PREDICT-GTN 2: Two-factor streamlined models match FIGO performance in gestational trophoblastic neoplasia.

OBJECTIVE: The International Federation of Gynecology and Obstetrics (FIGO) scoring system uses the sum of eight risk-factors to predict single-agent chemotherapy resistance in Gestational Trophoblastic Neoplasia (GTN). To improve ease of use, this study aimed to generate: (i) streamlined models that match FIGO performance and; (ii) visual-decision aids (nomograms) for guiding management. METHODS: Using training (n = 4191) and validation datasets (n = 144) of GTN patients from two UK specialist centres, logistic regression analysis generated two-factor models for cross-validation and exploration. Performance was assessed using true and false positive rate, positive and negative predictive values, Bland-Altman calibration plots, receiver operating characteristic (ROC) curves, decision-curve analysis (DCA) and contingency tables. Nomograms were developed from estimated model parameters and performance cross-checked upon the training and validation dataset. RESULTS: Three streamlined, two-factor models were selected for analysis: (i) M1, pre-treatment hCG + history of failed chemotherapy; (ii) M2, pre-treatment hCG + site of metastases and; (iii) M3, pre-treatment hCG + number of metastases. Using both training and validation datasets, these models showed no evidence of significant discordance from FIGO (McNemar's test p > 0.78) or across a range of performance parameters. This behaviour was maintained when applying algorithms simulating the logic of the nomograms. CONCLUSIONS: Our streamlined models could be used to assess GTN patients and replace FIGO, statistically matching performance. Given the importance of imaging parameters in guiding treatment, M2 and M3 are favoured for ongoing validation. In resource-poor countries, where access to specialist centres is problematic, M1 could be pragmatically implemented. Further prospective validation on a larger cohort is recommended.

PREDICT-GTN 1: Can we improve the FIGO scoring system in gestational trophoblastic neoplasia?

Gestational trophoblastic neoplasia (GTN) patients are treated according to the eight-variable International Federation of Gynaecology and Obstetrics (FIGO) scoring system, that aims to predict first-line single-agent chemotherapy resistance. FIGO is imperfect with one-third of low-risk patients developing disease resistance to first-line single-agent chemotherapy. We aimed to generate simplified models that improve upon FIGO. Logistic regression (LR) and multilayer perceptron (MLP) modelling (n = 4191) generated six models (M1-6). M1, all eight FIGO variables (scored data); M2, all eight FIGO variables (scored and raw data); M3, nonimaging variables (scored data); M4, nonimaging variables (scored and raw data); M5, imaging variables (scored data); and M6, pretreatment hCG (raw data) + imaging variables (scored data). Performance was compared to FIGO using true and false positive rates, positive and negative predictive values, diagnostic odds ratio, receiver operating characteristic (ROC) curves, Bland-Altman calibration plots, decision curve analysis and contingency tables. M1-6 were calibrated and outperformed FIGO on true positive rate and positive predictive value. Using LR and MLP, M1, M2 and M4 generated small improvements to the ROC curve and decision curve analysis. M3, M5 and M6 matched FIGO or performed less well. Compared to FIGO, most (excluding LR M4 and MLP M5) had significant discordance in patient classification (McNemar's test P < .05); 55-112 undertreated, 46-206 overtreated. Statistical modelling yielded only small gains over FIGO performance, arising through recategorisation of treatment-resistant patients, with a significant proportion of under/overtreatment as the available data have been used a priori to allocate primary chemotherapy. Streamlining FIGO should now be the focus.

Computed tomography chest imaging offers no advantage over chest X-ray in the initial assessment of gestational trophoblastic neoplasia.

BACKGROUND: The International Federation of Gynaecology and Obstetrics (FIGO) score identifies gestational trophoblastic neoplasia (GTN) patients as low- or high-risk of single-agent chemotherapy resistance (SACR). Computed tomography (CT) has greater sensitivity than chest X-ray (CXR) in detecting pulmonary metastases, but effects upon outcomes remain unclear. METHODS: Five hundred and eighty-nine patients underwent both CXR and CT during GTN assessment. Treatment decisions were CXR based. The number of metastases, risk scores, and risk category using CXR versus CT were compared. CT-derived chest assessment was evaluated as impact upon treatment decision compared to patient outcome, incidence of SACR, time-to-normal human chorionic gonadotrophin hormone (TNhCG), and primary chemotherapy resistance (PCR). RESULTS: Metastasis detection (p < 0.0001) and FIGO score (p = 0.001) were higher using CT versus CXR. CT would have increased FIGO score in 188 (31.9%), with 43 re-classified from low- to high-risk, of whom 23 (53.5%) received curative single-agent chemotherapy. SACR was higher when score (p = 0.044) or risk group (p < 0.0001) changed. Metastases on CXR (p = 0.019) but not CT (p = 0.088) lengthened TNhCG. Logistic regression analysis found no difference between CXR (area under the curve (AUC) = 0.63) versus CT (AUC = 0.64) in predicting PCR. CONCLUSIONS: CT chest would improve the prediction of SACR, but does not influence overall treatment outcome, TNhCG, or prediction of PCR. Lower radiation doses and cost mean ongoing CXR-based assessment is recommended.

Use of salivary cortisol and cortisone in the high- and low-dose synacthen test.

CONTEXT: Salivary cortisone reflects serum cortisol levels, is more sensitive than salivary cortisol at lower values of serum cortisol and is noninvasive. OBJECTIVE: To investigate the relationship between serum cortisol and salivary cortisol and cortisone following low- and high-dose synacthen. DESIGN AND SETTING: Prospective pharmacodynamic studies in clinical research facilities. PARTICIPANTS AND INTERVENTION: Thirty-five dexamethasone-suppressed, healthy adult males underwent an intravenous synacthen test: N = 23 low-dose (1 mcg), N = 12 high-dose (250 mcg). Paired serum and salivary samples were taken at 15 sampling points over 120 minutes. MAIN OUTCOME MEASURE: Serum cortisol and salivary cortisol and cortisone were analysed for correlations and by a mixed-effects model. RESULTS: At baseline, the correlation between serum cortisol and salivary cortisol was weak with many samples undetectable (r = .45, NS), but there was a strong correlation with salivary cortisone (r = .94, P < .001). Up to 50 minutes following synacthen, the correlation coefficient between serum cortisol and salivary cortisol and cortisone was <0.8, but both had a stronger correlation at 60 minutes (salivary cortisol r = .89, P < .001, salivary cortisone r = .85, P < .001). The relationship was examined excluding samples in the dynamic phase (baseline to 60 minutes). Salivary cortisol and cortisone showed a close relationship to serum cortisol. Salivary cortisone showed the stronger correlation: salivary cortisol r = .82, P < .001, salivary cortisone r = .96, P < .001. CONCLUSION: Following synacthen, both salivary cortisol and cortisone reflect serum cortisol levels, but there is a lag in their rise up to 60 minutes. The results support further research for possible future use of a 60-minute salivary cortisone measurement during the synacthen test.

Much of the variation in breast pathology quality assurance data in the UK can be explained by the random order in which cases arrive at individual centres, but some true outliers do exist.

AIMS: To investigate the role of random temporal order of patient arrival at screening centres in the variability seen in rates of node positivity and breast cancer grade between centres in the NHS Breast Screening Programme. METHODS AND RESULTS: Computer simulations were performed of the variation in node positivity and breast cancer grade with the random temporal arrival of patients at screening centres based on national UK audit data. Cumulative mean graphs of these data were plotted. Confidence intervals for the parameters were generated, using the binomial distribution. UK audit data were plotted on these control limit graphs. The results showed that much of the variability in the audit data could be accounted for by the effects of random order of arrival of cases at the screening centres. Confidence intervals of 99.7% identified true outliers in the data. CONCLUSIONS: Much of the variation in breast pathology quality assurance data in the UK can be explained by the random order in which cases arrive at individual centres. Control charts with confidence intervals of 99.7% plotted against the number of reported cases are useful tools for identification of true outliers.

Laboratory assessment of iron status in pregnancy.

BACKGROUND: Efforts to improve maternal nutrition during pregnancy prompted an observational study of the occurrence of maternal iron deficiency and its laboratory diagnosis in almost 500 pregnancies. METHODS: In this longitudinal study, the biochemical and haematological iron indices of women (n=492) attending a prenatal clinic in a Dublin maternity hospital were assessed at first booking (mean 15.9 weeks), and after 24 weeks, and 36 weeks of gestation. Full blood counts were measured. Serum ferritin (SF), zinc protoporphyrin (ZPP), and transferrin receptor (sTfR) concentrations were assayed and transferrin receptor index (sTfR-Index) was calculated. The occurrence of low values and their diagnostic values were considered. RESULTS: A high occurrence iron deficiency (ID) at first booking (SF<12 µg/L) had increased over six-fold by 24 weeks, and all biochemical iron indices reflected progressive iron depletion right up to term. The WHO recommended anaemia "cut-off" (Hb<110 g/L) was insensitive to biochemical iron deficiency at booking, missing over 90% of the low SF values (SF<12 µg/L) which were mostly associated with much higher Hb levels. CONCLUSIONS: This study stresses the importance of including a biochemical index of iron status in prenatal screening and supports SF as the best indicator of biochemical ID overall. sTfR was insensitive to iron deficiency in early pregnancy, whereas the sTfR-Index, as a ratio, has the potential to distinguish between ID and physiological anaemia, and may offer stability in the assessment of iron stores from early pregnancy to full term. A policy of early screening of both Hb and SF concentrations is recommended as the minimum requirement for surveillance of maternal iron status in pregnancy.

Salivary Cortisone to Estimate Cortisol Exposure and Sampling Frequency Required Based on Serum Cortisol Measurements.

Context: Population studies frequently measure cortisol as a marker of stress, and excess cortisol is associated with increased mortality. Cortisol has a circadian rhythm, and frequent blood sampling is impractical to assess cortisol exposure. We investigated measuring salivary cortisone and examined the sampling frequency required to determine cortisol exposure. Methods: Serum and saliva with cortisol and cortisone were measured by liquid chromatography-tandem mass spectrometry in independent cohorts. The relationship between serum cortisol and salivary cortisone was analyzed in cohort 1 using a linear mixed effects model. The resulting fixed effects component was applied to cohort 2. Saliva cannot easily be collected when a patient is sleeping, so we determined the minimum sampling required to estimate cortisol exposure [estimated area under the curve (eAUC)] using 24-hour cortisol profiles (AUC24) and calculated the relative error (RE) for eAUC. Results: More than 90% of variability in salivary cortisone could be accounted for by change in serum cortisol. A single serum cortisol measurement was a poor estimate of AUC24, especially in the morning or last thing at night (RE >68%); however, three equally spaced samples gave a median RE of 0% (interquartile range, -15.6% to 15.1%). In patients with adrenal incidentalomas, eAUC based on three serum cortisol samples showed a difference between those with autonomous cortisol secretion and those without (P = 0.03). Interpretation: Accepting that most people sleep 7 to 8 hours, ∼8-hourly salivary cortisone measurements provide a noninvasive method of estimating 24-hour cortisol exposure for population studies.

Automatic covariate selection in logistic models for chest pain diagnosis: a new approach.

A newly established method for optimizing logistic models via a minorization-majorization procedure is applied to the problem of diagnosing acute coronary syndromes (ACS). The method provides a principled approach to the selection of covariates which would otherwise require the use of a suboptimal method owing to the size of the covariate set. A strategy for building models is proposed and two models optimized for performance and for simplicity are derived via 10-fold cross-validation. These models confirm that a relatively small set of covariates including clinical and electrocardiographic features can be used successfully in this task. The performance of the models is comparable with previously published models using less principled selection methods. The models prove to be portable when tested on data gathered from three other sites. Whilst diagnostic accuracy and calibration diminishes slightly for these new settings, it remains satisfactory overall. The prospect of building predictive models that are as simple as possible for a required level of performance is valuable if data-driven decision aids are to gain wide acceptance in the clinical situation owing to the need to minimize the time taken to gather and enter data at the bedside.

Technology for Early Detection of Depression and Anxiety in Older People.

Under-diagnosis of depression and anxiety is common in older adults. This project took a mixed methods approach to explore the application of machine learning and technology for early detection of these conditions. Mood measures collected with digital technologies were used to predict depression and anxiety status according to the Geriatric Depression Scale (GDS) and the Hospital Anxiety and Depression Scale (HADS). Interactive group activities and interviews were used to explore views of older adults and healthcare professionals on this approach respectively. The results show good potential for using a machine learning approach with mood data to predict later depression, though prospective results are preliminary. Qualitative findings highlight motivators and barriers to use of mental health technologies, as well as usability issues. If consideration is given to these issues, this approach could allow alerts to be provided to healthcare staff to draw attention to service users who may go on to experience depression.

Resetting the Abnormal Circadian Cortisol Rhythm in Adrenal Incidentaloma Patients With Mild Autonomous Cortisol Secretion.

Context: Adrenal incidentalomas (AIs) are found commonly on axial imaging. Around 30% exhibit autonomous cortisol secretion (ACS) associated with increased cardiovascular events and death. Objective: We hypothesized that AI/ACS patients have an abnormal cortisol rhythm that could be reversed by use of carefully timed short-acting cortisol synthesis blockade, with improvement in cardiovascular disease markers. Design, Setting, and Participants: In a phase 1/2a, prospective study (Eudract no. 2012-002586-35), we recruited six patients with AI/ACS and two control groups of six sex-, age-, and body mass index-matched individuals: (1) patients with AI and no ACS (AI/NoACS) and (2) healthy volunteers with no AI [healthy controls (HC)]. Twenty-four-hour circadian cortisol analysis was performed to determine any differences between groups and timing of intervention for cortisol lowering using the 11ß-hydroxylase inhibitor metyrapone. Circadian profiles of serum interleukin-6 (IL-6) were assessed. Results: Serum cortisol levels in group AI/ACS were significantly higher than both group AI/NoACS and group HC from 6 pm to 10 pm [area under the curve (AUC) difference: 0.81 nmol/L/h; P = 0.01] and from 10 pm to 2 am (AUC difference: 0.86 nmol/L/h; P < 0.001). In light of these findings, patients with ACS received metyrapone 500 mg at 6 pm and 250 mg at 10 pm, and cortisol rhythms were reassessed. Postintervention evening serum cortisol was lowered, similar to controls [6 pm to 10 pm (AUC difference: -0.06 nmol/L/h; P = 0.85); 10 pm to 2 am (AUC difference: 0.10 nmol/L/h; P = 0.76)]. Salivary cortisone showed analogous changes. IL-6 levels were elevated before treatment [10 pm to 2 pm (AUC difference: 0.42 pg/mL/h; P = 0.01)] and normalized post treatment. Conclusions: In AI/ACS, the evening and nocturnal cortisol exposure is increased. Use of timed evening doses of metyrapone resets the cortisol rhythm to normal. This unique treatment paradigm is associated with a reduction in the cardiovascular risk marker IL-6.

Salivary Cortisone Reflects Cortisol Exposure Under Physiological Conditions and After Hydrocortisone.

In this study we tested the use of salivary cortisol and cortisone as alternatives to serum cortisol. Salivary cortisol is often undetectable and contaminated by hydrocortisone. Salivary cortisone strongly reflects serum cortisol.

Artificial neural network models for prediction of acute coronary syndromes using clinical data from the time of presentation.

STUDY OBJECTIVE: Clinical and ECG data from presentation are highly discriminatory for diagnosis of acute coronary syndromes, whereas definitive diagnosis from serial ECG and cardiac marker protein measurements is usually not available for several hours. Artificial neural networks are computer programs adept at pattern recognition tasks and have been used to analyze data from chest pain patients with a view to developing diagnostic algorithms that might improve triage practices in the emergency department. The aim of this study is to develop and optimize artificial neural network models for diagnosis of acute coronary syndrome, to test these models on data collected prospectively from different centers, and to establish whether the performance of these models was superior to that of models derived using a standard statistical technique, logistic regression. METHODS: The study used data from 3,147 patients presenting to 3 hospitals with acute chest pain. Data from hospital 1 were used to train the models, which were then tested on independent data from the other 2 hospitals. From 40 potential factors, variables were selected according to the logarithm of their likelihood ratios to produce models using 8, 13, 20, and 40 factors. Identical data were used for logistic regression and artificial neural network models. Calibration and performance were assessed, the latter using receiver operating characteristic (ROC) curve analysis. RESULTS: Although the performance of artificial neural network models generally increased with increasing numbers of factors, this was insignificant. The 13-factor model was therefore used for the rest of the study owing to its marginally improved calibration over the smallest model. Area under the ROC curve (with standard error) was 0.97 (0.006). The overall sensitivity and specificity of this model for acute coronary syndrome diagnosis using the training data was 0.93. ROC curves for logistic regression and artificial neural network models applied to data from the 3 hospitals were identical. For the 13-factor artificial neural network model tested on data from hospitals 2 and 3, area under the ROC curves (standard error) were 0.93 (0.006) and 0.95 (0.009), respectively. Investigation of the performance of the artificial neural network models throughout the range of predicted probabilities showed that they were well calibrated. CONCLUSION: This study confirms that artificial neural networks can offer a useful approach for developing diagnostic algorithms for chest pain patients; however, the exceptional performance and simplicity of the logistic model militates in favor of logistic regression for the present task. Our artificial neural network models were well calibrated and performed well on unseen data from different centers. These issues have not been addressed in previous studies. However, and unlike in previous studies, we did not find the performance of artificial neural network models to be significantly different from that of suitably optimized logistic regression models.

Hormonal circadian rhythms in patients with congenital adrenal hyperplasia: identifying optimal monitoring times and novel disease biomarkers.

OBJECTIVES: The treatment goal in congenital adrenal hyperplasia (CAH) is to replace glucocorticoids while avoiding androgen excess and iatrogenic Cushing's syndrome. However, there is no consensus on how to monitor disease control. Our main objectives were to evaluate hormonal circadian rhythms and use these profiles to identify optimal monitoring times and novel disease biomarkers in CAH adults on intermediate- and long-acting glucocorticoids. DESIGN: This was an observational, cross-sectional study at the National Institutes of Health Clinical Center in 16 patients with classic CAH. METHODS: Twenty-four-hour serum sampling for ACTH, 17-hydroxyprogesterone (17OHP), androstenedione (A4), androsterone, DHEA, testosterone, progesterone and 24-h urinary pdiol and 5ß-pdiol was carried out. Bayesian spectral analysis and cosinor analysis were performed to detect circadian rhythmicity. The number of hours to minimal (TminAC) and maximal (TmaxAC) adrenocortical hormone levels after dose administration was calculated. RESULTS: A significant rhythm was confirmed for ACTH (r(2), 0.95; P<0.001), 17OHP (r(2), 0.70; P=0.003), androstenedione (r(2), 0.47; P=0.043), androsterone (r(2), 0.80; P<0.001), testosterone (r(2), 0.47; P=0.042) and progesterone (r(2), 0.64; P=0.006). The mean (s.d.) TminAC and TmaxAC for 17OHP and A4 were: morning prednisone (4.3 (2.3) and 9.7 (3.5) h), evening prednisone (4.5 (2.0) and 10.3 (2.4) h), and daily dexamethasone (9.2 (3.5) and 16.4 (7.2) h). AUC0-24 h progesterone, androsterone and 24-h urine pdiol were significantly related to 17OHP. CONCLUSION: In CAH patients, adrenal androgens exhibit circadian rhythms influenced by glucocorticoid replacement. Measurement of adrenocortical hormones and interpretation of results should take into account the type of glucocorticoid and time of dose administration. Progesterone and backdoor metabolites may provide alternative disease biomarkers.

The development of IVF practice in Ireland: a personal view.

This paper traces the development of in vitro fertilization (IVF) in the Republic of Ireland from when the first attempts at using this technique were carried out in 1985 up to the present. Clinical changes are chronicled principally using the personal work of the author and his colleagues as representative of the day. The impact of the Catholic Church and the alterations in Medical Council governance guidelines over the years as these reflect societal changes are highlighted. The potential role of other regulators including Irish case law and the EU Tissue directive are discussed as well as the almost invariable private practice nature of the services provided and the various ways in which costs have been alleviated.

Psychosocial trends in couples prior to commencement of in vitro fertilisation (IVF) treatment.

Facing infertility and undergoing fertility treatment can create emotional turmoil in couples' lives. It is essential for fertility therapy providers to assess the coping and communication strategies of couples before treatment in order to provide appropriate support. We performed a two time point (year 2003 & year 2009) cross-sectional study of patients attending our services to undergo in vitro fertilisation. All couples attending the Human Assisted Reproduction Ireland Unit, a tertiary referral academic centre at the Rotunda Hospital, were requested to complete a psychosocial questionnaire before commencing the treatment. The questions assessed couples' understanding of their own infertility, family background and support, relationship traits and stress levels prior to commencing fertility treatment. A total of 180 patients participated in the study. Our findings showed that within a 6-year time span, couples' attitudes have changed significantly. Compared to 6 years ago, couples now have a better understanding of infertility and are seeking treatment quicker. Interestingly, we showed higher stress levels nowadays with fewer couples following routine stress management. We also identified specific gender differences in that women have a more open attitude in discussing infertility and seeking more support from friends and family compared to men.

Evaluation of machine-learning methods for ligand-based virtual screening.

Machine-learning methods can be used for virtual screening by analysing the structural characteristics of molecules of known (in)activity, and we here discuss the use of kernel discrimination and naive Bayesian classifier (NBC) methods for this purpose. We report a kernel method that allows the processing of molecules represented by binary, integer and real-valued descriptors, and show that it is little different in screening performance from a previously described kernel that had been developed specifically for the analysis of binary fingerprint representations of molecular structure. We then evaluate the performance of an NBC when the training-set contains only a very few active molecules. In such cases, a simpler approach based on group fusion would appear to provide superior screening performance, especially when structurally heterogeneous datasets are to be processed.

Virtual screening using binary kernel discrimination: analysis of pesticide data.

This paper discusses the use of binary kernel discrimination (BKD) for identifying potential active compounds in lead-discovery programs. BKD was compared with established virtual screening methods in a series of experiments using pesticide data from the Syngenta corporate database. It was found to be superior to methods based on similarity searching and substructural analysis but inferior to a support vector machine. Similar conclusions resulted from application of the methods to a pesticide data set for which categorical activity data were available.

Virtual screening using binary kernel discrimination: effect of noisy training data and the optimization of performance.

Binary kernel discrimination (BKD) uses a training set of compounds, for which structural and qualitative activity data are available, to produce a model that can then be applied to the structures of other compounds in order to predict their likely activity. Experiments with the MDL Drug Data Report database show that the optimal value of the smoothing parameter, and hence the predictive power of BKD, is crucially dependent on the number of false positives in the training set. It is also shown that the best results for BKD are achieved using one particular optimization method for the determination of the smoothing parameter that lies at the heart of the method and using the Jaccard/Tanimoto coefficient in the kernel function that is used to compute the similarity between a test set molecule and the members of the training set.

Modified Fuzzy ARTMAP Approaches Bayes Optimal Classification Rates: An Empirical Demonstration.

This paper investigates the effectiveness of the Fuzzy ARTMAP (FAM) neural network in classifying statistical data and compares the results with Bayesian decision theory. Binary classification problems are used to assess the performance of FAM operating autonomously and on-line in statistical settings. The results illustrate the limitations of FAM in this context. Novel modifications are, therefore, proposed for the category formation process and the category selection process of FAM, which allow the modified system to minimize the misclassification rates. A number of simulations with randomly generated data sets have been carried out. First, two continuous-valued Gaussian sources are used with various source (mean) separations, prior probabilities, and variances. Then, multi-dimensional discrete patterns are employed to examine the classification ability of modified FAM in both stationary and non-stationary environments. Simulation results consistently demonstrate that modified FAM is able to approach the Bayes optimal classification rates on-line, and thereby justify the rationale behind the modifications. Copyright 1997 Elsevier Science Ltd.