Predicting the outcome of pregnancies of unknown location: Bayesian networks with expert prior information compared to logistic regression

Conventional and modern markers of pregnancy of unknown location: Update and narrative review

Pregnancy of unknown location (PUL) is a temporary pathologic or physiologic phenomenon of early pregnancy that requires follow up to determine the final pregnancy outcome. Evidence indicated that PUL patients suffer a remarkably higher rate of adverse pregnancy outcomes, represented by ectopic gestation and early pregnancy loss, than the general population. In the past few decades, discussion about PUL has never stopped, and a variety of markers have been widely investigated for the early and accurate evaluation of PUL, including serum biomarkers, ultrasound imaging features, multivariate analysis, and the diagnosis of ectopic pregnancy based on risk stratification. So far, machine learning (ML) methods represented by M4 and M6 logistic regression have gained a level of recognition and are continually improving. Nevertheless, the heterogeneity of PUL markers, mainly caused by the limited sample size, the differences in population and technical maturity, etc., have hampered the management of PUL. With the advancement of multidisciplinary integration and cutting-edge technologies (e.g. artificial intelligence, prediction model development, and telemedicine), novel markers, and strategies for the management of PUL are expected to be developed. In this review, we summarize both conventional and novel markers (represented by artificial intelligence) for PUL assessment and management, investigate their advancements, limitations and challenges, and propose insights on future research direction and clinical application.

Predictive Analytic Model for Diagnosis of Ectopic Pregnancy

Objective: Ectopic pregnancy (EP) is a serious condition. Delayed diagnosis could lead to life-threatening outcomes. The study aimed to develop a diagnostic predictive model for EP to approach suspected cases with prompt intervention before the rupture occurred. Methods: A retrospective cross-sectional study enrolled 347 pregnant women presenting first-trimester complications (abdominal pain or vaginal bleeding) with diagnosis suspected of pregnancy of unknown location, who were eligible and underwent chart review. The data including clinical risk factors, signs and symptoms, serum human chorionic gonadotropin (hCG), and ultrasound findings were analyzed. The statistical predictive score was developed by performing logistic regression analysis. The testing data of 30 patients were performed to test the validation of predictive scoring. Results: From a total of 22 factors, logistic regression method-derived scoring model was based on five potent factors (history of pelvic inflammatory disease, current use of emergency pills, cervical motion tenderness, serum hCG ≥1,000 mIU/ml, and ultrasound finding of adnexal mass) using a cutoff score ≥3. This predictive index score was able to determine ectopic pregnancy with an accuracy of 77.8% [95% confidence interval (CI) = 73.1-82.1], specificity of 91.0% (95% CI = 62.1-72.0), sensitivity of 67.0% (95% CI = 88.0-94.0), and area under the curve of 0.906 (95% CI = 0.875-0.937). In the validation group, no patient with negative result of this score had an EP. Conclusion: Statistical predictive score was derived with high accuracy and applicable performance for EP diagnosis. This score could be used to support clinical decision making in routine practice for management of EP.

Diagnostic value of detection of serum β-HCG and CT-IgG combined with transvaginal ultrasonography in early tubal pregnancy

The diagnostic value of detection of serum β-human chorionic gonadotropin (β-HCG) and Chlamydia trachomatis immunoglobulin G (CT-IgG) combined with transvaginal ultrasonography in early tubal pregnancy was investigated. A total of 55 patients with early tubal pregnancy were selected as the tubal pregnancy group, while 55 subjects of normal intrauterine pregnancy were enrolled as the intrauterine pregnancy group. Transvaginal ultrasonography and quantitative detection of serum β-HCG and CT-IgG were performed for all patients, and the clinical examination results were analyzed and compared. The endometrial thickness and serum β-HCG level of patients with early tubal pregnancy were significantly lower than those of women with intrauterine pregnancy (6.7±1.5 vs. 11.6±1.2 mm; 776±109 vs. 5,598±187 U/l), and the differences were statistically significant (p<0.01); the serum CT-IgG antibody positive rate of patients in tubal pregnancy group (49.1%) was significantly higher than that in intrauterine pregnancy group (12.7%) (p<0.01); the serum CT-IgG antibody positive rates of patients with degree I, II and III of pelvic adhesion intubal pregnancy group were 28.6, 75.0 and 81.8%, respectively; the more severe the pelvic adhesion was, the higher the CT-IgG positive rate would be. The diagnostic coincidence rate of combined detection was significantly higher than that of single detection of serum β-HCG, progesterone and endometrial thickness. The detection of serum β-HCG and CT-IgG combined with transvaginal ultrasonography can diagnose the early tubal pregnancy soonest possible, and help choose the appropriate therapeutic methods depending on the situation to reduce the tubal damage of patients, so as to provide a reliable basis for the diagnosis, treatment and prognosis, and it has important clinical application value.

Ectopic pregnancy outcomes in patients discharged from the emergency department

OBJECTIVE

The objective of this study was to determine the proportion of women who had a ruptured ectopic pregnancy after being discharged from the emergency department (ED) where ectopic pregnancy had not yet been excluded.

METHODS

This was a retrospective chart review of pregnant (<12-week gestational age) women discharged home from an academic tertiary care ED with a diagnosis of ectopic pregnancy, rule-out ectopic pregnancy, or pregnancy of unknown location over a 7-year period.

RESULTS

Of the 550 included patients, 83 (15.1%) had a viable pregnancy, 94 (17.1%) had a spontaneous or missed abortion, 230 (41.8%) had an ectopic pregnancy, 72 (13.1%) had unknown outcomes, and 71 (12.9%) had other outcomes that included therapeutic abortion, molar pregnancy, or resolution of βHCG with no location documented. Of the 230 ectopic pregnancies, 42 (7.6%) underwent expectant management, 131 (23.8%) were managed medically with methotrexate, 29 (5.3%) were managed with surgical intervention, and 28 (5.1%) patients had a ruptured ectopic pregnancy after their index ED visit. Of the 550 included patients, 221 (40.2%) did not have a transvaginal ultrasound during their index ED visit, and 73 (33.0%) were subsequently diagnosed with an ectopic pregnancy.

CONCLUSION

These results may be useful for ED physicians counselling women with symptomatic early pregnancies about the risk of ectopic pregnancy after they are discharged from the ED.

Association between three-dimensional transvaginal sonographic markers and outcome of pregnancy of unknown location: a pilot study

OBJECTIVE

To assess the accuracy of three-dimensional (3D) transvaginal sonographic (TVS) parameters in predicting the evolution of a pregnancy of unknown location (PUL).

METHODS

This was a prospective observational study performed at the early pregnancy unit of a university hospital from September 2008 to June 2012. Women with a positive pregnancy test without any signs of intra- or extrauterine pregnancy at their first TVS examination were considered eligible and a 3D dataset containing the entire uterus was acquired. An experienced observer analyzed all 3D datasets for assessment of the following parameters: endometrial thickness, volume, mean gray-scale index and asymmetry. Women were followed until they were classified as having: (i) non-visualized pregnancy loss (NVPL); (ii) intrauterine pregnancy (IUP); or (iii) ectopic pregnancy or persistent PUL. We compared the values of the TVS parameters across the three groups. We also assessed the area under the receiver-operating characteristics curve of the 3D-TVS parameters in comparison to that for serum β-human chorionic gonadotropin (β-hCG) ratio (48 h/baseline) to predict PUL outcome. We then evaluated whether combining the 3D-TVS parameters with serum β-hCG ratio improved the predictive accuracy for PUL outcome by performing a logistic regression analysis.

RESULTS

During the study period 4939 consecutive pregnant women presented at the unit for their initial TVS examination and 325 (7%) were classified as having a PUL, of whom 161 women were enrolled and had a 3D scan of the uterus. However, 19 were excluded because of incomplete follow-up. Data from 142 women with PUL were therefore included in the analysis and the outcomes of these women were: NVPL in 98 (69%), IUP in 27 (19%) and ectopic pregnancy + persistent PUL in 14 + 3 = 17 (12%). Endometrial thickness, endometrial volume and the proportion of women with asymmetric endometrial shape differed significantly between the outcome groups. Endometrial thickness and volume could be used as reasonable predictors of both NVPL and IUP, whereas asymmetric endometrial shape and mean gray-scale index could be used as reasonable predictors of IUP only. The best single parameter to predict PUL outcomes was the β-hCG ratio. Regression analysis demonstrated that endometrial volume and endometrial shape asymmetry added significantly to the β-hCG ratio in predicting IUP but not NVPL.

CONCLUSIONS

3D-TVS markers have a low diagnostic accuracy in predicting PUL outcome. The addition of endometrial volume and shape asymmetry improves the accuracy of the β-hCG ratio in predicting IUP. Copyright © 2016 ISUOG. Published by John Wiley & Sons Ltd.

The clinical performance of the M4 decision support model to triage women with a pregnancy of unknown location as at low or high risk of complications

STUDY QUESTION

What are the adverse outcomes associated with using the M4 model in everyday clinical practice for women with pregnancy of unknown location (PUL)?

SUMMARY ANSWER

There were 17/835 (2.0%) adverse events and no serious adverse events associated with the performance of the M4 model in clinical practice.

WHAT IS KNOWN ALREADY

The M4 model has previously been shown to stratify women classified as a PUL as at low or high risk of complications with a good level of test performance. The triage performance of the M4 model is better than single measurements of serum progesterone or the hCG ratio (serum hCG at 48 h/hCG at presentation).

STUDY DESIGN, SIZE, DURATION

A prospective multi-centre cohort study of 1022 women with a PUL carried out between August 2012 and December 2013 across 2 university teaching hospitals and 1 district general hospital.

PARTICIPANTS/MATERIALS, SETTING, METHODS

All women presenting with a PUL to the early pregnancy units of the three hospitals were recruited. The final outcome for PUL was either a failed PUL (FPUL), intrauterine pregnancy (IUP) or ectopic pregnancy (EP) (including persistent PUL (PPUL)), with EP and PPUL considered high-risk PUL. Their hCG results at 0 and 48 h were entered into the M4 model algorithm. If the risk of EP was ≥5%, the PUL was predicted to be high-risk and the participant was asked to re-attend 48 h later for a repeat hCG and transvaginal ultrasound scan by a senior clinician. If the PUL was classified as 'low risk, likely failed PUL', the participant was asked to perform a urinary pregnancy test 2 weeks later. If the PUL was classified as 'low risk, likely intrauterine', the participant was scheduled for a repeat scan in 1 week. Deviations from the management protocol were recorded as either an 'unscheduled visit (participant reason)', 'unscheduled visit (clinician reason)' or 'differences in timing (blood test/ultrasound)'. Adverse events were assessed using definitions outlined in the UK Good Clinical Practice Guidelines' document.

MAIN RESULTS AND THE ROLE OF CHANCE

A total of 835 (82%) women classified as a PUL were managed according to the M4 model (9 met the exclusion criteria, 69 were lost to follow-up, 109 had no hCG result at 48 h). Of these, 443 (53%) had a final outcome of FPUL, 298 (36%) an IUP and 94 (11%) an EP. The M4 model predicted 70% (585/835) PUL as low risk, of which 568 (97%) were confirmed as FPUL or IUP. Of the 17 EP and PPUL misclassified as low risk, 5 had expectant management, 7 medical management with methotrexate and 5 surgical intervention.Nineteen PUL had an unscheduled visit (participant reason), 38 PUL had an unscheduled visit (clinician reason) and 68 PUL had deviations from protocol due to a difference in timing (blood test/ultrasound).Adverse events were reported in 26 PUL and 1 participant had a serious adverse event. A total of 17/26 (65%) adverse events were misclassifications of a high risk PUL as low risk by the M4 model, while 5/26 (19%) adverse events were related to incorrect clinical decisions. Four of the 26 adverse events (15%) were secondary to unscheduled admissions for pain/bleeding. The serious adverse event was due to an incorrect clinical decision.

LIMITATIONS, REASONS FOR CAUTION

A limitation of the study was that 69/1022 (7%) of PUL were lost to follow-up. A 48 h hCG level was missing for 109/1022 (11%) participants.

WIDER IMPLICATIONS OF THE FINDINGS

The low number of adverse events (2.0%) suggests that expectant management of PUL using the M4 prediction model is safe. The model is an effective way of triaging women with a PUL as being at high- and low-risk of complications and rationalizing follow-up. The multi-centre design of the study is more likely to make the performance of the M4 model generalizable in other populations.

STUDY FUNDING/COMPETING INTERESTS

None.

TRIAL REGISTRATION NUMBER

Not applicable.

Comparison of a Bayesian network with a logistic regression model to forecast IgA nephropathy

Models are increasingly used in clinical practice to improve the accuracy of diagnosis. The aim of our work was to compare a Bayesian network to logistic regression to forecast IgA nephropathy (IgAN) from simple clinical and biological criteria. Retrospectively, we pooled the results of all biopsies (n = 155) performed by nephrologists in a specialist clinical facility between 2002 and 2009. Two groups were constituted at random. The first subgroup was used to determine the parameters of the models adjusted to data by logistic regression or Bayesian network, and the second was used to compare the performances of the models using receiver operating characteristics (ROC) curves. IgAN was found (on pathology) in 44 patients. Areas under the ROC curves provided by both methods were highly significant but not different from each other. Based on the highest Youden indices, sensitivity reached (100% versus 67%) and specificity (73% versus 95%) using the Bayesian network and logistic regression, respectively. A Bayesian network is at least as efficient as logistic regression to estimate the probability of a patient suffering IgAN, using simple clinical and biological data obtained during consultation.

Optimizing the recovery efficiency of Finnish oil combating vessels in the Gulf of Finland using Bayesian Networks

Oil transport has greatly increased in the Gulf of Finland over the years, and risks of an oil accident occurring have risen. Thus, an effective oil combating strategy is needed. We developed a Bayesian Network (BN) to examine the recovery efficiency and optimal disposition of the Finnish oil combating vessels in the Gulf of Finland (GoF), Eastern Baltic Sea. Four alternative home harbors, five accident points, and ten oil combating vessels were included in the model to find the optimal disposition policy that would maximize the recovery efficiency. With this composition, the placement of the oil combating vessels seems not to have a significant effect on the recovery efficiency. The process seems to be strongly controlled by certain random factors independent of human action, e.g. wave height and stranding time of the oil. Therefore, the success of oil combating is rather uncertain, so it is also important to develop activities that aim for preventing accidents. We found that the model developed is suitable for this type of multidecision optimization. The methodology, results, and practices are further discussed.

Performance of human chorionic gonadotropin curves in women at risk for ectopic pregnancy: exceptions to the rules

OBJECTIVE

To investigate the accuracy of serial hCG to predict outcome of a pregnancy of unknown location in an ethnically and geographically diverse setting.

DESIGN

Multisite cohort study.

SETTING

University hospital.

PATIENT(S)

Women with a pregnancy of unknown location.

INTERVENTION(S)

None.

MAIN OUTCOME MEASURE(S)

Patients were followed until diagnosed with ectopic pregnancy (EP), intrauterine pregnancy (IUP), or miscarriage. To predict outcome, observed hCG level was compared with recommended thresholds to assess deviation from defined normal curves. Predicted outcome was compared with standard of care. Sensitivity, specificity, predictive value, and accuracy were calculated, stratified by diagnosis.

RESULT(S)

The final diagnosis of 1,005 patients included 179 EPs, 259 IUPs, and 567 miscarriages. The optimal balance in sensitivity and specificity used the minimal expected 2-day increase in hCG level of 35%, and the minimal 2-day decrease in hCG level of 36%-47% (depending on the level) achieving 83.2% sensitivity, 70.8% specificity to predict EP. However, 16.8% of EPs and 7.7% of IUPs would be misclassified solely using serial hCG levels. Consideration of a third hCG and early ultrasound decreased IUP misclassification to 2.7%.

CONCLUSION(S)

Solely using serial hCG values can result in misclassification. Clinical judgment should trump prediction rules and continued surveillance with a third hCG may be prudent, especially when initial values are low or when values are near suggested thresholds.

Improved modeling of clinical data with kernel methods

OBJECTIVE

Despite the rise of high-throughput technologies, clinical data such as age, gender and medical history guide clinical management for most diseases and examinations. To improve clinical management, available patient information should be fully exploited. This requires appropriate modeling of relevant parameters.

METHODS

When kernel methods are used, traditional kernel functions such as the linear kernel are often applied to the set of clinical parameters. These kernel functions, however, have their disadvantages due to the specific characteristics of clinical data, being a mix of variable types with each variable its own range. We propose a new kernel function specifically adapted to the characteristics of clinical data.

RESULTS

The clinical kernel function provides a better representation of patients' similarity by equalizing the influence of all variables and taking into account the range r of the variables. Moreover, it is robust with respect to changes in r. Incorporated in a least squares support vector machine, the new kernel function results in significantly improved diagnosis, prognosis and prediction of therapy response. This is illustrated on four clinical data sets within gynecology, with an average increase in test area under the ROC curve (AUC) of 0.023, 0.021, 0.122 and 0.019, respectively. Moreover, when combining clinical parameters and expression data in three case studies on breast cancer, results improved overall with use of the new kernel function and when considering both data types in a weighted fashion, with a larger weight assigned to the clinical parameters. The increase in AUC with respect to a standard kernel function and/or unweighted data combination was maximum 0.127, 0.042 and 0.118 for the three case studies.

CONCLUSION

For clinical data consisting of variables of different types, the proposed kernel function--which takes into account the type and range of each variable--has shown to be a better alternative for linear and non-linear classification problems.

Modulation of the endocannabinoid system in viable and non-viable first trimester pregnancies by pregnancy-related hormones

BACKGROUND

In early pregnancy, increased plasma levels of the endocannabinoid anandamide (AEA) are associated with miscarriage through mechanisms that might affect the developing placenta or maternal decidua.

METHODS

In this study, we compare AEA levels in failed and viable pregnancies with the levels of the trophoblastic hormones (beta-human chorionic gonadotrophin (beta-hCG), progesterone (P4) and (pregnancy-associated placental protein-A (PAPP-A)) essential for early pregnancy success and relate that to the expression of the cannabinoid receptors and enzymes that modulate AEA levels.

RESULTS

The median plasma AEA level in non-viable pregnancies (1.48 nM; n = 20) was higher than in viable pregnancies (1.21 nM; n = 25; P = 0.013), as were progesterone and beta-hCG levels (41.0 vs 51.5 ng/mL; P = 0.052 for P4 and 28,650 vs 6,560 mIU/L; P = 0.144 for beta-hCG, respectively, but were not statistically significant). Serum PAPP-A levels in the viable group were approximately 6.8 times lower than those in the non-viable group (1.82 vs 12.25 mg/L; P = 0.071), but again these differences were statistically insignificant. In the spontaneous miscarriage group, significant correlations between P4 and beta-hCG, P4 and PAPP-A and AEA and PAPP-A levels were observed. Simultaneously, immunohistochemical distributions of the two main cannabinoid receptors and the AEA-modifying enzymes, fatty acid amide hydrolase (FAAH) and N-acylphosphatidylethanolamine-phospholipase D (NAPE-PLD), changed within both the decidua and trophoblast.

CONCLUSIONS

The association of higher AEA levels with early pregnancy failure and with beta-hCG and PAPP-A, but not with progesterone concentrations suggest that plasma AEA levels and pregnancy failure are linked via a mechanism that may involve trophoblastic beta-hCG, and PAPP-A, but not, progesterone production. Although the trophoblast, decidua and embryo contain receptors for AEA, the main AEA target in early pregnancy failure remains unknown.

The term "pregnancy of unknown location" is here to stay

The term "pregnancy of unknown location" is an ultrasound classification and not a final diagnosis. The use of this terminology is here to stay and should continue as long as there is an appreciation for what it really means. It is the responsibility of the clinician, who follows up these women with a PUL, to ensure that a final diagnosis is achieved while preserving the well-being of these women.

Does a prediction model for pregnancy of unknown location developed in the UK validate on a US population?

BACKGROUND

A logistic regression model (M4) was developed in the UK to predict the outcome for women with a pregnancy of unknown location (PUL) based on the initial two human chorionic gonadotrophin (hCG) values, 48 h apart. The purpose of this paper was to assess the utility of this model to predict the outcome for a woman (PUL) in a US population.

METHODS

Diagnostic variables included log-transformed serum hCG average of two measurements, and linear and quadratic hCG ratios. Outcomes modeled were failing PUL, intrauterine pregnancy (IUP) and ectopic pregnancy (EP). This model was applied to a US cohort of 604 women presenting with symptomatic first-trimester pregnancies, who were followed until a definitive diagnosis was made. The model was applied before and after correcting for differences in terminology and diagnostic criteria.

RESULTS

When retrospectively applied to the adjusted US population, the M4 model demonstrated lower areas under the curve compared with the UK population, 0.898 versus 0.988 for failing PUL/spontaneous miscarriage, 0.915 versus 0.981 for IUP and 0.831 versus 0.904 for EP. Whereas the model had 80% sensitivity for EP using UK data, this decreased to 49% for the US data, with similar specificities. Performance only improved slightly (55% sensitivity) when the US population was adjusted to better match the UK diagnostic criteria.

CONCLUSIONS

A logistic regression model based on two hCG values performed with modest decreases in predictive ability in a US cohort for women at risk for EP compared with the original UK population. However, the sensitivity for EP was too low for the model to be used in clinical practice in its present form. Our data illustrate the difficulties of applying algorithms from one center to another, where the definitions of pathology may differ.

Development of a kernel function for clinical data

For most diseases and examinations, clinical data such as age, gender and medical history guides clinical management, despite the rise of high-throughput technologies. To fully exploit such clinical information, appropriate modeling of relevant parameters is required. As the widely used linear kernel function has several disadvantages when applied to clinical data, we propose a new kernel function specifically developed for this data. This "clinical kernel function" more accurately represents similarities between patients. Evidently, three data sets were studied and significantly better performances were obtained with a Least Squares Support Vector Machine when based on the clinical kernel function compared to the linear kernel function.

Comparing risks of alternative medical diagnosis using Bayesian arguments

This paper explains the role of Bayes Theorem and Bayesian networks arising in a medical negligence case brought by a patient who suffered a stroke as a result of an invasive diagnostic test. The claim of negligence was based on the premise that an alternative (non-invasive) test should have been used because it carried a lower risk. The case raises a number of general and widely applicable concerns about the decision-making process within the medical profession, including the ethics of informed consent, patient care liabilities when errors are made, and the research problem of focusing on 'true positives' while ignoring 'false positives'. An immediate concern is how best to present Bayesian arguments in such a way that they can be understood by people who would normally balk at mathematical equations. We feel it is possible to present purely visual representations of a non-trivial Bayesian argument in such a way that no mathematical knowledge or understanding is needed. The approach supports a wide range of alternative scenarios, makes all assumptions easily understandable and offers significant potential benefits to many areas of medical decision-making.

Prediction of outcome in women with symptomatic first-trimester pregnancy: focus on intrauterine rather than ectopic gestation

OBJECTIVE

Symptoms of vaginal bleeding and abdominal pain are common in cases of ectopic pregnancy (EP), spontaneous abortions (SAB), and complications of an intrauterine pregnancy (IUP). It is important to determine if efforts should focus on differentiating EP from an IUP (IUP + SAB) or a viable IUP from a nonviable gestation (EP + SAB) in women at risk for EP.

METHODS

This is a retrospective cohort study of women who presented with bleeding or pain or both during the first trimester of pregnancy. The cohort was divided into subjects diagnosed with IUP vs. (EP + SAB). The same cohort was then divided into subjects diagnosed with EP vs. (IUP + SAB). Logistic regression models based on risk factors for both outcomes (EP vs. [IUP + SAB] and IUP vs. [EP + SAB]) were obtained. ROC curves as well as Hosmer-Lemeshow goodness of fit and Akaike's information criterion (AIC) were used.

RESULTS

Overall, 18.1% (n = 367) of the women were diagnosed with EP, 58.8% (n = 1192) were diagnosed with an SAB, and 23.1% (n = 467) had an ongoing IUP. The area under the ROC curve for the model IUP vs. (EP + SAB) was statistically greater than the model EP vs. (IUP + SAB), p < 0.001. AIC and Hosmer-Lemeshow goodness of fit confirmed the better accuracy of the model comparing IUP vs. (EP + SAB).

CONCLUSIONS

Information collected at initial presentation from women at risk for EP to be used for building prediction rules should focus on differentiating a viable from a nonviable pregnancy rather than attempting to distinguish an extrauterine from an intrauterine pregnancy. However, this distinction should not affect current clinical care.

Predicting Outcomes in Pregnancies of Unknown Location

A pregnancy of unknown location (PUL) is a descriptive term used to classify a woman when she has a positive pregnancy test but no intra- or extra-uterine pregnancy is visualized on transvaginal sonography. Expectant management has been shown to be safe for the majority of women with a PUL. Serum progesterone and human chorionic gonadotrophin levels as well as mathematical models play a role in predicting the final outcomes of PULs, which include intrauterine pregnancy, failing PUL and ectopic pregnancy. Other possible predictors of outcome have been studied, but currently no factor has been identified that combines accuracy with reproducibility and simplicity. This article discusses the various aspects of the management of women with PULs. Future work should be aimed at prospectively testing current models in order to predict the outcome of a PUL and minimizing follow-up.

Query-driven module discovery in microarray data

MOTIVATION

Existing (bi)clustering methods for microarray data analysis often do not answer the specific questions of interest to a biologist. Such specific questions could be derived from other information sources, including expert prior knowledge. More specifically, given a set of seed genes which are believed to have a common function, we would like to recruit genes with similar expression profiles as the seed genes in a significant subset of experimental conditions.

RESULTS

We introduce QDB, a novel Bayesian query-driven biclustering framework in which the prior distributions allow introducing knowledge from a set of seed genes (query) to guide the pattern search. In two well-known yeast compendia, we grow highly functionally enriched biclusters from small sets of seed genes using a resolution sweep approach. In addition, relevant conditions are identified and modularity of the biclusters is demonstrated, including the discovery of overlapping modules. Finally, our method deals with missing values naturally, performs well on artificial data from a recent biclustering benchmark study and has a number of conceptual advantages when compared to existing approaches for focused module search.

miniTUBA: medical inference by network integration of temporal data using Bayesian analysis

MOTIVATION

Many biomedical and clinical research problems involve discovering causal relationships between observations gathered from temporal events. Dynamic Bayesian networks are a powerful modeling approach to describe causal or apparently causal relationships, and support complex medical inference, such as future response prediction, automated learning, and rational decision making. Although many engines exist for creating Bayesian networks, most require a local installation and significant data manipulation to be practical for a general biologist or clinician. No software pipeline currently exists for interpretation and inference of dynamic Bayesian networks learned from biomedical and clinical data.

RESULTS

miniTUBA is a web-based modeling system that allows clinical and biomedical researchers to perform complex medical/clinical inference and prediction using dynamic Bayesian network analysis with temporal datasets. The software allows users to choose different analysis parameters (e.g. Markov lags and prior topology), and continuously update their data and refine their results. miniTUBA can make temporal predictions to suggest interventions based on an automated learning process pipeline using all data provided. Preliminary tests using synthetic data and laboratory research data indicate that miniTUBA accurately identifies regulatory network structures from temporal data.

AVAILABILITY

miniTUBA is available at http://www.minituba.org.

Highly specific and sensitive rise in Days 14–17 pro-αC inhibin with clinical pregnancy after frozen embryo transfer with ovulatory cycles

BACKGROUND

Pro-alphaC inhibins are luteal derived analytes peaking in the maternal serum as early as Day 16 after conception. We set out to verify a previous post hoc analysis which suggested that pro-alphaC levels measured this early are extremely sensitive in predicting clinical pregnancy success after unstimulated IVF with ovulatory cycles.

METHODS

Prospective observational study of 246 women undergoing frozen embryo transfer with ovulatory cycles. Serum pro-alphaC and beta-HCG levels at 14-17 days after conception were measured by enzyme-linked immunosorbent assay and grouped according to whether a clinical pregnancy occurred (demonstrable cardiac activity at > or =6 weeks' gestation).

RESULTS

Of 34 (13.8%) women who achieved a clinical pregnancy, median (25th-75th centile) Days 14-17 pro-alphaC levels were 995 pg/ml (758-1463), 6- to 7-fold higher than levels observed in the remainder who did not fall pregnant (112.8 pg/ml (104-121); P < 0.0001). At a fixed 95% specificity, pro-alphaC was 100% sensitive in predicting clinical pregnancy. The best specificities achieved at 100% sensitivity were; 94.8% for pro-alphaC, 96.7% for beta-HCG and 98.1% when both analytes were combined.

CONCLUSIONS

Clinical pregnancy is always associated with a release of luteal derived pro-alphaC 14-17 days after conception. Pro-alphaC may play a possible biological role and be a useful clinical biomarker of luteal health.

Clinical information does not improve the performance of mathematical models in predicting the outcome of pregnancies of unknown location

OBJECTIVE(S)

To see if the incorporation of clinical variables can improve the diagnostic performance of logistic regression models in the prediction of pregnancy of unknown location (PUL) outcome.

DESIGN

Prospective observational study.

SETTING

Early Pregnancy Unit, St George's Hospital, University of London, London.

PATIENT(S)

All women with a PUL were included in the final analysis. This was defined on transvaginal ultrasonography (TVS) as there being no signs of either an intra- or an extrauterine pregnancy or retained products of conception in a woman with a positive pregnancy test.

INTERVENTION(S)

Noninterventional study; all women classified with a PUL were managed expectantly.

MAIN OUTCOMES MEASURE(S)

Data were collected prospectively from women classified as having a PUL. More than 30 clinical, ultrasonographic and biochemical end points were defined and recorded for analysis (these included risk factors for ectopic pregnancy [EP] and site-specific tenderness on TVS). Women were followed until the final diagnosis was established: failing PUL, intrauterine pregnancy (IUP), or EP. A multinomial logistic regression model (M5) was developed on 197 training cases and tested prospectively on a further 173 PUL cases. The performance of M5 was evaluated using receiver operating characteristic (ROC) curves and compared with logistic regression model M4 (hCG ratio [hCG 48 h/hCG 0 h], logarithm [log] of hCG average, and quadratic effect of the hCG ratio {[hCG ratio-1.17] x [hCG ratio-1.17]}), which was previously published.

RESULT(S)

Data from 376 consecutive women with a PUL were included in the analysis: 201 in the training set (109 [55.3%] failing PUL, 76 [38.6%] IUP, and 12 [6.1%] EP; four with a persisting PUL were excluded from the analysis) and 175 in the test set (94 [54.3%] with a failing PUL, 64 [37.0%] with an IUP, and 15 [8.7%] with an ectopic pregnancy; two with a persisting PUL were excluded from analysis). The most useful independent prognostic variables for the logistic regression model, M5, were as follows: log of serum hCG average, amount of vaginal bleeding, hCG ratio, and quadratic effect of the hCG ratio. On the test set, this model gave an area under the ROC curve of 0.979 for failing PUL, 0.979 for IUP, and 0.912 for EP. This model outperformed M4, which gave areas under the ROC curve of 0.978, 0.974, and 0.900, respectively; however, this was not significant.

CONCLUSION(S)

Clinical information does not significantly improve the performance of logistic regression models in the prediction of PUL outcome. On the basis of our results, we believe that historical, examination, and ultrasonographic factors are not essential input variables in logistic regression model building in the PUL population. When approaching women with a PUL, biochemical data alone, and in particular the hCG ratio, can be used to predict PUL outcome with a high degree of certainty.

The nonsurgical management of ectopic pregnancy

PURPOSE OF REVIEW

This review discusses the diagnosis and nonsurgical management of ectopic pregnancy.

RECENT FINDINGS

In the majority of cases the diagnosis of ectopic pregnancy should be made on transvaginal ultrasonography. Those for which the diagnosis is not made on the first scan may initially be classified as pregnancies of unknown location. There are now a number of strategies and mathematical models to predict ectopic pregnancy in this pregnancy of unknown location population. Reported success rates for expectant and medical management of ectopic pregnancy vary due to different inclusion criteria. A number of predictors of success have been studied: maternal age, previous obstetric history, gestational age, ultrasound features, human chorionic gonadotrophin levels, progesterone levels and the change in human chorionic gonadotrophin over time. At present the initial human chorionic gonadotrophin level probably remains the single most important predictor of success. Nonsurgical management is also particularly important for nontubal ectopic pregnancies: interstitial, cervical and caesarean section scar pregnancies.

SUMMARY

The majority of ectopic pregnancies can be visualized by ultrasound and so can be considered for conservative treatment. Nonsurgical management can be safe and effective. Appropriate selection criteria remain an issue, however, and a consensus needs to be reached on the predictors of success and failure to optimize management.