Law

A 3-Level Prognostic Classification in Septic Shock Based on Cortisol Levels and Cortisol Response to Corticotropin

Description
A 3-Level Prognostic Classification in Septic Shock Based on Cortisol Levels and Cortisol Response to Corticotropin
Categories
Published
of 8
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
  CARING FOR THECRITICALLY ILL PATIENT A 3-Level Prognostic Classificationin Septic Shock Based on Cortisol Levelsand Cortisol Response to Corticotropin Djillali Annane, MD, PhDVe´ronique Se´bille, PhDGilles Troche´, MDJean-Claude Raphae¨l, MDPhilippe Gajdos, MDEric Bellissant, MD, PhD S EPTIC SHOCK REMAINS THE MOST commoncauseofdeathinnon-coronary intensive care units(ICUs). 1 To tackle this prob-lem,numerousanti-inflammatorythera-pieshavebeentestedduringthepastde-cade, but all of them have been unabletoimprovesurvivalinpatientswithse-vere sepsis. 2 Thus, there is an urgentneed to better characterize septic pa-tients with the worst outcome. Severalclinicalprognosticfactorshavealreadybeen identified (ie, preexisting under-lying disease, presence of organ dys-function,andseverityofillnessscores). 3 Moreover,thehormonalprofilehasbeensuggestedtobeavalidpredictorofout-come in critically ill patients. 4 How-ever, a pathophysiologic derangementthat could help identify a group of pa-tientswhomightbenefitfromaparticu-lar treatment has not been character-ized yet.The integrity of the hypothalamic-pituitary-adrenal(HPA)axisisamajordeterminant of the host’s response tostress. 5,6 Duringsepsis,theactivationof the HPA axis is highlighted by in-creasedcorticotropinreleasefromthepi-tuitarygland, 7 enhancedadrenalsecre-toryactivity, 8,9 andhighplasmacortisollevels. 10-13 However, whether endoge-nous glucocorticoid levels are ad-equate or constitute an independentpredictor of death remains contro-versial. 10-15 For instance, several stud-ies showed that the higher the plasmacortisolconcentrations,theworsethepa-tient’soutcome. 4,7,10,16-18 Incontrast,otherstudiesreportedlowercortisollevelsin AuthorAffiliations:  DepartmentofIntensiveCareUnit,Hospital Raymond Poincare´, School of Medicine ofParis-Ouest, Garches, France (Drs Annane, Raphae¨l,andGajdos);DepartmentofIntensiveCareUnit,Hos-pitalAntoineBe´cle`re,Clamart,France(DrTroche´);andDepartment of Pharmacology, Hospital Pontchail-lou, Universite´ Rennes I, Rennes, France (Drs Se´billeand Bellissant). Corresponding Author and Reprints:  Djillali Annane,MD, PhD, Service de Re´animation Me´dicale, HoˆpitalR. Poincare´, 104 Blvd Raymond Poincare´, 92380Garches, France (e-mail: djillali.annane@rpc.ap-hop-paris.fr). Caring for the Critically Ill Patient Section Editor: Deborah J. Cook, MD, Consulting Editor,  JAMA . Advisory Board:  David Bihari, MD; Christian Brun-Buisson,MD;TimothyEvans,MD;JohnHeffner,MD;Norman Paradis, MD. Context  Thehypothalamic-pituitary-adrenalaxisisamajordeterminantofthehostre-sponsetostress.Therelationshipbetweenitsactivationandpatientoutcomeisnotknown. Objective  To evaluate the prognostic value of cortisol levels and a short corticotro-pin stimulation test in patients with septic shock. DesignandSetting  ProspectiveinceptioncohortstudyconductedbetweenOctober 1991 and September 1995 in 2 teaching hospital adult intensive care units in France. Participants  A total of 189 consecutive patients who met clinical criteria for septicshock. Intervention  A short corticotropin stimulation test was performed in all patients byintravenously injecting 0.25 mg of tetracosactrin; blood samples were taken immedi-ately before the test (T0) and 30 (T30) and 60 (T60) minutes afterward. Main Outcome Measures  Twenty-eight–day mortality as a function of variablescollected at the onset of septic shock, including cortisol levels before the corticotropintest and the cortisol response to corticotropin (  max, defined as the difference be-tween T0 and the highest value between T30 and T60). Results  The28-daymortalitywas58%(95%confidenceinterval[CI],51%-65%)andmedian time to death was 17 days (95% CI, 14-27 days). In multivariate analysis, inde-pendent predictors of death ( P  .001 for all) were McCabe score greater than 0, organsystemfailurescoregreaterthan2,arteriallactatelevelgreaterthan2.8mmol/L,ratioofPa O 2 tofractionofinspiredoxygennomorethan160mmHg,cortisollevelatT0greater than34µg/dLand  maxnomorethan9µg/dL.Threegroupsofpatientprognoseswereidentified: good (cortisol level at T0  34 µg/dL and  max  9 µg/dL; 28-day mortalityrate, 26%), intermediate (cortisol level at T0 34 µg/dL and   max  9 µg/dL or cortisollevelatT0  34µg/dLand  max  9µg/dL;28-daymortalityrate,67%),andpoor(cor-tisol level at T0  34 µg/dL and  max  9 µg/dL; 28-day mortality rate, 82%). Conclusion  Our data suggest that a short corticotropin test has a good prognosticvalueandcouldbehelpfulinidentifyingpatientswithsepticshockathighriskfordeath.  JAMA. 2000;283:1038-1045  www.jama.com 1038  JAMA,  February 23, 2000—Vol 283, No. 8  ©2000 American Medical Association. All rights reserved. Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013  nonsurvivors compared with survi-vors. 19-21 For this reason, in severe sep-sis, the evaluation of the appropriate-ness of the activation of the HPA axisrequires dynamic testing. In this re-spect, the most commonly used test istheshortcorticotropinstimulationtest,normal adrenal function being definedby a plasma cortisol level (before or at30 or 60 minutes after the injec-tionofcorticotropin)above20µg/dL. 22 However,basalplasmacortisollevelsarecommonlygreaterthan20µg/dLinse-vere sepsis and the use of the absoluteincrease in plasma cortisol levels aftertheintravenousinjectionofcorticotro-pin may be more useful to evaluate ad-renalfunction. 12,13 Indeed,occultadre-nal insufficiency (ie, an absoluteincrementofcortisolconcentrations  9µg/dL)aftercorticotropinmaybeasso-ciated with impaired pressor respon-sivenesstonorepinephrine 23 andahighmortalityrate. 24,25 Suchresultsmustbeconfirmedsinceotherinvestigatorshavenotfoundanyrelationshipbetweencor-tisolresponsetocorticotropinandsur-vival from sepsis. 26 Inthecontextofrenewedinterestincorticosteroids as therapy for septicshock, 14,15,21,23-25,27-30 weundertookapro-spective study to determine the inci-denceofoccultadrenalinsufficiencyinseptic shock patients and to assess thefactors associated with mortality, tak-ingspecialinterestincortisollevelsandcortisol response to corticotropin. METHODS Study Population All consecutive patients hospitalized inthe ICU of 2 teaching hospitals (Ray-mond Poincare´ hospital, Garches,France, and Antoine Be´cle`re hospital,Clamart,France)betweenOctober1991and September 1995 were prospec-tively enrolled in the study if they metthefollowingcriteriaforsepticshock 31 :(1) for less than 7 days, a systemic in-flammatoryresponseasdefinedby2ormore of the following: temperaturehigherthan38.5°Corlowerthan35.0°C,heartrateofmorethan90/min,respira-tory rate of more than 20/min or Pa CO 2 of less than 32 mm Hg or need for me-chanical ventilation, white blood cellcount of more than 12.0  10 9  /L or lessthan4.0  10 9  /Lorcontainingmorethan10% immature forms; (2) evidence foranidusofinfection;and(3)forlessthan24hours,systolicbloodpressureoflessthan90mmHg(foratleast1hour)de-spiteadequatefluidreplacementandper-fusion of 5 µg/kg/min or more of dopa-mine or dobutamine, and the presenceof at least 2 signs of perfusion abnor-malities (ie, lactic acidosis, oliguria, oranabruptalterationinthementalstatus).Patients were not eligible if they hadknown previous conditions that mayhavedisruptedtheHPAaxis. 5,6,13,22 Theprotocol was approved by our institu-tional review board and informed con-sentwasobtainedfromthepatient’snextof kin. Data Collection ClinicalEvaluation.  Attheonsetofsep-tic shock, the following variables wererecorded:(1)generalcharacteristicsin-cludingageandsex,dateofICUadmis-sion,medicalorsurgicaladmission,es-timated prognosis of any preexistingunderlyingdiseaseaccordingtotheclas-sificationofMcCabeandJackson 32 (non-fatal,ultimatelyfatal,orrapidlyfatal);(2)severityofillnessasassessedbythenum-ber of organ system failures (OSFscore), 33 Simplified Acute PhysiologyScore II, 34 and vital signs (temperature,mean arterial pressure, heart rate, uri-nary output); and (3) interventions (atphysiciandiscretion)includingvolumeoffluidinfusionper24hours,antibiot-ics, type and titration of vasopressors,corticosteroidtherapy,needformechani-cal ventilation, insertion of a Swan-Ganz catheter, and surgical procedure. Laboratory Variables.  At the onsetofsepticshock,bloodculturesandcul-tures of specimen drawn from the siteof infection, hematologic and chemis-try data, and arterial lactate and bloodgasdeterminationsweredonesystem-atically.Ashortcorticotropinstimula-tion test was performed with 0.25 mgof tetracosactrin (Synacthe`ne, Ciba,Rueil-Malmaison,France)givenintra-venously. Blood samples were takenimmediatelybeforethetest(T0)and30(T30) and 60 (T60) minutes after-ward. After centrifugation, plasmasamples were stored at 4°C and corti-sol(normalrange,6-28µg/dL)wasmea-sured by enzyme-linked fluorescentassay(VIDASCortisol,BioMe´rieuxSA,Lyon, France). The cortisol response(  max) was defined as the differencebetweenT0andthehighestoftheT30and T60 concentrations. Follow-up.  Allpatientswereevalu-ated for 28 days from inclusion in thestudy. The evaluation of the followingvariables was performed daily in eachpatientduringtheshock:vitalsigns,he-matocrit, total and differential leuko-cytecounts,plateletcount,plasmaelec-trolytes,glucoselevels,serumcreatinineand liver function test, arterial lactateand blood gases, and interventions, aspreviously defined. Statistical Analysis Statistical analyses were conducted us-ingSASsoftwarepackage(Version6.12,SAS Institute, Cary, NC). We investi-gatedtheprognosticvaluefortheprob-ability of dying based on patient char-acteristicscollectedattheonsetofsepticshock and on values obtained with theshort corticotropin test. We performedunivariate analyses in which the datawere compared between survivors andnonsurvivors using the  t  test for con-tinuousvariablesand  2 testforcategori-cal variables (or Fisher exact test asappropriate).Toperformsurvivalanaly-ses, continuous variables were dis-cretizedaccordingtotheirmedianvaluebut categorical variables remained un-changed. Survival was estimated by theKaplan-Meiermethodandcomparedbe-tween groups with the log-rank test forall the variables. Multivariate analyseswere performed using a logistic regres-sionmodeltoestimatetheoddsratioof dying(alongwiththe95%confidencein-terval [CI]). Calibration of the logisticmodel was assessed using the Hosmer-Lemeshowgoodness-of-fittest 35 toevalu-ate the importance of the discrepancybetweenobservedandexpectedmortal-ity.Discriminationwasassessedusingthearea under the receiver operating char-acteristic(ROC)curve 36 toevaluatehow CORTISOL RESPONSE TO CORTICOTROPIN IN SEPTIC SHOCK ©2000 American Medical Association. All rights reserved.  JAMA,  February 23, 2000—Vol 283, No. 8  1039 Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013  well the model distinguished patientswho lived from those who died. A Coxproportional hazards regression modelwas used to assess variables related todeath. This model assumes that the ef-fect of a variable on the instantaneousdeathrateisconstantovertime.Thisas-sumption was checked for all predictorvariables entered in the model. 37 Step-wise and backward selection proce-dureswereusedforbothregressionmod-els(logisticandCox)toiterativelyselectthe variables that were significantly re-lated to death, as assessed by the likeli-hoodratiotest.Foralltests, P  .05wasconsidered statistically significant. RESULTS Patient Characteristics Amongthe189patientsadmitteddur-ingthestudyperiod,96(51%)werere-cruited in the Garches center and 93(49%)intheClamartcenter.Ofthe189patients, 109 (58%; 95% CI, 51%-65%)diedwithinthe28-dayperiodfol-lowing the onset of septic shock, 3 pa-tientsdiedafter28days(theydiedafter31, 62, and 66 days, respectively). T ABLE  1  shows patient characteristicsattheonsetofsepticshockandresultsof the univariate analysis between thesurvivorandnonsurvivorgroups.TheMcCabe and OSF scores and the Sim-plified Acute Physiology Score II weresignificantly associated with mortal-ity. Among clinical and biological fac-tors, mean arterial pressure, plateletcount, arterial lactate and pH, the ra-tio of the Pa O 2  to the fraction of in-spired oxygen (F IO 2 ) were signifi-cantlydifferentbetweensurvivorsandnonsurvivors. Compared with survi-vors, nonsurvivors had significantlyhigherbasalplasmacortisollevels(T0)andlowercortisolresponsetocortico-tropin (  max). The mean maximum[SD] doses of dobutamine during thefirst6hoursfollowingtheonsetofsep-tic shock were significantly lower insurvivors compared with nonsurvi-vors (8.6 [4.5] vs 11.6 [6.4] µg/kg perminute; P  = .005).Treatmentwithhy-drocortisoneduringthefollow-upwasalso less frequent in survivors com-paredwithnonsurvivors(12%vs29%; P  = .006).Thenumberofpatientswhohad documented infection, sites of infection, and strains diagnosed at theonset of septic shock are shown in T ABLE  2 . Sites of infection were simi-lar among survivors and nonsurvivorswhereas gram-positive microorgan-isms were more common among non-survivors and gram-negative microor-ganisms were more common amongsurvivors ( P  = .008).All variables found to be signifi-cantlydifferentbetweenthesurvivorandnonsurvivor groups, according to theunivariate analysis performed on pa-tientcharacteristicsattheonsetofsep-tic shock (apart from physician’s inter-ventions, namely the administration of catecholaminesorhydrocortisone),wereentered into the logistic regressionmodel. Among those variables, the fol-lowing5remainedindependentlyasso-ciated with death: McCabe and OSFscores, arterial lactate, Pa O 2 :F IO 2 , and  max ( T ABLE  3 ). Increases in the Mc-Cabe and OSF scores were associatedwiththehighestoddsofdyingwith2.95(95%CI,1.56-5.59)and2.41(95%CI,1.51-3.84), respectively. The Hosmer-Lemeshowgoodness-of-fittestshowedthatthemodelwaswellcalibratedwith P  = .44 (a large  P  value indicating thatthere is not a large discrepancy be-tweenobservedandexpectedmortality).The area under the ROC curve was0.863,showingthatthemodeldiscrimi-nated well between patients who livedand those who died. Survival The median time to death was 17 days(95% CI, 14-27 days) for all patients. Table 1.  Clinical and Biological Data Collected at the Onset of Septic Shock *  VariableTotal(N = 189)Survivors(n = 77)Nonsurvivors(n = 112) P  Value  Age, y 63 (18) 62 (18) 64 (18) .45Sex†Male 117 (62) 46 (60) 71 (63).61Female 72 (38) 31 (40) 41 (37)McCabe†0 101 (54) 50 (65) 51 (46)1 65 (34) 22 (29) 43 (38) .022 23 (12) 5 (6) 18 (16)Organ system failure†0 6 (3) 6 (8) 0 (0)1 50 (26) 33 (43) 17 (15)2 64 (34) 27 (35) 37 (33)  .0013 54 (29) 9 (12) 45 (40)4 13 (7) 2 (2) 11 (10)5 2 (1) 0 (0) 2 (2)Simplified Acute Physiology Score II 58 (21) 49 (19) 64 (20)   .001 Temperature, °C 38.0 (1.9) 38.1 (1.9) 37.9 (1.9) .51Mean arterial pressure, mm Hg 59 (20) 65 (21) 55 (18)   .001Heart rate, beats/min 120 (28) 118 (27) 122 (29) .33Leukocytes,   10 9  /L 15.9 (9.8) 15.9 (8.8) 15.9 (10.5) .98Platelets,   10 9  /L 174 (129) 204 (140) 153 (117) .007Lactate, mmol/L 4.3 (4.1) 2.6 (2.0) 5.5 (4.7)   .001 Arterial pH 7.32 (0.13) 7.36 (0.10) 7.29 (0.13)   .001Pa O 2 :F IO 2 , mm Hg‡ 184 (112) 222 (114) 157 (102)   .001Cortisol, µg/dL§Level before test 34 (28) 28 (18) 39 (33) .002Maximum variation after test   11 (17) 14 (12) 8 (19) .01 *  Values are expressed as mean (SD) unless otherwise indicated.  P   values are for comparison of survivors vs nonsur-vivors.†Values are expressed as number (percentage).‡F IO 2  indicates fraction of inspired oxygen.§The short corticotropin test was used. To convert values for cortisol to nanomoles per liter, multiply by 27.6.  Indicates variation between pretest plasma highest level and the level 30 and 60 minutes after test. CORTISOL RESPONSE TO CORTICOTROPIN IN SEPTIC SHOCK 1040  JAMA,  February 23, 2000—Vol 283, No. 8  ©2000 American Medical Association. All rights reserved. Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013  Univariate analysis was performed tocomparesurvivaltimedistributionsofallvariablescollectedattheonsetofsepticshockusingthelog-ranktest.VariablesassociatedwithdeathwereMcCabescoreof more than 0 ( P  = .005), OSF scoregreater than 2 ( P  .001), SimplifiedAcute Physiology Score II greater than55 ( P  .001), mean arterial pressure of 60mmHgorless( P  .001),arteriallac-tate level greater than 2.8 mmol/L( P  .001), arterial pH of 7.33 or less( P  .001), Pa O 2 :F IO 2  of 160 mm Hg orless( P  = .002),T0greaterthan26µg/dL( P  = .003),and  maxof8µg/dLorless( P  .001). Among variables related tophysicianinterventions,higherdosesfordopamine( P  = .04)andtreatmentwithhydrocortisone ( P  = .04) were signifi-cantly associated with death.Variablesidentifiedbytheunivariateanalysis with the log-rank test (apartfromphysicianinterventions)wereen-tered in the Cox proportional hazardsregression model to identify the vari-ables that have an important effect onmortality. As shown in Table 3, 6 vari-ables were selected as being indepen-dently associated with mortality: Mc-Cabe score of more than 0, OSF scoregreater than 2, arterial lactate levelgreater than 2.8 mmol/L, Pa O 2 :F IO 2  of 160 mm Hg or less, T0 greater than 26µg/dL, and  max of 8 µg/dL or less. Cortisol Levels and CortisolResponse to Corticotropin  We further investigated the prognosticvalueoftheshortcorticotropintestus-ing univariate analyses (with   2 , log-rank tests, and ROC curves) and mul-tivariateanalyses(withlogisticandCoxmodels).The2variables,T0and  max,were first studied separately. The val-uesofT0and  maxwerediscretizedac-cordingtotheir25th,50th,and75thper-centiles as well as to their mean value.The reference value of 9 µg/dL 24 wasadded for  max.As shown in  T ABLE  4 , values of T0largerthan34µg/dL(mean)oreven45µg/dL (75th percentile) were signifi-cantly associated with death rates anddistribution of survival times, with thesmallest P value(  2 andlog-ranktests)for 34 µg/dL. With T0 greater than 26µg/dL (50th percentile), the differencein the proportion of deaths was almostsignificant (  2 test) whereas the differ-enceinthedistributionsofsurvivaltimeswas significant (log-rank test). All thethresholdvaluesofT0aredisplayedontheROCcurve( F IGURE 1 ).Theareaun-der the ROC curve was 0.620 and thehighestvaluereachedforsensitivityandspecificity,whichisusuallyclosetotheintersection point between the ROCcurveandthesecondbisectingline,wasthe threshold value of 26 µg/dL, which Table 2.  Patients Who Had Documented Infection, Sites of Infection, and StrainsDiagnosed at the Onset of Septic Shock *  Variable Total Survivors Nonsurvivors P  Value Patients (N = 189) (n = 77) (n = 112)Who had positive microbialdocumentation of infection153 (81) 58 (75) 95 (85) .10Who had positiveblood culture results36 (19) 12 (16) 24 (21) .32Sites of infection (N = 223) (n = 92) (n = 131)Lung 75 (34) 28 (30) 47 (36) Abdominal 84 (38) 32 (35) 52 (40)Cellulitis 21 (9) 9 (10) 12 (9) .33Genitourinary 23 (10) 14 (15) 9 (7)Others 20 (9) 9 (10) 11 (8)Microorganisms (N = 210) (n = 105) (n = 105)Gram-positive 85 (41) 33 (31) 52 (49)Gram-negative 105 (50) 62 (59) 43 (41).008Fungi 17 (8) 10 (10) 7 (7)Others 3 (1) 0 (0) 3 (3) *  Values are expressed as number (percentage). A patient could have more than 1 site of infection and microorganism. Table 3.  Multivariate Logistic and Cox Regression Analyses  VariableRegressionCoefficient (    ) SEOdds Ratio(95% Confidence Interval) P  ValueLogistic Regression * Intercept −1.69 0.66 NA .01McCabe 1.08 0.33 2.95 (1.56-5.59)   .001Organ system failure 0.88 0.24 2.41 (1.51-3.84)   .001Lactate, mmol/L 0.35 0.10 1.42 (1.16-1.73)   .001Pa O 2 :F IO 2 , mm Hg† −0.006 0.002 0.99 (0.98-0.99) .001Cortisol, µg/dL‡§Maximum variation after test −0.03 0.01 0.97 (0.95-0.99) .01 Cox Regression  McCabe   0 0.61 0.20 1.84 (1.24-2.72) .003Organ system failure   2 1.00 0.20 2.73 (1.84-4.06)   .001Lactate   2.8 mmol/L 0.60 0.20 1.83 (1.23-2.73) .003Pa O 2 :F IO 2   160 mm Hg† −0.60 0.20 0.55 (0.37-0.81) .003Cortisol, µg/dL‡Level before test   26 0.66 0.20 1.93 (1.30-2.85) .001Maximum variationafter test   8§−0.89 0.21 0.41 (0.27-0.61)   .001 * Results of stepwise and backward selection procedures. Other variables entered in the model were Simplified AcutePhysiology Score II, mean arterial pressure, platelets, arterial pH, and cortisol level before test. NA indicates not ap-plicable.†F IO 2  indicates fraction of inspired oxygen.‡To convert values for cortisol to nanomoles per liter, multiply by 27.6.§Indicates variation between pretest plasma level and the level 30 and 60 minutes after test.  Results of stepwise and backward selection procedures. All continuous variables are discretized according to theirmedianvalue.OthervariablesenteredinthemodelwereSimplifiedAcutePhysiologyScoreIIofmorethan55,meanarterial pressure of more than 60 mm Hg, and arterial pH of more than 7.33. CORTISOL RESPONSE TO CORTICOTROPIN IN SEPTIC SHOCK ©2000 American Medical Association. All rights reserved.  JAMA,  February 23, 2000—Vol 283, No. 8  1041 Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013  wasassociatedwithasensitivityof0.554and a specificity of 0.584.As shown in Table 4, all thresholdvalues chosen for   max were signifi-cantly associated with death rates anddistributionofsurvivaltimes,withthesmallest P value(  2 andlog-ranktests)for9µg/dL(referencevalue).Usingthisthreshold, the estimate of the inci-denceofoccultadrenalinsufficiencyis54% (95% CI, 47%-61%) in our septicshock patients. All the threshold val-uesof   maxaredisplayedontheROCcurve (Figure 1). The area under theROC curve was 0.686 and the highestvaluereachedforsensitivityandspeci-ficity was the reference value 9 µg/dL,which was associated with a sensitiv-ity of 0.679 and a specificity of 0.649.The highest values of the  2 and log-rankstatisticswerereachedfor34µg/dLfor T0 whereas the highest values forsensitivityandspecificitywerereachedfor 26 µg/dL. For  max, all results (  2 andlog-ranktests,andROCcurve)wereincloseagreement,leadingtothesamechoice for the threshold value, namely9µg/dL.Therefore,thefollowingcom-binations of T0 and   max were stud-ied:(1)T0of26or34µg/dLorlessand  maxgreaterthan9µg/dL;(2)T0of26or34µg/dLorlessand  maxof9µg/dLor less or a T0 greater than 26 or 34µg/dL and  max greater than 9 µg/dL;(3) T0 greater than 26 or 34 µg/dL and  max of 9 µg/dL or less. The informa-tion provided by T0 and   max to-gether, for both threshold values of T0(26and34µg/dL),wassignificantlyas-sociated with death rates and distribu-tion of survival times ( T ABLE  5 ). How-ever, the value of 34 µg/dL seems to beamoreinformativecut-offvaluethan26µg/dL. By using this threshold value,compared with 26 µg/dL, the propor-tion of survivors was a bit higher (70%vs68%)forcombination1andthepro-portionofsurvivorswasabitlower(18%vs 20%) for combination 3. Moreover,the highest values of the   2 and log-rank statistics were both reached with34µg/dL.Usingthisthreshold,thelike-lihood ratios for survival were 3.42 forT0of34µg/dLorlessand  maxgreaterthan9µg/dLand0.31forT0greaterthan34 µg/dL and  max of 9 µg/dL or less. We included, in a multivariate logis-tic regression model, T0 and   maxwhich were respectively discretized ac-cordingtotheirmeanandreferenceval-ues (34 µg/dL for T0 and 9 µg/dL for  max), the combination of T0 and  max, as well as the variables previ-ouslyidentifiedbytheunivariateanaly-sis(Table1).Asshownin T ABLE 6 ,highMcCabe and OSF scores, high arteriallactate,lowPa O 2 :F IO 2 ,T0greaterthan34µg/dL, and  max of 9 µg/dL or less re-mained independently and signifi-cantly associated with death. The Hos-mer-Lemeshow goodness-of-fit testshowed that the model was well cali-brated with  P  = .75. The area under theROCcurvewas0.884,showingthatthemodel discriminated well between pa-tients who lived and those who died. We also used Cox proportional haz-ards regression model by adding T0,  max, and their combination in thesamemanneraspreviouslydescribedto Figure 1.  Receiver Operating Characteristic(ROC) Curves for Basal Plasma CortisolLevels (T0) and Maximum Variation ofPlasma Cortisol Between T0 and 30 and 60Minutes After Corticotropin Test (  max) 10.90.80.70.60.50.40.30.20.100 0.40.30.20.1 0.5 0.6 0.7 0.8 0.9 1 1–   Specificity 1 0.60.70.80.9 0.5 0.4 0.3 0.2 0.1 0 Specificity    S  e  n  s   i   t   i  v   i   t  y Cutoff Points Basal Plasma Cortisol Level, µ g/dL ∆ max, µ g/dL 172634451611982 The ROC curves are generated by plotting the sensi-tivity against 1 minus the specificity for each value ofT0 and   max. The threshold values that are indi-cated for T0 and  max are the 25th, 50th, and 75thpercentilesandthemeanvalues.Thereferencevalueof 9 µg/dL also appears for    max. The diagonal linerepresents the second bisecting line. The areas under the ROC curves were 0.620 and 0.686 for T0 and  max, respectively. To convert values for cortisol tonanomoles per liter, multiply by 27.6. Table 4.   2 and Log-Rank Tests for Death Rates and Distribution of Survival Times for Different Values of Cortisol Levels Before Testand of Maximum Variation After Test * Threshold Values(Percentile or Other)PlasmaLevel, µg/dLTotal(N = 189)Survivors(n = 77)Nonsurvivors(n = 112)Death Rates Survival Times  2 P  Value Log-Rank   P  ValueCortisol Level Before Test 25   17 141 (75) 53 (69) 88 (79) 2.3 .13 2.3 .1350   26 94 (50) 32 (42) 62 (55) 3.5 .06 9.0 .003Mean   34 63 (33) 16 (21) 47 (42) 9.2 .002 20.0   .00175   45 48 (25) 12 (16) 36 (32) 6.6 .01 15.9   .001 Maximum Variation After Test 25   2 140 (74) 65 (84) 75 (67) 7.2 .007 12.0   .00150   8 94 (50) 51 (66) 43 (38) 14.1 .001 16.3   .001Reference value   9 86 (46) 50 (65) 36 (32) 19.8 .001 21.8   .001Mean   11 78 (41) 42 (55) 36 (32) 9.4 .002 11.1   .00175   16 47 (25) 31 (40) 16 (14) 16.5 .001 14.5   .001 *  To convert values for cortisol to nanomoles per liter, multiply by 27.6. CORTISOL RESPONSE TO CORTICOTROPIN IN SEPTIC SHOCK 1042  JAMA,  February 23, 2000—Vol 283, No. 8  ©2000 American Medical Association. All rights reserved. Downloaded From: http://jama.jamanetwork.com/ on 02/25/2013
Search
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks