Amniotic fluid white blood cell count: A rapid and simple test to diagnose microbial invasion of the amniotic cavity and predict preterm delivery Roberto Romero, MD, Ruben Quintero, MD, Jose Nores, MD, Cecilia Avila, MD, Moshe Mazor, MD, Shuichi Hanaoka, MD, Zion Hagay, MD, Lydia Merchant, MS, and John C. Hobbins, MD New Haven, Connecticut The purpose of this study was to determine the value of amniotic fluid white blood cell count in the diagnosis of microbial invasion of the amniotic cavity. Amniotic fluid was retrieved by amniocentesis from 195 patients with preterm labor and intact membranes. Fluid was cultured for aerobic and anaerobic bacteria, as well as for mycoplasmas. The prevalence of a positive amniotic fluid culture was 12.8% (25/195). Patients with a positive amniotic fluid culture had a significantly higher median amniotic fluid white blood cell count than did patients with a negative amniotic fluid culture (median, 6 cells/mm 3; range, o to 11,000 cells/mm 3 vs median, 320 cells/mm3; range, 1 to 4480 cells/mm3; p < 0.0001). An amniotic fluid white blood cell count 2:50 cells/mm 3 had a sensitivity of 80% (20125), a specificity of 87.64% (149/170), a positive predictive value of 48.78% (20/41), and a negative predictive value of 96.75% (149/154) in the detection of a positive amniotic fluid culture for microorganisms. Although the sensitivity of an amniotic fluid white blood cell count (2:50 cells/mm 3) in the detection of microbial invasion of the amniotic cavity was greater than that of the Gram stain of amniotic fluid (80% [20125] vs 48% [12/25]; P < 0.05), the specificity was lower (87.64% [149/170] vs 98.8% [168/170]; P < 0.05). However, 88% (15/17) of ali patients with an amniotic fluid white blood cell count 2:50 cells/mm 3 and a negative amniotic fluid culture had a spontaneous preterm delivery. We conclude that the amniotic fluid white blood cell count is a sensitive, simple, and inexpensive test for the detection of microbial invasion of the amniotic cavity. An elevated amniotic fluid white blood cell count is a risk factor for preterm delivery. (AM J OSSTET GVNECOL 1991 ;165:821-30.)

Key words: Neutrophils, leukocytes, amniotic fluid, preterm labor, prematurity, chorioamnionitis, tocolysis, labor, Gram stain, Mycoplasma, Ureaplasma urealyticum

The rapid and accurate diagnosis of microbial invasion of the amniotic cavity remains an important clinical challenge. Patients with pre term labor and intact membranes with a positive amniotic fluid culture for microorganisms are more likely to have preterm deliveryl7 and are at increased risk for both maternal (reviewed in reference 8) and neonatal infection-associated morbidity as compared with patients with a negative amniotic fluid cuiture. s-IO In centers using amniocentesis in the management of pre term labor, the Gram stain is used for the rapid detection of microbial invasion of the amniotic From the Departments of Obstetrics and Gynecology and Laboratory Medicine, Yale University School of Medicine. Supported by a grant from the Walter Scott Foundation for Medical Research. Dr. Romero is supported by a Physician Scientist Award from the N ationallnstitute of Child Health and Human Development. Presented at the Eleventh Annual Meeting of the Societ.y of Perinatal Obstetricians, San Francisco, California, January 28-February 2,

1991.

Reprint requests: Roberto Romero, MD, Department of Obstetrics and Gynecology, 333 Cedar St., PO Box 3333, New Haven, CT 06510.

616131200

cavity.I-IO However, the Gram stain detects microorganisms in only 50% of cases with a positive amniotic fluid cultureY Neonatal infection-related morbidity remains high in patients with a false-negative Gram stain l l ; therefore the search for other methods to identify microbial invasion of the amniotic cavity is justified. The white blood cell count has been used for the diagnosis of infection in other body fluids such as cerebrospinal fluid,12 synovial fluid,13 and pleural fluid. 11 Previous reports examining the value of white blood cells in amniotic fluid in the diagnosis of microbial invasion of the amniotic cavity have yielded conflicting results. l.IBl9 The purpose of this study was to determine the role of the amniotic fluid white blood cell count in the diagnosis of a positive amniotic fluid culture for microorganisms.

Material and methods Patient population. One hundred ninety-five consecutive patients admitted to Yale-New Haven Hospital over a 31- month period with the diagnosis of pre term labor and singleton gestation underwent amniocentesis

822

Romero et al.

October 1'l'll Am.J Obstet Gynecol

10000

•

•

•

•

1000

·•••

-.L I

• C')

E E

oco

S

••

&

• $

•

...

•

~

:ji

•

4• • • •

•

·RIIIRII

.11L-________~.--------------------~1---------Negative

Positive

Amniotic Fluid Culture Fig. 1. Amniotic fluid white blood cell counts (WBC/mm3) in patients in preterm labor with intact membranes. Patients with a positive amniotic fluid culture had significantly higher median amniotic fluid white blood cell count than did patients with a negative amniotic fluid culture (median, 320 cells/mm"; range, 1 to 4480 cells/mm" vs median, 6 cells/mm"; range, ().II,OO() cells/mm"; p = O.OOO()OOI, Mann-Whitney U test).

for the assessment of the microbiologic status of the amniotic cavity. Preterm labor was defined as the presence of regular uterine contractions with a frequency of at least two every 10 minutes. Rupture of membranes was excluded by testing for pooling, Nitrazine paper reaction, and ferning. Patients were hydrated for 1 hour with at least 1 L of lactated Ringer's solution, and if contractions persisted, the patient was started on a regimen of intravenous tocolysis with ritodrine, terbutaline, or magnesium sulfate. The protocol for tocolysis with f3-adrenergic agents followed the guidelines proposed by Caritis.'" Magnesium sulfate was given as a loading dose of 4 gm, followed by a rate of 2 mg/hr, which was increased to 3 to 4 gm/hr if needed. Failure of tocolysis was defined as progressive cervical dilatation beyond 5 cm or delivery of the infant. Retrieval of amniotic fluid. Amniotic fluid was retrieved by transabdominal amniocentesis, which was performed under antiseptic conditions, with a 22-gauge needle monitored with ultrasound. Microbiologic culture technique. Amniotic fluid was transported to the laboratory in a capped plastic syringe. Plating occurred within 30 minutes of collection. Amniotic fluid was cultured for aerobic and anaerobic bacteria , as well as for mycoplasmas, according to methods previously described."' Gram stain examination. Gram stain examination was performed with commercial reagents (crystal violet, saffranin, and Gram's iodine, Difco Laboratories, Detroit) under standard conditions. Stained slides were

examined by trained technologists, and the presence or absence of microorganisms was noted. The results of the Gram stain examinations were communicated to the clinicians. Patients with a negative Gram stain examination underwent tocolysis. Patients with a positive Gram stain examination of amniotic fluid for bacteria were given parenteral antibiotics and were delivered. A positive Gram stain for bacteria was an indication for discontinuing tocolysis. Amniotic fluid white blood cell count. An aliquot of amniotic fluid was transported to the hematology laboratory and examined under a hemocytometer (Neubauer ruled) for the presence of white blood cells. The absolute white blood cell count was calculated by multiplying the area examined by a factor of 10 per area and expressed as number of cells per cubic millimeter. Amniotic fluid white blood cell differential count. An aliquot of amniotic fluid was placed in a cytocentrifuge chamber (Cytospin 2, Shandon Scientific, London). The specimen was centrifuged for 10 minutes at 117g, allowed to air-dry, and stained with Wright's stain. The amniotic fluid white blood cell differential was calculated after 100 cells were counted. If the specimen had < I 00 cells, then 10, 25, or 50 cells were counted, and the differential count was calculated accordingly. The results of either the amniotic fluid white blood cell count or the differential count were not made available to the clinicians. Criteria for the diagnosis of chorioamnionitis and neonatal sepsis. Microbial invasion of the amniotic cav-

Amniotic fluid white blood cell count 823

Volume 165 Number 4, Part 1

Table I. Clinical characteristics of patients in preterm labor according to amniotic fluid culture results

Maternal age (yr, mean ± SD) Gestational age (wk, mean ± SD) Cervical dilatation (cm, median and range) Uterine contractions per 10 min (median and range)

Negative amniotic fluid culture (n = 170)

Positive amniotic fluid culture (n = 25)

P Value

25 ± 4,1 30 ± 5,2 1 (0-6) 2.3 (2-5)

27 ± 4,5 27.6 ± 3.1 3

NS 0.001 0.009

(1-8)

2.5 (2-5)

NS

NS, Not significant.

ity was defined as the presence of a positive amniotic fluid culture. 8 Clinical chorioamnionitis was defined according to the criteria proposed by Gibbs. 2 ' Neonatal sepsis was diagnosed by the presence of a positive culture of blood, urine, or cerebrospinal fluid. Statistical analysis. A Mann-Whitney test was used to compare the amniotic fluid white blood cell count in patients with a positive or negative amniotic fluid culture. Receiver-operator characteristic curve analysis was performed with the True Epistat software package (Epistat Services, Richardson, Tex.). Diagnostic indices (sensitivity, specificity, positive predictive value, and negative predictive value) were calculated for the Gram stain and different cutoff levels of amniotic fluid white blood cell count in the diagnosis of infection. Similarly, diagnostic indices were calculated for different levels of amniotic fluid white blood cell count in the prediction of response to tocolysis. Comparisons between sensitivity and specificity of the Gram stain and amniotic fluid white blood cell counts were performed with a modified t test for correlated samples, as described by Galen and Gambino. 2 " Results

One hundred ninety-five patients were included in the study; 43% (84/195) had delivered at term. The prevalence of a positive amniotic fluid culture was 12.8% (25/195). Table I describes the clinical characteristics of patients with positive and negative amniotic fluid cultures. Patients with a positive amniotic fluid culture had a significantly lower gestational age and a more advanced cervical dilatation than did patients with a negative amniotic fluid culture. Fig. I displays the amniotic fluid white blood cell count according to culture results. Patients with a positive amniotic fluid culture had a significantly higher median amniotic fluid white blood cell count than did patients with a negative amniotic fluid culture (median, 6 cells/mm"; range, 0 to 11,000 cells/mm' vs median, 320 cells/mm 3 ; range, 1 to 4480 cells/mm 3 ; p < 0.0000001, Mann-Whitney U test). To describe the relationship between the sensitivity

(true-positive rate) and the false-positive rate (I - specificity) of amniotic fluid white blood cell count in the detection of a positive amniotic fluid culture, a receiver-operator characteristic curve was constructed. Fig. 2 shows this receiver-operator characteristic curve (area under the curve, 0.823; SE, 0.04; z = 8.110; P < 0.0000001). The curve lies significantly above the 45-degree line, suggesting that a higher amniotic fluid white blood cell count increases the probability of a positive amniotic fluid culture. As with any other diagnostic test, there is a trade-off between the sensitivity and the false-positive rate (l - specificity). Table II displays the median percent amniotic fluid white blood cell differential count according to culture results. The median percent of neutrophils and macrophages was significantly higher in patients with a positive amniotic fluid culture than in patients with a negative amniotic fluid culture. Patients with a positive amniotic fluid culture had a significantly higher median absolute amniotic fluid neutrophil count than did patients with a negative amniotic fluid culture (median, 175 cells/mm"; range, 0 to 3853 cells/mm' vs median, 4 cells/mm'; range, 0 to 10,120 cells/mm3 ; p = 0.0001, Mann-Whitney U test). Fig. 3 displays the comparison between the receiver-operator characteristic curve of the amniotic fluid white blood cell count and that of the amniotic fluid absolute neutrophil count in the diagnosis of a positive amniotic fluid culture (area under the curve, 0.771; SE, 0.004; z = 6.68; P < 0.05). In the diagnosis of microbial invasion of the amniotic cavity the performance of the amniotic fluid absolute neutrophil count was similar to that of the total amniotic fluid white blood cell count. No difference between the two curves was detected (p = 0.34). There was no difference in the absolute red blood cell count between patients with a positive and a negative amniotic fluid culture (median, 168 cells/mm'; range, 8 to 98,000 cells/mm" vs median, 53; range, 0 to 116,000 cells/mm', respectively; p> 0.05). There was a correlation between amniotic fluid white blood cell and red blood cell count (r = 0.45, P < 0.05; Spearman's correlation coefficient). However, this correlation

824

Romero et al.

OClObc f 1'1'1 1 AI1I.I ObSlCl (; vnecol

1.0 10

•

20

50

0.8

0 .2

5000

0.0

10000

0.0

30

z = 8.110

P = < 0.00000001

0.2

0.4

0.6

0.8

1.0

1-Specificity Fig. 2. Rece iver-o pe ra tor cha racteristic curve analys is of amniotic fluid white blood ce ll co u nts in the diagnosis of a positive amniotic fluid culture. Numbers nex t to solid duts represe n t amn iotic fluid white blood cell COllnts (cells per cubic millim eter) (a rea un der curve, 0.823; SE , 0.04 ; z = 8. 110; P < 0.000000 I).

Table II. Comparison of amn iotic fluid white blood cell di fferential co unt between p atients with positive and negati ve amniotic fl u id cu ltu res Positive cliiture

M edian ( %)

Gra nulocy tes Lymp hocytes Macro phages Eosino phils

78 7 14 0

N egative cultu re

(%)

jHniilln (%)

0-100 0-30 0-90 0-10

60 4 3.5 0

Range

I

I

R ange ( %)

P Value

0-1 00 0-1 00 0- 100 0-100

0.04 NS 0.003 NS

Amnio tic fluid white blood cell diffe re ntia l count is ex pressed as perce ntage o f t ota l white b o l od cells. NS, Not significant.

is no t significant when the analysis is restricted to patients with a positive amniotic fluid culture (r = 0.37, P = 0.07). Ta ble III displays the diagnostic indices of the Gram stain and the amniotic flui d white blood cell count with d ifferent cuto ff levels. An am niotic fluid white blood cell count o f :=:: 50 cells / mm' had a geater r sensitivity than the G ram stain of amniotic fluid in the detection of a positi ve culture (80% [20 / 25 ] vs 4 8% [1 2 / 25 ]; P < 0. 05). When the Gram stain and amniotic fl uid white blood cell count were used in combination (a positive Gram stain o r an amniotic fluid wh ite blood c ell count :=::50 cell s/ mm' considered abn ormal), th e diagnostic indices we re a s f ollows: sensitivity, 84 % (21125); specificity,

88.82 % (1511170); positive predictive value, 52.50 % (21140); n egative predictive value, 97.4 1% (1511155). O ne hundred seventy-eigh t patien ts had a negative Gram stain and an amniotic fluid whi te blood cell count of < 50 cells ! mm '. Four of these p atien ts h ad a positive amniotic fluid culture (two with U1'eaplasma urealyticum, o ne w ith mixed anaerobic flora, and o ne with Candida albicans) . T ocolysis failed in all fo ur patients a nd all had p re mature delivery. T able IV displays the clinical da ta o f the 17 patients wi th a n amniotic fluid white b o l od cell count :=::50 cell s/ m m' and a negative a mniotic fluid culture. Eightyeigh t percent (15117) of these patients had refractory p l'ete rm labor that led to preter m delivery. Seventy-six percent (10 /13) of patients s howed histologic evidence

Amniotic fluid white blood cell count

Volum e Hi;; I\umbcr 4. Pan I

825

5

~ .s;

0.6

:e (/J

500

Q)

500 1000 1000

c

(f)

0.4

0.2

Z ~ 0.701 P ~ 0.483 • AF-WBC o FGR

5000

O.~~--~------~----~------L-----~

0.0

0.2

0.4

0.6

0.8

1.0

1 -Specificity

Fig. 3. Comparison of receiver-operator characteristic curve analysis of amniotic Allid white blood cell count (AF-WBC) and absolute granulocyte count (FGR) in t.he diagnosis of a positive amniotic Huid clllwre. NUlIlben next to solid £lois represent amniotic Huid wh ite blood cell counts (cells per cubic millimeter) and those next to clear datI represent absolute granuloc yte counts (ne lltrophils per cubic millimeter). ;\[0 significant difference between these two curves was found.

of chorioamnionitis. One patient was delivered by cesarean section at 35 weeks because of suspected abruptio placentae, which was not confirmed at placental examination. Two patients carried their pregnancies to term; one of them had a total amniotic fluid white blood cell count of 133 cells / mm' and 116 neutrophils / mm' , and the other had a total amniotic fluid white blood cell count. of 130 cells/mm:l and 98 l1eutrophils/mm". Table Vdisplays the clinical course and microbiologic findin gs of patients with a positive amniotic fluid culture. Twenty-eight percent of patients had polymicrobial infections. Thirteen patients had a negative Gram stain for microorganisms, and nine of these patients had an amniotic fluid white blood cell count >50 cells/mm 3 . Clinical chorioamnionitis occurred in 2% (4/195) of patients. Only one patient had a positive amniotic fluid culture (Escherichia coli); the Gram stain of amniotic fluid showed gram-negati ve rods, and the amnioti c fluid white blood cell count was 52 cells/mm'. The placenta showed histologic evidence of chorioamnioni tis. The amniotic fluid white blood cell counts of the other patients were 23, 16 , and 2 cells/mm '. Only one placenta from these three patients was examined, and no acute inflammatory lesions were found. No neonate had a positive blood, urine, or cerebrospinal fluid culture . However, two neonates were di-

agnosed as having presumed sepsis. One was born to a mother who had a positive amniotic fluid culture (En terococcus) at an outside hospital with a colony count of > 10" colony formin g units per milliliter and h ad been treated with ampicillin for a possible urinary tract infectio n before transfer to our institution . The amniotic fluid culture from repeat amniocentesis perform ed on ad mission was negative (Gram stain negative; amniotic fluid white blood cell count, 1 cell/mm'). In view of the early gestational age (26 weeks), it was elected to continue the mother on a regimen of antibiotic therapy in an effort to prolong the pregnancy. Cervical dilatation progressed to 6 cm, and the patient underwent a cesarean section because of breech presentation 1 week later. Intraoperative uterine cultures were positive for Klebsiella pneumoniae and Streptococcus viridans. The first organism was resistant to ampicillin whereas the second demonstrated an intermediate sensitivity to this antibiotic. The placenta showed severe histologic chorioamnionitis. The neonate had multiple compl ications of prematurity and died after a massive grade IV intraventricular hemorrhage. The second neonate with presumed sepsis was born to a mother with a positive cervical culture for Streptococcus agalactiae who was unresponsive to tocolysis (am niotic fluid white blood cell count, 22 cells/mm").

826

Romero et al.

October 1991 Am J Obslet Gynccol

Table III. Comparison of diagnostic indices of Gram stain and amniotic fluid white blood cell count Amniotic fluid white blood cell count ?50 cells 1mm3

Gram stain

%

Sensitivity Specificity Positive predictive value Negative predictive value

48 98.82 85.71 92.81

I

No.

%

12/25

80 87.64 48.78 96.75

1681170 12114 1681181

I

?J 00 cells 1mm3

No.

%

20/25

66 92.94 58.62 95.18

1491170 20/41

1491154

1

?5 00 cells 1mm3

No.

%

17/25

40 97.05 66.67 91.67

1581170 17/29 158/166

I

No. 10125

1651170 10115 165/180

Table IV. Clinical information and amniotic fluid white blood cell count in patients with false-positive amniotic fluid white blood cell results (?50 cells/mm') Patient No.

Amniotic fluid white blood cell count (cells 1mm3 )

Gestational age at amniocentesis (wk)

Gestational age at delivery (wk)

Previous antibiotic therapy

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

2,400 120 1,380 11,000 475 115 6,700 133 267 130 280 750 63 51 66 50 62

29 30.5 32.5 35 33 24 33 31.5 26 32.5 36 35 33.6 27.5 28.0 26.5 33

33.5 31.5 32.5 35 33 25 33 39 26.5 38.5 36 35 34 29 28 26.5 33

No No No No No No No No Yes No No No No No No No No

Histologic chorioamnionitis

No Yes Yes No Yes

N/A

Yes Yes Yes

N/A N/A

No Yes

N/A

Yes Yes Yes

NIA, Not available.

The placenta showed evidence of histopathologic chorioamnionitis. The neonate had a clinical course consistent with sepsis but recovered with antibiotic treatment. Of the 195 patients included in the study, 156 underwent a trial of tocolysis (12 patients had a positive Gram stain and 37 were delivered for maternal or fetal indications). Fig. 4 displays a receiver-operator characteristic curve for the value of an amniotic fluid white blood cell count in the prediction of preterm labor refractory to tocolysis (area under the curve, 0.625; SE, 0.044; z = 2.810; P < 0.01). Table VI displays the diagnostic indices of different amniotic fluid white blood cell counts in the identification of the patients who failed to respond to tocolysis. All patients with an amniotic fluid white blood cell count ?500 cells/mm 3 had delivery of a preterm neonate. The amniocentesis-to-delivery interval of patients with a positive amniotic fluid culture for mycoplasmas was significantly shorter than that of patients with sterile amniotic fluid (median, 0.8 days; range, 0 to 23.68

days vs median, 28 days; range, 0 to 122; Wilcoxon rank sum test).

P < 0.001;

Comment

Our data show a strong relationship between the amniotic fluid white blood cell count and microbial invasion of the amniotic cavity. This observation can be exploited for diagnostic purposes. In our study an amniotic fluid white blood cell count of ?50 cells/mm 3 had a higher sensitivity than did the Gram stain of amniotic fluid in the detection of a positive culture (80% vs 48%, P < 0.05). The greater sensitivity of the amniotic fluid white blood cell count may be attributed to the inability of the Gram stain to detect mycoplasmas (44% [11125] of patients with a positive amniotic fluid culture had only mycoplasmas isolated from the amniotic fluid). Indeed, only two of the 11 patients in whom mycoplasmas were the only amniotic fluid isolates had a positive Gram stain. Since mycoplasmas are not seen with a Gram stain examination, these cases represent either "false-positive" Gram stains or failure

Amniotic fluid white blood cell count 827

Volume 165 Number 4, Part 1

Table V. Amniotic fluid white blood cell count, microbiologic studies, and placental histopathologic findings in patients with positive amniotic fluid culture

Patient No.

Amniotic fluid white blood cell count (cells 1mm3 )

Total granulocyte (celZs1 mmJ)

Gram stain

1 2 3 4

330 30 10 162

303 21 14 39

+

5 6

54 1000

50 600

+

7 8 9 10

52 580 502 130

33 510 397 117

+

II

430 I

396 0

+

13 14 15

173 4480 835

1I8 3853 660

16 17 18 19 20 21

810 320 2150 1100 19 56

648 230 1763 748 18 0

22 23

9 1970

0 1793

24 25

114 1100

80 660

12

+

+ + + + + + +

Colony count (cfulml)

Organism Fusobacterium sp. Ureaplasma urealyticum Ureaplasma urealyticum Bacteroides fragilis Fusobacterium sp. Ureaplasma urealyticum Bacteroides fragilis Enterococcus fecalis Mycoplasma hominis Ureaplasma urealyticum Ureaplasma urealyticum

Mixed anaerobic organisms

Mycoplasma hominis

Mixed anaerobic organisms

Ureaplasma urealyticum P eptostreptococcus s p. Fusobacterium sp. Mycoplasma hominis Gardnerella vaginalis Mycoplasma hominis Staphylococcus aureus Fusobacterium sp. Streptococcus agalac liae Capnocytophaga Clostridium sp. Candida albicans Gardn(!'rella vaginalis Peptostreptococcus sp. Ureaplasma urealyticum Ureaplasma urealyticum

> 10' 1+ > 10' 1+

17,000 >10 5

>105 > 105 >10 5 1+ 4+ > 10' > 105 4+ 4+ > 10' 1+

Gestational age at amniocentesis (wk)

Gestational age at delivery (wk)

Histologic chorioamnionitis

25.5 31 30 26

25 .5 31 30.5 26

+ + + +

32 24

32 25

+

24.5 31 24.5 26

24.5 31 24.5 26

+ + + +

27 30.9

30 31

+

24.6 33.9 27.9

24.7 34 28

+ + +

27.9 23 32.9 30.9 25.4 25.9

28 24 33 31 25 26

+ + + +

24.4 23.9 23.9 28.8 27.9

26 24 24 29 28

N /A

+ + +

+ +

NIA, Not available.

to recover fastidious microorganisms in cases of mixed infections. On the other hand, an amniotic fluid white blood cell count ~ 50 cells / mm' was present in 81.8 % (9/11) of patients in whom mycoplasmas were the only isolates. The greater sensitivity of the amniotic fluid white blood cell count in the detection of a positive culture of amniotic fluid in comparison with the Gram stain was associated with a lower specificity (87.6% vs 98.8%, respectively; p < 0.05). The resulting higher falsepositive rate may be an important shortcoming of the amniotic fluid white blood cell count. However, the apparent false-positive results deserve comment. Eightyeight percent (15/17) of patients with an amniotic fluid white blood cell count ~ 50 cells / mm' and a negative amniotic fluid culture had premature delivery, and 66% (10113) of them showed histologic evidence of chorioamnionitis. Therefore, independent of the culture result, an elevated amniotic fluid white blood cell count identified a subset of patients at risk for failure to respond to tocolysis and impending preterm delivery.

Thesepatients may have had microbial invasion of the amniotic cavity that escaped detection with the microbiologic surveillance techniques used in our study or, alternatively, neutrophil recruitment into the amniotic cavity may have been driven by a noninfectious process. Five patients had false-negative amniotic fluid white blood cell counts «50 cells/mm' ) and a positive amniotic fluid culture. Three potential explanations may be invoked for these findin gs: (1) failure of the host to respond to microbial invasion of the amniotic cavity, (2) low virulence of microorganisms involved, or (3) a time-dependent phenomenon. If microbial invasion of the amniotic cavity is of short duration (i .e ., it occurred after the onset of preterm labor), the host may not have had enough time to mount an inflammatory response. Regrettably, the relative contribution of these mechanisms cannot be addressed without data. The white blood cell differential count has diagnostic value in peripheral blood and cerebrospinal fluid. Does it have value in amniotic fluid? Patients with a positive amniotic fluid culture had a significantly higher abso-

828

Romero et al.

October l'l'll Am.J 01>S('( Gynccol

1.0 5

0.8 10

~

0.6

"> '+=,

20

'00 C

Q)

(f)

30

0.4

0.2

00

400 500

1000

0.0

5000 10000 20000

0.0

0.2

0.4

0.6

0.8

1.0

1-Specificity Fig. 4. Receiver-operator characteristic curve analysis of amniotic fluid white blood cell count in identificaton of patient who will have delivery of preterm neonate. Only patients with negative Gram stain of amniotic fluid who received tocolysis were included in this analysis. Numbers next to solid dots represent amniotic fluid white blood cell counts (cells per cubic millimeter) (area under curve. 0.625; SE. 0.044; z = 2.810; P < 0.01).

Table VI. Diagnostic indices of amniotic Huid white blood cell count in prediction of spontaneous preterm delivery Amniotic fluid white blood cell count (cells / mm')

Sensitil)ity (%)

Specificity (%)

Positive predictive value (%)

Negative predictive value (%)

2:500 2: 100 2:50

12.5 (9/72) 27.7 (20172) 31.9 (23/72)

100 (84/84) 98.8 (83/84) 98.8 (83/84)

100 (9/9) 95.2 (20/21) 95.8 (23/24)

57.1 (841147)

lute neutrophil count than did patients with a negative amniotic Huid culture. However, we could not demonstrate any difference between the receiver-operator characteristic curves of the absolute neutrophil count and the total amniotic Huid white blood cell count (Fig. 3). The most likely explanation for this observation is that in preterm labor the neutrophil is the white blood cell most commonly recruited into the amniotic cavity (see Table II). The practical consequence of this observation is that a simple amniotic Huid white blood cell count has diagnostic value comparable to that of the absolute neutrophil count, which requires a centrifugation step, staining, and reading of the smear. However, these data do not negate the potential value of the amniotic Huid white blood cell differential. Further studies are required to address the frequency and value of other white blood cell types in the diagnosis of other

61.4 (83/135) 62.8 (83/1 32)

intrauterine pathologic conditions (e.g., viral disease, immune-mediated phenomena, etc.). An elevated amniotic Huid white blood cell count was not associated with clinical chorioamnionitis. This suggests that the inHammatory response was localized to the uterine cavity and did not elicit a clinically demonstrable systemic response (e.g., fever). Indeed, preterm parturition may be considered as an efficient mechanism to limit the dissemination of intrauterine infection and to prevent systemic maternal disease. We were surprised to find a low amniotic Huid white blood cell count in three of the four patients with clinical chorioamnionitis. These cases may represent extraamniotic infections (e.g., deciduitis) or clinical diagnostic errors. The absence of proved neonatal sepsis in our study must be interpreted with caution since patients with a positive Gram stain for microorganisms were treated

Volume 165 Number 4, Part l

antepartum with antibiotics, and this may have altered neonatal culture results. Moreover, neonatal cultures for mycoplasmas, the most common microbial isolates in our study, were not performed . It is possible that a fraction of the suspected but unproved neonatal sepsis may be due to these organisms.2u 5 An important additional observation of this study is that patients with microbial invasion of the amniotic cavity as a result of mycoplasmas had a significantly shorter amniocentesis-to-delivery interval than did patients with a negative amniotic fluid culture. Moreover, 81.8% (9111) of patients with a positive amniotic fluid culture and genital mycoplasmas had an amniotic fluid white blood cell count ~50 cells/mm 3 and also had histologic evidence of chorioamnionitis (Table V). These observations indicate that microbial invasion of the amniotic cavity with mycoplasmas is associated with recruitment of white blood cells into the amniotic cavity and also with acute inflammatory lesions of the placenta. These findings, coupled with the isolation of Ureaplasma urealyticum and Mycoplasma hominis from the cerebrospinal fluid 24 and bronchial secretions 25 of preterm neonates who had meningitis and chronic lung disease, support a pathogenic role of these microorganisms. However, our results are at variance with those of Gravett et al.," who found no difference in amniocentesis-to-delivery interval between patients with sterile amniotic fluid and those with a positive amniotic fluid culture for genital mycoplasmas. Moreover, they found virtually no amniotic fluid white blood cells in patients with positive amniotic fluid cultures for genital mycoplasmas. We have no explanation for the discrepancy between the two reports. We have examined the combined use of the amniotic fluid white blood cell count and Gram stain examination of amniotic fluid in the rapid diagnosis of microbial invasion of the amniotic cavity. Simultaneous use of both tests was associated with an increased sensitivity in comparison to that of the Gram stain alone (84% vs 48 %, P < 0.05). However, this was associated with an increased false-positive rate (1.2 % to 1l.2%; p < 0.05). We believe this false-positive rate is more apparent than real because 88% of patients with an elevated amniotic fluid white blood cell count and a negative culture had a spontaneous preterm delivery. These findings call for caution in the use of the amniotic fluid culture as the gold standard in this type of study. A positive amniotic fluid culture requires a minimum inoculum size, viability of the microorganisms in the specimen, and successful recovery of the microorganisms in the microbiology laboratory. Moreover, microorganisms unidentified by current laboratory techniques also may be responsible for some intrauterine infections. Our data indicate that an elevated amniotic fluid white blood cell count identifies patients with microbial invasion of the

Amniotic fluid white blood cell count 829

amniotic cavity and a population of patients at risk for preterm delivery despite a negative amniotic fluid culture. It remains to be established if the inflammatory reaction in these patients is caused by an unidentified infectious agent or a non-infection-driven mechanism. Simultaneous use of the Gram stain and the amniotic fluid white blood cell count takes advantage of the high specificity of the Gram stain for the detection of microbial invasion of the amniotic cavity and of the high specificity of amniotic fluid white blood cell count for the prediction of preterm delivery. The patient with a negative Gram stain of amniotic fluid and a white blood cell count