Bleeding in the first trimester of pregnancy - CHAPTER 3 Threatened miscarriage in general practice diagnostic value of history taking and physical exemination

Bleeding in the first trimester of pregnancy

Wieringa-de Waard, M.

Publication date

2002

Link to publication

Citation for published version (APA):

Wieringa-de Waard, M. (2002). Bleeding in the first trimester of pregnancy.

General rights

It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons).

Disclaimer/Complaints regulations

If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: https://uba.uva.nl/en/contact, or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible.

Threatenedd miscarriage in general practice

diagnosticdiagnostic vaiue of history taking and physical

examination examination

Margreett Wieringa-de Waard, Gouke J. Bonsel, Willem M. Ankum, Jeroenn Vos, Patrick J. E. Bindels

Submittedd for publication

Abstract t

OBJECTIVESS To determine the diagnostic value of history taking and physical exam-inationn in first-trimester bleeding, for differentiating between patients requiring im-mediatee further diagnostic examination from those in whom an expectant policy will initiallyy suffice.

DESIGNN Prospective population based cohort study. SETTINGG 74 General practices in Amsterdam.

PARTICIPANTSS Pregnant women with first-trimester bleeding.

RESULTSS 225 patients with first-trimester vaginal bleeding were referred for an early pregnancyy assessment and the data of 204 were analysed.

Threee approaches (models) were explored to identify diagnostic subgroups relevant too clinical practice. Model 1 distinguishes between viable pregnancies and complete miscarriagess on the one hand (expectant policy possible) and all other diagnoses on thee other (further diagnostics required). By using this model the pre-test probability off viable pregnancies and complete miscarriages combined, increased from 72% to 77%% post-test. Model 2, which separates viable pregnancies from other diagnoses, enabledd an increase from a 47% pre-test probability to a post-test probability of 70%.. Model 3, enabling the identification of complete miscarriages, showed a post-testt probability of 4 1 % of a complete miscarriage, given a pre-test sample probabili-tyy of 25%. The tentative diagnosis of the GP, based entirely on clinical judgement, turnedd out to be a poor predictor for the ultrasonographically confirmed diagnosis (pre-testt probability of 47% changed to a post-test probability of 58%).

CONCLUSIONSS This study shows that in first-trimester bleeding neither statistical predictionn models based on signs and symptoms, nor clinical judgement can validly replacee ultrasonographic assessment in establishing a diagnosis.

Introduction n

First-trimesterr vaginal bleeding is the most common problem in pregnancy, al-mostt always leading to a consultation in the general or gynaecological practice. Thee problem occurs in about one-fifth of all pregnancies, with an estimated miscarriagee rate of about 50%, while the other 50% of pregnancies continue withoutt serious problems.1 Ultrasonography, the gold standard for establishing aa diagnosis in first-trimester vaginal bleeding, is not always immediately available.. In these circumstances, the patient's medical history and a gynae-cologicall examination are used to make a provisional diagnosis and to decide whetherr referral for any further clinical assessment, i.e. ultrasonographic examination,, is needed. To date virtually no information is available on the discriminativee capacity of clinical symptoms and gynaecological examination in differentiatingg between the various underlying diagnoses, e.g. viable or non-viablee pregnancy, complete or incomplete miscarriage or ectopic pregnancy.2 Inn this paper we present a prospective population-based cohort study of oth-erwisee unselected women, presenting in general practice with first-trimester bleeding,, who were referred to our unit for ultrasonographic assessment. We analysedd whether the patient's history and physical examination were of any usee in predicting the final diagnosis. In addition, we analysed the skills of the attendingg GPs in making a provisional diagnosis of a viable pregnancy on clinicall grounds before referring the patient.

Methods s

Patients Patients

Ourr study was conducted as part of a randomised trial on the management of miscarriagess in two Amsterdam hospitals, the Academic Medical Center and thee Onze Lieve Vrouwe Gasthuis, between April 1998 and September 2000. Seventy-fourr GPs working in the health district covered by these two hospi-talss participated in the study. They referred all women with vaginal bleeding, aa positive urinary pregnancy test result, and a gestational age of less than 16 weeks,, for ultrasonographic assessment. Oral informed consent was obtained inn all patients.

Thee study was approved by the medical ethics committees of both hospitals.

DataData collection

Duringg the first consultation, the participating GPs collected base line data onn a structured form on clinical signs and symptoms, obstetric history and gestationall age. Additional findings from the gynaecological examination weree recorded. With regard to the gynaecological examination GPs had the possibilityy to fill out 'not performed' on the form. A provisional diagnosis was thenn made, and the GP was asked to estimate the probability of a viable

preg-nancyy being present on a 0-100% scale. Information on sociodemographic data wass collected by means of a patient questionnaire.

UltrasoundUltrasound examination

AA standardised transvaginal sonographic examination was performed by trained physicianss using a transvaginal 6.5 MHz sonographic probe (Hitachi Corpo-ration,, Tokyo, Japan). A viable pregnancy was diagnosed whenever fetal heart activityy could be detected. A non-viable pregnancy was diagnosed in case of ann immeasurable embryonic pole at a gestational sac diameter > 15 mm, an embryoo without cardiac activity, or an empty gestational sac at a diameter <

155 mm not showing any growth after a seven-day interval.3 , 4 An incomplete miscarriagee was diagnosed in case of sonographic evidence of retained prod-uctss of conception (RPOC) > 15 mm anteroposterior (AP) diameter. AA complete miscarriage was diagnosed in case of sonographic evidence of RPOCC < 15 mm (AP diameter) in combination with previous sonographic documentationn of an intra-uterine pregnancy (IUP) or with decreasing serum betaa human chorionic gonadotrophin (hCG) concentrations during follow-up. AA diagnosis of ectopic pregnancy was considered if no intra-uterine gestation-all sac was seen, or if an ectopic gestational sac was seen, or if an ectopic mass wass visualised, together with a serum hCG concentration > 1.500 IU/L.5 Thee ultrasonographic diagnosis of hydatidiform mole had to be confirmed by histopathologicall examination.

StatisticalStatistical analysis

Ourr prospective approach implies that we analysed all cases, including rela-tivelyy rare conditions, e.g. ectopic pregnancy and hydatidiform mole, which shouldd be considered in differential diagnosis of first-trimester bleeding. Wee analysed three models, which were constructed to classify women pre-sentingg with first-trimester bleeding into a group in whom an expectant policy cann be justified and a group in whom further ultrasonographic assessment is required. .

Modell 1 aims as the first step to separate viable pregnancies together with completee miscarriages from the rest of the study population, thereby assum-ingg that these women do not require immediate sonographic assessment. Afterr the first step model 1 would be further explored to separate viable pregnanciess from complete miscarriages.

Modell 2 tries to predict the presence of a viable pregnancy, while model 3 triess to separate complete miscarriages from all remaining diagnoses. Thee relationship between patient characteristics, clinical variables, findings fromm gynaecological examination and the diagnosis was analysed, starting withh an univariate regression analysis. All variables with a p-value < 0.10 weree selected for further multivariate analysis. Prior to multivariate mod-elling,, the multicollinearity of potential predictive variables was checked

Tablee 3 . 1 . Patient characteristics according to the sonographically confirmed

diagnosis.3 3

Characteristic c

AgeAge - yr, mean

Parity Parity 0 0 1 1 > 1 1 PriorPrior miscarriage 0 0 1 1 > 1 1 Prior Prior ectopicectopic pregnancy GestationalGestational age < 88 wk 8-122 wk 12-166 wk Uncertain n Viable e pregnancy y ( ( / / 56 6 25 5 15 5 54 4 28 8 14 4 1 1 35 5 37 7 12 2 12 2 nn = 96) 29.3 3 (58.3) ) (26.0) ) (15.6) ) (56.3) ) (29.2) ) (14.6) ) (1.0) ) (36.5) ) (38.5) ) (12.5) ) (12.5) ) Complete e miscarriage e (nn = 51) 30.9 9 300 (58.8) 144 (27.5) 7(13.7) ) 388 (74.5) 77 (13.7) 66 (11.8) --211 (41.2) 211 (41.2) 55 (9.8) 44 (7.8) Ectopic c pregnancy y (nn = 8) 34.0 0 33 (37.5) 44 (50.0) 11 (12.5) 33 (37.5) 22 (25.0) 33 (37.5) --66 (75.0) 11 (12.5) --11 (12.5) No o n-viable e pregnancy0 0 ( ( n = 4 9 ) ) 31.4 4 25 5 14 4 10 0 33 3 10 0 6 6 10 0 23 3 8 8 8 8 (51.0) ) (28.6) ) (20.4) ) (67.3) ) (20.4) ) (12.2) ) --(20.4) ) (46.9) ) (16.3) ) (16.3) ) NativeNative country Western-Europe e andd USA 48(50.0) 29(56.9) 3(37.5) 2 7 ( 5 5 . 1 ) Africann country 11 (11.5) 5 (9.8) 1 (12.5) 5 (10.2) Surinamm and Antilles 26(27.1) 11(21.6) 2(25.0) 9 ( 1 8 . 4 ) Otherr and unknown 11(11.5) 6(11.8) 2(25.0) 8 ( 1 6 . 3 )

aa

Values are numbers with percentages in parentheses.

bb

Non-viable pregnancies; two hydatidiform mole pregnancies included.

withh a conventional correlation analysis (threshold 0.4). In case of multi-collinearityy the variable with the highest clinical face validity and/or ease of determinationn in primary care was selected. The level of significance for the multivariatee model was set at 0.05. With the variables, identified with multi-variatee regression analyses as independent predictors, Receiver-Operating-Characteristicc (ROC) curves were made, describing the relationship between sensitivityy and specificity for each model. We regarded the point with the

lowestt sum of false-positives plus false-negatives (highest accuracy) as the best thresholdd in the ROC curve.

Inn addition, the diagnostic performance of model 2 was compared with the GPs'' estimation of the probability of a viable pregnancy being present. Thee Statistical Package of the Social Sciences (SPSS, version 9.0) was used forr all analyses.

Results s

Outt of 225 women referred by GPs for an early pregnancy assessment, 217 weree eligible for the study; eight women with a negative hCG pregnancy test weree excluded. Of these 217 women 13 were excluded; two were lost to fol-low-upp before a final diagnosis was made and 11 because of an unacceptably highh number of missing data. Of the remaining 204 women, 96 (47%) had a viablee pregnancy, 51 (25%) a complete miscarriage, 47 (23%) a non-viable pregnancy,, two (1%) a hydatidiform mole, while eight women (4%) had an ectopicc pregnancy.

Noo significant differences in patient characteristics were observed between womenn with viable pregnancies, complete miscarriages and non-viable preg-nanciess (Table 3.1). The number of women with ectopic pregnancies was too smalll for a meaningful comparison.

Inn Table 3.2 the univariate analysis according to the three specified pre-dictivee models are presented. For gynaecological examination the possibility wass given to fill out 'not performed'. Vaginal examination was performed in 57.4%% and speculum examination in 61.3% of the cases. Performed or not performedd was not different between the diagnosis groups.

Tablee 3.3 shows the results of the multivariate regression analyses.

Forr model 1, we found both stable bleeding (OR: 0.3) and increasing bleeding (OR:: 0.4) to be independent predictors of the group of viable pregnancies and completee miscarriages.

Inn model 2, independent predictors of a viable pregnancy were stable and increasingg bleeding (OR: 0.4 and OR: 0.1 respectively). The absence of blood att speculum examination was also negatively associated with a viable preg-nancyy (OR: 0.4). A negative history of passing blood clots was a predictor of a viablee pregnancy (OR: 2.2).

Forr model 3, independent predictors of a complete miscarriage were stable or increasingg bleeding (positively associated) and a negative history of passing bloodd clots (negatively associated).

Inn Figure 3.1, the ROC curves of the three models and the ROC curve for the predictionn of the GPs are shown.

Tablee 3.2. Univariate logistic regression analyses in three models3. Onlyy Odds ratios with p-values < 0.10 are reported.0

AgeAge (per year)

PriorPrior miscarriage 0 0 1 1 >1 1 Bleeding Bleeding decreasing g stable e increasing g Bleeding Bleeding

lesss than period similarr to period moree than period HistoryHistory of passing

yes s no o

BloodBlood colour brown n

redd and brown

Modell 1 0.99 (0.9-1 NS S NS S NS S 0) ) 1 1 0.33 (0.1-0.7) 0.44 (0.2-1.0) NS S NS S NS S bloodblood clots NS S NS S NS S NS S Modell 2 0.99 (0.9-1.0) 1 1 2.00 (1.0-4.0) 1.33 (0.6-2.9) 1 1 0.33 (0.2-0.6) 0.11 (0.0-0.2) 1 1 0.22 (0.1-0.5) 0.6(0.3-1.5) ) 1 1 3.33 (1.7-6.3) NS S NS S Modell 3 NSC C 1 1 0.44 (0.2-1.0) 0.66 (0.2-1.6) 1 1 1.77 (0.7-4.1) 7.11 (2.3-17.3) 1 1 6.22 (2.6-14.4) 3.66 (1.4-8.9) 1 1 0.22 (0.1-0.5) 1 1 2.44 (1.0-5.5) HistoryHistory of passing tissue mass

yess NS noo NS AbdominalAbdominal pain

yes s no o

CourseCourse of abdominal pain noo pain decreasing g stable e increasing g NS S NS S 1 1 2.55 (0.9-7.1) 1.00 (0.5-2.2) 2.55 (0.7-9.4) 11 1 4 . 3 ( 1 . 4 - 1 3 . 3 )) 0 . 2 ( 0 . 1 - 0 . 6 ) NS S NS S NS S NS S NS S NS S 1 1 0.5(0.3-1.0) ) 1 1 1.55 (0.6-3.8) 2.11 (1.0-4.5) 1.99 (0.7-5.8)

LocationLocation of abdominal pain noo pain

nott in the middle (unilateral) ) middle e difficultt to say SpeculumSpeculum examination done e nott done Modell 1 NS S NS S NS S NS S NS S NS S Modell 2 NS S NS S NS S NS S 1 1 1.66 (0.9-2.9) Modell 3 1 1 2.9(1.1-7.8) ) 1.66 (0.8-3.4) 1.4(0.4-4.8) ) NS S NS S SpeculumSpeculum examination nott done done,, no blood done,, blood SpeculumSpeculum examination nott done 1 1 0.88 (0.3-2.0) 0.55 (0.3-1.1) 1 1 2.00 (0.8-4.8) 0.44 (0.2-0.8) 1 1 0.11 (0.0-0.9) 1.66 (0.8-3.0) done,, no clots done,, clots SpeculumSpeculum examination nott done

done,, no erosion of the cervi> done,, erosion of the cervix

VaginalVaginal digital examination done e

nott done

VaginalVaginal digital examination nott done

done,, cervix open done,, cervix closed

0.66 (0.3-1.1) 0.44 (0.1-1.6) 1 1 [0.55 (0.2-1.0) 0.88 (0.3-2.3) 1 1 1.77 (0.9-3.3) 1 1 0.66 (0.2-2.3) 0.55 (0.3-1.0) 0.7(0.4-1.2) ) 0.11 (0.0-0.8) 1 1 0.66 (0.3-1.1) 1.11 (0.4-2.8) NS S NS S 1 1 0.66 (0.2-1.8) 0.66 (0.3-1.1) 0.99 (0.5-1.9) 3.2 2 (0.8-12.1) ) NS S NS S NS S NS S NS S NS S NS S NS S aa 95% CI in parentheses. bb

Model 1: viable pregnancy and complete miscarriage versus ectopic and non-viable pregnancy.. Model 2: viable pregnancy versus all other diagnoses. Model 3: complete miscarriagee versus all other diagnoses. The three models included all characteristics (Tablee 3.1) as well as the variables feeling pregnant and period of pain and bleeding untill consultation of the GP.

cc

Tablee 3.3. Multivariate logistic regression analyses in three models. a'b

Modell 1 Model 2 Model 3

Bleeding Bleeding

decreasingg 1 1 1 stablee 0.3(0.2-0.7) 0.4(0.2-0.8) 1.5(0.6-3.8) increasingg 0.4(0.2-1.0) 0.1(0.0-0.3) 4.9(1.9-12.7) HistoryHistory of passing blood clots

yes s 11 1 noo 2.2(1.0-4.6) 0.4(0.2-0.8) SpeculumSpeculum examination nott done 1 done,, no blood 1.0(0.4-2.9) done,, blood 0.4 (0.2-0.8) aa

Results expressed in Odds ratio (95% CI).

bb

Model 1: viable pregnancy and complete miscarriage versus ectopic and non-viable pregnancy.. Univariately significant variables used as input were: age, amount of bleeding,, course of abdominal pain, and blood at speculum examination. Model 2: viablee pregnancy versus all other diagnoses. As input were used the variables age, priorr miscarriage, amount of bleeding, history of passing blood clots, and blood at speculumm examination. Model 3: complete miscarriage versus all other diagnoses. Variabless used as input were prior miscarriage, amount of bleeding, history of pass-ingg blood clots, blood colour, abdominal pain, and blood at speculum examination. Duee to multicollinearity we removed four, six and six variables, respectively, from the variabless selected with univariate analyses.

Modell 1: The optimal point in the ROC curve of this model, which tried to distinguishh between viable pregnancies together with complete miscarriages andd other diagnoses, is the point with a sensitivity of 80% and a specificity of 40%,, which resulted in a positive likelihood ratio (LR + ) of 1.3. The pre-test probabilityy of 72% (percentage of viable pregnancies and complete miscar-riagess in the study) changed to a post-test probability of 77%.

Modell 2: The optimal point in the ROC curve of this model, which predicted thee presence of viable pregnancies, showed a sensitivity of 68% and a speci-ficityy of 74% (LR+ 2.6). By using this model, the pre-test probability of a viablee pregnancy changed from 47% to a post-test value of 70%.

Modell 3: The best threshold in the ROC curve of this model, separating com-pletee miscarriage from all other diagnoses, has a sensitivity of 78% and a specificityy of 65% (LR 2.2 + ). The probability of a complete miscarriage increasess from 25% pre-test to a post-test value of 4 1 % .

in in 100-1 1

s o

--ar r

--// i; /

Figuree 3 . 1 . ROC curves of the pre-specified models and of the GP prediction. Model 1

aimss to distinguish between viable pregnancies and complete miscarriages on the onee hand and all other diagnoses on the other, model 2 tries to predict viable preg-nanciess and model 3 distinguishes complete miscarriages from all other diagnoses. Thee GP model shows the prediction of a viable pregnancy by GPs.

beingg present, the best point, with a sensitivity of 89% and a specificity of 42%,, gives an LR+ of 1.5. This changes the pre-test probability of 47% to a post-testt probability of 58%, which is worse than the post-test probability of 70%% in model 2.

Thee weak relation between the prediction by model 2 and by the GPs' esti-matess on a viable pregnancy being present, is illustrated in a scatter plot (Figuree 3.2).

Discussion n

Inn this prospective study of pregnant women with first-trimester bleeding we triedd to develop diagnostic rules based on the patient's history and physical examinationn to enhance the efficiency of referral for ultrasonographic assess-ment.. ROC curve analysis showed that even the best of the pre-specified

a. a. Ü Ü c c o o u u T5 5 0) ) i --Q. . 0.88 O—A-0 . 6 1 . 0 O A O

nr®

ok n no

--a --a

-a--a-a -a--a-a

o o

Figuree 3.2. Scatter plot of the prediction of a viable pregnancy by GPs versus

pre-dictionn of a viable pregnancy by model 2.

modelss had insufficient discriminative power to be of any use in clinical practice. T h ee possibility t h a t GPs might be able to discriminate between viable pregnan-ciess and other diagnoses in early pregnancy on clinical grounds, was rejected. Inn model 1, which aims to identify p a t i e n t s for whom no referral is n e e d e d ( t h o s ee with c o m p l e t e miscarriages a n d viable p r e g n a n c i e s ) , s y m p t o m s r e l a t e d too t h e course and a m o u n t of bleeding have opposite significance for t h e two d i a g n o s e s ,, which most likely explains t h e insufficient d i s c r i m i n a t i o n . So a f u r t h e rr e x p l o r a t i o n of this model to distinguish b e t w e e n viable p r e g n a n c y and c o m p l e t ee m i s c a r r i a g e was not possible. H o w e v e r , models 2 and 3 , designed to identifyy e i t h e r viable pregnancies or complete m i s c a r r i a g e s , also t u r n e d out too b e of little v a l u e .

A l t h o u g hh t h e c o u r s e of b l e e d i n g was an i n d e p e n d e n t p r e d i c t o r in all t h r e e m o d e l s ,, t h e p r e d i c t i v e p o w e r was low and not discriminative enough to rely u p o nn in clinical decisions.

gynaecologicall examination in our models. Apparently this examination is of noo diagnostic significance in first-trimester bleeding. However, we still regard thiss examination as important for evaluating the degree of bleeding (to rule outt emergency situations) and for finding products of conception in the vagi-naa and the cervical os. In our study, in which GPs were asked to perform as theyy did in normal practice, vaginal examination was only performed in 57.4%% and speculum examination in 61.3% of the cases. In a survey in Wessexx about management of bleeding in early pregnancy 24% of the GPs reportedd that they never performed a vaginal examination. A speculum examinationn was never done by 38%.2

Iff in our study gynaecological examination was performed in all cases, we can nott exclude that the multivariate analysis had given more support for this examination. .

AA potential drawback of our study was the delay between the measurements off classifying variables and the ultrasonographic verification with a median timee of two days (25-75 percentiles: 1-3 days), according to normal clinical practicee in the Netherlands. Perhaps a better diagnostic performance could bee achieved, especially in case of a complete miscarriage (model 3), without delayy between GP consultation and sonographic assessment. However, it is nott likely that the very low post-test probability of model 3 will increase in suchh a way that clinical relevance might be reached. Also such a procedure settingg may be feasible for research purposes, but has little relevance for practicess without direct access to ultrasonographic facilities.

Wee were unable to pay specific attention to ectopic pregnancies, because of thee limited number of cases. Earlier studies showed that a physical examina-tionn is insufficient too for diagnosing an ectopic pregnancy in the majority of cases.6 77 Highly suspect cases were usually referred directly to the gynaecolo-gyy department and thus not included in our study. Of the eight ectopic preg-nanciess diagnosed at our unit, only two were referred for suspected ectopic pregnancy.. Three ectopic pregnancies were treated surgically (including the twoo suspected cases), four self-limiting cases with declining serum hCG con-centrationss were managed expectantly, and one ectopic pregnancy was treat-edd with a single dose of methotrexate. The question as to whether it is accep-tablee to withhold an immediate ultrasonographic assessment from patients withh first-trimester bleeding, thereby postponing the diagnosis of ectopic pregnancy,, can not be answered from our data.

Chungg found discrepancy in uterine size to be the most significant predictor off non-viable pregnancies.8 We also included this item in our Case Record Form,, but could not confirm its prognostic value. In our study GPs did not carryy out a vaginal examination in 37% of the cases. In the remaining 63% of thee cases, GPs reported the uterine size to be according to gestational age in 32%,, indeterminate in 23%, while a discrepancy in uterine size was reported inn only 8%. Our study differs from Chung's in both its setting (primary care

versuss secondary care) and in its analyses. We analysed all cases, including ectopicc and molar pregnancies in order to resemble as much as possible the situationn in which the physician has to make a decision in daily practice. The variabless analysed in our study all derive from the standard management of womenn with first-trimester bleeding and can easily be obtained.

Wee conclude that in a primary care setting clinical symptoms and physical examinationn are not useful in predicting a diagnosis in first-trimester bleed-ing.. To establish a certain diagnosis, sonographic assessment is required. References s

1.. Everett C. Incidence and outcome of bleeding before the 20th week of pregnancy: prospectivee study from general practice. BMJ 1997;315:32-4.

2.. Everett C, Ashurst H, Chalmers I. Reported management of threatened miscarriage by generall practitioners in Wessex. BMJ 1987;295:583-6.

3.. Coulam CB, Goodman C, Dorfmann A. Comparison of ultrasonographic findings in spontaneouss abortions with normal and abnormal karyotypes. Hum Reprod 1997;12:823-6. .

4.. Deaton JL, Honore GM, Huffman CS, Bauguess P. Early transvaginal ultrasound followingg an accurately dated pregnancy: the importance of finding a yolk sac or fetal

heartt motion. Hum Reprod 1997;12:2820-3.

5.. Mol BW, Hajenius PJ, Engelsbel S, et al. Serum human chorionic gonadotropin measurementt in the diagnosis of ectopic pregnancy when transvaginal sonogra-phyy is inconclusive. Fertil Steril 1998;70:972-81.

6.. Buckley RG, King KJ, Disney JD, Gorman JD, Klausen JH. History and physical examinationn to estimate the risk of ectopic pregnancy: validation of a clinical pre-dictionn model. Ann Emerg Med 1999;34:589-94.

7.. Mol BW, Hajenius PJ, Engelsbel S, et al. Should patients who are suspected of hav-ingg an ectopic pregnancy undergo physical examination? Fertil Steril 1999;71:155-7. 8.. Chung TK, Sahota DS, Lau TK, Mongelli JM, Spencer JA, Haines CJ. Threatened

abor-tion:: prediction of viability based on signs and symptoms. Aust NZ J Obstet Gynaecol 1999;39:443-7. .