High blood pressure is a common problem in obstetrics, complicating 6–10% of pregnancies.1,2 When a pregnant patient presents with hypertension, the clinician's initial task is to determine which hypertensive disorder is present in order to stratify risk for mother and fetus and to guide treatment. If preeclampsia is strongly suspected or confirmed, the next task is to determine disease severity.
Establishing the diagnosis of preeclampsia and determining its severity depend on history, physical examination, and laboratory testing. However, which tests should be ordered is uncertain largely because of a lack of data defining the clinical usefulness of the many available tests. Guidelines from both the United States National High Blood Pressure Education Program Working Group Report on High Blood Pressure in Pregnancy1 and Australasian Society for the Study of Hypertension in Pregnancy3 recommend a complete blood count, platelet count, urinalysis as well as serum creatinine, uric acid, hepatic transaminase, and albumin. The United States group further recommends a lactate dehydrogenase test, whereas the Australasians also suggest a serum bilirubin. Despite these recommendations, we believe that there is much variability in the use of laboratory tests to evaluate hypertensive pregnant patients. Specifically, we observed that many clinicians obtain prothrombin time (PT), partial thromboplastin time (PTT), fibrinogen levels, or a combination of them when evaluating patients with preeclampsia despite the lack of evidence of clinical usefulness or expert panel recommendations to do so. This practice is particularly problematic because these tests are labor intensive and relatively expensive.
The current study was undertaken to provide data that would help clinicians improve the appropriateness of diagnostic testing in the evaluation of pregnant patients suspected of, or known to have, preeclampsia. Specific objectives were to obtain an estimate of the extent of laboratory testing in the evaluation of hypertensive patients and to determine whether abnormalities in commonly ordered coagulation tests could be predicted from the results of other commonly obtained, less expensive, tests so that coagulation testing could be omitted in many patients.
Materials and Methods
This study was done in two phases. A preliminary data set was collected to examine relationships between laboratory variables of interest. Results were used to generate hypotheses tested in a subsequent, confirmatory data set.
For preliminary data, we identified patients at the University of Chicago Hospitals who had laboratory tests done to evaluate hypertension during pregnancy between May and November 1993. Computerized records were searched for gravidas who had the combination of a platelet count plus a uric acid or a platelet count plus a serum creatinine within any 24-hour period. We reasoned that one of these combinations would be ordered on virtually all patients evaluated for preeclampsia, and that these combinations would be ordered on few gravidas being assessed for other conditions.
During the 6-month period, 295 pregnant patients cared for at the University of Chicago were identified by computer search as having potentially undergone evaluation for a hypertensive disorder. Of these 295 subjects, 28 had at least one abnormal PT, activated partial thromboplastin time (aPTT), or fibrinogen test. Of these, 25 charts were available for review and 12 (48%) patients with no confounding disorders were identified as having had tests ordered for the evaluation of hypertension. With regard to the 267 subjects without abnormal PT, aPTT, or fibrinogen tests, we determined that to provide a reasonable estimate of test specificity (within 7% using a 95% confidence interval) a sample of 58 would be required. To assure an adequate number of medical records, 117 subjects were selected with a computer-generated random number table. Of these, 73 charts were available for review and 61 (84%) subjects had had testing for some hypertensive disorder. Thus, the preliminary data set consisted of results on 73 gravidas (12 with at least abnormal coagulation test and 61 with no abnormal coagulation test) who were evaluated for hypertension and who had no concomitant disorder that might produce coagulopathy.
The confirmatory data set, identified with the computer search procedure described above, included laboratory test results on patients cared for on the inpatient or labor and delivery services of the University of Chicago Hospitals from April 1994 through December 1996 and on the inpatient obstetric service at Loyola University Medical Center from October 1994 to October 1996. At the University of Chicago, 1341 subjects were identified by the search strategy. Of these, 697 charts were available for review and 562 gravidas had laboratory testing for a hypertensive disorder. At Loyola 879 subjects were identified by the search strategy, 378 charts were available for review, and 170 gravidas were tested for hypertension. Thus, the final data set included 562 gravidas at the University of Chicago and 170 at Loyola without complicating conditions who were tested for hypertension complicating gestation.
During both phases of the study, structured chart reviews were done to determine whether tests were ordered for the evaluation of a hypertensive disorder and whether any confounding conditions (ie, placental abruption, sepsis, or prolonged intrauterine fetal death) were present that might alter coagulation tests. In the preliminary phase, all charts were reviewed by one of the investigators (WMB.) and subsequently by one of the authors, or a nurse, physician, or medical student trained by one of the authors. During training for chart review, agreement between the investigator and the trainee was at least 95% before the trainee was permitted to review charts independently.
Abnormal platelet count was defined as less than 150 × 109/L. Transaminase and lactate dehydrogenase were considered abnormal if they exceed the upper limit of normal in each institution for the adult, non-pregnant population. Analyses of PT, PTT, and fibrinogen were done at two thresholds of abnormality. Abnormal PTs and PTTs were defined as either exceeding the upper limit of normal for that institution's adult, nonpregnant population or over 18 seconds (PT) and over 40 seconds (aPTT). The latter were determined on the basis of discussions with our hematology consultant as levels that present a significant risk of bleeding. Fibrinogen levels were defined as abnormal at less than 200 mg/dL or less than 250 mg/dL.
Because organ dysfunction progresses rapidly in some preeclamptics, we assessed the relationship only between tests ordered simultaneously, defined as those obtained within 2 hours of one another. Two-by-two contingency tables were constructed to describe relationships between predictor and outcome test results, where outcomes were PT, aPTT, and fibrinogen. Because of the likelihood of correlation between serial tests on the same patient, only the group of tests containing the first abnormal outcome was analyzed. Sensitivity, specificity, and their respective 95% confidence intervals were calculated using STATXACT Version 2 statistical software program (CYTEL Software Corp., Cambridge, MA). Positive and negative predictive values and their respective 95% confidence intervals were calculated using a Taylor's series approximation.4 The χ2 test was used to compare rates of selected variables at the two institutions studied and to assess whether abnormal coagulation parameters were associated with transfusion of fresh frozen plasma platelets. Because of multiple testing, a P value < .01 was considered significant.
Table 1 summarizes coagulation testing in the initial 73 gravidas, many of whom had a PT, aPTT, or fibrinogen measured. Of those tested, only a few had at least one abnormal result. Equally important, the most abnormal values for PT, aPTT, and fibrinogen were 14.6 seconds, 44.7 seconds, and 212 mg/dL, respectively, results of questionable clinical significance.
Multiple exploratory analyses were done to identify a predictor test or combination of tests that would accurately identify abnormal PT, aPTT, and fibrinogen results. Because the primary clinical concern was the failure to detect a clinically significant coagulation disturbance (false-negative result), analyses and discussion focus on sensitivity (sensitivity = 1−false-negative rate) and negative predictive value. Results demonstrated that an abnormal platelet count (less than 150 × 109/L) or an abnormal lactate dehydrogenase (above the laboratory reference standard for nonpregnant patients) resulted in the fewest false-negative results. A single, minimally abnormal PT (14.6 seconds) was not accurately predicted by an abnormal platelet count or an abnormal lactate dehydrogenase. All four abnormal aPTTs were accurately identified by an abnormal platelet count or an abnormal lactate dehydrogenase, and three of four abnormal fibrinogen levels (all between 200 and 250 mg/dL) were correctly identified. The ability of an abnormal platelet count or an abnormal lactate dehydrogenase to detect abnormalities of PT, aPTT, and fibrinogen was then assessed in a second data set.
During the next stage, 562 gravidas at the University of Chicago and 170 at Loyola were identified to have had laboratory testing for the evaluation of a hypertensive disorder and to have no confounding condition that might produce coagulopathy. During this study period there were 8007 deliveries at the University of Chicago and 2493 at Loyola. Table 2 shows that the two populations differed in race, as well as in the prevalence of renal disease, type I diabetes, and preeclampsia in the index pregnancy.
Table 3 summarizes coagulation and liver function testing. Approximately 30% of subjects had determinations of PT and aPTT; 25% of those at the University of Chicago and 45% of those at Loyola also had fibrinogen testing. More than 90% had measurement of lactate dehydrogenase and transaminase levels.
Among subjects who had coagulation testing, few had abnormal results and very few had levels denoting significant risk of hemorrhage. Four of 212 subjects (1.9%) had an abnormal PT, of which none exceeded 18 seconds. Twenty-six (12%) of 213 subjects who had aPTT test had an abnormal result, but only two (1%) exceeded 40 seconds. Similarly, only 18 (8%) of the 215 subjects who had fibrinogen measurements had a result less than 250 mg/dL, and only three (1%) had a level less than 200 mg/dL. These data indicate that in a population of gravidas being tested for a hypertensive disorder, the prevalence of clinically significant perturbations of coagulation studies was extremely low, ie, 1% or less.
Table 4 summarizes operating characteristics for an abnormal platelet count or an abnormal lactate dehydrogenase as a predictor of coagulation test outcome. The combination of these two tests had a sensitivity greater than 70% to detect values of PT and aPTT above the nonpregnant normal range. Similar sensitivity was observed for detection of fibrinogen levels less than 250 mg/dL. Because of the very low prevalence of abnormal coagulation results, the associated negative predictive values are high, ie, 92–100%.
To explore the clinical significance of false-negative results of an abnormal platelet count or an abnormal lactate dehydrogenase, abnormal coagulation tests not detected by this combination of tests were studied in detail. Only a single abnormal PT value of 13.9 seconds was missed; simultaneous aPTT, fibrinogen, and platelet count were normal. Five abnormal aPTT values were not identified by an abnormal platelet count or abnormal lactate dehydrogenase; all were less than 36 seconds in patients with normal PT, fibrinogen, and platelet counts. Finally, three patients with decreased fibrinogen levels of 239, 242, and 192 mg/dL were undetected by the predictor test set. Only one patient had simultaneous PT and aPTT determinations and results were normal (all had normal platelet counts).
When the threshold for abnormality of coagulation tests was adjusted to levels that put patients at significant risk of bleeding (Table 4), the sensitivity and negative predictive value of an abnormal platelet count or an abnormal lactate dehydrogenase increased to nearly 100% in almost all instances. Sensitivities of 50% and 67% for detection of fibrinogen less than 200 mg/dL at the University of Chicago and in the combined group, respectively, were due to a failure of the test combination to identify a single fibrinogen level of 192 mg/dL in a patient with a normal PT, aPTT, and platelet count. These analyses indicate that the combination of a normal platelet count and a normal lactate dehydrogenase make it extremely unlikely that a patient will have a clinically significant perturbation of the PT, aPTT, or fibrinogen.
To further explore the potential significance of abnormal coagulation tests, the relationship between such tests and receiving a transfusion of platelets or blood products designed to replenish coagulation factors was studied in the second phase. Eight patients (six from University of Chicago; two from Loyola) received platelets, cryoprecipitate, or fresh frozen plasma. There was no correlation between the results of PT, aPTT, or fibrinogen and receiving transfusion of any coagulation factor. Only the result of the platelet count was associated with such an event (P = .001).
These results confirm that a substantial amount of diagnostic laboratory testing was done during the evaluation of hypertension in pregnant women at the two institutions studied. The true extent of such testing is likely to be larger than current results indicate for the following two reasons: our method did not count tests obtained on patients cared for solely in the outpatient clinic, and a significant number of charts could not be retrieved for evaluation. We are unaware of similar data, but several reports suggest that, at least at academic medical centers, coagulation testing is common in the evaluation of hypertensive pregnant women.5–9
Data regarding the prevalence of abnormal coagulation tests in gravidas with hypertensive disorders are highly discrepant partly because of variance in populations evaluated and different definitions of abnormal values. In addition, studies often include small numbers of highly selected patients, uniformly fail to provide confidence limits for reported point estimates, and often do not separately analyze patients with concomitant disorders that could produce coagulopathy.5,8–12 However, taken together these studies indicate that the prevalence of abnormalities of PT, aPTT, and fibrinogen among patients with preeclampsia is very low.
If it were possible to identify, before coagulation testing, patients in whom the probability of an abnormal PT, aPTT, or fibrinogen was extremely unlikely, it might be justifiable not to order those tests. Previous investigators used the platelet count and liver function tests to predict abnormalities in coagulation tests. Leduc et al6 found that all severely preeclamptic gravidas with an abnormal PT, aPTT, or fibrinogen had thrombocytopenia, and Kramer et al9 reported that each of the small number of preeclamptic gravidas with abnormal PT or aPTT had either a low platelet count or elevated liver function test results. In contrast, Metz et al7 reported that eight of 12 subjects with abnormal aPTT had normal platelet counts. However, most subjects with an abnormal aPTT had a normal PT and platelet count, hence the clinical significance of a prolonged aPTT was uncertain.
A major methodologic concern about each of those studies is that coagulation testing was evaluated in populations with established hypertensive disorders of pregnancy of one type or another. In clinical practice, however, tests are ordered because the diagnosis is uncertain. Test results and further observation often rule out clinically significant hypertensive disease. Principles of diagnostic testing require that an evaluation of the clinical usefulness of coagulation tests must utilize a study population that includes individuals with symptoms and signs that could be confused with preeclampsia, in addition to patients with mild and severe disease, both treated and untreated.13 Failure to include such a study population yields estimates of test sensitivity, specificity, and predictive value that are not representative of the circumstances in which the test is actually utilized in clinical practice.
A strength of the current study is that the population of gravidas was considered by their doctors to warrant laboratory testing. In such patients the prevalence of coagulation tests sufficiently abnormal to put the patient at risk of significant hemorrhage, ie, PT over 18 seconds, aPTT over 40 seconds, and fibrinogen less than 200 mg/dL, was between zero and 1.4%. Even minimal abnormalities of these coagulation tests were infrequent, ranging from 2% for the PT to 12% for the aPTT. Said otherwise, in the two institutions studied, coagulation tests are ordered for many patients even though the pretest likelihood (prevalence) of a clinically significant abnormal value is extremely low. Of importance is that the low pretest likelihood of an abnormal coagulation result leads directly to a very high predictive value of a negative test. Consequently all negative predictive values in this study exceeded 90%.
Clinicians might be concerned about diagnostic tests that fail to detect 5–10% of cases in which a serious, clinically relevant abnormality is present. In this regard the negative predictive value of the platelet count plus a lactate dehydrogenase in our combined population was only 92% for aPTT above the nonpregnant reference range and 96% for a fibrinogen less than 250 mg/dL. However, in each of the few cases in which the tests failed to identify an abnormal aPTT or fibrinogen, all other coagulation tests were normal and no hemorrhagic complications occurred. Furthermore, when we examined the ability of a platelet count plus a lactate dehydrogenase to identify abnormalities of PT, aPTT, or fibrinogen that might pose a direct, substantial threat of bleeding, the negative predictive value was 100% for both PT and aPTT, and 98% for fibrinogen. The test combination failed to identify only one patient with fibrinogen less than 200 mg/dL, and in this case the value was 192 mg/dL, other coagulation tests were normal, and no hemorrhagic complications ensued.
A key attribute of a clinically useful test is the demonstration of its effect on patient treatment or outcomes. Thus, the finding of an abnormal aPTT in a pregnant woman being evaluated for preeclampsia does not in and of itself justify ordering the test. We attempted to explore this issue further by examining the relationship between abnormal coagulation parameters and administration of clotting factors or platelets. We found no correlation between an abnormal PT or aPTT and such transfusion; however, thrombocytopenia was highly correlated with use of platelets or fresh frozen plasma.
We emphasize that our study was not designed to determine definitively whether the availability of PT, aPTT, or fibrinogen results modify patient treatment or outcome. However, our results in combination with other investigations strongly support the position that a slightly abnormal aPTT or fibrinogen test, in the absence of other clinical or laboratory evidence of severe disease, would be extremely unlikely to modify treatment.
The current study has several limitations. Because many charts were not reviewed (approximately 50% of eligible), the results might be biased if these charts differed systematically from those evaluated. We have no a priori reason to believe this is the case. Second, our results represent findings at two academic medical centers in Chicago and might not be generalizable to other types of hospitals in other locations. However, the racially diverse mix of patients studied as well as the urban and suburban locations of the two study sites could increase the external validity of our findings. Bias could have been introduced if the case finding strategy (platelet plus uric acid or platelet plus creatinine on the same day) did not sensitively detect patients being evaluated for hypertension in pregnancy. Finally, bias could have resulted from analyzing only the first set of predictor-outcome tests in which there was an abnormal outcome; however, statistical validity required such an approach.
The present findings support published guidelines, including those of the National High Blood Pressure Education Program Working Group on High Blood Pressure in Pregnancy,1 ACOG,2 and the Australasian Society for the Study of Hypertension in Pregnancy.3 None of these guidelines call for routine use of PT, aPTT, or fibrinogen in the evaluation of women with hypertension complicating gestation.
Current and published data indicate that measurement of PT, aPTT, or fibrinogen are unnecessary when evaluating a patient with suspected or known preeclampsia in whom there is no clinical evidence of bleeding or of a condition that could produce coagulopathy and in whom platelet count and lactate dehydrogenase levels are normal. Because the vast majority of platelet counts and lactate dehydrogenase levels in this study were normal, such an approach would eliminate the need for most of the coagulation tests obtained. The economic and operational benefits of a major reduction in coagulation testing of patients evaluated for preeclampsia is magnified when one considers that many, if not most, of these tests are ordered on a rush basis. Finally, although not directly addressed by the current study, available data justify the position that other coagulation tests, such as fibrin-fibrinogen degradation products, D-dimer, and antithrombin III, have not been shown to be clinically necessary in most gravidas being evaluated for preeclampsia and should not be ordered routinely until further studies demonstrate their specific influence on patient treatment or outcome.
1. National High Blood Pressure Education Program Working Group. Report on High Blood Pressure in Pregnancy. Am J Obstet Gynecol 1990;163:1689–712.
2. American College of Obstetricians and Gynecologists. ACOG technical bulletin no. 219. Hypertension in pregnancy. Washington DC: American College of Obstetricians and Gynecologists, 1996.
3. Australasian Society for the Study of Hypertension in Pregnancy. Management of hypertension in pregnancy: Executive summary. Med J Aust 1993;158:700–2.
4. Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic research: Principles and quantitative methods. New York: Van Nostrand Reinhold Company Inc., 1982:298–9.
5. Trofatter KF, Howell ML, Greenberg CS, Hage ML. Use of the fibrin d-dimer in screening for coagulation abnormalities in preeclampsia. Obstet Gynecol 1992;73:435–40.
6. Leduc L, Wheeler JM, Kirshon B, Mitchell P, Cotton DB. Coagulation profile in severe preeclampsia. Obstet Gynecol 1992;79:14–8.
7. Metz J, Cincotta R, Francis M, DeRosa L, Balloch A. Screening for consumptive coagulopathy in preeclampsia. Int J Gynecol Obstet 1994;46:3–9.
8. Prieto JA, Mastrobattista JM, Blanco JD. Coagulation studies in patients with marked thrombocytopenia due to severe preeclampsia. Am J Perinatal 1995;12:220–2.
9. Kramer RL, Izquierdo LA, Gilson GJ, Curet LB, Quails R. “Preeclamptic labs” for evaluating hypertension in pregnancy. J Reprod Med 1997;42:223–8.
10. Pritchard JA, Cunningham FG, Mason RA. Coagulation changes in eclampsia: Their frequency and pathogenesis. Am J Obstet Gynecol 1976;124:855–64.
11. Löpez-Llera M, de la Luz Espinosa M, Diaz de Leön M, Linares GR. Abnormal coagulation and fibrinolysis in eclampsia. A clinical and laboratory correlation study. Am J Obstet Gynecol 1976;124:681–7.
12. Weinstein L. Preeclampsia/eclampsia with hemolysis, elevated liver enzymes, and thrombocytopenia. Obstet Gynecol 1985;66:657–60.
13. Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical epidemiology. A basic science for clinical medicine. 2nd ed. Boston, Massachusetts: Little, Brown and Company, 1991.