are effectivelyone-off prototypes, means thatthe clinicaland laboratory pro-cedures to be used for zirconia-based polycrystalline ceramicrestorations to achieve adhesive cementation should be prac-tical and easily handled. We may note that while hydrouoricacid (HF) etching is an easy if hazardous surface treatmentfor silicate-containing ceramics [8], some oxide ceramics, andin particular zirconia, whichcontain less than 15 mass% silicaand little or no glass phase [9], cannot be so treated. How-ever, the application of a fused glass-ceramic on the intagliosurfaces of zirconia restorations, which glaze is then etchableand so capable of adhesive cementation, has been described

[5,10,11]. The effectiveness of this bonding depends on inti-mate contact and micromechanical key with grain boundarytopography, rather than chemical interaction as has some-times been suggested [12], there being no chemical reactionor elemental migration or diffusion at the interface [13]. Somestudies suggest that compressive stresses arising from a dif-ference in coefcient of thermal expansion (CTE) between aceramic andzirconia have a positive correlation with thebondstrength [14]. This is a general principle that should applyin the present context. Ordinary silane chemistry also maynot improve the afnity of resin cements for zirconia sinceit is more stable then silica-based ceramics and cannot behydrolysed easily [3]. Zirconate coupling agents have been

introduced with the aim of improving the bonding of zirconia

to resin cements [3]. It has been reported that the phosphategroups of one of these primers, which is organophosphate-carboxylic acid monomer based, bond to metal oxides such aszirconia,whiletheorganic moiety canbe co-polymerizedwiththe monomers of lled resins and therefore might be feasibleas a surface treatment for zirconia [15].

The greatvariabilityin thedataobtained from some studiesof dental materials stems from a variety of reasons: inap-propriate set-ups for the relevant clinical problem, unsuitabletest methods, inappropriate specimen geometries, as well asseveral variables that cannot be standardized [16,17]. Bondstrength is commonly treated in direct tension since brittle

dental materials are more

sensitive to tensile stresses thancompression [18], but there are great difculties in doing thiswith ceramics because of that brittleness: risk of introductionof surface aws during the machining of the specimens, andparasitic stresses introduced in mounting small misalign-ments are enough to be problematic [17,19]. Tensile strengthas such is then often determined by a bending (or exural)test [20], which is less sensitive to such problems, althoughalignment of the supports remains critical. Shear testing isassociated with much greater difculties [17,21]. Simplesheartest designs, as commonly used in dental research, do notgive a uniform stress at the interface, especially with elasticmodulus mismatch, and withstress concentrations at contact

points [22].Flexural strength can be determined using three- andfour-point loading. However, the two modes commonly yielddifferent results [23]. Although the equations for the twoapproaches consider the differences in load applicationand stress distribution, this difference between results isexplained by the volume of the test piece affected: in a three-point test the effective volume of peak stress (or location fora critical aw) is very small and lies under the center load-ing point, while in a four-point test the region between theinner span loading points (which may be half the outer spanlength) is uniformly stressed in pure bending such that therisk of a large critical aw being present in the critical zone

is much greater. However, if a bending test is to be usedinstead of direct tension for a bonding effectiveness assess-ment, logically the bonded surface must be central. However,in three-point bending this then requires great (and, rou-tinely, essentially unobtainable) precision with the alignmentof the surface directly under the central load roller, but if thesubstrate-superstructure model is used, whether in a three-or four-point test, thespecimen is then asymmetrical in mod-ulus of elasticity and so in deection, and thus the stress stateis not accurately represented by the usual equations [24,25].Furthermore, to place a load singularity on the adhesive layeris itself problematic. Despite this, such asymmetric systemshave been used in four-point bending of zirconia-resin-dentin

and zirconia-resin [24,26]. Thus, although it could be argued

http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004

7/23/2019 1-s2.0-S0109564115000810-main epr14.pdf

3/12

d en ta l m at er ia ls 3 1 ( 2 0 1 5 ) 657668 659

that concentrated loading at the interface (i.e. in tension onthe lower surface) is relevant there would be practical dif-culties. This is borne out perhaps by the large coefcients of variation found with this approach, some 20 28%[24].

Elementary beam theory is based on the idea that thegeometry of the system is unaffected under load. In addition,the shear stress in the specimen between the outer and innersupport(s) contributes to the deection, and if this is large itmust be takenintoaccount [20], especiallyfor deectioncalcu-lations [27]. Even slight malalignment in the three-point casewill thus superimpose a shear stress on the bonded regionwhich will compromise the interpretation of the results, andthis applies especially when the bonded region has a denitethickness(as it must), and especiallywith intermediary layers:the stress causing failure is unknown. In any case, a uniformstress state is required in order to identify the weakest link,and this cannot be attained in three-point bending. This doesnotapply in thefour-pointcase.To control shear deformation,and thus without correction, a span to depth ratio ( S /d) of 16is considered acceptable for most materials, although somerequire S /d = 3264 in four-pointtests [27]. Fordental ceramics,therecommended ratio is 10:1 or greater to provide consistentstrengthvalues [28]. Four-pointbending is used for porous [29]materials such as spinel, zirconia and graphite, according toASTM C167411, a standard exural test method for advancedceramics with engineered porosity (e.g. honeycomb cellularchannels) at ambient temperatures.

The goal of laboratory testing ought to be to provide inter-pretable data through a mode of loading relevant to theserviceconditions. Certainly, it is difcult to envisage a pure ten-sile stress state in any oro-dental system (likewise for pureshear, a fortiori). On the other hand, bending, especially at thepontic of a xed dental prosthesis (FDP), is much more com-mon, and occlusal loads there cause traction forces that havebeen reported to cause a lever effect on the terminal edges of the restorations [30] rather similar to the stress state in theload axis plane of a exural test specimen. This might there-fore provide an appropriate model for simulating (or at leastrepresenting) intraoral loading conditions. Indeed, a exuralstrength test is widely used and proposed for strength testing for ceramics [27], although the symmetrical four-point modehas not previously been used for dental ceramic-resin bondstrength testing.

In microtensile tests, since the resin, treated surfaces andzirconia have differing elastic moduli, Poisson ratios andstrengths, the materials differing behaviors mutually inter-act and a pure, uniform stress eld is unattainable [25]. Thissituation cannot be controlled. In the usual so-called sheartests, uncontrolled and unmeasurable parasitic stresses occursince thebending of the material cannot be prevented despiteall theprecautions taken in test designs [17,21,31]. In contrast,three-point bending and microtensile tests are recommendedfor homogeneous materials, while four-point bending testsshould be preferred for heterogeneous materials, where fea-sible [25].

The primary motivation for the present work was to beable to study the behavior of the interface between zirco-nia and resin cement or surface treatments (and, secondarily,between resin cement and surface treatment). The unionbetween adhesive and tooth tissue was not of concern and

thus could be eliminated from thetest system. This suggestedthat a symmetrical zirconia-surface treatment-resin cement-surface treatment-zirconia test system could be used. Giventhen that all interfaces should be loaded identically, and alsogiven the problems of direct tension and shear, four-pointbending appears to provide all necessary conditions.

Thus, with a view to identifying an effective procedurefor bonding zirconia, and a suitable test method free of com-monproblems, the effect of various surface treatments on thestrength of the union to luting resins was investigated using a symmetrical specimen, bonding zirconia to zirconia, using four-point bending, enabling easier fabrication and simplerinterpretation of results.

2. Materials and methods

2.1. Specimen preparation

Presintered zirconia blocks (Vita In-Ceram YZ-40/19; Vita Zah-

nfabrik, Bad Sckingen, Germany) were cut with a low-speed(400rpm) diamond saw under water cooling (Isomet 1000;Buehler, Lake Bluff, IL, USA) to give bars with the dimensionsof 31.25 mm 6.25mm 2.50mm (each aspect measured atthree places with a digital caliper) using an acrylic jig. Thesewere then sintered at 1530 C for 7.5 h in a furnace (MOS160/1; Protherm, Ankara, Turkey) giving nal dimensionsof 25.0mm 5.0mm 2.0 mm after shrinkage. These dimen-sions were decided according to previous remarks on S/d ratio[27,28]f or this test method [29], given the size of the zirconiablocks.

The zirconia specimens were then ultrasonically cleaned(Sonorex RK102 Transistor; Bandelin, Walldorf, Germany) for

15min

in 96% ethanol (Smyras, Bornova, Izmir, Turkey) and15min in distilled water to ensure the absence of particu-late debris immediately before surface treatment. Calculationindicated that 10 specimens per subgroup would give a power>90%.

2.2. Surface treatments

The following treatments were used, with 20 pieces per groupfor n = 10 luted assemblies.

2.2.1. Control (C)No surface treatment.

2.2.2. Gritblasting (GB)Gritblasting (EasyBlast; Bego, Bremen, Germany) was per-formed normal to the luting surface with 50 m Al2O3 (SheraWerkstoff Technologie, Hannover, Germany) for 13s at a pres-sure of 2.8 bar from a distance of 10mm [32].

2.2.3. Glaze and hydrouoric acid etching (GHF)Theluting surface received two coats of sprayed glaze(CrystallGlaze spray; Ivoclar Vivadent, Schaan, Liechtenstein) yield-ing a continuous, thin layer. Glaze ring was conducted witha ceramic furnace (Programat P90; Ivoclar Vivadent), follow-ing the manufacturers detailed program, but broadly thisinvolved starting at 403 C, the temperature then being raised

http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004

7/23/2019 1-s2.0-S0109564115000810-main epr14.pdf

4/12

660 d en ta l m at er ia ls 3 1 ( 2 0 1 5 ) 657668

at 60 C/min to 725 C, which was held for 1h before cool-ing. Glaze that extended beyond the cementation surface wasremoved, by grinding at on 600-grit SiC paper under water,to avoid any supportive effect during the fracture test.

The glazed surfaces were then treated at room tempera-ture with 9.5% hydrouoric acid gel (Porcelain Etch; UltradentProducts, South Jordan, UT, USA) for 60s, rinsed with deion-ized water for 90s. They were then neutralized with a slurryof neutralizing powder (CaCO 3, Na2CO3) (IPS Ceramic Neutral-izing Powder, Ivoclar Vivadent) for 5 min, washed thoroughlyfor 20s with distilled water, and air dried [5].

2.2.4. Fusion glass-ceramic (CC)To mix the material (Crystall.Connect; Ivoclar Vivadent), thesupplied closed capsule was lightly pressed onto a vibrating plate (Ivomix; Ivoclar Vivadent) for 10s for agitation. Themix-ture was applied with a brush tip on the cementation surface,the specimen held against the vibrating plate for 5 s, and thenallowed to dryfor 60s. After cleaning awaytheexcess materialwith a clean brush tip, ring was conducted using a ceramicfurnace (Programat P90; Ivoclar Vivadent), again following themanufacturers detailed program; broadly, this involved start-ing at 403 C, the temperature then being raised at 30 C/minto 820 C for 2 min,then to840 C for 7min before cooling. Theprepared luting surface was then etched, washed, and driedas above.

2.2.5. Zirconia primer (Z)Aftercleaningthe cementationsurfaces byrinsing withdeion-ized water and air drying, two coats of primer (Z-Prime Plus;Bisco, Schaumburg, IL, USA) were applied uniformly, wetting the bondable surface, and dried with an air syringe for 35 s toremove solvent, according to the manufacturers instructions.

2.3. Luting

The zirconia bars were luted end-to-end on the preparedsurfaces with one of two chemically distinct luting resins(Panavia F 2.0: Kuraray, Tokyo, Japan [P]; Rely-X U 200: 3MESPE,St.Paul,USA[R])(Table 1).A custom-made stainlesssteelmold was used to standardize the luting material thicknessto 0.10 0.05mm. Specimens were placed in the mold end toend having applied the cement to one end of each using acotton pellet, the molds screw was tightened to approximatethe pieces and decrease the cement gap thickness to the tar-get value, as indicated by the separation of the mold faces. A4 mm gap between the molds approximating faces was left toallow for the removal of excess cement. Prior to the cementa-tion thelengthof each half-specimen ( 25mm) wasmeasuredbefore placing in the mold. The gap between the molds edgeswas measured using a digital caliper to calculate the closurerequired. After closure, excess cement was removed and themold face separation rechecked.

For both resins, catalyst and base pastes in equalamounts were mixedaccording to the manufacturers instruc-tions, applied to both luting surfaces of the zirconia bars,which were then brought together. Under magnication(loupe, 20 ), excess cement was removed using micro-applicators (Disposable micro applicators, ne, Premium PlusInternational, Hong Kong, China) to avoid dragging the resin

from the interface. For resin P, the material was then coveredwithan oxygen-inhibitingmaterial (Oxyguard; Kuraray, Tokyo, Japan),and irradiatedfor 20 s eachfrom the top and the bottom(LED Bluephase C5; Ivoclar Vivadent),. Resin R was similarlyirradiated from top andbottomfor 10s each, to prevent move-ment, then left undisturbed for 5 min (25 C, 5065% relativehumidity) to complete self-curing.

2.4. Four-point bending tests

Specimens were mounted with the luted interface centralizedbetween the upper load points. The load was applied throughrods of radius 2.0mm at a crosshead speed of 0.5 mm/min ona computer-controlled universal testing machine (AutographModel AG-5 kN; Shimadzu, Kyoto, Japan). The load at fracturewas recorded. Four-point exural strength ( 4) was calculatedfrom:

4 =3F(L Li)

2wh 2

where F is the load at fracture, Li is the distance betweenthe centers of the loading rollers (20.0 mm), L is the distancebetween the centers of the supporting rollers (40.0 mm), w isthe specimen width, and h the mean thickness (2.0mm) [29],giving S /d =20.

2.5. Fractographic analysis

After fracture, the surfaces of the specimens were evaluatedunder light microscopy (LM) (Eclipse ME600, Nikon, Melville,NY, USA) at 10 magnication. Failure was classied as: adhe-sive (a) more than 85% between the framework and the

cement; cohesive (c) more than 85% debonding within thezirconia, lute, or both; and mixed (m) other combination.This was on the grounds that neither adhesive nor cohesivefailure was seen pure, and the arbitrary cut was taken to tryto assess the dominant behavior.

Two specimens for each group were taken for furtheranalysis of the fractured surfaces using Environmental Scan-ning Electron Microscope-Energy Dispersive Spectroscopy(ESEM-EDS) (Quanta 250 FEG SEM, FEI, Hillsboro, OR, USA).Low-vacuum imaging mode was used, with no need fora con-ductive specimen coating. Specimens were evaluated under56 and 10,000 magnications and analysed with energy-dispersive X-ray spectroscopy (EDS) at 43 magnication and

10 keV.

2.6. Statistical analysis

Flexural strengths were subjected to two-way analysis of variance (2 AoV), Resin Surface Treatment, subsequentlyreducing to 1 AoV on Surface Treatment, with checksfor homogeneity of variance and normality (ShapiroWilk)(SigmaPlot 12.5, Systat Software, San Jose, CA, USA). Compar-isons were then made using the HolmSidak method.

Fracture-type distribution was examined by means of log-linear analysisof thethree-way contingency table (3 CT)[33],then by exact 2 (StatXact v.9, Cytel Software, Cambridge, MA,USA) on the then-indicated collapsed 2 CT.

http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004

7/23/2019 1-s2.0-S0109564115000810-main epr14.pdf

5/12

d en ta l m at er ia ls 3 1 ( 2 0 1 5 ) 657668 661

Table 1 Manufacturers information on products used.Product Manufacturer Composition description/mass% BatchPanavia F 2.0a Kuraray Medical, Tokyo,

JapanA paste: BPEDMA/MDP/DMAB paste:*Ba-*B-Si-glass/silica-containing composite

00037A00020A

Oxyguard II Glycerin gel 00471A

Rely-X U200bBase(Clicker Dispenser)

3 M/EspeSeefeld, Germany

Glass powder, silica, calciumhydroxide, pigment, substitutedpyrimidine, peroxy compound,initiator

315751

Rely-X U200Catalyst(Clicker Dispenser)

Methacrylated, phosphoric esters,dimethacrylates, acetate,stabilizers, self-cure initiators,light-cure initiators

Vita In-ceramYZ-40/19Zirconia Block

Vita Zahnfabrik, BadSckingen, Germany

9194%*ZrO2, 46%*Y2O,24%*HfO2,

7/23/2019 1-s2.0-S0109564115000810-main epr14.pdf

6/12

662 d en ta l m at er ia ls 3 1 ( 2 0 1 5 ) 657668

Table 2 Analysis of variance of four-point bending strength: resin surface treatment (log-transformed data).

Source of variation DF SS MS F pResin 1 0.000841263 0.000841263 0.0467 0.829Treatment 4 1.108007 0.277002 15.385 1.27 10 9Resin treatment 4 0.209951 0.0524878 2.915 0.0256Residual 90 1.620414 0.0180046

Total 99

Table 3 Analysis of variance of four-point bending strength: surface treatment (log-transformed data).

Source of variation DF SS MS F pTreatment 4 1.108007 0.277002 14.370 3.28 10 9Residual 95 1.831206 0.0192759

Total 99 2.939213

Table 4 HolmSidak comparisons of means of four-point bending strength by surface treatment.

Comparison Diff. of means t p p

7/23/2019 1-s2.0-S0109564115000810-main epr14.pdf

7/12

d en ta l m at er ia ls 3 1 ( 2 0 1 5 ) 657668 663

Table 7 Failure type discrepancy evidence. Y means: (under A) in an area that appears to be simply an adhesivefailure, there is evidence of remnants of cement; (under C) in an area that appears to be simply a cohesive failure, there isevidence of the underlying zirconia showing through. N means: not detected in the elds examined. E: as determined by high magnication ESEM examination. S: as determined by EDX spectrum.Resin P RTreatment A C A CC Y (E, S) Y (E) Y (E,S) Y (E)GB N Y (E,S) Y (E,S) Y (E)Z Y (E) Y (E,S) Y (E) Y (E,S)GHF Y (E,S) N Y (E,S) Y (E,S)CC N Y (E,S) Y (E,S) Y (E)

suggesting an interpenetrating network of lute, glaze and zir-conia for both resins. For CC, sharply distinct areas of thelute, glaze (with visible voids), and the zirconia surface werefound, with apparently relatively clean separation at theseinterfaces, while for Z remnants of primer and cement werefound, EDS showing Si from the cement, indicative again of a degree of cohesive failure of the cement as well as adhe-sive failure between cement and primer. The results of the

ESEM and EDS checks of the categorization of failure modeare shown in Table 7, indicating where contrary evidence wasfound.

4. Discussion

The symmetrical four-point test was found to be a straightfor-ward and effective means of testing bond strength, with thecoefcient of variation for the log-transformed data (8.6 2.0)being acceptable for this class of test ( Fig. 1, top) and indi-cating useful discriminatory power. Technically, the set-upcould be improvedto control more precisely theluteorcement

Fig. 1 (bottom) Four-point bending test fracture stressresults for the two resins, with means and standarddeviation error bars calculated from log values for thepooled resins. (top) Coefcient of variation for log data by

resin and treatment.

thickness, as would be true elsewhere [24,26], although it isnoted that in clinical practice such control is also problematic[24,26]so that some variation is realistic. It is also a conser-vative test in that two interfaces are loaded, with the weakestcontrolling the outcome.

It would seem self-evident that bond-strength testing should use a system that properly stresses theadhesive inter-face [34], yet while shear has been imagined to represent a

relevantfailure mode, it hasnot yet been shown,to ourknowl-edge, that this occurs in practice. Even so, shear tests have anumber of problems, including inhomogeneity of the stresseld and blatant stress concentrations and parasitic stresses[18,22,25,34,35]. In contrast, it has been said that interfacialtension is preferable [18]. However, direct tension also has dif-culties in specimen preparation, alignment and xation of thespecimens,and so againa tendencyfor aninhomogeneousstress distribution [25,34]. Flexural strength tests are, in con-trast, relatively easy to set up (complex specimen preparationincreases the likelihood of damage), and specimen xationproblems are avoided [3537]. Nevertheless, it needs to be rec-ognized that the stress eld in the load axis is a gradient, with

the tensile maximum on the convex surface. We would arguenow that this is more likely to represent the clinical situationthan the improbable direct (uniform) tension and the franklyimpossible pure shear. As we argue above, three-point testing is problematic because of the small critical volume in rela-tion to the (intentionally) narrow interfacial region. The purebending zone of a four-point test [35]reduces the criticality of (lateral) alignment and removes the stress singularity causedby the central load-point of the three-point test [36]. That isnot to say that the need for care is removed [28]. Specimenand testing apparatus designs vary in existing standards (e.g.DIN ENV 843, ISO 6872, ASTM C167411) for this test[29,37].Therefore, the properties of the material to be tested should

be very well known. Here, ASTM C167411 was preferred [29].Although zirconia for dentistry may contain from 0 to 15%

silica [38] or other glass phase, becoming very dense andhomogeneous on sintering [9], conventional adhesive proce-dures cannot work [3]because being very unreactive it is notetched by phosphoric or hydrouoric acids and so cannotprovide any roughness for simple micromechanical adhesion[3,911], unlike gritblasting. A silica coating, however, is reac-tive and etchable, while zirconia primer (see below) hasbeen said to increase chemical adhesion [3]. Although theusual resin cements provide good mechanical retention onrough surfaces, those containing phosphate monomer arepreferred [3] f or zirconia since they are said to be hydrolyt-

ically stable in the long term because of the strength of the

http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004

7/23/2019 1-s2.0-S0109564115000810-main epr14.pdf

8/12

664 d en ta l m at er ia ls 3 1 ( 2 0 1 5 ) 657668

Fig. 2 ESEM images for typical fracture surfaces, low and high magnications.

http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004

7/23/2019 1-s2.0-S0109564115000810-main epr14.pdf

9/12

d en ta l m at er ia ls 3 1 ( 2 0 1 5 ) 657668 665

chemical interaction between the phosphate ester monomersof 10-methacryloyloxydecyl dihydrogen phosphate (MDP) andhydroxyl groups on the zirconia surface [3,39]. On the otherhand, the lm thickness of MDP-containing resin cements istwice that of other resin cements and this is a disadvantagefor adaptation [40,41]. It was thought, therefore, that theresinstested here, Panavia F 2.0 (with MDP), and Rely-X U200 1 (withan alternative phosphoricestermethacrylatemonomer),weremost appropriate for the present study.

In a study on the effect of metal primers on the microten-sile bond strength to zirconia of phosphate-containing resin cements (Panavia F: 8.8 5.1MPa; Rely-X Unicem:7.2 3.2MPa), thedifference was notsignicant [39]. Likewise,an asymmetric four-point bend test of lled-resin compositeto gritblasted (125 m alumina) zirconia found no signif-icant difference between three resin cements (Panavia F:43.9 12.2; RelyX Unicem: 41.3 8.1MPa; Multilink Automix:39.0 8.9MPa) [24]. These results are echoed now (in termsof relative magntude) in symmetric four-point bend, but withrather smaller coefcients of variation.

From the results, we may say that simple gritblasting (GB)was the most effective treatment because of its simplicity(given the complexity and hazard of GHF), while the fusionglass-ceramic coating (CC) failed to achieve any detectableimprovement. Gritblasting with alumina has been widelystudied. It hasgiven goodresults withMDP-containing cement[42], has been reported to increase signicantly resin-zirconiabond strength in a shear test, whether before or after thesintering stage (4.5 to 6.5MPa) [43], and with priming alsoto increase tensile bond strength, and to improve long-termbonding, using bothhigh- and low-pressure blasting [44]. Grit-blasting is intended to remove contamination and increasethe roughness and hence the specic surface area, therebydecreasing theeffective contact angle and increasing thewet-tability of the luted material [32,45,46]. While GB was found togive the best bond strength here, there is a caveat . It is knownthat this can induce the tetragonal monoclinic phase trans-formation, despite stabilization [47]; the affected layer may besome 10 m thick [48]. While the conditions now used are notfar removed from the optimum indentied then [47], it willrequire further work to ascertain what the implications arefor bond strength. It would appear that the present methodwould provide the necessary discriminatory power.

If surface hydroxyl groups are important for bonding withphosphate-containing cements, the increased specic areaof a blasted surface should also be benecial, even withoutappreciable micromechanical key [49].

However, gritblasting has been reported to cause impact-induced aws which may affect the longevity of zirconiarestorations [24,50], although this may be controlled to someextentby decreasing theblastingpressure and grit size [34,41].Even so, shear-bond strength has been reported to increase[51], while biaxial exural strength has been found to increasealong with the monoclinic phase induced by gritblasting, anddecrease likewise when the monoclinic phase was removedby heat treatment [48]. However, while these initial cracksmay have little or no effect, under cyclic loading such cracks

1 This product is known in other regions as RelyX Unicem 2.

propagate and have been found to decrease bond strength by2030%[49,52,53].

For zirconia-supported xed partial dentures (FPDs), phos-phate monomer-containing primers, bonding agents andzirconatebinding agents have beenproposed, intended to givea similar effect as silanes on silica-based ceramics [15,49].A copolymerizable functional group is supposed to establishcontinuity with the resin system. Magne et al. investigated anexperimentalprimer of this type in a shear test and reportedthat with Panavia F it gave the highest strength [15]. A similartreatment here (Z) gave a marginal effect.

A simple fused ceramic glaze layer is meant to inltrateroughness and porosity, but its adhesion to zirconia dependsonly onvander Waals andelectrostatic forces [54]. Etchingthisglass with hydrouoric acid (to which zirconia is not suscep-tible) then gives a micromechanically retentive surface, saidto yield a durable and stable resin bond [55,56], which sur-face may also be treated with silane. This has been reportedto decrease low-temperature degradation and damage [56],hence the inclusion now of such a process. Derand et al.reported signicantly increasedbond strengthcompared withthenon-treated surface for such a treatment [54], a result nowechoed (GHF). However, while Ntala et al. found that a lithiumdisilicate glaze was better than gritblasting [57], here therewas no signicant difference between GB and GHF, albeit witha feldspathic ceramic. The differing results might stem fromeither test mode, glaze chemistry, or that fact that the glaze istwo-phase, or all three.

The thickness of the glaze layer has ranged between 20and 40 m [58,59], although a new approach reduced this tosome 79 m [5]. For comparison, the cement thickness forFPDs with current CAD-CAM systems has been reported tobe 3050 m, but 39502 m for zirconia-supported devices,with marginal openings of 8272 m [60]. While the glazelayer for GHF would not affect internal adaptation, the thick-ness should be carefully controlled for zirconia prostheses(CAD-CAM may offer a means of doing this in practice[61]). Here, control of the glaze for CC was compromisedsince the wet mixture showed shear-thinning under vibra-tion. A reformulation or a method better to control this isrequired.

Considering the distribution of fracture types, that modewasaffected by treatment canbe expected, but the signicantinteraction Resin Mode|Treatment is noteworthy ( Table 6). Itcan be seen from Table 5 that the proportion of adhesive fail-ures for resin R is appreciably higher. Whether this conrmsthe value of MDP (as in resin P), or identies some deciencyin resin R, requires further investigation.

Thus, using LM to estimate percentage area coverage inorder to classify the failure type is a simple enough proce-dure for this kind of relatively crude analysis of thebehavior of cements. However, it is clear from thedetection byEDS of rem-nants of cement in areas that were apparently plain adhesivefailure (that is, seeming to be clean zirconia substrate), whenviewed by LM or low-magnication ESEM, that such simplecategorization may be misleading ( Table 7). Thus, for exam-ple, in thecaseof P-Z, remnants of cement werefound in areasthat had apparently failed adhesively, while in the case of R-GHFzirconiawasfoundcleanlyexposed whencohesive failureof the cement had been the LM and low-magnication ESEM

http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004

7/23/2019 1-s2.0-S0109564115000810-main epr14.pdf

10/12

666 d en ta l m at er ia ls 3 1 ( 2 0 1 5 ) 657668

Table 8 Comparison of t of failure data to Normal andWeibull Probability plots ( r 2 values).

Ordinate: Normal Normal Weibullabscissa: linear log log C 0.9334 0.9812 0.9071CC 0.8815 0.9613 0.8839GB 0.9299 0.9838 0.9076GHF 0.9415 0.9709 0.9263Z 0.9255 0.9497 0.8884

conclusion, i.e. a more mixed condition. Thus, LM alone maybe unreliable in this respect.

The thickness of the glass-ceramic layer (CC) is clear whilethat for the glazed specimens, GHF was both thinner andadherent to both the substrate and cement ( Fig. 2). Both of these treatments involve a high-temperature process, GHF:725 C max, CC: 840 C max. The thermal properties of the

coatings are similar [62] but it has been suggested that suchheattreatmentcoulddecreasethe strengthof thezirconia [63].This might account for the rather more disrupted appearanceof the fracture surfaces in these cases, which effect would beworth further study.

The distribution of failures regarding the surface treat-ment procedures support the bond strength test ndings (C:lowest 4: 26MPa and highest adhesive failure: 65%; GB: high-est 4: 52MPa and lowest adhesive failure: 5%; Z, GHF andCC gave similar 4 values and adhesive and mixed failurerates) with the exception that cohesive failures for fusionglass-ceramic (CC) were more frequent (35%). This nding points to this material being a potential alternative for zir-

conia surface treatment for enhancing adhesive cementationif the thickness concern can be solved and its strengthimproved.

Theneed to use thelog transformation to stabilizethe vari-ance is, as indicated above, often necessary for strength data.This is shown quite clearly in Fig. 3 (top), where a sensibleapproach to linearity is found (with the possible exceptionof CC), which linearity is lost on an untransformed abscissa.Further, it is commonly assumed that that the Weibull dis-tribution applies to such strength data in general; it is plainfrom Fig. 3 (bottom) that this is not the case here the curva-ture is pronounced for each surface treatment. Comparison of the coefcients of determination in each case bears out both

points ( Table 8) where the Normal distribution for log datais uniformly the best of the three, and the Weibull plot theworst except for CC. It has been said that it is often difcultto discriminate between the distributions for small samplesizes (i.e.

7/23/2019 1-s2.0-S0109564115000810-main epr14.pdf

11/12

d en ta l m at er ia ls 3 1 ( 2 0 1 5 ) 657668 667

simple gritblastingalone (GB) may be thecurrent optimizedprocedure.

Although the fusion glass-ceramic application (CC) proveddifcult to control, the concept could be viable if the formu-lation can be improved.

Zirconia primer (Z) and etched glaze (GHF) gave someimprovement in adhesionover thecontrol(C),buta protocolfor controlling the thickness of the glaze layer is required.

Light and scanning electron microscopy might not providefull information on fractured surfaces, but EDS analysis canbe recommended as a complementary method for evalua-tion of the chemistry of fracture surfaces in more detail.

Acknowledgements

This work was submitted in partial fulllment of the require-ments for the degree of Ph D for SS. It is dedicated to thememory of Professor Dr. Cenk Cura (19652011), under whose

tutelage this work was commenced. We gratefully acknowl-edge theassistance of theIzmir Institute of Technology, Izmir,Turkey, in the ESEM-EDS analysis.

r e f e r e n c e s

[1] Kelly JR, Denry I. Stabilized zirconia as a structural ceramic:an overview. Dent Mater 2008;24:28998.

[2] Denry I, Kelly JR. State of the art of zirconia for dentalapplications. Dent Mater 2008;24:299 307.

[3] Thompson JY, Stoner BR, Piascik JR, Smith R.Adhesion/cementation to zirconia and other non-silicateceramics: where are we now? Dent Mater 2011;27:71 82.

[4] Vagkopoulou T, Koutayas SO, Koidis P, Strub JR. Zirconia indentistry: Part 1. Discovering the nature of an upcoming bioceramic. Eur J Esthet Dent 2009;4:130 51.

[5] Cura C, zcan M, Isik G, Saracoglu A. Comparison of alternative adhesive cementation concepts for zirconiaceramic: glaze layer vs. zirconia primer. J Adhes Dent2012;14:7582.

[6] Zhang S, Kocjan A, Lehmann F, Kosma c T, Kern M. Inuenceof contamination on resin bond strength to nano-structuredalumina-coated zirconia ceramic. Eur J Oral Sci2010;118:396403.

[7] Foxton RM, Cavalcanti AN, Nakajima M, Sherriff M, Melo L,Watson TF. Durability of resin cement bond to aluminiumoxide and zirconia ceramics after air abrasion and lasertreatment. J Prosthodont 2011;20:84 92.

[8] zcan M, Allahbeickaraghi A, Dndar M. Possible hazardouseffects of hydrouoric acid and recommendations fortreatment approach: a review. Clin Oral Investig 2012;16:1523.

[9] zcan M, Dndar M, Cmleko glu ME. Adhesion concepts indentistry: tooth and material aspects. J Adhes Sci Technol2012;26:2281661.

[10] Chen JH, Matsumura H, Atsuta M. Effect of etchant, etching period, and silane priming on bond strength to porcelain of composite resin. Oper Dent 1998;23:250 7.

[11] Bailey LF, Bennet RJ. DICOR surface treatments for enhancedbonding. J Dent Res 1988;67:92531.

[12] Tada K, Sato T, Yoshinari M. Inuence of surface treatmenton bond strength of veneering ceramics fused to zirconia.Dent Mater 2012;31:28796.

[13] Kwon JE, Lee SH, Lim HN, Kim HS. Bonding characteristicsbetween zirconia core and veneering porcelain. Dent Mater2009;25(5):e42.

[14] Gstemeyer G, Jendras M, Dittmer MP, Bach FW, Stiesch M,Kohorst P. Inuence of cooling rate on zirconia/veneerinterfacial adhesion. Acta Biomater 2010;6:4532 8.

[15] Magne P, Paranhos MPG, Burnett LH. New zirconia primerimproves bond strength of resin-based cements. Dent Mater2010;26:34552.

[16] Anusavice KJ, Kakar K, Ferree N. Which mechanical andphysical testing methods are relevant for predicting theclinical performance of ceramic-based dental prostheses?Clin Oral Implants Res 2007;18:21831.

[17] Scherrer SS, Cesar PF, Swain MV. Direct comparison of thebond strength results of the different test methods: a criticalliterature review. Dent Mater 2010;26(2):e78 93.

[18] Ban S, Anusavice KJ. Inuence of test method on failurestress of brittle dental materials. J Dent Res 1990;69:1791 9.

[19] Armstrong S, Geraldeli S, Maia R, Raposo LHA, Soares CJ,Yamagawa J. Adhesion to tooth structure: a critical review of micro bond strength test methods. Dent Mater2010;26:e5062.

[20] Timoshenko S. Strength of materials. Part I. Elementarytheory and problems. New York: D. Van Nostrand; 1940. p.137.

[21] Placido E, Meira JB, Lima RG, Muench A, de Souza RM,Ballester RY. Shear versus micro-shear bond strength test: anite element stress analysis. Dent Mater 2007;23:1086 92.

[22] Darvell BW. Adhesion strength testing time to failor awaste of time? J Adhes Sci Technol 2009;23:935 44.

[23] Della Bona A, Donassollo TA, Demarco FF, Barrett AA,Mecholsky JJ. Characterization and surface treatment effectson topography of a glass-inltratedalumina/zirconia-reinforced ceramic. Dent Mater2007;23:76975.

[24] Mirmohammadi H, Aboushelib MN, Kleverlaan CJ, Jager N,Feilzer AJ. The inuence of rotating fatigue on the bondstrength of zirconia-composite interfaces. Dent Mater2010;26:62733.

[25] Granjon H. Fundamentals of welding metallurgy. Cambridge,UK: Abington Publishing; 1991. p. 192206.

[26] Staninec M, Kim P, Marshall GW, Ritchie RO, Marshall SJ.Fatigue of dentin composite interfaces with four-pointbend. Dent Mater 2008;24:799 803.

[27] Carter CB, Norton MG. Ceramic materials: science andengineering. Springer; 2007. p. 297 8.

[28] Jones DW, Jones PA, Wilson HJ. The relationship betweentransverse strength and testing methods for dentalceramics. J Dent 1972;1:8591.

[29] ASTM International. Designation: C1674-11. Standard testmethod for exural strength of advanced ceramics withengineered porosity (Honeycomb Cellular Channels) atambient temperatures. West Conshohocken, PA, USA: ASTMInternational; 2011.

[30] Song HY, Yi YJ, Cho LR, Park DY. Effects of two preparationdesigns and pontic distance on bending and fracturestrength of ber-reinforced composite inlay xed partialdentures. J Prosthet Dent 2003;90:347 53.

[31] Van Noort R. An introduction to dental materials. 4th ed.London: Elsevier Mosby; 2013. p. 317.

[32] Amaral R, zcan M, Valandro LF, Balducci I, Bottino MA.Effect of conditioning methods on the microtensile bondstrength of phosphate monomer-based cement on zirconiaceramic in dry and aged conditions. J Biomed Mater Res BAppl Biomater 2008;85:19.

[33] Lowry R. Log-linear analysis for an A B C contingencytable; 2001. http://vassarstats.net/abc.html [accessed01.03.14].

http://localhost/var/www/apps/conversion/tmp/scratch_7/dx.doi.org/10.1016/j.dental.2015.03.004http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0325http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0325http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0325http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0325http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0330http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0330http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0330http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0335http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0335http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0335http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0335http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0340http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0340http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0340http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0340http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0345http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0345http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0345http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0345http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0345http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0350http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0350http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0350http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0350http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0350http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0350http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0350http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0350http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0350http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0355http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0355http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0355http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0355http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0355http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0360http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0360http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0360http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0360http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0360http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0365http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0365http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0365http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0365http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0365http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0365http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0370http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0370http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0370http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0370http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0375http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0375http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0375http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0380http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0380http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0380http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0380http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0380http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0385http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0385http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0385http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0390http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0390http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0390http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0390http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0395http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0395http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0395http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0395http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0395http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0395http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0395http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0400http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0400http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0400http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0400http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0400http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://vassarstats.net/abc.htmlhttp://vassarstats.net/abc.htmlhttp://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0480http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0475http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0470http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0465http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0460http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0455http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0450http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0445http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0440http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0435http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0430http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0425http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0420http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0415http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0410http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0405http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0400http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0400http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0400http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0400http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0400http://refhub.elsevier.com/S0109-5641(15)00081-0/sbref0400http://refhub.elsevier.com/S0109-5641(15