Printer Friendly

Concordance of Endotoxemia With Gram-Negative Bacteremia.

A Meta-analysis Using Receiver Operating Characteristic Curves

Endotoxin is a cell wall component found exclusively in gram-negative bacteria. In a range of body fluids other than blood, the detection of endotoxin with the limulus amebocyte lysate (LAL) test can be reliably used as a marker of infection with gram-negative bacteria. For example, numerous published studies of body fluids as diverse as bronchoalveolar lavage,[1] cerebrospinal fluid,[2] and urine[3] have found that endotoxin detection with the LAL test can be used as a reliable marker of gram-negative infection.[4]

By contrast, despite numerous studies[5-58] in patient groups with suspected sepsis, the value of endotoxin detection using the LAL assay as a diagnostic test remains unclear,[4] with some early studies reaching overtly negative conclusions.[10,15] In 1983, the assay was modified by the addition of a chromogenic substrate (CLAL) as a more sensitive indicator of LAL activation by endotoxin. The CLAL assay is sensitive to endotoxin in plasma samples spiked in vitro at concentrations approximately 10 times lower than the previous gelation end point (GLAL) assay.[59] Although many studies have examined the value of CLAL and GLAL in patients with suspected gram-negative bacteremia, I found no direct comparison study of the 2 types of LAL assay using the same patient population.

Endotoxemia is detected in approximately half of those patients with gram-negative bacteremia and, similarly, gram-negative bacteremia is detected in approximately half of patients with endotoxemia.[60] However, the concordance between gram-negative bacteremia and endotoxemia varies with the type of gram-negative organism causing bacteremia. The rate of endotoxemia with Enterobacteriaceae is low (eg, as low as 41% and 46% for Escherichia coli and Enterobacter species, respectively) versus the rate with non-Enterobacteriaceae (eg, 63% and 74% for Pseudomonas aeruginosa and Neisseria meningitidis, respectively).[60]

The purpose of this review is to examine the diagnostic experience with endotoxemia detection using gram-negative bacteremia as the basis for comparison. The statistical techniques of meta-analysis based on a summary receiver operating characteristic (SROC) curve are applied[61-63] as a basis for quantitative comparisons. Using this method of analysis, it is possible to pool the experience from studies of a diagnostic test among diverse populations in which test thresholds were unequal. The value of the more recently introduced CLAL assay, the prevalence of non-Enterobacteriaceae gram-negative bacteremias in the patient population, and the comparison of the results of studies that used differing criteria for patient inclusion, such as current criteria of sepsis,[64] were 3 factors identified for attention.

METHODS

Study Selection

The following inclusion criteria were used: (1) comparison of the LAL assay with blood cultures in patients with suspected gram-negative bacteremia, (2) minimum sample size of 10 patients, and (3) minimum of 2 patients with gram-negative bacteremia. An additional requirement was the availability of data extractable into a 2 x 2 contingency table format. In some cases, these data became available after seeking clarifications from authors of potentially eligible studies.

Data Sources

A comprehensive search of the literature from 1966 to June 1997 was performed using the search strategy detailed previously.[60] This search was in addition to publications obtained from a library of several hundred publications related to clinical aspects of endotoxemia that I have accumulated by repeatedly searching the literature during a 10-year period. More than 400 of these publications were cited in an earlier publication.[4] A total of 261 articles were found, of which 31 articles met the inclusion criteria. An additional 15 articles were obtained by scanning the reference lists in relevant articles. Authors of 14 studies published since 1990 were approached to provide clarifications of data in cases of potentially eligible studies to enable their inclusion, and responses were obtained from 9 authors (10 studies).

Publications that were excluded and the numbers and reasons for exclusion are as follows: publications without specific data (eg, description of methodology, letter or editorial), 75; publications pertaining to nonbacteremic infections, 48; data inadequate or not available (including 11 studies published since 1990 for which responses to requests for clarifications were not received from the authors), 45; duplicate publications, 2; fewer than 10 patients studied, 6; fewer than 2 patients with bacteremia, 18; and other (including publication unavailable), 37 publications. A partial catalog of excluded publications that have been cited previously is given in the Appendix.

Studies included in this analysis were similar to those included in the previous analysis[60] with the following variations: 19 new studies were included, and 8 studies from the previous analysis that had fewer than 10 patients studied (3 studies) or fewer than 2 patients with gram-negative bacteria (5 studies) were excluded. Because the meta-analytic method employed here is not weighted for study size, the inclusion criteria were slightly modified from those used previously[60] to allow the inclusion of 5 studies[15,32,33,36,40] that reported results on a per episode rather than per patient basis and that had been excluded from the previous analysis.

Data Extraction and Subgroups

Three subgroup analyses were undertaken as predetermined at the time the analysis was designed. Studies were stratified into those that used GLAL versus those that used CLAL. Where data were available for types of gram-negative bacteremias, studies were next stratified into 3 levels, depending on the proportion of isolates that were non-Enterobacteriaceae. Gram-negative isolates were classified as Enterobacteriaceae based on their listing in a standard microbiological text[65] and as non-Enterobacteriaceae if they were not in this list.

Among the CLAL studies, 10 were limited to patients meeting the criteria of sepsis syndrome (SS),[64] and this subgroup was compared with the other CLAL studies (non-SS studies). Because the criteria for SS were not defined until after 1990, only 1 of the GLAL studies applied these criteria and, hence, this study was omitted from this subgroup analysis. Finally, because the analytical method used here is not weighted for study size, each of the analyses were repeated, excluding studies with fewer than 25 patients.

Data Analysis

The method used in this meta-analysis employs a logistic transformation of sensitivity and specificity so that an SROC curve can be fitted with linear regression.[62] A correction factor of 0.5 was added to the data to allow for zero counts in the logistic transformation. The SROC curve was then determined by back transformation of the fitted linear regression line. A single number summary (Q*) was obtained for each SROC curve, according to the method described by Moses et al.62 Q* corresponds to the point on the SROC curve where sensitivity and specificity are equal. Testing for differences between SROC curves was based on Q* values and their associated standard errors.

RESULTS

Fifty-six studies derived from 54 publications[5-58] were identified, reporting the results for a total of 4134 patients with suspected gram-negative bacteremia. These studies (published between 1970 and 1997) represent only 54 publications, as 2 publications[16-56] each provided 2 discrete studies. Twenty-eight studies used the GLAL (Table 1) and 28 studies used the CLAL (Table 2).

Table 1. Endotoxemia ([E.sup.+]) in the Presence or Absence of Gram-Negative Bacteremia: Studies Using the Gelation Limulus Amebocyte Lysate Assay
 Gram-Negative
 Bacteremia

 Detected,
 [E.sup.+]/
Source, y N Total (%)

Levin et al,[5] 1970 98 10/15 (67)
Levin et al,[6] 1972 218 17/31 (55)
Butler et al,[7] 1973([dagger]) 10 2/2 (100)
Das et al,[8] 1973 54 9/11 (82)
Martinez-G et al,[9] 1973 83 1/16 (6)
Stumacher et al,[10] 1973 139 28/65 (43)
Fossard and [Kakkar,[11] 1974 25 2/2 (100)
Obele et al,[12] 1974 23 3/3 (100)
Feldman and [Pearson,[13] 1974 89 2/30 (7)
Lirillo et al,[14] 1977 10 1/2 (50)
Elin et al,[15] 1975([double dagger]) 237 8/23 (35)
Magliulo et al,[16] 1976([dagger]) 25 4/4 (100)
Butler et al,[17] 1976([dagger]) 10 3/5 (60)
van Wieringen et al,[18] 1976 35 6/6 (100)
Magliulo et al,[16] 1976([dagger]) 14 8/12 (67)
Clumeck et [at,[19] 1977 48 9/12 (75)
Buffer et al,[20] 1978([dagger]) 21 0/13 (0)
Usawattanakul et al,[21] 1979 58 34/40 (85)
Scheifele et al,[22] 1981 63 5/8 (62)
Kelsey et al,[23] 1982 30 1/2 (50)
McCartney et al,[24] 1983 31 15/15 (100)
Scheifele et al,[25] 1985 47 3/5 (60)
Cooperstock and [Riegle,[26] 1985 43 10/11 (91)
Usawattanakul et al,[27] 1985 43 9/13 (69)
Adinolfi et al,[28] 1987([dagger]) 21 7/14 (50)
Shenep et al,[29] 1988 26 9/10 (90)
McGladdery et al,[30] 1993([dagger]) 22 0/16 (0)
Ng et al,[31] 1995([sections]) 56 34/38 (89)

 Gram-Negative
 Bacteremia

 Not Detected,
Source, y E+/Total (%)

Levin et al,[5] 1970 7/83 (8)
Levin et al,[6] 1972 19/18 (10)
Butler et al,[7] 1973([dagger]) 7/8 (88)
Das et al,[8] 1973 7/43 (16)
Martinez-G et al,[9] 1973 1/67 (1)
Stumacher et al,[10] 1973 26/74 (35)
Fossard and [Kakkar,[11] 1974 22/23 (96)
Obele et al,[12] 1974 6/20 (30)
Feldman and [Pearson,[13] 1974 0/53 (0)
Lirillo et al,[14] 1977 4/8 (50)
Elin et al,[15] 1975([double dagger]) 40/214 (19)
Magliulo et al,[16] 1976([dagger]) 9/21 (43)
Butler et al,[17] 1976([dagger]) 0/5 (0)
van Wieringen et al,[18] 1976 1/29 (3)
Magliulo et al,[16] 1976([dagger]) 1/2 (50)
Clumeck et [at,[19] 1977 2/36 (6)
Buffer et al,[20] 1978([dagger]) 0/8 (0)
Usawattanakul et al,[21] 1979 2/18 (11)
Scheifele et al,[22] 1981 10/55 (18)
Kelsey et al,[23] 1982 0/28 (0)
McCartney et al,[24] 1983 16/16 (100)
Scheifele et al,[25] 1985 20/42 (48)
Cooperstock and [Riegle,[26] 1985 6/32 (19)
Usawattanakul et al,[27] 1985 7/30 (23)
Adinolfi et al,[28] 1987([dagger]) 2/7 (29)
Shenep et al,[29] 1988 3/16 (19)
McGladdery et al,[30] 1993([dagger]) 0/6 (0)
Ng et al,[31] 1995([sections]) 5/18 (28)

 Non-Enterobacteriaceae,(*)
Source, y N (%)

Levin et al,[5] 1970 3 (23)
Levin et al,[6] 1972 7 (23)
Butler et al,[7] 1973([dagger]) 0 (0)
Das et al,[8] 1973 NA
Martinez-G et al,[9] 1973 2 (14)
Stumacher et al,[10] 1973 6 (9)
Fossard and [Kakkar,[11] 1974 1 (100)
Obele et al,[12] 1974 2 (67)
Feldman and [Pearson,[13] 1974 13 (43)
Lirillo et al,[14] 1977 0 (0)
Elin et al,[15] 1975([double dagger]) NA
Magliulo et al,[16] 1976([dagger]) 0 (0)
Butler et al,[17] 1976([dagger]) 0 (0)
van Wieringen et al,[18] 1976 2 (40)
Magliulo et al,[16] 1976([dagger]) 0 (0)
Clumeck et [at,[19] 1977 2 (17)
Buffer et al,[20] 1978([dagger]) 0 (0)
Usawattanakul et al,[21] 1979 8 (25)
Scheifele et al,[22] 1981 0 (0)
Kelsey et al,[23] 1982 1 (50)
McCartney et al,[24] 1983 9 (36)
Scheifele et al,[25] 1985 0 (0)
Cooperstock and [Riegle,[26] 1985 5 (50)
Usawattanakul et al,[27] 1985 4 (31)
Adinolfi et al,[28] 1987([dagger]) 0 (0)
Shenep et al,[29] 1988 7 (70)
McGladdery et al,[30] 1993([dagger]) 0 (0)
Ng et al,[31] 1995([sections]) NA


(*) In some studies, the identification of gram-negative bacteremia isolates was incomplete and the percentage is of the total for which identification was given.

([dagger]) Studies of patients with microbiological documentation of infection as an inclusion criteria. Studies reporting data on a per episode basis.

([sections]) Studies of patients meeting the sepsis syndrome criteria.

Table 2. Endotoxemia (E+) in the Presence or Absence of Gram. Negative Bacteremia: Studies Using the Chromogenic LAL (CLAL) Assay
 Gram-Negative Bacteremia

 Detected,
Source, y N E+/Total (%)

Harris et al,[32] 1984 ([dagger]) 43 5/6 (83)
Thomas et al,[33] 1984 ([dagger]) 400 14/23 (61)
Pearson et al,[34] 1985 10 4/5 (80)
Hass et al,[35] 1986 36 3/3 (100)
McCartney et al,[36] 1987 ([dagger]) 28 1/3 (33)
Dolan et al,[37] 1987 153 7/19 (37)
Van Deventer et al,[38] 1987 ([double dagger]) 76 15/38 (39)
Van Deventer et al,[39] 1988 473 17/53 (32)
Thompson et al,[40] 1988 ([dagger]) 40 0/5 (0)
Brandtzaeg et al,[41] 1989 ([double dagger]) 42 24/35 (69)
Fugger et al,[42] 1990 24 5/5 (100)
Danner et al,[43] 1991([subsections]) 100 11/19 (58)
Dofferhoff et al,[44] 1992([subsections]) 18 4/6 (67)
Wortel et al,[45] 1992([subsections]) 82 18/32 (56)
Yoshida et al,[46] 1994 136 21/30 (70)
Van Dissel et al,[47] 1993([subsections]) 14 4/4 (100)
Bion et al,[48] 1994 52 1/2 (50)
Billard et al,[49] 1994([subsections]) 18 6/6 (100)
Guidet et al,[50] 1994([subsections]) 93 24/33 (73)
Prins et al,[51] 1995 ([double dagger]) 30 5/10 (50)
Hynninen et al,[52] 1995 123 3/27 (11)
Engervall et al,[53] 1995 26 6/9 (67)
Goldie et al,[54] 1995([subsections]) 133 9/12 (75)
Lau et al,[55] 1996 40 8/11 (73)
Massignon et al,[56] 1996([subsections]) 52 14/15 (93)
Massignon et al,[56] 1996([subsections]) 52 12/15 (80)
Engervall et al,[57] 1997 24 2/6 (33)
Ketchum et al,[58] 1997([subsections]) 356 10/49 (20)

 Gram-Negative Bacteremia

 Not Detected,
Source, y E+/Total (%)

Harris et al,[32] 1984 ([dagger]) 12/37 (32)
Thomas et al,[33] 1984 ([dagger]) 12/377 (3)
Pearson et al,[34] 1985 2/5 (40)
Hass et al,[35] 1986 8/33 (24)
McCartney et al,[36] 1987 ([dagger]) 8/25 (32)
Dolan et al,[37] 1987 28/134 (21)
Van Deventer et al,[38] 1987 ([double dagger]) 0/38 (0)
Van Deventer et al,[39] 1988 14/420 (3)
Thompson et al,[40] 1988 ([dagger]) 71/272 (26)
Brandtzaeg et al,[41] 1989 ([double dagger]) 0/7 (0)
Fugger et al,[42] 1990 17/19 (89)
Danner et al,[43] 1991([subsections]) 32/81 (40)
Dofferhoff et al,[44] 1992([subsections]) 6/12 (50)
Wortel et al,[45] 1992([subsections]) 9/50 (18)
Yoshida et al,[46] 1994 35/106 (33)
Van Dissel et al,[47] 1993([subsections]) 7/10 (70)
Bion et al,[48] 1994 31/50 (62)
Billard et al,[49] 1994([subsections]) 4/12 (33)
Guidet et al,[50] 1994([subsections]) 20/60 (33)
Prins et al,[51] 1995 ([double dagger]) 2/20 (10)
Hynninen et al,[52] 1995 0/96 (0)
Engervall et al,[53] 1995 9/17 (53)
Goldie et al,[54] 1995([subsections]) 83/121 (69)
Lau et al,[55] 1996 20/29 (69)
Massignon et al,[56] 1996([subsections]) 21/37 (57)
Massignon et al,[56] 1996([subsections]) 16/37 (43)
Engervall et al,[57] 1997 4/18 (22)
Ketchum et al,[58] 1997([subsections]) 109/307 (36)

 Non-Enterobacteriaceae,(*)
Source, y N (%)

Harris et al,[32] 1984 ([dagger]) NA
Thomas et al,[33] 1984 ([dagger]) NA
Pearson et al,[34] 1985 NA
Hass et al,[35] 1986 1 (l00)
McCartney et al,[36] 1987 ([dagger]) 1 (33)
Dolan et al,[37] 1987 NA
Van Deventer et al,[38]
 1987 ([double dagger]) 0 (0)
Van Deventer et al,[39] 1988 1 (8)
Thompson et al,[40] 1988 ([dagger]) NA
Brandtzaeg et al,[41]
 1989 ([double dagger]) 35 (100)
Fugger et al,[42] 1990 NA
Danner et al,[43] 1991([sections]) 3 (21)
Dofferhoff et al,[44]
 1992([sections]) 2 (33)
Wortel et al,[45] 1992([sections]) NA
Yoshida et al,[46] 1994 18 (60)
Van Dissel et al,[47]
 1993([sections]) 0 (0)
Bion et al,[48] 1994 2 (100)
Billard et al,[49] 1994([sections]) NA
Guidet et al,[50] 1994([sections]) 3 (10)
Prins et al,[51]
 1995 ([double dagger]) 1 (10)
Hynninen et al,[52] 1995 0 (0)
Engervall et al,[53] 1995 1 (11)
Goldie et al,[54] 1995([sections]) 4 (33)
Lau et al,[55] 1996 0 (0)
Massignon et al,[56] 1996([sections]) 5 (33)
Massignon et al,[56] 1996([sections]) 5 (33)
Engervall et al,[57] 1997 3 (50)
Ketchum et al,[58] 1997([sections]) 8 (16)


(*) In some studies, the identification of gram-negative bacteremia isolates was incomplete and the percentage cited is of the total for which identification was given.

([dagger]) Studies reporting data on a per episode basis.

([dagger]) Studies of patients with microbiological documentation of infection as an inclusion criteria.

([sections]) Studies of patients meeting the sepsis syndrome criteria.

Of the 28 CLAL studies, 10 studies[43-45,47,49,50,54,56,58] used SS criteria as the basis for patient inclusion. In the 18 other CLAL studies, and in all but 1 of the GLAL studies, patients had been included usually on the basis of a clinical suspicion of gram-negative bacteremia. As an aggregate result, of the 4134 patients from the 56 studies, gram-negative bacteremia was detected in 856 patients (21%). However, the range in the pretest probability for gram-negative bacteremia in individual studies was much broader in the case of non-SS studies (4% to 84%) than for SS studies (9% to 39%). Fourteen studies (8 GLAL and 6 CLAL) included fewer than 25 patients.

The concordance between the detection of endotoxemia and gram-negative bacteremia in these 2 groups of studies is illustrated in Figure 1. The results of SROC analysis demonstrated no difference in the performance of studies using CLAL versus those using GLAL (Figure 1). The Q* values were 0.71 (95% CI, 0.64--0.77) for the GLAL studies and 0.66 (95% CI, 0.61--0.72) for the CLAL studies (Table 3). The analysis was repeated excluding studies with fewer than 25 patients with the same result found, that is, Q* values of 0.74 (95% CL 0.68--0.81) and 0.65 (95% CI, 0.58-0.72) for GLAL studies and CLAL studies, respectively.

Ten studies (7 GLAL and 3 (CLAL) were limited to specific infection types (plague, 2 studies[7,17]; typhoid, 4 studies[16,20,26,30] salmonellosis, 1 study[16]; meningococcal infections, 1 study[41]; and urinary tract origin of infection, 2 studies[38,51]. In these studies, the inclusion criteria were supplemented by a requirement for confirmatory microbiological documentation, either with a positive culture from blood or another site, or with serology. To ascertain whether the application of this supplementary inclusion criterion might have biased the assessment of the concordance between endotoxemia and bacteremia among these studies, the analysis was repeated with these studies excluded and was also repeated with the studies with fewer than 25 patients excluded. The Q* values were unchanged (Table 3). Five studies had reported data on a per episode rather than per patient basis. These 5 studies were also excluded, and the Q* values were recalculated; again, the Q* values were unchanged.

Among the CLAL studies, the results of SROC analysis demonstrated no difference in the performance of the 10 studies in which SS criteria were used for patient inclusion versus the 18 studies that did not use these criteria. The Q* values were 0.65 (95% CI, 0.57-0.73) for non-SS CLAL studies and 0.62 (95% CI, 0.54-0.70) for CLAL studies that used SS criteria. For the studies with more than 25 patients, the Q* values were 0.64 (95% CI, 0.52-0.77) for non-SS CLAL studies and 0.58 (95% CI, 0.48-0.69) for CLAL studies that used SS criteria.

Proportion of Non-Enterobacteriaceae

Data on the proportion of non-Enterobacteriaceae among the gram-negative bacteremia isolates were available for 45 studies (25 GLAL studies and 20 CLAL studies). These 45 studies were stratified into those with 0% non-Enterobacteriaceae (14 studies), those with 1% to 35% non-Enterobacteriaceae (18 studies), and those with more than 35% non-Enterobacteriaceae (13 studies). The stratification of studies was designed to give 3 approximately equal groups. All of the SS studies were within the middle strata. Among the 13 studies with more than 35% non-Enterobacteriaceae among the gram-negative bacteremia isolates, Pseudomonas species (including Stenotrophomonas species) were the most common isolates, followed by N meningitidis and Haemophilus influenzae. Among the 14 studies with 0% non-Enterobacteriaceae, the 3 most common gram-negative isolates were E coli, Klebsiella species, and Salmonella species. The Q* values derived from the 3 respective SROC curves (Figure 2), disclosed a trend toward higher concordance for studies with a higher proportion of non-Enterobacteriaceae: 0% non-Enterobacteriaceae, 0.63 (95% CI, 0.53-0.73); 1% to 35% non-Enterobacteriaceae, 0.67 (95% CI, 0.60--0.74); and more than 35% non-Enterobacteriaceae, 0.74 (95% CI, 0.65--0.84). The analysis was repeated excluding studies with fewer than 25 patients and the same results were found, that is, 0% non-Enterobacteriaceae, 0.69 (95% CI, 0.58--0.80); 1% to 35% non-Enterobacteriaceae, 0.68 (95% CI, 0.61--0.75); and more than 35% non-Enterobacteriaceae, 0.75 (95% CI, 0.65--0.86).

COMMENT

This analysis attempts to survey the concordance between the detection of endotoxemia with the LAL test and gram-negative bacteremia in published studies using SROC meta-analytic methods. It should be noted that the LAL assay is not an assay for gram-negative bacteria. Despite this, gram-negative bacteremia has many desirable qualities as a gold standard, and it is commonly used as the standard for comparison with the LAL assay for endotoxin in plasma. It is objective and always clinically significant.[66] The subgroup of bacteremic patients is commonly examined retrospectively in the analysis of the results of trials of novel therapies for sepsis to identify possible unequal therapeutic responses. A method that was able to assist in the rapid identification of this patient group prospectively could be useful in targeting these therapies.[67] Moreover, evidence is emerging that the presence of endotoxemia may have more prognostic significance when it is detected in the presence of gram-negative bacteremia than when it is detected in the absence of gram-negative bacteremia.[68,69]

In the general experience of diagnostic tests, there is an inevitable trade-off between sensitivity and specificity. Sensitivity and specificity, in turn, vary as a function of the test threshold, which is the breakpoint between positive and negative test results.[61-62] This trade-off is apparent in an ROC plot, which is a plot of sensitivity (TPR) versus 1 -- specificity (FPR) as observed in multiple studies. A meta-analytic method that uses linear regression to combine data from independent studies can be used to provide an SROC curve.[62] This has emerged as the method of choice for the meta-analysis of studies of diagnostic tests for 3 important reasons. First, it can provide a single number descriptor, the maximal conjoint sensitivity and specificity (Q*), as the basis for subgroup comparisons.[63]

The second reason is that this method is able to accommodate differences in test performance characteristics that result from variable test thresholds associated with the different tests in the comparison. Among the studies of the LAL assay, the threshold breakpoint used to separate positive from negative test results varies considerably among the studies. The sensitivity limit for GLAL studies was usually in the range of 0.5 to 10 ng/mL and for CLAL studies, in the range of 10 to 500 pg/mL. However, the test breakpoint is determined by the investigator and need not necessarily be the same as the sensitivity limit of the LAL assay employed. Moreover, the efficiency of detection of bacteremia varied considerably among the studies. For example, even within the SS studies, the proportion of patients with gram-negative bacteremia varied between 9% and 39%.

Finally, this method is sufficiently robust to accommodate studies of a broad range of patient groups that were found previously[60] to have statistically significant heterogeneity in the reported results. This method of analysis enables an exploration of factors that may or may not affect the utility of the assay and that would not otherwise be recognized in the individual studies. For example, in the comparison of CLAL and GLAL in this analysis, the findings are not altered when small studies, studies that included documented infections, studies reporting per episode data, or all 3 of these categories of studies are excluded from the analysis. Likewise, the Q* derived from the CLAL studies was not altered when the Q* was calculated separately for SS and non-SS studies. The main finding of this meta-analysis is the degree of influence that the proportion of non-Enterobacteriaceae isolates has on the clinical utility of the assay.

This analysis of the LAL assay yielded Q* values between 0.62 and 0.75, which compare poorly to Q* values derived in published meta-analyses of other diagnostic tests in widespread clinical use. For example, in the radiological evaluation of lymph node metastases in patients with cervical cancer, the Q* values have been found to be 0.75 for lymphangiography, 0.80 for computed tomography, and 0.85 for magnetic resonance imaging methods.[70] Other published Q* values include 0.80[62] for the radiological evaluation with computed tomographic scanning of mediastinal lymph node metastases in patients with non-small cell lung cancer and 0.85 for thallium scintography for the evaluation of angiographic coronary artery disease.[61] Extremely high Q* values, 0.97 to 0.98, are obtained in polymerase chain reaction techniques for the detection of human immunodeficiency virus.[71] By contrast, a diagnostic test that were, for example, to give results no more accurate than chance would have plots in an ROC approximating the TPR = FPR line and would have an ROC curve with a Q* value approximating 0.50.

The Q* value found in this analysis for the CLAL as a predictor of gram-negative bacteremia in patients with SS is no better than that found with prediction rules based on clinical criteria. In a multi-center validation study of clinical prediction rules for bacteremia in 1342 episodes of SS, the Q* values associated with the ROC curves were approximately 0.65 for the prediction of gram-negative bacteremia[72] versus 0.62 with the CLAL in this analysis.

The main finding of this meta-analysis is that the clinical utility of the LAL test would appear not to depend on whether a CLAL or GLAL version of the assay was employed. Moreover, there was no evidence that the more recently developed and more sensitive CLAL assay gave superior concordance. This meta-analysis reconfirms the findings of the previous meta-analysis[60] using current meta-analytic methods applied to a larger panel of studies, which included more recently published studies. While the meta-analysis presented here is unweighted for study size, the main conclusions are not altered when smaller studies are excluded from the analysis.

While the LAL assay for the detection of endotoxin can be used as a reliable marker of infection with gram-negative bacteria in various body fluids other than blood,[1-3] it is unlikely the application of the assay to blood samples will ever match this utility. The levels of endotoxin in infected fluids other than blood are well above the limits of detection with either LAL assay. By contrast, not only are the levels of endotoxemia in patients with sepsis at the limits of detection by the CLAL assay,[39,44,47,50] but also the plasma levels in patients with and patients without gram-negative bacteremia are similar. For example, in one study, levels in such subgroups were 52 and 61 pg/mL, respectively.[50] Similar levels are also seen in patient groups without sepsis. For example, as many as 21.5% of apparently healthy elderly individuals who would not be expected to have gram-negative bacteremia have been found to have endotoxin levels close to the sensitivity limit of the CLAL assay (10 pg/mL).[73] Moreover, there are multiple interactions between the LAL assay and plasma that are complex and easily overlooked in the CLAL assay.[74] As a consequence, for plasma, any gain in TPR with the CLAL assay over that found with the GLAL assay will be offset by an increase in FPR. Moreover, the type of bacteremia, whether Enterobacteriaceae or non-Enterobacteriaceae bacteremia, emerges as an important determinant of the concordance between bacteremia and endotoxemia at the limits of endotoxin detectability. However, no single published study of those included in this meta-analysis would have had a sufficient number of bacteremic patients to examine this question specifically.

The proportion of gram-negative bacteremia isolates that were non-Enterobacteriaceae varied extensively (0-100%) in the studies included in this analysis. In the study most critical of the value of the LAL assay as a clinical test, this proportion was only 9% in the patient group under investigation.[10] By contrast, the proportion of non-Enterobacteriaceae was 60%[46] and 100%[41] in studies with conclusions more favorable to the clinical application of the LAL test. A patient population with a high likelihood of a non-Enterobacteriaceae gram-negative bacteremia, for example, neutropenic patients or patients in the setting of a meningococcal epidemic, would be more appropriate for further evaluation of the LAL assay. The proportion of non-Enterobacteriaceae among the gram-negative bacteremia isolates of the studies that used the sepsis criteria for selecting patients and that were included in this meta-analysis was never more than 33%. Hence, it is unlikely that the results of testing with the LAL assay could be used as the basis for the rapid identification of patients with gram-negative bacteremia in this population.

There are several possible explanations why the association between endotoxemia and bacteremia is not uniform for different types of gram-negative bacteria. The LAL assay is a measure of not only the presence of endotoxin, but also its potency, which is not uniform for endotoxins derived from different bacteria.[75] Other possible explanations discussed previously[60] include differential rates of production and clearance of endotoxin from the blood, which in turn depends on factors such as type and timing of antibiotic therapy, and possibly on the quantitative level of bacteremia. At least in the case of meningococcemia, it appears that the levels of endotoxemia may be higher in patients with higher levels of bacteremia.[41]

This analysis did not incorporate any formal attempt to stratify the publications on the basis of study quality. However, the data analysis presented in Table 3 is an attempt at focusing on those studies which have study designs that most closely resemble a `real-life' use of the assay to detect gram-negative bacteremia in patients with suspected bacteremia. The Q* values were unchanged when only the selected studies were analyzed, suggesting that clinical utility did not track with study quality.

Table 3. Q* Values Derived from Summary Receiver Operating Characteristic Curves for Various Subgroups of Studies [dagger]
 GLAL Studies
 Subgroup Q* (95% CI) No.

All studies 0.71 (0.64-0.77) 28

All studies, excluding studies
 of patients with documented
 infections 0.73 (0.67-0.80) 21

All studies, excluding studies
 with <25 patients 0.74 (0.68-0.81) 20

All studies, excluding studies
 with <25 patients and studies
 of patients with documented
 infections 0.74 (0.67-0.81) 19

All studies, excluding studies
 with <25 patients, studies of
 patients with documented infections,
 and studies reporting data on a
 per episode basis 0.75 (0.68-0.81) 18

 CLAL Studies
 Subgroup Q* (95% CI) No.

All studies 0.66 (0.61-0.72) 28

All studies, excluding studies
 of patients with documented
 infections 0.66 (0.59-0.71) 25

All studies, excluding studies
 with <25 patients 0.65 (0.58-0.72) 22

All studies, excluding studies
 with <25 patients and studies
 of patients with documented
 infections 0.64 (0.56-0.71) 19

All studies, excluding studies
 with <25 patients, studies
 patients with documented infections,
 and studies reporting data on a
 per episode basis 0.64 (0.57-0.71) 15


([dagger]) Q* indicates the point on the summary receiver operating characteristic curve where sensitivity and specificity are equal. GLAL indicates gelation limulus amebocyte lysate assay; CLAL, chromogenic limulus amebocyte lysate assay.

Examinations of other subgroups, for example, patient groups of different ages, underlying disease, source of infection, or type of antibiotic treatment, would be of interest as a guide to further studies with the LAL assay. This was not done for this study because the information on these factors was not available in many of the studies. In any case, post hoc analysis of subgroups in a meta-analysis is potentially misleading, as the likelihood of a significant finding increases by chance alone with the number of subgroups examined.

I thank R. Danner, MD, National Institutes of Health, Bethesda, Md[43]; M. Yoshida, MD, Jichi Medical School, Japan[46]; B. Guidet, MD, Hopital Saint-Antoine, Paris, France[50]; J. M. Prins, MD, PhD, Academic Medical Center, Amsterdam, The Netherlands[51]; P. Engervall, MD, PhD, Karolinska Hospital, Stockholm, Sweden[53]; K. C. H. Fearon, MD, Royal Infirmary, Edinburgh, Scotland[54]; J. Lau, MD, The Chinese University of Hong Kong, Hong Kong[55]; D. Massignon, MD, Centre Hospitalier Lyon-Sud, Lyon, France[56]; and R Ketchum, PhD, Associates of Cape Cod, Wood's Hole, Mass,[58] for the clarification of previously published data.

References

[1.] Pugin J, Auckenthaler R, Delaspre O, van Gessel E, Suter PM. Rapid diagnosis of gram negative pneumonia by assay of endotoxin in bronchoalveolar lavage fluid. Thorax. 1992;47:547-549.

[2.] Saubolle MA, Jorgensen JH. Use of the limulus amebocyte lysate test as a cost effective screen for Gram negative agents of meningitis. Diagn Microbiol Infect Dis. 1987;7:177-183.

[3.] Nachum R, Berzofsky RN. Chromogenic Limulus amoebocyte lysate assay for rapid detection of Gram negative bacteriuria. J Clin Microbiol. 1985;21:759-763.

[4.] Hurley JC. Endotoxemia: methods of detection and clinical correlates. Clin Microbiol Rev. 1995;8:268-292.

[5.] Levin J, Poore TE, Zauber NP, Oser RS. Detection of endotoxin in the blood of patients with sepsis due to gram-negative bacteria. N Engl J Med. 1970;283: 1313-1316.

[6.] Levin J, Poore TE, Young NS, et al. Gram-negative sepsis: detection of endotoxemia with the limulus test: with studies of associated changes in blood coagulation, serum lipids and complement. Ann Intern Med. 1972;76:1-7.

[7.] Butler T, Levin J, Cu DQ, Walker RI. Bubonic plague: detection of endotoxemia with the limulus test. Ann Intern Med. 1973;79:642-646.

[8.] Das J, Schwartz AA, Folkman J. Clearance of endotoxin by platelets: role in increasing the accuracy of the Limulus gelation test and in combating experimental endotoxemia. Surgery. 1973;74:235-240.

[9.] Martinez-G LA, Quintiliani R, Tilton RC. Clinical experience on the detection of endotoxemia with the limulus test. J Infect Dis. 1973;127:102-105.

[10.] Stumacher RJ, Kovnat MJ, McCabe WR. Limitations of the usefulness of the limulus assay for endotoxin. N Engl J Med. 1973;288:1261-1264.

[11.] Fossard DP, Kakkar VV. The limulus test in experimental and clinical endotoxemia. Br J Surg. 1974;61:798-804.

[12.] Oberle MW, Graham GG, Levin J. Detection of endotoxemia with the limulus test: preliminary studies in severely malnourished children. J Pediatr. 1974;85:570-573.

[13.] Feldman S, Pearson TA. The limulus test and gram-negative bacillary sepsis. Am J Dis Child. 1974;128:172-174.

[14.] Jirillo E, Fumarola D, Marcuccio L, et al. Endotossemia e malnutrizione. Minerva Pediatr. 1977;29:139-142.

[15.] Elin RJ, Robinson RA, Levine AS, Wolff SM. Lack of clinical usefulness of the limulus test in the diagnosis of endotoxemia. N Engl J Med. 1975;293:521-524.

[16.] Magliulo E, Scevola D, Fumarola D, Vaccaro R, Bertotto A, Burberi S. Clinical experience in detecting endotoxemia with the limulus test in typhoid fever and other salmonella infections. Infection. 1976;4:21-24.

[17.] Butler T, Levin J, Linh NN, Chau DM, Adickman M, Arnold K. Yersinia pestis infection in Vietnam, II: quantitative blood cultures and detection of endotoxin in the cerebrospinal fluid of patients with meningitis. J Infect Dis. 1976; 133:493-499.

[18.] van Wieringen PMV, Monnens LAH, Bakkeren JAJM. Hemolytic-uremic syndrome: absence of circulating endotoxin. Pediatrics. 1976;58:561-563.

[19.] Clumeck N, Lauwers S, Kahn A, Mommens M, Butzler J-P. Apport du "test limule" au diagnostic des endotoxinemies et des meningites a germes Gramnegatif. Nouv Presse Med. 1977;6:1451-1454.

[20.] Butler T, Bell WR, Levin J, Linh NN, Arnold K. Typhoid fever: studies of blood coagulation, bacteremia, and endotoxemia. Arch Intern Med. 1978;138: 407-410.

[21.] Usawattanakul W, Tharavanij S, Limsuwan A. Tachypleus lysate test for endotoxin in patients with gram negative bacterial infections. Southeast Asian J Trop Med Public Health. 1979;10:13-17.

[22.] Scheifele DW, Melton P, Whitchelo V. Evaluation of the limulus test for endotoxemia in neonates with suspected sepsis. J Pediatr. 1981;98:899-903.

[23.] Kelsey MC, Lipscomb AP, Mowles JM. Limulus amoebocyte lysate test: an aid to the diagnosis in the septic neonate? J Infect. 1982;4:69-72.

[24.] McCartney AC, Banks JG, Clements GB, Sleigh JD, Tehrani M, Ledingham IMcA. Endotoxemia in septic shock: clinical and post mortem correlations. Intensive Care Med. 1983;9:117-122.

[25.] Scheifele DW, Olsen EM, Pendray MR. Endotoxinemia and thrombocytopenia during neonatal necrotizing enterocolitis. Am J Clin Pathol. 1985;83:227-229.

[26.] Cooperstock M, Riegle L. Plasma limulus gelation assay in infants and children: correlation with gram negative bacterial infection and evidence for "intestinal endotoxemia." Prog Clin Biol Res. 1985;189:329-345.

[27.] Usawattanakul W, Tharavanij S, Warrell DA, et al. Factors contributing to the development of cerebral malaria, II: endotoxin. Clin Exp Immunol. 1985;61: 562-568.

[28.] Adinolfi LE, Utili R, Gaeta GB, Perna P, Ruggiero G. Presence of endotoxemia and its relationship to liver dysfunction in patients with typhoid fever. Infection. 1987;15:359-362.

[29.] Shenep JL, Flynn PM, Barrett FF, Stidham GL, Westenkirchner DE Serial quantitation of endotoxemia and bacteremia during therapy for gram-negative bacterial sepsis. J Infect Dis. 1988;157:565-568.

[30.] McGladdery S, Larasati R, Silitonga N, et al. Acute inflammatory cytokine responses in typhoid fever: abstracts of the 1993 IDSA annual meeting [abstract 284]. Clin Infect Dis. 1993;17:578.

[31.] Ng KP, Bhanumathy M, Ong GSY, et al. Endotoxin tests in patients with sepsis. J Endotoxin Res. 1995;2:387-393.

[32.] Harris RI, Stone PCW, Evans GR, Stuart J. Endotoxaemia as a cause of fever in immunosuppressed patients. J Clin Pathol. 1984;37:467-470.

[33.] Thomas LL, Sturk A, Buller HR, ten Cate JW, Spijker RE, ten Cate H. Comparative investigation of a quantitative chromogenic endotoxin assay and blood cultures. Am J Clin Pathol. 1984;82:203-206.

[34.] Pearson FC, Dubczak J, Weary M, Bruszer G, Donohue G. Detection of endotoxin in the plasma of patients with gram-negative bacterial sepsis by the Limulus amoebocyte lysate assay. J Clin Microbiol. 1985;21:865-868.

[35.] Hass A, Rossberg MI, Hodes HL, Hyatt AC, Hodes DS. Endotoxin levels in immunocompromised children with fever. J Pediatr. 1986;109:265-269.

[36.] McCartney AC, Robertson MRI, Piotrowicz BI, Lucie NP. Endotoxemia, fever and clinical status in immunosuppressed patients: a preliminary study. J Infect. 1987;15:201-206.

[37.] Dolan SA, Riegle L, Berzofsky R, Cooperstock M. Clinical evaluation of the plasma chromogenic Limulus assay. Prog Clin Biol Res. 1987;231:405-416.

[38.] van Deventer SJH, de Vries I, Statius van Eps LW, et al. Endotoxemia, bacteremia and urosepsis. Prog Clin Biol Res. 1988;272:213-244.

[39.] van Deventer SJH, Buller HR, ten Cate JW, Sturk A, Pauw W. Endotoxemia: an early predictor of septicemia in febrile patients. Lancet. 1988;1:605-608.

[40.] Thompson JN, Cohen J, Moore RH, et al. Endotoxemia in obstructive jaundice: observations on cause and clinical significance. Am J Surg. 1988;155:314-321.

[41.] Brandtzaeg P, Kierulf P, Gaustad P, et al. Plasma endotoxin as a predictor of multiple organ failure and death in systemic meningococcal disease. J Infect Dis. 1989;159:195-204.

[42.] Fugger R, Hamilton G, Rogy M, et al. Prognostic significance of endotoxin determination in patients with severe intra-abdominal infection. J Infect Dis. 1990;161:1314-1315.

[43.] Danner RL, Elin RJ, Hosseini JM, Wesley RA, Reilly JM, Parillo JE. Endotoxemia in human septic shock. Chest. 1991;99:169-175.

[44.] Dofferhoff ASM, Bom VJJ, de Vries-Hospers HG, et al. Patterns of cytokines, plasma endotoxin, plasminogen activator inhibitor, and acute phase proteins during the treatment of severe sepsis in humans. Crit Care Med. 1992;20:185-192.

[45.] Wortel CH, von der Mohlen MAM, van Deventer SJH, et al. Effectiveness of a human monoclonal anti-endotoxin antibody (HA-1A) in gram-negative sepsis: relationship to endotoxin and cytokine levels. J Infect Dis. 1992;166:1367-1374.

[46.] Yoshida M, Obayashi T, Tamura H, et al. Diagnostic and prognostic significance of plasma endotoxin determination in febrile patients with haematological malignancies. Eur J Cancer. 1994;30A:145-147.

[47.] van Dissel JT, van Furth R, Compier BA, Feuth HDM, Frolich M. Survival in selected patients with gram-negative sepsis after adjunctive therapy with HA-1A. Lancet. 1993;341:959-960.

[48.] Bion JF, Badger I, Crosby HA, et al. Selective decontamination of the digestive tract reduces gram negative pulmonary colonization but not systemic endotoxemia in patients undergoing elective liver transplantation. Crit Care Med. 1994;22:40-49.

[49.] Billard J-L, Berthier-Berrada S, Page Y, et al. Endotoxemia in human septic shock: relation to gastric intra-mucosal pH. Crit Care Med. 1994;22:A113.

[50.] Guidet B, Barakett V, Vassal T, Petit JC, Offenstadt G. Endotoxemia and bacteremia in patients with sepsis syndrome in the intensive care unit. Chest. 1994;106:1194-1201.

[51.] Prins JM, van Agtmael MA, Kuijper EJ, van Deventer SJH, Speelman P. Antibiotic induced endotoxin release in patients with gram-negative urosepsis: a double blind study comparing imipenem and ceftazidime. J Infect Dis. 1995;172: 886-891.

[52.] Hynninen M, Valtonen M, Vaara M, et al. Plasma endotoxin and cytokine levels in neutropenic and non-neutropenic bacteremic patients. Eur J Clin Microbiol Infect Dis. 1995;14:1039-1045.

[53.] Engervall P, Granstrom M, Andersson B, Bjorkholm M. Monitoring of endotoxin, interleukin-6 and C-reactive protein serum concentrations in neutropenic patients with fever. Eur J Haematol. 1995;54:226-234.

[54.] Goldie AS, Fearon KCH, Ross JA, et al. Natural cytokine antagonists and endogenous anti-endotoxin core antibodies in sepsis syndrome. JAMA. 1995;274: 172-177.

[55.] Lau JYW, Ip SM, Chung SC, et al. Endoscopic drainage aborts endotoxaemia in acute cholangitis. Br J Surg. 1996;83:181-184.

[56.] Massignon D, Lepape A, Debize G, et al. Detection of gram-negative bacteraemia in early sepsis by a quantitative chromogenic and kinetic endotoxin assay. Eur J Clin Invest. 1996;26:596-601.

[57.] Engervall P, Granstrom M, Andersson B, Kalin M, Bjorkholm M. Endotoxemia in febrile patients with hematological malignancies: relationship of type of bacteremia, clinical findings and serum cytokine pattern. Infection. 1997;25:2-7.

[58.] Ketchum PA, Parsonnet J, Stotts LS, et al. Utilization of a chromogenic Limulus amebocyte lysate blood assay in a multi-center study of sepsis. J Endotoxin Res. 1997;4:9-16.

[59.] Thomas LLM, Sturk A, Kahle LH, ten Cate JW. Quantitative endotoxin determination in blood with a chromogenic substrate. Clin Chim Acta. 1981;116: 63-68.

[60.] Hurley JC. Concordance of endotoxaemia with gram-negative bacteremia in patients with gram-negative sepsis: a meta-analysis. J Clin Microbiol. 1994;32: 2120-2127.

[61.] Irwig L, Tosteson ANA, Gatsonis C, et al. Guidelines for meta-analyses evaluating diagnostic tests. Ann Intern Med. 1994;120:667-676.

[62.] Moses LE, Shapiro D, Littenberg B. Combining independent studies of a diagnostic test into a summary ROC curve: data-analytic approaches and some additional considerations. Stat Med. 1993;12:1293-1316.

[63.] Vamvakas EC. Meta-analysis of studies of the diagnostic accuracy of laboratory tests: a review of the concepts and methods. Arch Pathol Lab Med. 1998; 122:675-686.

[64.] American College of Chest Physicians/Society of Critical Care Medicine Consensus Committee. Definitions for sepsis and organ failures and guidelines for the use of innovative therapies in sepsis. Chest. 1992;101:1658-1662.

[65.] Farmer JJ 3rd. Enterobacteriaceae: introduction and identification. In: Murray PR, Baron EJ, Pfaller MA, Tenover FC, Yolken RH, eds. Manual of Clinical Microbiology. 6th ed. Washington, DC: ASM Press; 1995:438-449.

[66.] Young LS. Sepsis syndrome. In: Mandell GL, Bennet JE, Dolin R, eds. Principles and Practice of Infectious Diseases. 4th ed. New York, NY: Churchill Livingstone; 1995:690-705.

[67.] Bates DW, Lee TH. Projected impact of monoclonal anti-endotoxin antibody therapy. Arch Intern Med. 1994;154:1241-1249.

[68.] Hurley JC. Reappraisal with meta-analysis of bacteremia, endotoxemia and mortality in gram negative sepsis. J Clin Microbiol. 1995;33:1278-1282.

[69.] Hurley JC, Levin J. The relevance of endotoxin detection in sepsis. In: Brade H, Opal S, Vogel S, Morrison D, eds. Endotoxin in Health and Disease. New York, NY: Marcel Dekker; 1999:841-854.

[70.] Scheidler J, Hricak H, Yu KK, Subak L, Segal MR. Radiological evaluation of lymph node metastases in patients with cervical cancer: a meta-analysis. JAMA. 1997;278:1096-1101.

[71.] Owens DK, Holodniy M, Garber AM, et al. Polymerase chain reaction for the diagnosis of HIV infection in adults: a meta-analysis with recommendations for clinical practice and study design. Ann Intern Med. 1996;124:803-815.

[72.] Bates DW, Sands K, Miller E, et al. Predicting bacteremia in patients with sepsis syndrome. J Infect Dis. 1997;176:1538-1551.

[73.] Goto T, Eden S, Nordenstam G, Sundh V, Svanborg-Eden C, Mattsby-Baltzer I. Endotoxin levels in sera of elderly individuals. Clin Diagn Lab Immunol. 1994;1:684-688.

[74.] Hurley JC, Tosolini FA, Louis WJ. Quantitative limulus lysate assay for endotoxin and the effects of plasma. J Clin Pathol. 1991; 44:849-854.

[75.] Elin RJ, Sandberg AL, Rosenstreich DL. Comparison of the pyrogenicity, Limulus activity, mitogenicity and complement reactivity of several bacterial endotoxins and related compounds. J Immunol. 1976;117:1238-1242.

Appendix. Catalog of 151 Excluded Studies Cited in Hurley[4](*)
Publications without specific data (n = 45)

 33, 35, 37, 45, 54, 55, 59, 61-63, 73, 76, 89, 111, 117,
 118, 130, 145, 153-157, 161, 166, 171, 183, 202, 249,
 263, 264, 270, 273, 274, 283, 294, 310, 313, 329, 331,
 359, 363, 365, 368, 385

Studies of nonbacteremic infections (n = 25)

 7, 9, 11, 29, 39, 40, 46, 51, 77, 109, 110, 119, 131, 179,
 217, 218, 240, 319, 327, 353, 364, 369, 370, 371, 408

Studies with inadequate data (n = 40)

 2, 8, 19, 26, 30, 41, 47, 65, 74, 91-93, 105, 112, 116,
 133, 134, 139, 169, 170, 182, 190, 199, 201, 211, 245,
 254, 256, 286, 304, 311, 315, 334, 337, 338, 343, 377,
 382, 417, 421

Fewer than 10 patients studied (n = 6)

 28, 36, 95, 141, 196, 239

Fewer than 2 patients with bacteremia (n = 10)

 6, 12, 49, 114, 140, 147, 177, 217, 225, 318

Other, including unavailable (n = 31)

 13, 21, 31, 38, 48, 78, 84, 85,176, 178, 205, 206, 210,
 213, 226, 228, 230, 235, 237,241, 278, 279, 293a, 297,
 298, 335, 367, 381, 389, 400,404


(*) Reference numbers pertain to literature cited list of Hurley.

Accepted for publication January 12, 1999.

From the Division of Medicine, Ballarat Base Hospital, Ballarat, Australia.

Reprints: James C. Hurley, Mbbs, PhD, Division Of Medicine, Ballarat Base Hospital, Drummond St N, Ballarat, Australia, 3350.
COPYRIGHT 2000 College of American Pathologists
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2000 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Hurley, James C.
Publication:Archives of Pathology & Laboratory Medicine
Date:Aug 1, 2000
Words:7812
Previous Article:Decreased CD10-Positive Mature Granulocytes in Bone Marrow From Patients With Myelodysplastic Syndrome.
Next Article:Prevalence and Pathogenesis of Pancreatic Acinar Tissue at the Gastroesophageal Junction in Children and Young Adults.
Topics:


Related Articles
FDA panel backs septic shock treatment.
Multidrug-resistant Pseudomonas aeruginosa bloodstream infections: analysis of trends in prevalence and epidemiology. (Letters).
Antimicrobial drug-resistant Salmonella Typhimurium (Reply to Dahl).
Neisseria meningitidis endotoxin and capsule transmission by transplantation.
THE BUG BUSTER; Brave Joe gets TEN NHS killer infections.. and LIVES.
THE BUG BUSTER; Brave Joe gets TEN hospital infections.. and LIVES.
Sphingomonas mucosissima bacteremia in patient with sickle cell disease.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters