Contrast-associated acute kidney injury in the critically ill: systematic review and meta-analysis
Stephan EHRMANN, Andrew QUARTIN, Brian Paul HOBBS, Vincent ROBERT-EDAN, Cynthia CELY, Cynthia BELL, Genevieve LYONS, Tai PHAM, Roland SCHEIN, Yimin GENG, Karim LAKHAL, Chaan S. NG
Electronic supplementary material
Complementary methods:
Systematic review:We searched Ovid MEDLINE and Embase,PubMed, Cochrane Library, Scopus and Web of Science databases from the inception of each database to December 31st, 2015.The search strategies are detailed in Table 4-7. No language restriction was applied. All references were screened based on title and abstract by two independent researchers (V.R.E. and K.L.) following a four step process according to the PICO system: (1) exclusion of studies that were obviously not clinical studies related to contrast associated AKI, editorials and conference abstracts; (2) evaluation of the imaging procedure requiring contrast, and exclusion of studies which only involved coronary angiography; (3) exclusion of studies not including a control group (i.e. those not comprising a contemporaneous group of patients not receiving iodinated contrast); (4) exclusion of studies not dealing with critically ill patients admitted to ICUs. At each step, discrepancies were addressed independently by a third researcher (S.E.); when uncertainty remained, the full text of articles were evaluated. The full text of all studies passing step 4 (i.e., ICU studies with a control group) were evaluated for their AKI outcome reporting and group matching.
Table 4: MEDLINE database search strategy
1 / contrast media/ or expacetrizoic acid/ or expdiatrizoate/ or expdiatrizoatemeglumine/ or expiodamide/ or expiodipamide/ or expiodopyracet/ or expioglycamic acid/ or expiohexol/ or expiopamidol/ or expiothalamatemeglumine/ or expiothalamic acid/ or expioxaglic acid/ or expmetrizamide/ or expmetrizoate/2 / ((contrast* or radiocontrast or iodin* or radiopaque*) adj3 (media or medium* or agent* or material*)).ti,ab.
3 / (Iodixanol* or Visipaque* or Iopromide* or Ultravist* or Ioversol* or Optiray* or Iohexol* or Omnipaque* or Iopamidol* or Isovue* or Ioxilan* or Oxilan* or Conray* or Iothalamate* or Hexabrix* or Cholografin* or Iodipamide* or Meglumine* or "MD-76 R" or diatrizoate* or Ioxaglate*).mp.
4 / ("Triiodobenzoic Acid*" or "Acetrizoic Acid" or Iodamide* or "Ioglycamic Acid" or "Iothalamic Acid" or "Ioxaglic Acid" or Metrizamide* or Metrizoate*).mp.
5 / expIodobenzoates/
6 / or/1-5 [Contrast media]
7 / exp Kidney/ab, de, in, me, pa, pp, to [Abnormalities, Drug Effects, Injuries, Metabolism, Pathology, Physiopathology, Toxicity]
8 / exp Kidney Diseases/
9 / exp Renal Insufficiency/
10 / Kidney Function Tests/
11 / exp Renal Replacement Therapy/
12 / exp Acute Kidney Injury/
13 / Renal Dialysis/
14 / Creatinine/an, bl, ch, du, ur [Analysis, Blood, Chemistry, Diagnostic Use, Urine]
15 / Cystatin C/bl, ch, du, ur [Blood, Chemistry, Diagnostic Use, Urine]
16 / (creatinine or cystatin).ti.
17 / ((kidney or renal) adj2 (failure* or insufficienc* or dysfunction or injur* or impairment or function or impaired)).ti,ab.
18 / (nephropath* or nephrotoxic*).ti,ab.
19 / (dialyses or dialysis).ti.
20 / or/7-19 [renal insufficiency]
21 / 6 and 20 [contrast media+ renal insufficiency]
22 / ((contrast* or contrast-induced) adj3 (nephropath* or nephrotoxic*)).ti,ab.
23 / 21 or 22 [Contrast-induced nephropathy]
24 / ((contrast* or radiocontrast or iodin* or radiopaque*) adj5 (AKI or kidney or renal or creatinine or cystatin or nephropath* or nephrotoxic* or media or medium* or agent* or material*)).ti,ab.
25 / (contrast* or radiocontrast* or iodin* or radiopaque*).ti.
26 / *contrast media/ and exp *Acute Kidney Injury/
27 / or/24-26
28 / 23 and 27
29 / (AKI or creatinine or cystatin or nephropath* or nephrotoxic* or dialyses or dialysis).ti.
30 / ((kidney or renal) adj2 (failure* or insufficienc* or dysfunction or injur* or impairment or function or impaired)).ti.
31 / exp *Acute Kidney Injury/
32 / Renal Dialysis/
33 / (dialyses or dialysis).ti.
34 / or/29-33
35 / 28 and 34
36 / (animals not (humans and animals)).sh.
37 / 35 not 36 [human study]
38 / remove duplicates from 37
39 / case reports.pt. or "case report*".ti.
40 / 38 not 39
Table 5: Embase database search strategy
1 / contrast medium/2 / iodinated contrast medium/
3 / exp ionic contrast medium/
4 / exp nonionic contrast medium/
5 / ((contrast* or radiocontrast* or radiopaque*) adj3 (media* or medium* or agent* or material*)).ti,ab.
6 / (Iodixanol* or Visipaque* or Iopromide* or Ultravist* or Ioversol* or Optiray* or Iohexol* or Omnipaque* or Iopamidol* or Isovue* or Ioxilan* or Oxilan* or Conray* or Iothalamate* or Hexabrix* or Cholografin* or Iodipamide* or Meglumine* or "MD-76 R" or diatrizoate* or Ioxaglate*).mp.
7 / expiodobenzoic acid derivative/
8 / ("Triiodobenzoic Acid*" or "Acetrizoic Acid" or Iodamide* or "Ioglycamic Acid" or "Iothalamic Acid" or "Ioxaglic Acid" or Metrizamide* or Metrizoate*).mp.
9 / or/1-8 [Contrast media]
10 / exp kidney disease/
11 / exp kidney function test/
12 / ((kidney or renal) adj2 (failure* or insufficienc* or dysfunction or injur* or impairment or function or impaired)).ti,ab.
13 / exp kidney function/
14 / exp renal replacement therapy/
15 / (nephropath* or nephrotoxic*).ti,ab.
16 / exp creatinine blood level/
17 / exp creatinine clearance/
18 / exp creatinine urine level/
19 / exp cystatin C/
20 / exp glomerulus filtration rate/
21 / (creatinine or cystatin).ti.
22 / or/10-21 [renal insufficiency]
23 / 9 and 22
24 / ((contrast* or contrast-induced) adj3 (nephropath* or nephrotoxic*)).ti,ab.
25 / 23 or 24 [contrast media + renal insufficiency]
26 / ((contrast* or radiocontrast* or iodin* or radiopaque*) adj5 (AKI or kidney or renal or creatinine or cystatin or nephropath* or nephrotoxic* or media or medium* or agent* or material*)).ti,ab.
27 / (contrast* or radiocontrast* or iodin* or radiopaque*).ti.
28 / 26 or 27
29 / 25 and 28
30 / (AKI or kidney or renal or creatinine or cystatin or nephropath* or nephrotoxic*).ti.
31 / 29 and 30
32 / Human/
33 / Nonhuman/ or ANIMAL/ or Animal Experiment/
34 / 33 not 32
35 / 31 not 34 [human study only]
36 / exp case report/ or "case report*".ti.
37 / 35 not 36
38 / remove duplicates from 37
39 / conference abstract.pt.
40 / 38 and 39
41 / 38 not 39 [exclude conference abstracts]
42 / ("acetrizoic acid" or iodamide* or iodopyracet* or "ioglycamic acid" or "iothalamic acid" or "ioxaglic acid" or metrizamide* or metrizoate* or Iodixanol* or Visipaque* or Iopromide* or Ultravist* or Ioversol* or Optiray* or Iohexol* or Omnipaque* or Iopamidol* or Isovue* or Ioxilan* or Oxilan* or Conray* or Iothalamate* or Hexabrix* or Cholografin* or Iodipamide* or Meglumine* or "MD-76 R" or diatrizoate* or Ioxaglate* or "Triiodobenzoic Acid*" or Iodobenzoate*).sh.
43 / (contrast* or radiocontrast* or iodin* or radiopaque*).ab.
44 / 42 and 43
45 / *contrast medium/
46 / *iodinated contrast medium/
47 / exp *ionic contrast medium/
48 / exp *nonionic contrast medium/
49 / *expiodobenzoic acid derivative/
50 / ("acetrizoic acid" or iodamide* or iodopyracet* or "ioglycamic acid" or "iothalamic acid" or "ioxaglic acid" or metrizamide* or metrizoate* or Iodixanol* or Visipaque* or Iopromide* or Ultravist* or Ioversol* or Optiray* or Iohexol* or Omnipaque* or Iopamidol* or Isovue* or Ioxilan* or Oxilan* or Conray* or Iothalamate* or Hexabrix* or Cholografin* or Iodipamide* or Meglumine* or "MD-76 R" or diatrizoate* or Ioxaglate* or "Triiodobenzoic Acid*" or Iodobenzoate*).ti.
51 / 27 or 44 or 45 or 46 or 47 or 48 or 49 or 50
52 / 25 and 51
53 / 30 and 52
54 / 53 not 34
55 / 54 not 36
Table 6: Scopus database search strategy
(((TITLE(contrast* OR radiocontrast* ORiodin* OR radiopaque*)) OR (TITLE("acetrizoic acid" ORiodamide* ORiodopyracet* OR "ioglycamic acid" OR "iothalamic acid" OR "ioxaglic acid" ORmetrizamide* ORmetrizoate* ORiodixanol* ORvisipaque* ORiopromide* ORultravist* ORioversol* ORoptiray* ORiohexol* ORomnipaque* ORiopamidol* ORisovue* ORioxilan* ORoxilan* ORconray* ORiothalamate* ORhexabrix* ORcholografin* ORiodipamide* ORmeglumine* OR "MD-76 R" ORdiatrizoate* ORioxaglate* OR "Triiodobenzoic Acid*" ORiodobenzoate*))) AND ((((TITLE-ABS-KEY((contrast OR radiocontrast ORiodin* OR radiopaque*) W/4 (aki OR kidney OR renal OR creatinine OR cystatin ORnephropath* OR nephrotoxic* OR media OR medium* OR agent* OR material*))) OR (TITLE(contrast* OR radiocontrast* ORiodin* OR radiopaque*))) AND ((((TITLE-ABS-KEY(((contrast OR radiocontrast ORiodin* OR radiopaque*) PRE/3 (media OR medium* OR agent* OR material*)))) OR (TITLE-ABS-KEY("acetrizoic acid" ORiodamide* ORiodopyracet* OR "ioglycamic acid" OR "iothalamic acid" OR "ioxaglic acid" ORmetrizamide* ORmetrizoate* ORiodixanol* ORvisipaque* ORiopromide* ORultravist* ORioversol* ORoptiray* ORiohexol* ORomnipaque* ORiopamidol* ORisovue* ORioxilan* ORoxilan* ORconray* ORiothalamate* ORhexabrix* ORcholografin* ORiodipamide* ORmeglumine* OR "MD-76 R" ORdiatrizoate* ORioxaglate* OR "Triiodobenzoic Acid*" ORiodobenzoate*))) AND ((TITLE-ABS-KEY(nephropath* OR nephrotoxic*)) OR (TITLE-ABS-KEY((kidney OR renal) W/3 (failure* ORinsufficienc* OR dysfunction ORinjur* OR impairment OR function OR impaired))) OR (TITLE((creatinine OR cystatin))))) OR (TITLE-ABS-KEY((contrast OR contrast-induced) PRE/3 (nephropath* OR nephrotoxic*))))) AND (TITLE(aki OR kidney OR renal OR creatinine OR cystatin ORnephropath* OR nephrotoxic*)))) AND NOT (TITLE(rat* OR rabbit* OR mice OR mouse OR cat* OR dog*))Table 7: Web of science database search strategy
25 / #22 NOT #2424 / #21
Refined by: DOCUMENT TYPES:(MEETING ABSTRACT )
21 / #19 NOT #20
20 / TI=(rat* OR rabbit* OR mice OR mouse OR cat* or dog*)
19 / #18 AND #17
18 / TI=(AKI or kidney OR renal OR creatinine OR cystatin OR nephropath* OR nephrotoxic*)
17 / #16 AND #14
16 / #15 OR #12
15 / TI=("acetrizoic acid" or iodamide* or iodopyracet* or "ioglycamic acid" or "iothalamic acid" or "ioxaglic acid" or metrizamide* or metrizoate* or Iodixanol* or Visipaque* or Iopromide* or Ultravist* or Ioversol* or Optiray* or Iohexol* or Omnipaque* or Iopamidol* or Isovue* or Ioxilan* or Oxilan* or Conray* or Iothalamate* or Hexabrix* or Cholografin* or Iodipamide* or Meglumine* or "MD-76 R" or diatrizoate* or Ioxaglate* or "Triiodobenzoic Acid*" or Iodobenzoate*)
14 / #13 AND #10
13 / #12 OR #11
12 / TI=(contrast* OR radiocontrast* OR iodin* OR radiopaque*)
11 / TS=((contrast or radiocontrast or iodin* or radiopaque*) SAME (AKI or kidney or renal or creatinine or cystatin or nephropath* or nephrotoxic* or media or medium* or agent* or material*))
10 / #9 OR #8
9 / TS=((contrast or contrast-induced) SAME (nephropath* or nephrotoxic*))
8 / #7 AND #3
7 / #6 OR #5 OR #4
6 / TI=((creatinine or cystatin))
5 / TS=((kidney or renal) SAME (failure* or insufficienc* or dysfunction or injur* or impairment or function or impaired))
4 / TS=(nephropath* or nephrotoxic*)
3 / #2 OR #1
2 / TS=("acetrizoic acid" or iodamide* or iodopyracet* or "ioglycamic acid" or "iothalamic acid" or "ioxaglic acid" or metrizamide* or metrizoate* or Iodixanol* or Visipaque* or Iopromide* or Ultravist* or Ioversol* or Optiray* or Iohexol* or Omnipaque* or Iopamidol* or Isovue* or Ioxilan* or Oxilan* or Conray* or Iothalamate* or Hexabrix* or Cholografin* or Iodipamide* or Meglumine* or "MD-76 R" or diatrizoate* or Ioxaglate* or "Triiodobenzoic Acid*" or Iodobenzoate*)
1 / TS=((contrast or radiocontrast or iodin* or radiopaque*) SAME (media or medium* or agent* or material*))
Bayesian meta-analysis:Given the factors believed to contribute to the risk of AKI, both matching and multiple regression were utilized in the statistical model to adjust for imbalances due to enrollment trends in potentially important patient-level prognostic risk factors for AKI that were measured consistently among studies. Included studies matched patients receiving intravascular iodinated contrast with control patients on a one-to-one basis. Furthermore, multiple linear regression was used to adjust the baseline log-odds of AKI for each patient as a function of the partial effects of three prognostic factors: age, contrast volume and serum creatinine at the time of contrast administration. Thus, matching and multiple regression were used for AKI risk adjustment. Accounting for the prognostic features using the aforementioned methods enhances the precision of the estimator used for inferring AKI risk attribution to contrast infusion.
Implementation details of the Bayesian hierarchical modeling used to characterize the baseline log-odds of AKI while accounting for both inter-study and within-study, inter-pair heterogeneity in AKI risk:
Let and denote observed values of the three patient-level prognostic covariates. Denoting the probability of AKI for the ith patient of the jth pair nested within study h by the statistical model for the baseline log-odds of AKI was characterized by the following linear combination:
where represents the random intercept for study h, represents the nest random effect for pair j within study h, and the s represent partial linear regression coefficients that adjust the baseline log-odds for the effects of the prognostic covariates. Using pair indices to denote the presence of contrast and to denote its absence, let , denote the log-odds of AKI for the noncontrast member of the jth matched pair within study h and similarly let denote the baseline log-odds of AKI for the concordant pair member receiving contrast. The purpose of the meta-analysis was to conduct posterior inference on the relative effect of contrast as estimated by the ratio of odds of AKI (with contrast versus without contrast), , which was assumed to be identical for all studies h and pairs j.
The study-specific random intercepts, were assumed to be exchangeable using a hierarchical Gaussian prior centered at mean with standard deviation . Given the relatively low incidence of AKI, which was 8% in the studied population, hyperparameter assumed Gaussian distribution with mean fixed at the log-odds of 0.1. Hyperparameter assumed a uniform prior distribution on the interval (0, 10). Vectors of random effects were estimated independently by study using zero-centered hierarchical Gaussian priors with study-specific standard deviation hyperparameters that assumed uniform distribution on the interval (0,). The regression coefficients assumed independent Gaussian priors with mean = 0 and precision = 0.1.
All study-specific Bayesian analyses were implemented assuming a non-hierarchical Gaussian prior for the intercept parameter with mean fixed at the log-odds of 0.1 and standard deviation = 5. The approach to Bayesian sensitivity analysis follows the recommendations put forth by Spiegelhalter et al. [E1]. Bayesian computation used Markov chain Monte Carlo sampling implemented by Open-BUGS software (version 3.2.3) with the BRugs package in statistical software R (version 2.12.2).
Complementary results:
Details about patient-level data of the three studies included in the meta-analysis are provided in Table 8.
Table 8: Summary of datasetsEhrmann et al.,
France [28] / Celyet al.,
Florida [26] / Ng et al.,
Texas [25]
N=292 / N=106 / N=162
Age (years) / 61 (14-87) / 68 (47-91) / 58 (18-82)
Female sex / 104 (37%) / 1 (1%) / 72 (44%)
Diabetes / 39 (13%) / 30 (28%)
SOFA score at inclusion / 5 (0-17) / 4 (0-10) / 2 (0-8)
Serum creatinine level at inclusion (µmol/L) / 80 (31-587) / 76 (38-206) / 71 (35-123)
Iodinated Contrast Media
Volume (mL) / 100 (47-210) / 150 (98-180) / 150 (100-150)
Agent
Iohexol 510 mosm/kg / 2 (1%)
Iohexol 640 mosm/kg / 45 (31%)
Iohexol 780 mosm/kg / 25 (17%)
Iomeprol 521 mosm/kg / 1 (1%)
Iomeprol 618 mosm/kg / 51 (35%)
Iomeprol 726 mosm/kg / 3 (2%)
Iopamidol 630 mosm/kg / 1 (1%)
Iopamidol 790 mosm/kg / 7 (5%)
Iopamidol 835 mosm/kg / 2 (1%)
Ioversol 645 mosm/kg / 9 (6%)
Ioversol 702 mosm/kg / 165 (100%)
Iodixanol 290 mosm/kg / 22 (42%)
Iopromide 610 mosm/kg / 31 (59%)
Peak serum creatinine level after inclusion (µmol/L) / 80 (34-572) / 76 (38- 450) / 71 (35-477)
AKI as originally defined in papers / 16 (6%)
AKIN 48h / 33 (31%)
33% decline in measured GFR over 72h / 28 (17%)
AKIN over 7 days
AKI as defined in meta-analysis (AKIN over 72h) / 21 (7%) / 44 (42%) / 13 (8%)
Renal Replacement Therapy / 13 (5%) / 2 (2%) / 3 (2%)
Hospital death rate / 69 (24%) / 27 (26%) / 31 (19%)
SOFA score: sequential organ failure assessment score [E2]; AKI: acute kidney injury; AKIN: acute kidney injury network definition of AKI: serum creatinine concentration absolute increase of 26.4 µmol/L or more, relative increase of 50% or more [E3]; GFR: glomerular filtration rate. Quantitative variables are indicated as median (range) and qualitative variables as absolute counts (%).
Objective meta-analysis:
A priori and a posteriori distributions of the OR for the objective meta-analysis are presented in Figure 3.
Figure 3: A priori and a posteriori distributions of odds ratios (OR) in the objective meta-analysis.
The a priori distribution was centered on an OR value of 1 and was given a 95% highest posterior density (HPD) interval of 0-4.33. This distribution models a neutral state of a priori belief, making no hypothesis concerning an increased or decreased acute kidney injury (AKI) risk attributable to iodinated contrast media (contrast) as the area under the curve, representative of the cumulative probability density, is equal below and above the null value 1. The resulting a posterioridistribution is therefore intrinsically only supported by the data observed in the clinical studies, i.e., not influenced by any a priori belief that contrast causes AKI. The observed data of the three studies included in the meta-analysis did not significantly alter this a priori objective hypothesis, as the meta-analysis a posteriori OR distribution remained centered on 1 indicating a lack of evidence for a significant AKI risk attributable to contrast.
Subjective meta-analysis:
The probability density distributions of the subjective a priorihypothesis of an OR = 1.37, representing prior belief of the clinical community, for the three studies included in the meta-analysis individually, as well as the a posterioridistribution resulting from their combination are depicted in Figure 4.
Figure 4:A priori and a posteriori distributions of odds ratios (OR) in the subjective meta-analyses
The a priori distribution was centered on an odds ratio (OR) value of 1.37 to reflect the commonly-held belief of a significant acute kidney injury (AKI) risk attributable to iodinated contrast media (contrast). In order to observe an a posteriori distribution with a 95% highest posterior density interval (HPD) >1.0, i.e., supporting statistically such an attributable risk, given the data observed in the studies, modelling a strong a priori strengths of belief in this 1.37 OR was necessary. Increasing a priori strengths of belief was modelled using a priori OR distributions with increasing effective sample size (RESS). The a priori distribution corresponding to the minimum RESS value needed to observe an a posteriori significant OR distribution is represented (OR distribution centered on 1.37 with a 95% highest posterior density interval of 1.06-1.72). This RESS value is 4.8-fold higher than that of the objective meta-analysis (main manuscript). A posteriori distribution of OR resulting of the combination of this subjective a priori distribution with the observed data was centered on a median a posteriori OR of 1.31 with an HPD of 1.00-1.61, effectively demonstrating minimal learning from the objective data observed in the clinical studies.
Sensitivity analysis:
Results of the sensitivity analysis, consisting in bracketing the 1.37 OR considered to reflect prior belief of the clinical community with a more conservative value of 1.20 and a higher value of 1.50, to test lower and higher increased risks of AKI the clinical community may attribute to contrast are presented in Table 9 with corresponding Forest plots in Figure 5 and Figure 6.
Table 9: Sensitivity analysisA priori hypotheses / A posteriori results
Type / OR / 95% HPD / RESS / OR / 95% HPD / RESS
Objective Meta-analysis / Objective / 1.0 / (0.00, 4.33) / 1.0 / 0.95 / (0.45, 1.62) / 14.6 a
Low attributable risk hypothesis / Subjective / 1.20 / (1.01, 1.40) / 176.5 b / 1.18 / (1.00, 1.37) / 187.9
Principal subjective meta-analysis / Subjective / 1.37* / (1.06, 1.72) / 70.0 / 1.31 / (1.00, 1.61) / 79.1
High attributable risk hypothesis / Subjective / 1.50 / (1.08, 1.95) / 42.9 c / 1.37 / (1.00, 1.78) / 51.3
Results of the objective meta-analysis are presented in the upper panel for reference. The a posteriori relative effective sample size (RESS)a reflects the amount of information carried by this “impartial” analysis (no a priori opinion about AKI risk attributable to iodinated contrast media (contrast): a priori OR =1. In the lower panel sensitivity analysis is presented taking as an a priori hypothesis an OR value of 1.20 to reflect a lower a priori AKI risk attributable to contrast than the principal subjective analysis (OR=1.37) and an a priori OR value of 1.50 to reflect a higher attributable risk. These values address the uncertainty concerning the putative consensus of the clinical community regarding AKI risk attributable to contrast. Similarly to the principal subjective meta-analysis a minimum a priori RESS value of, respectively 176.5b and 42.9cneeded to be introduced in the model to observe a posteriori OR distributions statistically supporting an attributable risk (HPD interval not comprising 1.0); RESS values, respectively, 12-foldb and 3-fold c higher than the RESS of the objective meta-analysis a
Results of the subjective meta-analysis modelling an a priori belief of a high magnitude of the AKI risk attributable to contrast (a priori OR=1.50) showed that the minimum a priori RESS value needed in the Bayesian model in order to observe an a posteriori OR distribution supporting the conclusion of a significant AKI risk attributable to contrast was 42.9. Although this value is lower than the corresponding a priori minimum RESS of 70 observed in the principal subjective meta-analysis it is still 3-fold higher than the RESS of the objective meta-analysis (Figure 5). This indicates that practitioners who believe in a very high magnitude of AKI risk attributable to contrast, would still need to have a very high degree of confidence in their belief for not abandoning this subjective opinion when confronted with the observed data. At the other end of the spectrum, the sensitivity analysis with a low a priori attributable risk hypothesis (OR=1.20) showed a minimum a priori RESS value needed of more than 10-times the RESS value of the objective meta-analysis to observe a statistically significant a posteriori distribution (Figure 6). This extreme result raises the question if such practitioners would be considered to be evidence based observers given the extraordinary faith they would show in their a priori subjective belief as compared to the amount of available information gathered in matched clinical studies comprising a control group.