Algorithm for Assessing Renal Dysfunction Risk in Critically Ill Trauma Patients Receiving Aminoglycosides Bradley A. Boucher, PharmD, Bridgett C. Coffey, PharmD, David A. Kuhl, PharmD, Elizabeth A. Tolley, PhD, Timothy C. Fabian, MD, Memphis,Tennessee
A recent retrospective study proposed that the following screening criteria be used in identifying critically ill trauma patients receiving aminoglycosides who are at significant risk to develop renal dysfunction: (1) post-admission shock, (2) minimum serum concentration more than 2 mg/L, and (3) diagnosis of septicemia. The major purpose of the present study was to validate these criteria and design a corresponding algorithm for clinical use. All patients admitted to a trauma intensive care unit and receiving an aminoglycoside were prospectively studied over a 7-month period. A control group not receiving aminoglycosides was also studied. All patients were evaluated for the presence of renal dysfunction (i.e., serum creatinine increase greater than or equal to 0.5 m g / d L ) . Univariate and multivariate statistical analyses were used to compare potential associated risk factors. The overall renal dysfunction incidence was 10% in the treatment patients (n = 9 3 ) versus 5% in the control patients (n = 199) (p = 0 . 1 3 ) . Sensitivity and specificity of the screening criteria were 67% and 92%, respectively. The predictive values of a positive and negative test relative to correctly labeling patients at high risk or low risk to develop renal dysfunction were 46% and 96%, respectively. Major risk factors associated with renal dysfunction in the treatment group were post-admission shock, minimum serum concentration more than 2 mg/L, and liver dysfunction. Use of three major risk factors has excellent predictive value in identifying severely traumatized patients at low risk for developing renal dysfunction while receiving aminoglycosides. The modest predictive value of a positive test results in conservative management of patients by avoidance of aminoglycosides, i.e., use of alternative antimicrobial agents.
he aminoglycosides are an important and frequently T used class of antibiotics for serious gram-negative infections. The potential for serious adverse effects, most
notably nephrotoxicity, limits their use from being even more extensive. The ability to predict subsets of patients likely to experience renal dysfunction while receiving an aminoglycoside would be an invaluable aid in screening patients to receive or not receive these antibiotics. Moore et al [1] determined that initial calculated creatinine clearance, female gender, 1-hour post-dose aminoglycoside concentration, patient age, shock, and the presence of liver disease were significantly associated with nephrotoxicity in general medical service patients receiving aminoglycosides. A follow-up report by the same investigative group in medical and surgical patients revealed duration of therapy as a significant risk factor also [2]. Shock was not identified as a risk factor in the latter study. Jaresko et al [3] recently identified several risk factors associated with renal dysfunction in critically ill trauma patients receiving aminoglycosides based on a retrospective analysis of 179 patients. Three major associated risk factors identified in this study using multivariate analysis were: (1) post-admission shock (termed "shock during therapy" in the original study), (2) minimum serum concentration exceeding 2 mg/L, and (3) diagnosis of septicemia [3]. Patients with the presence of any of these factors independent of the other 2 factors were over 7, 10, and 3 times more likely, respectively, to develop renal dysfunction than patients without the risk factor being present [3]. The investigators suggest that use of the three major risk factors could be used as screening criteria for initiating and/or continuing aminoglycoside therapy in critically ill trauma patients relative to development of renal dysfunction. The merits of any predictive clinical test or set of criteria can only be assessed by validation in an independent patient population. No study to date has provided such a validation for the renal dysfunction screening criteria proposed by Jaresko et al [3]. Furthermore, since that study did not evaluate a control population, the contribution of the aminoglycoside to the renal dysfunction in severely traumatized patients could not be ascertained. Fromthe DepartmentsofClinicalPharmacy,Surgery,and Biostatistics A prospective study was therefore conducted with the and Epidemiology,Universityof Tennessee,Memphis, the Regional following major objectives: evaluate the utility of the Medical Center at Memphis, Memphis,Tennessee. screening criteria for renal dysfunction in critically ill Requests for reprints shouldbe addressedto BradleyA. Boucher, trauma patients receiving aminoglycosides, and design a PharmD, Departmentof Clinical Pharmacy,26 South Dunlap, Room clinically oriented algorithm incorporating the criteria. 202, Memphis,Tennessee38163. Manuscript submitted September5, 1989,and acceptedin revised Additional objectives were to determine the contribution formJanuary 4, 1990. of aminoglycosides to the renal dysfunction by comparing THE AMERICAN JOURNAL OFSURGERY VOLUME160 NOVEMBER 1990 473
BOUCHER ET AL
this patient subset with a control group matched tbr admission and post-admission shock status, and to identify other potential risk factors associated with renal dysfunction in this patient population. PATIENTS AND M E T H O D S All patients admitted to the trauma intensive care unit over the 7-month period from October 1988 through April 1989 with a length of stay greater than 24 hours who received at least four consecutive doses of an aminoglycoside (treatment group) were prospectively studied. A control population of all other patients admitted to the trauma intensive care unit over the 6-month period between October 1988 through March 1989 was also prospectively studied. Using a random number table, a subset of patients were randomly chosen from the control population matched to the aminoglycoside treatment group for admission and post-admission shock status. The rationale for matching patients according to shock status was based on the recognition of hypotension as a major cause of acute renal insufficiency in critically ill patients [4]. Shock was defined as systolic blood pressure (SBP) less than 80 mm Hg with a urine output of less than 500 mL/day or a decrease in SBP by at least 50 mm Hg with a final SBP less than 100 mm Hg. Initial aminoglycoside dosing regimens were prescribed by the surgical service with or without clinical pharmacy consultation. All patients were thereafter pharmacokinetically monitored by the clinical pharmacy investigators. Dosing adjustment recommendations were made using the Sawchuk-Zaske method [5] when predose and post-dose steady-state aminoglycoside serum concentrations were available. Pre-dose and post-dose concentrations were considered to be at steady state if scheduled dosing occurred for at least 3.5 half-lives prior to the serum sampling. Pre-dose concentrations were obtained within 30 minutes of administration. An estimate of minimum serum concentration was calculated by extrapolation of the measured pre-dose concentration to the beginning of the infusion. Post-dose concentrations were obtained between 30 and 60 minutes following a 30minute infusion of the aminoglycoside. An estimate of maximum serum concentration was calculated by backextrapolation of the post-dose concentration to the end of the infusion assuming a one-compartment model [5]. All aminoglycoside assays were performed by the hospital clinical laboratory using fluorescence polarization immunoassay within 18 hours following collection. Patients were evaluated for the presence or absence of renal dysfunction both during and 3 to 5 days after discontinuation of aminoglycoside therapy in the treatment group. In the control group, patients were evaluated for the presence or absence of renal dysfunction from admission until their transfer from the trauma intensive care unit. Renal dysfunction was defined as an increase in serum creatinine of greater than or equal to 0.5 m g / d L or an increase of at least 50% above the patient's baseline serum creatinine value measured on day 1 of therapy if greater than 1.5 mg/dL. This definition is consistent with a large number of other studies investigating aminoglycoside-induced nephrotoxicity [4,6-16]. Serum creatinine 474
was measured daily during the trauma intensive care unit stay and as clinically indicated following transfer (i.e., usually every day to every other day). Other antimicrobials were administered as clinically indicated to both the treatment and control groups. Agents included as potential risk factors were cephalosporins, vancomycin, and amphotericin B, based on a review of the literature [15,17-19]. Renal dysfunction risk factors and other definitions used in this study were consistent with those used by Jaresko et al [3] in the original study deriving major risk factors of renal dysfunction in a trauma intensive care unit patient population receiving aminoglycosides. Potential risk factors recorded for all patients upon admission to the trauma intensive care unit were patient age by decade, gender, shock on admission, and baseline serum creatinine level. The Injury Severity Score was also calculated to compare injury severity with the risk of renal dysfunction [20]. Other factors recorded for all patients for the entire length of stay in the trauma intensive care unit were presence of liver dysfunction; development of shock post-admission; vancomycin, amphotericin B, or cephalosporin use; and type of infection (pneumonia, septicemia, or other) when present. Additional risk factors recorded in the treatment group were type of aminoglycoside used (gentamicin, tobramycin, amikacin), number of aminoglycoside therapeutic courses, total dose, maximum serum concentration, and minimum serum concentration. If more than one set of aminoglycoside pre-dose and post-dose concentrations were obtained, the highest value for each was used for the purpose of analysis. Also, if a patient was changed from one type of aminoglycoside to another during hospitalization, the patient was categorized as receiving the most recent type for the purpose of analysis. Liver dysfunction was defined as the presence of any three of the following six criteria: elevation of serum glutamic oxaloacetic transaminase (SGOT) or serum glutamic pyruvic transaminase (SGPT) greater than twice normal (i.e., 80 U / L ) , alkaline phosphatase more than 100 IU/L, total bilirubin level more than 2.5 mg/ dL, serum albumin level less than 3 g/dL, or prothrombin time more than 15 seconds. Clinical diagnoses of pneumonia and septicemia were made by documenting increased temperature and white blood cell count and positive sputum and blood cultures, respectively. In equivocal cases of pneumonia based on these criteria, review of the daily physician assessment notes was used as the deciding factor. Patients with clinical diagnoses of pneumonia and septicemia were categorized as having septicemia for analysis purposes. Statistical analysis: The major risk factors associated with renal dysfunction in critically ill trauma patients receiving aminoglycosides identified by Jaresko et al [3] were: (1) post-admission shock, (2) minimum serum concentration more than 2 mg/L, and (3) diagnosis of septicemia. These three factors served as the screening criteria being validated in this study. Patients in the aminoglycoside treatment group having any of these three major risk factors were categorized as having a "positive" screening test result, i.e., at high risk to develop renal dysfunction.
THE AMERICAN JOURNAL OF SURGERY VOLUME160 NOVEMBER 1990
NEPHROTOXICITY IN TRAUMA PATIENTS RECEIVING AMINOGLYCOSIDES
Analogously, patients with none of the three risk factors were categorized as having a "negative" screening test result, i.e., at low risk to develop renal dysfunction. Patients in the treatment group with a positive test result and evidence of renal dysfunction were categorized as having true-positive results. Patients in the treatment group having none of these factors present without evidence of renal dysfunction were categorized as having true-negative results. Analogous reasoning was used in classifying false-positive and false-negative results in the treatment group. The utility of the screening criteria was assessed by calculating sensitivity, specificity, and predictive values of the criteria [21,22]. For the aminoglycoside treatment group, categorical and dichotomous risk factors were identified by comparing patients with stable renal function and patients with renal dysfunction using either chi-square or Fisher's exact tests. Continuous risk factors were identified by comparing these two patient subsets within the treatment group using the Mann-Whitney U test. A p value of less than 0.05 was considered significant for the univariate analyses. The control groups (total and matched) and the treatment group were compared for the presence or absence of renal dysfunction using chi-square analysis. Differences in patient demographics, shock status, baseline serum creatinine level, liver dysfunction status, length of intensive care unit stay, and Injury Severity Score were tested by comparing the matched control and treatment groups. Similar analyses were performed using the overall treatment group and the original study treatment group to identify potential differences in the populations, including comparison of Injury Severity Scores [20]. The purpose of the univariate analyses was to compare individual potential risk factors associated with the development of renal dysfunction in the validation study with those factors identified in the original study. In order to assess the stability of the original test, the treatment group was combined with the total control group. Stepwise multiple logistic regression analysis was used to construct statistical models for the combined data sets. Details of this analysis are described in the original study by Jaresko et al [3]. A p value of less than 0.10 was considered the significance level for retaining variables in the model. A final stepwise multiple logistic regression analysis was performed on the combined data sets from the original study and the validation study including controls. A minimum serum concentration value of 0 was given to each of the control patients. The strength of association for each risk factor was measured by estimating the odds ratio (i.e., approximate relative risk) [23]. RESULTS A total of 94 patients in the trauma intensive care unit met the study inclusion criteria for the aminoglycoside treatment group during the enrollment period. One treatment group patient was excluded from data analysis because of receiving hemodialysis prior to initiation of aminoglycoside therapy, thereby factitiously affecting the serum creatinine concentration. An additional 348 patients were admitted to the trauma intensive care unit
TABLE I
Comparison of Trauma Patients Receiving Amlnoglycosldes (Treatment PaUents) and Matched Control Patients* Patient Groups t Characteristics Gender Male Female Age by decade 10-19 20-29 30-39 40-49 50-59 >60 Liver dysfunction Baselinecreatinine (mg/dL) Injury Severity Score
Treatment (n = 92)
Control (n = 92)
64 (70) 28 (30)
75 (82) 17 (18)
16 (17) 23 (25) 26 (28) 12 (13) 6 (7) 9 (10) 11 (12) 1.2 (0.4)
15 (16) 34 (37) 20 (22) 8 (9) 3 (3) 12 (13) 23 (25) 1.3 (0.5)
28 -I- 13
19 4- 10
p Valuer 0.059 0,396
0.023 0.25 60 Shock on admission Shock post-admission Liver dysfunction Infectious disease diagnosis Pneumonia Septicemia II Other II None Baseline creatinine (mg/dL) Concurrent vancomycin Concurrent cephalosporin Aminoglycoside type Gentamicin Tobramycin Number of courses 1 ----,2 Duration of therapy (d) Total dose (mg) Maximum serum concentration (mg/L) Minimum serum concentration (mg/L) Injury Severity Score
Control Group
p Value t
Stable Renal Function (n -- 86)
Renal Dysfunction (n = 6)
1.000 w
58 (70)
6 (67)
25 (30)
3 (33)
1,000 w
70 (81)
5 (83)
18 (19)
1 (17)
15 32 19 7 3 10 23 7 9
0 2 1 1 0 2 4 4 2
0.773 14 (17) 22 (27) 22 (27) 11 (13) 6 (7) 8 (9) 23 (28) 6 (7) 17 (21) (n = 82)
2 (22) 1 (11) 4 (45) 1 (11) 0 1 (11) 4 (44) 5 (55) 6 (66)
29 (35) 4( 5 50 (60) NA 1.2 (0.3) 12 (14) 75 (90)
5 (56)
0.441w 0.001w 0*r007w
0.571 (17) (37) (22) (8) (4) (12) (27) (8) (11)
(33) (17) (17) (33) (66) (66) (33)
0,034
35 (42) 48 (58)
)
2 (22) NA 1.3 (0.6) 2 (22) 8 (89) 5 (56) 4 (44)
0.059 w 0.002 w 0.150 w 0,229
0.96 0.622w 1.000w
2 (2) 1 (1) 51 (59) 32 (38) 1.2 (0.3) 3 (3) 50 (58)
! (17) 0 4 (66) 1 (17) 1.9 (1.5) 0 4 (66)
0.495w
NA
NA
0.99 0.72 0.040
NA NA NA NA NA
NA NA NA NA NA
0.014 0.58
NA 18 (i0)
NA 24 (i)
(22)
2
p Value
0.96 1.000 w 1.000 w
0.637 70 (84) 13 (16) 7.2 (4.4) 2,044 (1596) 5.0 (2.0) (n = 60) 0.65 (0.37) 28 (14)
7 (78) 2 (22) 6.5 (2.7) 1,813 (1304) 6.8 (2.1) (n = 7) 1.7 (1.2) 30 (12)
0.58
* Patients matched by admission and post-admission shock status. t Values represent mean (4- SD) and frequency (%) for continuous and categorical data, respectively. t Statistical analysis by chi-square test or Mann Whitney U test unless otherwise noted. Fisher's exact test. II One patient with combined diagnoses of septicemia and pneumonia. 82Category includes the following isolated indications for treatment and control groups, respectively: empiric therapy following open fractures (n = 3i, n = 8), urinary tract infections (n = 0, n = 1), miscellaneous (n = 4, n = 1), postoperative prophylaxis (n = 17, n = 45). NA = Not applicable.
Sensitivity of the screening criteria in predicting renal dysfunction for the overall treatment group (n = 94) utilizing each of the risk factors sequentially was 66.7%. Specificity under the same conditions was 91.7%. The predictive value of a positive screening test and the predictive value of a negative screening test were 46% and 96%, respectively. The incidence of renal dysfunction in the group testing negative (i.e., 1-predictive value of a negative test) was 4%. Corresponding values for the matched treatment group (n = 92) were virtually identical. Table II summarizes the mean value or percentage value for each of the potential associated risk factors for the matched treatment and control groups, including the total aminoglycoside dose and serum concentrations for the treatment group. Each group was further divided into 476
THE AMERICAN JOURNAL OFSURGERY
patients with stable renal function and patients with renal dysfunction. Results of the univariate analysis for each risk factor are also included in Table II. Concurrent amphotericin B therapy and amikacin use were not compared since only one patient in the treatment group received amphotericin B and no patients received amikacin. Factors significantly associated with decreased renal function in the matched treatment group (p
A recent retrospective study proposed that the following screening criteria be used in identifying critically ill trauma patients receiving aminoglycosides who are at significant risk to develop renal dysfunction: (1) post-admission shock, (2) minimum serum concentration more than 2 mg/L, and (3) diagnosis of septicemia. The major purpose of the present study was to validate these criteria and design a corresponding algorithm for clinical use. All patients admitted to a trauma intensive care unit and receiving an aminoglycoside were prospectively studied over a 7-month period. A control group not receiving aminoglycosides was also studied. All patients were evaluated for the presence of renal dysfunction (i.e., serum creatinine increase greater than or equal to 0.5 m g / d L ) . Univariate and multivariate statistical analyses were used to compare potential associated risk factors. The overall renal dysfunction incidence was 10% in the treatment patients (n = 9 3 ) versus 5% in the control patients (n = 199) (p = 0 . 1 3 ) . Sensitivity and specificity of the screening criteria were 67% and 92%, respectively. The predictive values of a positive and negative test relative to correctly labeling patients at high risk or low risk to develop renal dysfunction were 46% and 96%, respectively. Major risk factors associated with renal dysfunction in the treatment group were post-admission shock, minimum serum concentration more than 2 mg/L, and liver dysfunction. Use of three major risk factors has excellent predictive value in identifying severely traumatized patients at low risk for developing renal dysfunction while receiving aminoglycosides. The modest predictive value of a positive test results in conservative management of patients by avoidance of aminoglycosides, i.e., use of alternative antimicrobial agents.
he aminoglycosides are an important and frequently T used class of antibiotics for serious gram-negative infections. The potential for serious adverse effects, most
notably nephrotoxicity, limits their use from being even more extensive. The ability to predict subsets of patients likely to experience renal dysfunction while receiving an aminoglycoside would be an invaluable aid in screening patients to receive or not receive these antibiotics. Moore et al [1] determined that initial calculated creatinine clearance, female gender, 1-hour post-dose aminoglycoside concentration, patient age, shock, and the presence of liver disease were significantly associated with nephrotoxicity in general medical service patients receiving aminoglycosides. A follow-up report by the same investigative group in medical and surgical patients revealed duration of therapy as a significant risk factor also [2]. Shock was not identified as a risk factor in the latter study. Jaresko et al [3] recently identified several risk factors associated with renal dysfunction in critically ill trauma patients receiving aminoglycosides based on a retrospective analysis of 179 patients. Three major associated risk factors identified in this study using multivariate analysis were: (1) post-admission shock (termed "shock during therapy" in the original study), (2) minimum serum concentration exceeding 2 mg/L, and (3) diagnosis of septicemia [3]. Patients with the presence of any of these factors independent of the other 2 factors were over 7, 10, and 3 times more likely, respectively, to develop renal dysfunction than patients without the risk factor being present [3]. The investigators suggest that use of the three major risk factors could be used as screening criteria for initiating and/or continuing aminoglycoside therapy in critically ill trauma patients relative to development of renal dysfunction. The merits of any predictive clinical test or set of criteria can only be assessed by validation in an independent patient population. No study to date has provided such a validation for the renal dysfunction screening criteria proposed by Jaresko et al [3]. Furthermore, since that study did not evaluate a control population, the contribution of the aminoglycoside to the renal dysfunction in severely traumatized patients could not be ascertained. Fromthe DepartmentsofClinicalPharmacy,Surgery,and Biostatistics A prospective study was therefore conducted with the and Epidemiology,Universityof Tennessee,Memphis, the Regional following major objectives: evaluate the utility of the Medical Center at Memphis, Memphis,Tennessee. screening criteria for renal dysfunction in critically ill Requests for reprints shouldbe addressedto BradleyA. Boucher, trauma patients receiving aminoglycosides, and design a PharmD, Departmentof Clinical Pharmacy,26 South Dunlap, Room clinically oriented algorithm incorporating the criteria. 202, Memphis,Tennessee38163. Manuscript submitted September5, 1989,and acceptedin revised Additional objectives were to determine the contribution formJanuary 4, 1990. of aminoglycosides to the renal dysfunction by comparing THE AMERICAN JOURNAL OFSURGERY VOLUME160 NOVEMBER 1990 473
BOUCHER ET AL
this patient subset with a control group matched tbr admission and post-admission shock status, and to identify other potential risk factors associated with renal dysfunction in this patient population. PATIENTS AND M E T H O D S All patients admitted to the trauma intensive care unit over the 7-month period from October 1988 through April 1989 with a length of stay greater than 24 hours who received at least four consecutive doses of an aminoglycoside (treatment group) were prospectively studied. A control population of all other patients admitted to the trauma intensive care unit over the 6-month period between October 1988 through March 1989 was also prospectively studied. Using a random number table, a subset of patients were randomly chosen from the control population matched to the aminoglycoside treatment group for admission and post-admission shock status. The rationale for matching patients according to shock status was based on the recognition of hypotension as a major cause of acute renal insufficiency in critically ill patients [4]. Shock was defined as systolic blood pressure (SBP) less than 80 mm Hg with a urine output of less than 500 mL/day or a decrease in SBP by at least 50 mm Hg with a final SBP less than 100 mm Hg. Initial aminoglycoside dosing regimens were prescribed by the surgical service with or without clinical pharmacy consultation. All patients were thereafter pharmacokinetically monitored by the clinical pharmacy investigators. Dosing adjustment recommendations were made using the Sawchuk-Zaske method [5] when predose and post-dose steady-state aminoglycoside serum concentrations were available. Pre-dose and post-dose concentrations were considered to be at steady state if scheduled dosing occurred for at least 3.5 half-lives prior to the serum sampling. Pre-dose concentrations were obtained within 30 minutes of administration. An estimate of minimum serum concentration was calculated by extrapolation of the measured pre-dose concentration to the beginning of the infusion. Post-dose concentrations were obtained between 30 and 60 minutes following a 30minute infusion of the aminoglycoside. An estimate of maximum serum concentration was calculated by backextrapolation of the post-dose concentration to the end of the infusion assuming a one-compartment model [5]. All aminoglycoside assays were performed by the hospital clinical laboratory using fluorescence polarization immunoassay within 18 hours following collection. Patients were evaluated for the presence or absence of renal dysfunction both during and 3 to 5 days after discontinuation of aminoglycoside therapy in the treatment group. In the control group, patients were evaluated for the presence or absence of renal dysfunction from admission until their transfer from the trauma intensive care unit. Renal dysfunction was defined as an increase in serum creatinine of greater than or equal to 0.5 m g / d L or an increase of at least 50% above the patient's baseline serum creatinine value measured on day 1 of therapy if greater than 1.5 mg/dL. This definition is consistent with a large number of other studies investigating aminoglycoside-induced nephrotoxicity [4,6-16]. Serum creatinine 474
was measured daily during the trauma intensive care unit stay and as clinically indicated following transfer (i.e., usually every day to every other day). Other antimicrobials were administered as clinically indicated to both the treatment and control groups. Agents included as potential risk factors were cephalosporins, vancomycin, and amphotericin B, based on a review of the literature [15,17-19]. Renal dysfunction risk factors and other definitions used in this study were consistent with those used by Jaresko et al [3] in the original study deriving major risk factors of renal dysfunction in a trauma intensive care unit patient population receiving aminoglycosides. Potential risk factors recorded for all patients upon admission to the trauma intensive care unit were patient age by decade, gender, shock on admission, and baseline serum creatinine level. The Injury Severity Score was also calculated to compare injury severity with the risk of renal dysfunction [20]. Other factors recorded for all patients for the entire length of stay in the trauma intensive care unit were presence of liver dysfunction; development of shock post-admission; vancomycin, amphotericin B, or cephalosporin use; and type of infection (pneumonia, septicemia, or other) when present. Additional risk factors recorded in the treatment group were type of aminoglycoside used (gentamicin, tobramycin, amikacin), number of aminoglycoside therapeutic courses, total dose, maximum serum concentration, and minimum serum concentration. If more than one set of aminoglycoside pre-dose and post-dose concentrations were obtained, the highest value for each was used for the purpose of analysis. Also, if a patient was changed from one type of aminoglycoside to another during hospitalization, the patient was categorized as receiving the most recent type for the purpose of analysis. Liver dysfunction was defined as the presence of any three of the following six criteria: elevation of serum glutamic oxaloacetic transaminase (SGOT) or serum glutamic pyruvic transaminase (SGPT) greater than twice normal (i.e., 80 U / L ) , alkaline phosphatase more than 100 IU/L, total bilirubin level more than 2.5 mg/ dL, serum albumin level less than 3 g/dL, or prothrombin time more than 15 seconds. Clinical diagnoses of pneumonia and septicemia were made by documenting increased temperature and white blood cell count and positive sputum and blood cultures, respectively. In equivocal cases of pneumonia based on these criteria, review of the daily physician assessment notes was used as the deciding factor. Patients with clinical diagnoses of pneumonia and septicemia were categorized as having septicemia for analysis purposes. Statistical analysis: The major risk factors associated with renal dysfunction in critically ill trauma patients receiving aminoglycosides identified by Jaresko et al [3] were: (1) post-admission shock, (2) minimum serum concentration more than 2 mg/L, and (3) diagnosis of septicemia. These three factors served as the screening criteria being validated in this study. Patients in the aminoglycoside treatment group having any of these three major risk factors were categorized as having a "positive" screening test result, i.e., at high risk to develop renal dysfunction.
THE AMERICAN JOURNAL OF SURGERY VOLUME160 NOVEMBER 1990
NEPHROTOXICITY IN TRAUMA PATIENTS RECEIVING AMINOGLYCOSIDES
Analogously, patients with none of the three risk factors were categorized as having a "negative" screening test result, i.e., at low risk to develop renal dysfunction. Patients in the treatment group with a positive test result and evidence of renal dysfunction were categorized as having true-positive results. Patients in the treatment group having none of these factors present without evidence of renal dysfunction were categorized as having true-negative results. Analogous reasoning was used in classifying false-positive and false-negative results in the treatment group. The utility of the screening criteria was assessed by calculating sensitivity, specificity, and predictive values of the criteria [21,22]. For the aminoglycoside treatment group, categorical and dichotomous risk factors were identified by comparing patients with stable renal function and patients with renal dysfunction using either chi-square or Fisher's exact tests. Continuous risk factors were identified by comparing these two patient subsets within the treatment group using the Mann-Whitney U test. A p value of less than 0.05 was considered significant for the univariate analyses. The control groups (total and matched) and the treatment group were compared for the presence or absence of renal dysfunction using chi-square analysis. Differences in patient demographics, shock status, baseline serum creatinine level, liver dysfunction status, length of intensive care unit stay, and Injury Severity Score were tested by comparing the matched control and treatment groups. Similar analyses were performed using the overall treatment group and the original study treatment group to identify potential differences in the populations, including comparison of Injury Severity Scores [20]. The purpose of the univariate analyses was to compare individual potential risk factors associated with the development of renal dysfunction in the validation study with those factors identified in the original study. In order to assess the stability of the original test, the treatment group was combined with the total control group. Stepwise multiple logistic regression analysis was used to construct statistical models for the combined data sets. Details of this analysis are described in the original study by Jaresko et al [3]. A p value of less than 0.10 was considered the significance level for retaining variables in the model. A final stepwise multiple logistic regression analysis was performed on the combined data sets from the original study and the validation study including controls. A minimum serum concentration value of 0 was given to each of the control patients. The strength of association for each risk factor was measured by estimating the odds ratio (i.e., approximate relative risk) [23]. RESULTS A total of 94 patients in the trauma intensive care unit met the study inclusion criteria for the aminoglycoside treatment group during the enrollment period. One treatment group patient was excluded from data analysis because of receiving hemodialysis prior to initiation of aminoglycoside therapy, thereby factitiously affecting the serum creatinine concentration. An additional 348 patients were admitted to the trauma intensive care unit
TABLE I
Comparison of Trauma Patients Receiving Amlnoglycosldes (Treatment PaUents) and Matched Control Patients* Patient Groups t Characteristics Gender Male Female Age by decade 10-19 20-29 30-39 40-49 50-59 >60 Liver dysfunction Baselinecreatinine (mg/dL) Injury Severity Score
Treatment (n = 92)
Control (n = 92)
64 (70) 28 (30)
75 (82) 17 (18)
16 (17) 23 (25) 26 (28) 12 (13) 6 (7) 9 (10) 11 (12) 1.2 (0.4)
15 (16) 34 (37) 20 (22) 8 (9) 3 (3) 12 (13) 23 (25) 1.3 (0.5)
28 -I- 13
19 4- 10
p Valuer 0.059 0,396
0.023 0.25 60 Shock on admission Shock post-admission Liver dysfunction Infectious disease diagnosis Pneumonia Septicemia II Other II None Baseline creatinine (mg/dL) Concurrent vancomycin Concurrent cephalosporin Aminoglycoside type Gentamicin Tobramycin Number of courses 1 ----,2 Duration of therapy (d) Total dose (mg) Maximum serum concentration (mg/L) Minimum serum concentration (mg/L) Injury Severity Score
Control Group
p Value t
Stable Renal Function (n -- 86)
Renal Dysfunction (n = 6)
1.000 w
58 (70)
6 (67)
25 (30)
3 (33)
1,000 w
70 (81)
5 (83)
18 (19)
1 (17)
15 32 19 7 3 10 23 7 9
0 2 1 1 0 2 4 4 2
0.773 14 (17) 22 (27) 22 (27) 11 (13) 6 (7) 8 (9) 23 (28) 6 (7) 17 (21) (n = 82)
2 (22) 1 (11) 4 (45) 1 (11) 0 1 (11) 4 (44) 5 (55) 6 (66)
29 (35) 4( 5 50 (60) NA 1.2 (0.3) 12 (14) 75 (90)
5 (56)
0.441w 0.001w 0*r007w
0.571 (17) (37) (22) (8) (4) (12) (27) (8) (11)
(33) (17) (17) (33) (66) (66) (33)
0,034
35 (42) 48 (58)
)
2 (22) NA 1.3 (0.6) 2 (22) 8 (89) 5 (56) 4 (44)
0.059 w 0.002 w 0.150 w 0,229
0.96 0.622w 1.000w
2 (2) 1 (1) 51 (59) 32 (38) 1.2 (0.3) 3 (3) 50 (58)
! (17) 0 4 (66) 1 (17) 1.9 (1.5) 0 4 (66)
0.495w
NA
NA
0.99 0.72 0.040
NA NA NA NA NA
NA NA NA NA NA
0.014 0.58
NA 18 (i0)
NA 24 (i)
(22)
2
p Value
0.96 1.000 w 1.000 w
0.637 70 (84) 13 (16) 7.2 (4.4) 2,044 (1596) 5.0 (2.0) (n = 60) 0.65 (0.37) 28 (14)
7 (78) 2 (22) 6.5 (2.7) 1,813 (1304) 6.8 (2.1) (n = 7) 1.7 (1.2) 30 (12)
0.58
* Patients matched by admission and post-admission shock status. t Values represent mean (4- SD) and frequency (%) for continuous and categorical data, respectively. t Statistical analysis by chi-square test or Mann Whitney U test unless otherwise noted. Fisher's exact test. II One patient with combined diagnoses of septicemia and pneumonia. 82Category includes the following isolated indications for treatment and control groups, respectively: empiric therapy following open fractures (n = 3i, n = 8), urinary tract infections (n = 0, n = 1), miscellaneous (n = 4, n = 1), postoperative prophylaxis (n = 17, n = 45). NA = Not applicable.
Sensitivity of the screening criteria in predicting renal dysfunction for the overall treatment group (n = 94) utilizing each of the risk factors sequentially was 66.7%. Specificity under the same conditions was 91.7%. The predictive value of a positive screening test and the predictive value of a negative screening test were 46% and 96%, respectively. The incidence of renal dysfunction in the group testing negative (i.e., 1-predictive value of a negative test) was 4%. Corresponding values for the matched treatment group (n = 92) were virtually identical. Table II summarizes the mean value or percentage value for each of the potential associated risk factors for the matched treatment and control groups, including the total aminoglycoside dose and serum concentrations for the treatment group. Each group was further divided into 476
THE AMERICAN JOURNAL OFSURGERY
patients with stable renal function and patients with renal dysfunction. Results of the univariate analysis for each risk factor are also included in Table II. Concurrent amphotericin B therapy and amikacin use were not compared since only one patient in the treatment group received amphotericin B and no patients received amikacin. Factors significantly associated with decreased renal function in the matched treatment group (p
