Kukatharmini Tharmaratnam

Cancer incidence and survival in Lynch syndrome patients receiving colonoscopic and gynaecological surveillance: first report from the prospective Lynch syndrome database.

Objective Estimates of cancer risk and the effects of surveillance in Lynch syndrome have been subject to bias, partly through reliance on retrospective studies. We sought to establish more robust estimates in patients undergoing prospective cancer surveillance. Design We undertook a multicentre study of patients carrying Lynch syndrome-associated mutations affecting MLH1, MSH2, MSH6 or PMS2. Standardised information on surveillance, cancers and outcomes were collated in an Oracle relational database and analysed by age, sex and mutated gene. Results 1942 mutation carriers without previous cancer had follow-up including colonoscopic surveillance for 13 782 observation years. 314 patients developed cancer, mostly colorectal (n=151), endometrial (n=72) and ovarian (n=19). Cancers were detected from 25 years onwards in MLH1 and MSH2 mutation carriers, and from about 40 years in MSH6 and PMS2 carriers. Among first cancer detected in each patient the colorectal cancer cumulative incidences at 70 years by gene were 46%, 35%, 20% and 10% for MLH1, MSH2, MSH6 and PMS2 mutation carriers, respectively. The equivalent cumulative incidences for endometrial cancer were 34%, 51%, 49% and 24%; and for ovarian cancer 11%, 15%, 0% and 0%. Ten-year crude survival was 87% after any cancer, 91% if the first cancer was colorectal, 98% if endometrial and 89% if ovarian. Conclusions The four Lynch syndrome-associated genes had different penetrance and expression. Colorectal cancer occurred frequently despite colonoscopic surveillance but resulted in few deaths. Using our data, a website has been established at http://LScarisk.org enabling calculation of cumulative cancer risks as an aid to genetic counselling in Lynch syndrome.

Incidence of and survival after subsequent cancers in carriers of pathogenic MMR variants with previous cancer: a report from the prospective Lynch syndrome database

Objective Today most patients with Lynch syndrome (LS) survive their first cancer. There is limited information on the incidences and outcome of subsequent cancers. The present study addresses three questions: (i) what is the cumulative incidence of a subsequent cancer; (ii) in which organs do subsequent cancers occur; and (iii) what is the survival following these cancers? Design Information was collated on prospectively organised surveillance and prospectively observed outcomes in patients with LS who had cancer prior to inclusion and analysed by age, gender and genetic variants. Results 1273 patients with LS from 10 countries were followed up for 7753 observation years. 318 patients (25.7%) developed 341 first subsequent cancers, including colorectal (n=147, 43%), upper GI, pancreas or bile duct (n=37, 11%) and urinary tract (n=32, 10%). The cumulative incidences for any subsequent cancer from age 40 to age 70 years were 73% for pathogenic MLH1 (path_MLH1), 76% for path_MSH2 carriers and 52% for path_MSH6 carriers, and for colorectal cancer (CRC) the cumulative incidences were 46%, 48% and 23%, respectively. Crude survival after any subsequent cancer was 82% (95% CI 76% to 87%) and 10-year crude survival after CRC was 91% (95% CI 83% to 95%). Conclusions Relative incidence of subsequent cancer compared with incidence of first cancer was slightly but insignificantly higher than cancer incidence in patients with LS without previous cancer (range 0.94–1.49). The favourable survival after subsequent cancers validated continued follow-up to prevent death from cancer. The interactive website http://lscarisk.org was expanded to calculate the risks by gender, genetic variant and age for subsequent cancer for any patient with LS with previous cancer.

Cancer risk and survival in path_MMR carriers by gene and gender up to 75 years of age: a report from the Prospective Lynch Syndrome Database

Background Most patients with path_MMR gene variants (Lynch syndrome (LS)) now survive both their first and subsequent cancers, resulting in a growing number of older patients with LS for whom limited information exists with respect to cancer risk and survival. Objective and design This observational, international, multicentre study aimed to determine prospectively observed incidences of cancers and survival in path_MMR carriers up to 75 years of age. Results 3119 patients were followed for a total of 24 475 years. Cumulative incidences at 75 years (risks) for colorectal cancer were 46%, 43% and 15% in path_MLH1, path_MSH2 and path_MSH6 carriers; for endometrial cancer 43%, 57% and 46%; for ovarian cancer 10%, 17% and 13%; for upper gastrointestinal (gastric, duodenal, bile duct or pancreatic) cancers 21%, 10% and 7%; for urinary tract cancers 8%, 25% and 11%; for prostate cancer 17%, 32% and 18%; and for brain tumours 1%, 5% and 1%, respectively. Ovarian cancer occurred mainly premenopausally. By contrast, upper gastrointestinal, urinary tract and prostate cancers occurred predominantly at older ages. Overall 5-year survival for prostate cancer was 100%, urinary bladder 93%, ureter 85%, duodenum 67%, stomach 61%, bile duct 29%, brain 22% and pancreas 0%. Path_PMS2 carriers had lower risk for cancer. Conclusion Carriers of different path_MMR variants exhibit distinct patterns of cancer risk and survival as they age. Risk estimates for counselling and planning of surveillance and treatment should be tailored to each patient’s age, gender and path_MMR variant. We have updated our open-access website www.lscarisk.org to facilitate this.

Identification of predicted individual treatment effects in randomized clinical trials

In most medical research, treatment effectiveness is assessed using the average treatment effect or some version of subgroup analysis. The practice of individualized or precision medicine, however, requires new approaches that predict how an individual will respond to treatment, rather than relying on aggregate measures of effect. In this study, we present a conceptual framework for estimating individual treatment effects, referred to as predicted individual treatment effects. We first apply the predicted individual treatment effect approach to a randomized controlled trial designed to improve behavioral and physical symptoms. Despite trivial average effects of the intervention, we show substantial heterogeneity in predicted individual treatment response using the predicted individual treatment effect approach. The predicted individual treatment effects can be used to predict individuals for whom the intervention may be most effective (or harmful). Next, we conduct a Monte Carlo simulation study to evaluate the accuracy of predicted individual treatment effects. We compare the performance of two methods used to obtain predictions: multiple imputation and non-parametric random decision trees. Results showed that, on average, both predictive methods produced accurate estimates at the individual level; however, the random decision trees tended to underestimate the predicted individual treatment effect for people at the extreme and showed more variability in predictions across repetitions compared to the imputation approach. Limitations and future directions are discussed.

Monotone splines lasso

The important problems of variable selection and estimation in nonparametric additive regression models for high-dimensional data are addressed. Several methods have been proposed to model nonlinear relationships when the number of covariates exceeds the number of observations by using spline basis functions and group penalties. Nonlinear monotone effects on the response play a central role in many situations, in particular in medicine and biology. The monotone splines lasso (MS-lasso) is constructed to select variables and estimate effects using monotone splines (I-splines). The additive components in the model are represented by their I-spline basis function expansion and the component selection becomes that of selecting the groups of coefficients in the I-spline basis function expansion. A recent procedure, called cooperative lasso, is used to select sign-coherent groups, i.e. selecting the groups with either exclusively non-negative or non-positive coefficients. This leads to the selection of important covariates that have nonlinear monotone increasing or decreasing effect on the response. An adaptive version of the MS-lasso reduces both the bias and the number of false positive selections considerably. The MS-lasso and the adaptive MS-lasso are compared with other existing methods for variable selection in high dimensions by simulation and the methods are applied to two relevant genomic data sets. Results indicate that the (adaptive) MS-lasso has excellent properties compared to the other methods both in terms of estimation and selection, and can be recommended for high-dimensional monotone regression.

Tilting the lasso by knowledge-based post-processing

BackgroundIt is useful to incorporate biological knowledge on the role of genetic determinants in predicting an outcome. It is, however, not always feasible to fully elicit this information when the number of determinants is large. We present an approach to overcome this difficulty. First, using half of the available data, a shortlist of potentially interesting determinants are generated. Second, binary indications of biological importance are elicited for this much smaller number of determinants. Third, an analysis is carried out on this shortlist using the second half of the data.ResultsWe show through simulations that, compared with adaptive lasso, this approach leads to models containing more biologically relevant variables, while the prediction mean squared error (PMSE) is comparable or even reduced. We also apply our approach to bone mineral density data, and again final models contain more biologically relevant variables and have reduced PMSEs.ConclusionOur method leads to comparable or improved predictive performance, and models with greater face validity and interpretability with feasible incorporation of biological knowledge into predictive models.