Objectives
To test whether dynamic contour tonometry yields ocular pulse amplitude (OPA) measurements that are independent of corneal thickness and curvature, and to assess variables of observer agreement.
Methods
In a multivariate cluster analysis on 223 eyes, the relationship between central corneal thickness, corneal curvature, axial length, anterior chamber depth, intraocular pressure, sex, age, and OPA measurements was assessed. Intraobserver and interobserver variabilities were calculated from repeated measurements obtained from 8 volunteers by 4 observers.
Results
The OPA readings were not affected by central corneal thickness (P = .08), corneal curvature (P = .47), anterior chamber depth (P = .80), age (P = .60), or sex (P = .73). There was a positive correlation between OPA and intraocular pressure (0.12 mm Hg/1 mm Hg of intraocular pressure; P<.001) and a negative correlation between OPA and axial length (0.27 mm Hg/1 mm of length; P<.001). Intraobserver and interobserver variabilities were 0.08 and 0.02 mm Hg, respectively, and the intraclass correlation coefficient was 0.89.
Conclusions
The OPA readings obtained with dynamic contour tonometry in healthy subjects are not influenced by the structure of the anterior segment of the eye but are affected by intraocular pressure and axial length. We found a high amount of agreement within and between observers.
Most of the recent studies on pulsatile ocular hemodynamics are based on the developments of the recording pneumatonometer of Langham and McCarthy,1 such as the ocular blood flow tonometer (OBF) (OBF Laboratories Ltd, Malmesbury, England) or the ocular blood flow analyzer (Paradigm Medical Industries, Salt Lake City, Utah). The reliability of pressure waves recorded by pneumatonometers has repeatedly been questioned. Intraocular pressure (IOP) measurements obtained by the OBF system have been shown to be even more dependent on central corneal thickness (CCT)2-4 and corneal curvature3 than those of Goldmann applanation tonometry. A similar effect of CCT on IOP measurements has been shown for the OBF analyzer system,5 and the repeatability of the OBF system compared unfavorably with that of Goldmann applanation tonometry.6
Dynamic contour tonometry (DCT) (Pascal; Swiss Microtechnology AG, Port, Switzerland) represents a novel type of continuously recording tonometry giving a reading of IOP and ocular pulse amplitude (OPA) (Figure 1). According to the working principles of DCT, matching up the concave pressure sensor with the cornea provides direct measurements independently of corneal properties.7 In eyes that have undergone laser in situ keratomileusis, DCT has been shown to measure IOP independently of CCT.8-10 In normal eyes, IOP measurements assessed with DCT are not affected by corneal curvature, and intraobserver and interobserver variabilities compare favorably with those of Goldmann applanation tonometry.11 Furthermore, IOP values obtained with DCT are closer to the manometric reference pressure than those measured with Goldmann applanation tonometry.12 However, in a given tonometer, IOP and OPA do not necessarily correlate with regard to dependency on corneal variables. With the OBF analyzer, CCT affects IOP readings but not OPA measurements, whereas corneal curvature affects OPA recordings but not IOP readings.5 Therefore, the aim of this study was to investigate the factors affecting OPA measurements done using DCT in healthy subjects. In addition, we assessed the reference range for OPA and variables of observer agreement.
We enrolled 150 healthy hospital staff members. The study group consisted of the same cohort as presented in an earlier report11 comparing IOP measurements using Goldmann applenation tonometry and DCT. All study participants gave informed consent to be enrolled in the study, which was approved by the local medical ethics committee and was performed according to the tenets of the Declaration of Helsinki. All participants were asked for bilateral measurements, but the option to have only 1 eye measured was offered as well. Seventy-four volunteers agreed to unilateral measurements only. None of the subjects had any history of ocular disease, trauma, or surgery. Slitlamp findings were unremarkable. Owing to the different calibrations of DCT and Goldmann applenation tonometry, reference IOP values as measured by DCT are 1.0 to 2.6 mm Hg higher.13-16 Because the upper threshold for reference IOP measured with DCT has so far not been defined, all of the IOP values recorded were included in the study. Axial length, corneal curvature, and anterior chamber depth were obtained by means of an optical biometry system (IOL Master; Carl Zeiss AG, Feldbach, Switzerland), followed by ultrasonic pachymetry (SP2000; Tomey Corp, Cambridge, Mass). The pachymeter probe was placed on the center of the cornea, and the mean of 3 readings was calculated for each eye. Finally, DCT was performed in a sitting position according to the manufacturer's guidelines. A technically identical prototype of the model launched in November 2003 was used; however, it was devoid of auxiliary routines to calculate the grade of measurement quality (Q value). The smallest possible recording time of about 5 seconds per single measurement was used. Every measurement furnished an IOP and an OPA reading on a digital display, with a precision of 1 decimal place. Three consecutive readings were taken and the mean of the values was recorded. In the absence of objective Q values, a set of measurements was excluded from the study in the event that the output routine failed to calculate the OPA for any of the 3 individual readings.
To study the intraobserver and interobserver variability, the OPA was measured in 8 participants by 4 fully masked observers 3 times each, resulting in a total of 96 measurements. The delay between different observers was kept as short as possible (<30 seconds).
For statistical analysis, OPA values were transformed into their logarithm to obtain a normally distributed dependent variable. To avoid the problem of colinearity and overparameterization, we used the strategy proposed by Kleinbaum et al.17
The variables investigated were CCT, IOP, anterior chamber depth, and axial length. The variables considered to be confounders were sex, age, eye (left or right), and keratometric reading. Data were analyzed to determine whether a linear relationship existed between the OPA measurements and the different independent continuous variables. Dummy variables (indicator variables) were used if nonlinearity was observed. This set of variables was used to fit a first model. In a second step, interaction terms were entered to see whether their inclusion would give a statistically significant better fit of the model (using the log-likelihood ratio test,18 considering P<.05 as statistically significant). Finally, new models were fitted, leaving out 1 confounder at a time to see how the coefficients of the other characteristics would change.
Because the final data set contained some measurements (n = 75) in both eyes of the same subject, a clustered analysis was applied in which each participant was considered as a cluster to adjust for the dependency of measurements in the fellow eye.
Analyses of variance were performed to study the variability between the different observers and the OPA readings. The contribution of an independent variable (observer, subjects, observer × subject interaction, and residual error) to the variance of the dependent variable (OPA measurement) was estimated using the variance components procedure. On the basis of the variance components, we calculated the intraclass correlation coefficient and the interobserver and intraobserver variabilities. The intraclass correlation coefficient (also referred to as reliability coefficient) represents the proportion of the total variability in a given measure that can be attributed to the true variability among subjects. It assumes values from 0.0 to 1.0, approaching 1.0 when the variation in measurements is solely due to subject-to-subject variability.
Statistical analysis was performed using commercially available software (Stata, version 8.2 [Stata Corp LP, College Station, Tex], and the SPSS statistical software package, version 10 [SPSS Inc, Chicago, Ill]).
We excluded 1 volunteer from the unilateral and 1 volunteer from the bilateral group (total of 3 eyes excluded) because the output routine failed to calculate an accurate OPA from these eyes. Thus, the study included 148 healthy participants (64 women and 84 men) contributing 223 normal eyes. We evaluated measurements taken from 115 right and 108 left eyes. The characteristics of the study group are given in Table 1.
The OPA readings ranged from 0.9 to 7.2 mm Hg (median, 3.0 mm Hg; 10th-90th intercentile range, 1.8-4.3 mm Hg). The OPA difference in the 75 pairs of eyes examined ranged from 0.0 to 2.5 mm Hg (median, 0.4 mm Hg; 10th-90th intercentile range, 0.1-1.2 mm Hg). There was no significant difference in the OPA between right and left eyes (P = .71). Multiregression analysis showed no significant effect of CCT (P = .08) (Figure 2), keratometry (P = .47), anterior chamber depth (P = .80), or sex (P = .73) on OPA readings. The mean OPA was 0.2 mm Hg higher in the group older than 50 years compared with the group younger than 30 years. However, this difference was not significant (P = .60).
There was a positive correlation between OPA and IOP (0.12 mm Hg/1 mm Hg of IOP; 95% confidence interval, 0.07-0.18; P<.001) and a negative correlation between OPA and axial length (0.27 mm Hg/1 mm of length; 95% confidence interval, −0.14 to 0.41; P<.001).
When the OPA was measured repeatedly in the same eye, the intraclass correlation coefficient was 0.89, the intraobserver variability was 0.08 mm Hg, and the interobserver variability was 0.02 mm Hg (Table 2).
The purpose of continuously measuring tonometry such as DCT is to record IOP oscillations as an indirect measure for intraocular pulsatile hemodynamics. Corneal curvature and CCT have been shown to interfere with the pristine detection of this signal by means of pneumatonometry.2,3,5 However, the present study is the first, to our knowledge, to describe OPA measurements unaffected by these corneal variables. This finding reproduces what has already been demonstrated for IOP measurements with the same instrument on the same cohort.11
The independence of corneal dimensions is attributed to the concave shape of the dynamic contour tonometer tip. When in touch with the tip, the contour-matched cornea is thought to be tension free, allowing for direct IOP measurements across its wall.7
Other factors altering the signal of IOP oscillations may not be neutralized by contour matching alone. Filling of the orbital vessels causes a pulsatile protrusion of the whole bulbus, allowing for pressure wave recordings even from enucleated eye sockets.19 Furthermore, the IOP response to the increase in ocular volume during cardiac systole depends on the elastic properties of the ocular shell,20 a phenomenon that may explain our finding of the positive correlation between OPA and IOP. With higher IOP levels, scleral wall tensions increase, and the injection of a given volume of blood results in a distinct pressure increase rather than a further elastic extension of already prestressed bulbus walls.19 Similar results have been published,21-23 but there are reports indicating no such correlation.24,25 In this study, there was a tendency toward higher OPA in thinner and therefore potentially more elastic corneas.
Moreover, OPA and IOP are not only linked by the elastic properties of the bulbus walls but also by ocular blood flow. The difference between the pressure in the ophthalmic artery and IOP represents the pressure gradient maintaining ocular perfusion. Elevated IOP levels may therefore affect pulsatile ocular blood flow, although the direction of the net effect (increase or decrease) is difficult to predict owing to regulatory mechanisms.26
Unlike DCT, the OBF and OBF analyzer use instantaneous IOP measurements to mathematically derive an estimate of pulsatile ocular blood flow (in microliters per second). Based on experimental pressure-volume relations and the heart rate, pulsatile ocular blood flow is approximated by transforming the OPA measurements into ocular volume changes.27
The heart rate itself represents another variable affecting the OPA, and such an inverse correlation has been described.28,29 It is a limitation of the present study that the effect of the heart rate has not been investigated. The second correlation found in this study, the OPA's dependence on axial length, has already been documented in the literature.22-25 These findings are generally attributed to the size of the large myopic bulbus for which a given inflow of blood represents a smaller relative volume change than for a shorter emmetropic eye. Myopia is also associated with thinning of the sclera,30 which consequently offers less resistance to expansion brought about by a pulsatile volume change. In addition, there is evidence suggesting that ocular blood flow in myopic eyes is decreased, probably owing to the reduced diameter of straightened vessels.31-33
In systemic hemodynamics, transcutaneous tonometer readings derived from peripheral arteries show an elevated pulse amplitude with age secondary to an increase in artery stiffness and pulse-wave velocity.34,35 In the ophthalmological literature, however, the correlation between OPA and age shows a contrasting trend23,36 or is not significant at all,24,37 as confirmed by the present study. This apparent paradox might be explained by the vascular resistance of the bulbus that acts as a hemodynamic transducer38 between peripheral arteries and the tonometer head. A reduced inflow of blood due to an age-related loss and hyalinosis of arterioles39 is likely to dampen the pulse transmitted via the ophthalmic artery.
Drawing comparisons between OPA standard values in the literature must take into account the different techniques by which the data were obtained. With the original pneumatonometer of Langham and McCarthy,1 the reference mean ± SD OPA was found to be 1.5 ± 0.11 mm Hg.40 With the more recent OBF and OBF analyzer pneumatonometers, mean ± SD values ranging from 2.2 ± 0.8 to 3.0 ± 1.2 mm Hg5,41 were found. The present study demonstrated a median OPA value of 3.0 mm Hg (10th-90th intercentile range, 1.8-4.3 mm Hg), which is in agreement with a previous report on DCT that described a mean ± SD OPA value of 3.08 ± 0.92 mm Hg in a smaller set of 19 healthy subjects aged 45 to 73 years.42
In the present study, we found no significant difference between an individual's eyes, which is in keeping with earlier reports that described significant differences only in cases of anisometropia of 2 or more diopters.23-25,43
Knowledge about the reference range of OPA in healthy individuals permits one to suspect systemic or bilateral disease, even if no difference between sides is detected.44 However, a significant asymmetry in OPA may indicate vascular stenosis45,46 or arteriovenous fistulas.47,48
Intraobserver and interobserver variabilities in this study were 0.08 and 0.02 mm Hg, respectively, indicating that repeated readings within and between observers are reproducible. Accordingly, we calculated an intraclass correlation coefficient of 0.89, suggesting that the total variation in OPA measurements depends mainly on subject-to-subject variability, ie, on actual differences in OPA among volunteers.
Reports about observer agreements with respect to OPA are sparse, and measuring conditions differ. With a modified slitlamp-mounted OBF tonometer, 2 observers performed 272 measurements on 34 healthy volunteers on different days and reported an intraclass correlation coefficient of 0.66.41 With a handheld dynamic observing tonometer (SmartLens; Ophthalmic Development Company AG, Zürich, Switzerland), a coefficient of 0.86 was found by 3 examiners performing 5 measurements on 10 volunteers within 2 hours. Intraobserver and interobserver variability were 0.39 and 0.40 mm Hg, respectively.49
While the indices of observer agreement that we found for DCT compare favorably with those earlier reports, some limitations of this study must be taken into consideration. Although there is no agreed-upon method to determine a sufficient sample size, the range of data derived from the 8 subjects included in this subanalysis may not be representative of a large collective. Furthermore, the nonrandomized sequence of observers may have led to unequal measuring conditions owing to a tonographic effect associated with repeated corneal distortion. Finally, although it seems unlikely that the corneal contact during ultrasonic pachymetry induced a significant tonographic effect, it has been reported that a decline in pressure may also result from accommodation or extraocular muscular or vascular influences.50 Therefore, it cannot be excluded that even biometry and pachymetry have interfered to some extent with the subsequent detection of the OPA signal.
The present study did not show any statistically significant influence of corneal curvature, CCT, or anterior chamber depth on OPA readings obtained using DCT in healthy subjects. However, OPA readings are affected by axial length and IOP. With these limitations in mind, DCT might not only provide clinically relevant information in an individual patient, but its independence of anterior segment structure might prove to be useful in determining differences between groups. As the estimation of pulsatile ocular blood flow is mainly based on the OPA, the implementation of a routine that links OPA to volume change may provide blood flow estimates that are largely independent of corneal variables.
Correspondence: Michael A. Thiel, MD, PhD, Department of Ophthalmology, University of Zürich, Frauenklinikstr 24, CH- 8091 Zürich, Switzerland (michael.thiel@usz.ch).
Submitted for Publication: November 1, 2005; final revision received February 24, 2006; accepted March 1, 2006.
Financial Disclosure: Dr Robert is a consultant to Swiss Microtechnology AG.
Funding/Support: This study was supported by the Schweizerische Fonds zur Verhuetung und Bekaempfung der Blindheit (Dr Kaufmann) and by grants 3233B0-103182 and 3200B0-103183 from the Swiss National Science Foundation (Dr Bachmann).
1.Langham
MEMcCarthy
E A rapid pneumatic applanation tonometer: comparative findings and evaluation.
Arch Ophthalmol 1968;79389- 399
PubMedGoogle ScholarCrossref 2.Bhan
ABrowning
ACShah
SHamilton
RDave
DDua
HS Effect of corneal thickness on intraocular pressure measurements with the pneumotonometer, Goldmann applanation tonometer, and Tono-Pen.
Invest Ophthalmol Vis Sci 2002;431389- 1392
PubMedGoogle Scholar 3.Gunvant
PBaskaran
MVijaya
L
et al. Effect of corneal parameters on measurements using the pulsatile ocular blood flow tonograph and Goldmann applanation tonometer.
Br J Ophthalmol 2004;88518- 522
PubMedGoogle ScholarCrossref 4.Ko
YCLiu
CJHsu
WM Varying effects of corneal thickness on intraocular pressure measurements with different tonometers.
Eye 2005;19327- 332
PubMedGoogle ScholarCrossref 5.Morgan
AJHarper
JHosking
SLGilmartin
B The effect of corneal thickness and corneal curvature on pneumatonometer measurements.
Curr Eye Res 2002;25107- 112
PubMedGoogle ScholarCrossref 7.Kanngiesser
HEKniestedt
CRobert
YC Dynamic contour tonometry: presentation of a new tonometer.
J Glaucoma 2005;14344- 350
PubMedGoogle ScholarCrossref 8.Kaufmann
CBachmann
LMThiel
M Intraocular pressure measurements using dynamic contour tonometry after laser in situ keratomileusis.
Invest Ophthalmol Vis Sci 2003;443790- 3794
PubMedGoogle ScholarCrossref 9.Duba
IWirthlin
AC Dynamic contour tonometry for post-LASIK intraocular pressure measurements.
Klin Monatsbl Augenheilkd 2004;221347- 350
PubMedGoogle ScholarCrossref 10.Siganos
DSPapastergiou
GIMoedas
C Assessment of the Pascal dynamic contour tonometer in monitoring intraocular pressure in unoperated eyes and eyes after LASIK.
J Cataract Refract Surg 2004;30746- 751
PubMedGoogle ScholarCrossref 11.Kaufmann
CBachmann
LMThiel
MA Comparison of dynamic contour tonometry with Goldmann applanation tonometry.
Invest Ophthalmol Vis Sci 2004;453118- 3121
PubMedGoogle ScholarCrossref 12.Kniestedt
CNee
MStamper
RL Dynamic contour tonometry: a comparative study on human cadaver eyes.
Arch Ophthalmol 2004;1221287- 1293
PubMedGoogle ScholarCrossref 13.Pache
MWilmsmeyer
SLautebach
SFunk
J Dynamic contour tonometry versus Goldmann applanation tonometry: a comparative study.
Graefes Arch Clin Exp Ophthalmol 2005;243763- 767
PubMedGoogle ScholarCrossref 14.Ku
JYDanesh-Meyer
HVCraig
JPGamble
GDMcGhee
CN Comparison of intraocular pressure measured by Pascal dynamic contour tonometry and Goldmann applanation tonometry.
Eye 2006;20191- 198
PubMedGoogle ScholarCrossref 15.Kniestedt
CLin
SChoe
JBostrom
ANee
MStamper
RL Clinical comparison of contour and applanation tonometry and their relationship to pachymetry.
Arch Ophthalmol 2005;1231532- 1537
PubMedGoogle ScholarCrossref 16.Doyle
ALachkar
Y Comparison of dynamic contour tonometry with Goldman applanation tonometry over a wide range of central corneal thickness.
J Glaucoma 2005;14288- 292
PubMedGoogle ScholarCrossref 17.Kleinbaum
DGKupper
LLMuller
KEAzhar
N Multiple regression analysis: general considerations. In: Applied Regression Analysis and Multivariable Methods Pacific Grove, Calif Brooks/Cole Publishing Co1998;111- 135
18.Kleinbaum
DGKlein
M The likelihood ratio test: logistic regression—a self-learning text. In: Statistics for Biology and Health 2nd New York, NY Springer-Verlag NY Inc2002;130- 134
19.Wegner
W Neue Ergebnisse ueber die pulsatorischen Schwankungen des menschlichen Bulbus und seiner Huellen.
Arch Augenheilkd 1930;1021- 32
Google Scholar 21.Bynke
HG Influence of intraocular pressure on the amplitude of the corneal pulse: a study on man and rabbits.
Acta Ophthalmol (Copenh) 1968;461135- 1145
PubMedGoogle ScholarCrossref 22.To’mey
KFFaris
BMJalkh
AENasr
AM Ocular pulse in high myopia: a study of 40 eyes.
Ann Ophthalmol 1981;13569- 571
PubMedGoogle Scholar 25.James
CBTrew
DRClark
KSmith
SE Factors influencing the ocular pulse–axial length.
Graefes Arch Clin Exp Ophthalmol 1991;229341- 344
PubMedGoogle ScholarCrossref 27.Silver
DMFarrell
RALangham
MEO’Brien
VSchilder
P Estimation of pulsatile ocular blood flow from intraocular pressure.
Acta Ophthalmol Suppl 1989;19125- 29
PubMedGoogle Scholar 28.Trew
DRJames
CBThomas
SHSutton
RSmith
SE Factors influencing the ocular pulse: the heart rate.
Graefes Arch Clin Exp Ophthalmol 1991;229553- 556
PubMedGoogle ScholarCrossref 29.Gekkieva
MOrgul
SGherghel
DGugleta
KPrunte
CFlammer
J The influence of sex difference in measurements with the Langham Ocular Blood Flow System.
Jpn J Ophthalmol 2001;45528- 532
PubMedGoogle ScholarCrossref 30.McBrien
NAGentle
A Role of the sclera in the development and pathological complications of myopia.
Prog Retin Eye Res 2003;22307- 338
PubMedGoogle ScholarCrossref 31.Avetisov
ESSavitskaya
NF Some features of ocular microcirculation in myopia.
Ann Ophthalmol 1977;91261- 1264
PubMedGoogle Scholar 32.Akyol
NKukner
ASOzdemir
TEsmerligil
S Choroidal and retinal blood flow changes in degenerative myopia.
Can J Ophthalmol 1996;31113- 119
PubMedGoogle Scholar 33.Shih
YFFitzgerald
MENorton
TTGamlin
PDHodos
WReiner
A Reduction in choroidal blood flow occurs in chicks wearing goggles that induce eye growth toward myopia.
Curr Eye Res 1993;12219- 227
PubMedGoogle ScholarCrossref 34.Kelly
RHayward
CAvolio
AO’Rourke
M Noninvasive determination of age-related changes in the human arterial pulse.
Circulation 1989;801652- 1659
PubMedGoogle ScholarCrossref 35.Franklin
SSGustin
W
IVWong
ND
et al. Hemodynamic patterns of age-related changes in blood pressure: the Framingham Heart Study.
Circulation 1997;96308- 315
PubMedGoogle ScholarCrossref 36.Ravalico
GToffoli
GPastori
GCroce
MCalderini
S Age-related ocular blood flow changes.
Invest Ophthalmol Vis Sci 1996;372645- 2650
PubMedGoogle Scholar 38.Schilder
P Ocular blood flow responses to pathology of the carotid and cerebral circulations.
Surv Ophthalmol 1994;38
((suppl))
S52- S58
PubMedGoogle ScholarCrossref 40.Langham
ME Pneumatonography: a new methodological approach to the analysis of intraocular pressure and aqueous humor dynamics in human eyes. Paper presented at: 22nd International Glaucoma Symposium May 20-24, 1974 Albi, France
41.Spraul
CWLang
GERonzani
MHogel
JLang
GK Reproducibility of measurements with a new slit lamp–mounted ocular blood flow tonograph.
Graefes Arch Clin Exp Ophthalmol 1998;236274- 279
PubMedGoogle ScholarCrossref 42.Hoffmann
EMGrus
FHPfeiffer
N Intraocular pressure and ocular pulse amplitude using dynamic contour tonometry and contact lens tonometry.
BMC Ophthalmol 2004;44
PubMedGoogle ScholarCrossref 43.Lam
AKChan
STChan
HChan
B The effect of age on ocular blood supply determined by pulsatile ocular blood flow and color Doppler ultrasonography.
Optom Vis Sci 2003;80305- 311
PubMedGoogle ScholarCrossref 45.Perkins
ES The ocular pulse and intraocular pressure as a screening test for carotid artery stenosis.
Br J Ophthalmol 1985;69676- 680
PubMedGoogle ScholarCrossref 46.Kaufmann
CFierz
AKollias
SSRobert
YC Ocular pulse amplitude in a case of innominate steal syndrome.
Am J Ophthalmol 2002;133155- 156
PubMedGoogle ScholarCrossref 47.Golnik
KCMiller
NR Diagnosis of cavernous sinus arteriovenous fistula by measurement of ocular pulse amplitude.
Ophthalmology 1992;991146- 1152
PubMedGoogle ScholarCrossref 48.Kaufmann
CThiel
MAValavanis
ALandau
K Normalisation of ocular pulse amplitude after embolisation of dural cavernous sinus arteriovenous fistula.
Klin Monatsbl Augenheilkd 2004;221431- 434
PubMedGoogle ScholarCrossref 49.Vogel
ABeck
SSchwenn
OGrus
FKrummenauer
FPfeiffer
N Reproducibility of measurement of ocular pulse amplitude and intraocular pressure using Smartlens [in German].
Ophthalmologe 2001;98944- 949
PubMedGoogle ScholarCrossref