Elizabeth Williamson | London School of Hygiene & Tropical Medicine (original) (raw)

Papers by Elizabeth Williamson

Statistical Methods in Medical Research, 2012

Estimation of the effect of a binary exposure on an outcome in the presence of confounding is oft... more Estimation of the effect of a binary exposure on an outcome in the presence of confounding is often carried out via outcome regression modelling. An alternative approach is to use propensity score methodology. The propensity score is the conditional probability of receiving the exposure given the observed covariates and can be used, under the assumption of no unmeasured confounders, to estimate the causal effect of the exposure. In this article, we provide a non-technical and intuitive discussion of propensity score methodology, motivating the use of the propensity score approach by analogy with randomised studies, and describe the four main ways in which this methodology can be implemented. We carefully describe the population parameters being estimated - an issue that is frequently overlooked in the medical literature. We illustrate these four methods using data from a study investigating the association between maternal choice to provide breast milk and the infant's subsequent neurodevelopment. We outline useful extensions of propensity score methodology and discuss directions for future research. Propensity score methods remain controversial and there is no consensus as to when, if ever, they should be used in place of traditional outcome regression models. We therefore end with a discussion of the relative advantages and disadvantages of each.

Pharmaceutical Statistics, 2015

In cost-effectiveness analyses of drugs or health technologies, estimates of life years saved or ... more In cost-effectiveness analyses of drugs or health technologies, estimates of life years saved or quality-adjusted life years saved are required. Randomised controlled trials can provide an estimate of the average treatment effect; for survival data, the treatment effect is the difference in mean survival. However, typically not all patients will have reached the endpoint of interest at the close-out of a trial, making it difficult to estimate the difference in mean survival. In this situation, it is common to report the more readily estimable difference in median survival. Alternative approaches to estimating the mean have also been proposed. We conducted a simulation study to investigate the bias and precision of the three most commonly used sample measures of absolute survival gain - difference in median, restricted mean and extended mean survival - when used as estimates of the true mean difference, under different censoring proportions, while assuming a range of survival patterns, represented by Weibull survival distributions with constant, increasing and decreasing hazards. Our study showed that the three commonly used methods tended to underestimate the true treatment effect; consequently, the incremental cost-effectiveness ratio (ICER) would be overestimated. Of the three methods, the least biased is the extended mean survival, which perhaps should be used as the point estimate of the treatment effect to be inputted into the ICER, while the other two approaches could be used in sensitivity analyses. More work on the trade-offs between simple extrapolation using the exponential distribution and more complicated extrapolation using other methods would be valuable. Copyright © 2015 John Wiley & Sons, Ltd.

Acupuncture in medicine : journal of the British Medical Acupuncture Society, 2015

ABSTRACT Drs White and Cummings state1 that we negatively reported findings of our ‘positive stud... more ABSTRACT Drs White and Cummings state1 that we negatively reported findings of our ‘positive study’—namely, our trial evaluating acupuncture for chronic knee osteoarthritis (OA).2 Our study was not a ‘positive study’ as they suggest. Rather, our study showed no benefit of needle or laser acupuncture compared with sham laser acupuncture on either of our two primary outcomes (pain and function), and small clinically irrelevant benefits compared with no acupuncture for both laser and needle acupuncture. In accordance with best practice in the conduct and interpretation of clinical trials, we (1) a priori specified our primary outcome measures, including the minimum clinically important difference (MCID) on the scale of measurement for each outcome measure; (2) determined and recruited the sample size needed to show meaningful between-group differences in primary outcomes based on our chosen MCIDs and estimated SDs with adequate power; and (3) interpreted our findings based on both of these primary outcomes, their MCIDs and considering both control conditions included in our study design (ie, sham laser acupuncture and no acupuncture). Accordingly, the conclusion stating that our findings do not support acupuncture for these patients is appropriate and is the only conclusion that can (and should) be drawn from our study. We powered our study to detect the absolute MCID between groups in pain (1.8 units) and function (6 units), which have been recommended specifically for our chosen primary …

International journal of cancer. Journal international du cancer, Jan 13, 2015

The associations between intake of or circulating fatty acids and risk of colorectal cancer (CRC)... more The associations between intake of or circulating fatty acids and risk of colorectal cancer (CRC) are unclear. We examined prospectively the associations between dietary or biomarker fatty acids and CRC. For 41,514 men and women, aged 40-69 years, baseline (1990-94) dietary intakes of fatty acids were estimated using a food frequency questionnaire and plasma phospholipid (PPL) fatty acids were measured for 4,205 participants including 395 CRC cases, according to a case-cohort design. Hazard ratios were computed using Cox regression adjusting for education, alcohol intake, smoking status, physical activity and total energy intake; and stratified for gender, ethnicity and family history of cancer, with age as the time scale. We assessed the heterogeneity of associations with colon and rectal cancers. PPL saturated fatty acids (SFAs) were positively associated with CRC risk, while total n-3 polyunsaturated fatty acids (PUFA) and long chain marine n-3 PUFAs showed inverse associations, ...

Recognition that vitamin D might be associated with many chronic diseases has led to large-scale ... more Recognition that vitamin D might be associated with many chronic diseases has led to large-scale epidemiological and clinical studies. Dried blood spots (DBS) are a useful resource for these studies. Consequently, accurate, efficient, and inexpensive assays to quantify 25-hydroxyvitamin D (25OHD) in DBS are required. This study evaluated the validity and reliability of a liquid chromatography-tandem mass spectrometry assay for measuring 25OHD in archived DBS and compared measurements of 25OHD in DBS with those in plasma. Sixty-two participants in the Melbourne Collaborative Cohort Study who had plasma and matching DBS stored since study entry in the early 1990s were randomly selected for a study calibrating 25OHD concentrations in DBS with plasma. As part of a study of vitamin D and mortality, cancer, and diabetes, we also assessed the reliability of measurements from DBS using 500 replicates placed randomly within 31 batches run over 15 months. 25OHD concentrations were measured by liquid chromatography-tandem mass spectrometry. There was good agreement between measurements of 25OHD from DBS and plasma; R(2) = 0.73 from a regression of plasma concentration on DBS concentration. The within-batch and between-batch intraclass correlations from the 500 replicate measurements were 0.82 (95% confidence interval, 0.80, 0.85) and 0.73 (95% confidence interval, 0.68, 0.78), respectively. Measuring 25OHD in DBS is a valid and reliable alternative to measuring 25OHD in sera or plasma. A simple calibration model was developed to convert measurements from DBS to equivalent plasma measurements, thus enabling comparisons against clinical reference ranges and with studies using sera or plasma samples.

Although randomization provides a gold-standard method of assessing causal relationships, it is n... more Although randomization provides a gold-standard method of assessing causal relationships, it is not always possible to randomly allocate exposures. Where exposures are not randomized, estimating exposure effects is complicated by confounding. The traditional approach to dealing with confounding is to adjust for measured confounding variables within a regression model for the outcome variable. An alternative approach--propensity scoring--instead fits a regression model to the exposure variable. For a binary exposure, the propensity score is the probability of being exposed, given the measured confounders. These scores can be estimated from the data, for example by fitting a logistic regression model for the exposure including the confounders as explanatory variables and obtaining the estimated propensity scores from the predicted exposure probabilities from this model. These estimated propensity scores can then be used in various ways-matching, stratification, covariate-adjustment or inverse-probability weighting-to obtain estimates of the exposure effect. In this paper, we provide an introduction to propensity score methodology and review its use within respiratory health research. We illustrate propensity score methods by investigating the research question: 'Does personal smoking affect the risk of subsequent asthma?' using data taken from the Tasmanian Longitudinal Health Study.

Human Pathology, 2014

Adenocarcinomas of the colon and rectum are graded using a 2-tiered system into histologic low-gr... more Adenocarcinomas of the colon and rectum are graded using a 2-tiered system into histologic low-grade and high-grade tumors based on the proportion of gland formation. The current grading system does not apply to subtypes of carcinomas associated with a high frequency of microsatellite instability (MSI), such as mucinous and medullary carcinomas. We investigated the combined effect of histologic grade and MSI status on survival for 738 patients with colorectal carcinoma (48% female; mean age at diagnosis 68.2 years). The proportion of high-grade adenocarcinoma was 18%. MSI was observed in 59 adenocarcinomas (9%), with higher frequency in high-grade tumors compared with low-grade tumors (20% versus 6%; P < .001). Using Cox regression models, adjusting for sex and age at diagnosis and stratifying by the American Joint Committee on Cancer stage, microsatellite stable (MSS) high-grade tumors were associated with increased hazard of all-cause and colorectal cancer-specific mortality: hazard ratio 2.09 (95% confidence interval [CI], 1.58-2.77) and 2.54 (95% CI, 1.86-3.47), respectively, both P < .001. A new grading system separating adenocarcinoma into low grade (all histologic low grade and MSI high grade) and high grade (MSS histologic high grade) gave a lower Akaike information criterion value when compared with the current grading system and thus represented a better model fit to stratify patients according to survival. We found that patients with a high-grade adenocarcinoma had significantly shorter survival than patients with low-grade adenocarcinoma only if the tumor was MSS, suggesting that the grading of colorectal adenocarcinoma with high-grade histologic features should be made according to the MSI status of the tumor.

Statistics in Medicine, 2013

In individually randomised controlled trials, adjustment for baseline characteristics is often un... more In individually randomised controlled trials, adjustment for baseline characteristics is often undertaken to increase precision of the treatment effect estimate. This is usually performed using covariate adjustment in outcome regression models. An alternative method of adjustment is to use inverse probability-of-treatment weighting (IPTW), on the basis of estimated propensity scores. We calculate the large-sample marginal variance of IPTW estimators of the mean difference for continuous outcomes, and risk difference, risk ratio or odds ratio for binary outcomes. We show that IPTW adjustment always increases the precision of the treatment effect estimate. For continuous outcomes, we demonstrate that the IPTW estimator has the same large-sample marginal variance as the standard analysis of covariance estimator. However, ignoring the estimation of the propensity score in the calculation of the variance leads to the erroneous conclusion that the IPTW treatment effect estimator has the same variance as an unadjusted estimator; thus, it is important to use a variance estimator that correctly takes into account the estimation of the propensity score. The IPTW approach has particular advantages when estimating risk differences or risk ratios. In this case, non-convergence of covariate-adjusted outcome regression models frequently occurs. Such problems can be circumvented by using the IPTW adjustment approach.

Respirology, 2014

In respiratory health research, interest often lies in estimating the effect of an exposure on a ... more In respiratory health research, interest often lies in estimating the effect of an exposure on a health outcome. If randomization of the exposure of interest is not possible, estimating its effect is typically complicated by confounding bias. This can often be dealt with by controlling for the variables causing the confounding, if measured, in the statistical analysis. Common statistical methods used to achieve this include multivariable regression models adjusting for selected confounding variables or stratification on those variables. Therefore, a key question is which measured variables need to be controlled for in order to remove confounding. An approach to confounder-selection based on the use of causal diagrams (often called directed acyclic graphs) is discussed. A causal diagram is a visual representation of the causal relationships believed to exist between the variables of interest, including the exposure, outcome and potential confounding variables. After creating a causal diagram for the research question, an intuitive and easy-to-use set of rules can be applied, based on a foundation of rigorous mathematics, to decide which measured variables must be controlled for in the statistical analysis in order to remove confounding, to the extent that is possible using the available data. This approach is illustrated by constructing a causal diagram for the research question: 'Does personal smoking affect the risk of subsequent asthma?'. Using data taken from the Tasmanian Longitudinal Health Study, the statistical analysis suggested by the causal diagram approach was performed.

Optometry and Vision Science, 2013

Population-based studies on abnormal binocular vision and low literacy are rare. The aim is to de... more Population-based studies on abnormal binocular vision and low literacy are rare. The aim is to determine the prevalence of poor stereoacuity among children with low literacy; to identify the characteristics associated with poor stereoacuity among children with low literacy; and to determine the agreement between poor stereoacuity as measured by graded stereocircles with a computerized assessment. A total of 490 children attending primary school in the greater Hobart region, Tasmania, aged 7 to 14 years, with literacy results below the 10th percentile for Tasmanian students at grade 3 level of the NAPLAN (National Assessment Program in Literacy and Numeracy) testing completed a vision screen. Poor stereoacuity was defined as more than 100 seconds of arc as measured by Titmus stereocircles. The prevalence of poor stereoacuity was 16.8% (95% confidence interval [95% CI], 13.6 to 20.4%). Children with poor stereoacuity had a higher frequency of symptom report using the Convergence Insufficiency Symptom Survey. Factors associated with poor stereopsis were prematurity (adjusted odds ratio [AOR], 2.19; 95% CI, 1.09 to 4.42) and bottom shuffling (AOR, 2.39; 95% CI, 1.17 to 4.88). Features associated with poor stereopsis included squint (AOR, 6.05; 95% CI, 3.02 to 12.12), migraine (AOR, 2.25; 95% CI, 1.05 to 4.83), and attention deficit disorder (AOR, 1.92; 95% CI, 1.01 to 3.65). In this low-literacy sample, one-sixth had low stereoacuity. The associations reported require further investigation.

Ophthalmic Epidemiology, 2009

To assess, among children with low literacy and poor stereoacuity, the efficacy of two interventi... more To assess, among children with low literacy and poor stereoacuity, the efficacy of two intervention programs on child vision and education compared to a control program. Eighty-eight children aged 8 to 13 years who had reading problems, and demonstrated poor stereoacuity as measured by the Titmus stereocircle test (> 100 seconds arc) or computerized assessment were randomized to one of two intervention programs: Lawson vision or Phono-Graphix, or a control group: Parental Literacy Support. Vision (Lang test, visual acuity, convergence insufficiency symptom survey) and education assessments (Woodcock Reading Mastery Tests-Revised) were conducted at baseline, intervention end (10 weeks), and 36 weeks. Analysis used intention to treat multi-level models. Compared to the parental literacy support group, convergence insufficiency symptoms were reduced 36 weeks post-randomization amongst those receiving the Lawson orthoptic intervention (mean difference -5.55; 95% confidence interval (CI): -11.1 to -0.05, P < 0.05). Stereoacuity, measured by the Lang test, improved for both the Lawson and Phono-Graphix interventions compared to the parental literacy support group (-1.01; 95% CI: -1.6 to -0.4, P = 0.001, and -0.77; 95% CI: -1.4 to -0.2, P = 0.01). At the 36 week follow-up assessment, word identification had also improved for the Lawson and Phono-Graphix groups but other educational outcomes did not improve. A formal randomized control trial was feasible in this setting. Intervention among children with poor stereoacuity and low literacy produced small improvements in stereopsis and convergence insufficiency symptom scores. Further randomized control trials should be conducted to clarify the role of orthoptic intervention on literacy in selected child populations.

Ophthalmic and Physiological Optics, 2006

Modern Pathology, 2013

Mucinous differentiation is associated with both CpG island methylator phenotype and microsatelli... more Mucinous differentiation is associated with both CpG island methylator phenotype and microsatellite instability in colorectal cancer. The mucinous phenotype derives from abundant expression of the colonic goblet cell mucin, MUC2, and de novo expression of gastric foveolar mucin, MUC5AC. We, therefore, investigated the protein expression levels of MUC2 and MUC5AC, as well as MUC5B and MUC6, in molecular subtypes of colorectal cancer. Seven-hundred and twenty-two incident colorectal carcinomas occurring in 702 participants of the Melbourne Collaborative Cohort Study were characterized for methylator status, MLH1 methylation, somatic BRAF and KRAS mutations, microsatellite-instability status, MLH1, MSH2, MSH6, and PMS2 mismatch repair, and p53 protein expression, and their histopathology was reviewed. Protein expression levels of MUC2, MUC5AC, MUC5B, MUC6, and the putative mucin regulator CDX2 were compared with molecular and clinicopathological features of colorectal cancers using odds ratios and corresponding 95% confidence intervals. MUC2 overexpression (>25% positive tumor cells) was observed in 33% colorectal cancers, MUC5B expression in 53%, and de novo MUC5AC and MUC6 expression in 50% and 39%, respectively. Co-expression of two or more of the mucins was commonly observed. Expression of MUC2, MUC5AC and MUC6 was strongly associated with features associated with tumorigenesis via the serrated neoplasia pathway, including methylator positivity, somatic BRAF p.V600E mutation, and mismatch repair deficiency, as well as proximal location, poor differentiation, lymphocytic response, and increased T stage (all…

Statistical Methods in Medical Research, 2012

Estimation of the effect of a binary exposure on an outcome in the presence of confounding is oft... more Estimation of the effect of a binary exposure on an outcome in the presence of confounding is often carried out via outcome regression modelling. An alternative approach is to use propensity score methodology. The propensity score is the conditional probability of receiving the exposure given the observed covariates and can be used, under the assumption of no unmeasured confounders, to estimate the causal effect of the exposure. In this article, we provide a non-technical and intuitive discussion of propensity score methodology, motivating the use of the propensity score approach by analogy with randomised studies, and describe the four main ways in which this methodology can be implemented. We carefully describe the population parameters being estimated - an issue that is frequently overlooked in the medical literature. We illustrate these four methods using data from a study investigating the association between maternal choice to provide breast milk and the infant's subsequent neurodevelopment. We outline useful extensions of propensity score methodology and discuss directions for future research. Propensity score methods remain controversial and there is no consensus as to when, if ever, they should be used in place of traditional outcome regression models. We therefore end with a discussion of the relative advantages and disadvantages of each.

Pharmaceutical Statistics, 2015

In cost-effectiveness analyses of drugs or health technologies, estimates of life years saved or ... more In cost-effectiveness analyses of drugs or health technologies, estimates of life years saved or quality-adjusted life years saved are required. Randomised controlled trials can provide an estimate of the average treatment effect; for survival data, the treatment effect is the difference in mean survival. However, typically not all patients will have reached the endpoint of interest at the close-out of a trial, making it difficult to estimate the difference in mean survival. In this situation, it is common to report the more readily estimable difference in median survival. Alternative approaches to estimating the mean have also been proposed. We conducted a simulation study to investigate the bias and precision of the three most commonly used sample measures of absolute survival gain - difference in median, restricted mean and extended mean survival - when used as estimates of the true mean difference, under different censoring proportions, while assuming a range of survival patterns, represented by Weibull survival distributions with constant, increasing and decreasing hazards. Our study showed that the three commonly used methods tended to underestimate the true treatment effect; consequently, the incremental cost-effectiveness ratio (ICER) would be overestimated. Of the three methods, the least biased is the extended mean survival, which perhaps should be used as the point estimate of the treatment effect to be inputted into the ICER, while the other two approaches could be used in sensitivity analyses. More work on the trade-offs between simple extrapolation using the exponential distribution and more complicated extrapolation using other methods would be valuable. Copyright © 2015 John Wiley & Sons, Ltd.

Acupuncture in medicine : journal of the British Medical Acupuncture Society, 2015

ABSTRACT Drs White and Cummings state1 that we negatively reported findings of our ‘positive stud... more ABSTRACT Drs White and Cummings state1 that we negatively reported findings of our ‘positive study’—namely, our trial evaluating acupuncture for chronic knee osteoarthritis (OA).2 Our study was not a ‘positive study’ as they suggest. Rather, our study showed no benefit of needle or laser acupuncture compared with sham laser acupuncture on either of our two primary outcomes (pain and function), and small clinically irrelevant benefits compared with no acupuncture for both laser and needle acupuncture. In accordance with best practice in the conduct and interpretation of clinical trials, we (1) a priori specified our primary outcome measures, including the minimum clinically important difference (MCID) on the scale of measurement for each outcome measure; (2) determined and recruited the sample size needed to show meaningful between-group differences in primary outcomes based on our chosen MCIDs and estimated SDs with adequate power; and (3) interpreted our findings based on both of these primary outcomes, their MCIDs and considering both control conditions included in our study design (ie, sham laser acupuncture and no acupuncture). Accordingly, the conclusion stating that our findings do not support acupuncture for these patients is appropriate and is the only conclusion that can (and should) be drawn from our study. We powered our study to detect the absolute MCID between groups in pain (1.8 units) and function (6 units), which have been recommended specifically for our chosen primary …

International journal of cancer. Journal international du cancer, Jan 13, 2015

The associations between intake of or circulating fatty acids and risk of colorectal cancer (CRC)... more The associations between intake of or circulating fatty acids and risk of colorectal cancer (CRC) are unclear. We examined prospectively the associations between dietary or biomarker fatty acids and CRC. For 41,514 men and women, aged 40-69 years, baseline (1990-94) dietary intakes of fatty acids were estimated using a food frequency questionnaire and plasma phospholipid (PPL) fatty acids were measured for 4,205 participants including 395 CRC cases, according to a case-cohort design. Hazard ratios were computed using Cox regression adjusting for education, alcohol intake, smoking status, physical activity and total energy intake; and stratified for gender, ethnicity and family history of cancer, with age as the time scale. We assessed the heterogeneity of associations with colon and rectal cancers. PPL saturated fatty acids (SFAs) were positively associated with CRC risk, while total n-3 polyunsaturated fatty acids (PUFA) and long chain marine n-3 PUFAs showed inverse associations, ...

Recognition that vitamin D might be associated with many chronic diseases has led to large-scale ... more Recognition that vitamin D might be associated with many chronic diseases has led to large-scale epidemiological and clinical studies. Dried blood spots (DBS) are a useful resource for these studies. Consequently, accurate, efficient, and inexpensive assays to quantify 25-hydroxyvitamin D (25OHD) in DBS are required. This study evaluated the validity and reliability of a liquid chromatography-tandem mass spectrometry assay for measuring 25OHD in archived DBS and compared measurements of 25OHD in DBS with those in plasma. Sixty-two participants in the Melbourne Collaborative Cohort Study who had plasma and matching DBS stored since study entry in the early 1990s were randomly selected for a study calibrating 25OHD concentrations in DBS with plasma. As part of a study of vitamin D and mortality, cancer, and diabetes, we also assessed the reliability of measurements from DBS using 500 replicates placed randomly within 31 batches run over 15 months. 25OHD concentrations were measured by liquid chromatography-tandem mass spectrometry. There was good agreement between measurements of 25OHD from DBS and plasma; R(2) = 0.73 from a regression of plasma concentration on DBS concentration. The within-batch and between-batch intraclass correlations from the 500 replicate measurements were 0.82 (95% confidence interval, 0.80, 0.85) and 0.73 (95% confidence interval, 0.68, 0.78), respectively. Measuring 25OHD in DBS is a valid and reliable alternative to measuring 25OHD in sera or plasma. A simple calibration model was developed to convert measurements from DBS to equivalent plasma measurements, thus enabling comparisons against clinical reference ranges and with studies using sera or plasma samples.

Although randomization provides a gold-standard method of assessing causal relationships, it is n... more Although randomization provides a gold-standard method of assessing causal relationships, it is not always possible to randomly allocate exposures. Where exposures are not randomized, estimating exposure effects is complicated by confounding. The traditional approach to dealing with confounding is to adjust for measured confounding variables within a regression model for the outcome variable. An alternative approach--propensity scoring--instead fits a regression model to the exposure variable. For a binary exposure, the propensity score is the probability of being exposed, given the measured confounders. These scores can be estimated from the data, for example by fitting a logistic regression model for the exposure including the confounders as explanatory variables and obtaining the estimated propensity scores from the predicted exposure probabilities from this model. These estimated propensity scores can then be used in various ways-matching, stratification, covariate-adjustment or inverse-probability weighting-to obtain estimates of the exposure effect. In this paper, we provide an introduction to propensity score methodology and review its use within respiratory health research. We illustrate propensity score methods by investigating the research question: 'Does personal smoking affect the risk of subsequent asthma?' using data taken from the Tasmanian Longitudinal Health Study.

Human Pathology, 2014

Adenocarcinomas of the colon and rectum are graded using a 2-tiered system into histologic low-gr... more Adenocarcinomas of the colon and rectum are graded using a 2-tiered system into histologic low-grade and high-grade tumors based on the proportion of gland formation. The current grading system does not apply to subtypes of carcinomas associated with a high frequency of microsatellite instability (MSI), such as mucinous and medullary carcinomas. We investigated the combined effect of histologic grade and MSI status on survival for 738 patients with colorectal carcinoma (48% female; mean age at diagnosis 68.2 years). The proportion of high-grade adenocarcinoma was 18%. MSI was observed in 59 adenocarcinomas (9%), with higher frequency in high-grade tumors compared with low-grade tumors (20% versus 6%; P < .001). Using Cox regression models, adjusting for sex and age at diagnosis and stratifying by the American Joint Committee on Cancer stage, microsatellite stable (MSS) high-grade tumors were associated with increased hazard of all-cause and colorectal cancer-specific mortality: hazard ratio 2.09 (95% confidence interval [CI], 1.58-2.77) and 2.54 (95% CI, 1.86-3.47), respectively, both P < .001. A new grading system separating adenocarcinoma into low grade (all histologic low grade and MSI high grade) and high grade (MSS histologic high grade) gave a lower Akaike information criterion value when compared with the current grading system and thus represented a better model fit to stratify patients according to survival. We found that patients with a high-grade adenocarcinoma had significantly shorter survival than patients with low-grade adenocarcinoma only if the tumor was MSS, suggesting that the grading of colorectal adenocarcinoma with high-grade histologic features should be made according to the MSI status of the tumor.

Statistics in Medicine, 2013

In individually randomised controlled trials, adjustment for baseline characteristics is often un... more In individually randomised controlled trials, adjustment for baseline characteristics is often undertaken to increase precision of the treatment effect estimate. This is usually performed using covariate adjustment in outcome regression models. An alternative method of adjustment is to use inverse probability-of-treatment weighting (IPTW), on the basis of estimated propensity scores. We calculate the large-sample marginal variance of IPTW estimators of the mean difference for continuous outcomes, and risk difference, risk ratio or odds ratio for binary outcomes. We show that IPTW adjustment always increases the precision of the treatment effect estimate. For continuous outcomes, we demonstrate that the IPTW estimator has the same large-sample marginal variance as the standard analysis of covariance estimator. However, ignoring the estimation of the propensity score in the calculation of the variance leads to the erroneous conclusion that the IPTW treatment effect estimator has the same variance as an unadjusted estimator; thus, it is important to use a variance estimator that correctly takes into account the estimation of the propensity score. The IPTW approach has particular advantages when estimating risk differences or risk ratios. In this case, non-convergence of covariate-adjusted outcome regression models frequently occurs. Such problems can be circumvented by using the IPTW adjustment approach.

Respirology, 2014

In respiratory health research, interest often lies in estimating the effect of an exposure on a ... more In respiratory health research, interest often lies in estimating the effect of an exposure on a health outcome. If randomization of the exposure of interest is not possible, estimating its effect is typically complicated by confounding bias. This can often be dealt with by controlling for the variables causing the confounding, if measured, in the statistical analysis. Common statistical methods used to achieve this include multivariable regression models adjusting for selected confounding variables or stratification on those variables. Therefore, a key question is which measured variables need to be controlled for in order to remove confounding. An approach to confounder-selection based on the use of causal diagrams (often called directed acyclic graphs) is discussed. A causal diagram is a visual representation of the causal relationships believed to exist between the variables of interest, including the exposure, outcome and potential confounding variables. After creating a causal diagram for the research question, an intuitive and easy-to-use set of rules can be applied, based on a foundation of rigorous mathematics, to decide which measured variables must be controlled for in the statistical analysis in order to remove confounding, to the extent that is possible using the available data. This approach is illustrated by constructing a causal diagram for the research question: 'Does personal smoking affect the risk of subsequent asthma?'. Using data taken from the Tasmanian Longitudinal Health Study, the statistical analysis suggested by the causal diagram approach was performed.

Optometry and Vision Science, 2013

Population-based studies on abnormal binocular vision and low literacy are rare. The aim is to de... more Population-based studies on abnormal binocular vision and low literacy are rare. The aim is to determine the prevalence of poor stereoacuity among children with low literacy; to identify the characteristics associated with poor stereoacuity among children with low literacy; and to determine the agreement between poor stereoacuity as measured by graded stereocircles with a computerized assessment. A total of 490 children attending primary school in the greater Hobart region, Tasmania, aged 7 to 14 years, with literacy results below the 10th percentile for Tasmanian students at grade 3 level of the NAPLAN (National Assessment Program in Literacy and Numeracy) testing completed a vision screen. Poor stereoacuity was defined as more than 100 seconds of arc as measured by Titmus stereocircles. The prevalence of poor stereoacuity was 16.8% (95% confidence interval [95% CI], 13.6 to 20.4%). Children with poor stereoacuity had a higher frequency of symptom report using the Convergence Insufficiency Symptom Survey. Factors associated with poor stereopsis were prematurity (adjusted odds ratio [AOR], 2.19; 95% CI, 1.09 to 4.42) and bottom shuffling (AOR, 2.39; 95% CI, 1.17 to 4.88). Features associated with poor stereopsis included squint (AOR, 6.05; 95% CI, 3.02 to 12.12), migraine (AOR, 2.25; 95% CI, 1.05 to 4.83), and attention deficit disorder (AOR, 1.92; 95% CI, 1.01 to 3.65). In this low-literacy sample, one-sixth had low stereoacuity. The associations reported require further investigation.

Ophthalmic Epidemiology, 2009

To assess, among children with low literacy and poor stereoacuity, the efficacy of two interventi... more To assess, among children with low literacy and poor stereoacuity, the efficacy of two intervention programs on child vision and education compared to a control program. Eighty-eight children aged 8 to 13 years who had reading problems, and demonstrated poor stereoacuity as measured by the Titmus stereocircle test (> 100 seconds arc) or computerized assessment were randomized to one of two intervention programs: Lawson vision or Phono-Graphix, or a control group: Parental Literacy Support. Vision (Lang test, visual acuity, convergence insufficiency symptom survey) and education assessments (Woodcock Reading Mastery Tests-Revised) were conducted at baseline, intervention end (10 weeks), and 36 weeks. Analysis used intention to treat multi-level models. Compared to the parental literacy support group, convergence insufficiency symptoms were reduced 36 weeks post-randomization amongst those receiving the Lawson orthoptic intervention (mean difference -5.55; 95% confidence interval (CI): -11.1 to -0.05, P < 0.05). Stereoacuity, measured by the Lang test, improved for both the Lawson and Phono-Graphix interventions compared to the parental literacy support group (-1.01; 95% CI: -1.6 to -0.4, P = 0.001, and -0.77; 95% CI: -1.4 to -0.2, P = 0.01). At the 36 week follow-up assessment, word identification had also improved for the Lawson and Phono-Graphix groups but other educational outcomes did not improve. A formal randomized control trial was feasible in this setting. Intervention among children with poor stereoacuity and low literacy produced small improvements in stereopsis and convergence insufficiency symptom scores. Further randomized control trials should be conducted to clarify the role of orthoptic intervention on literacy in selected child populations.

Ophthalmic and Physiological Optics, 2006

Modern Pathology, 2013

Mucinous differentiation is associated with both CpG island methylator phenotype and microsatelli... more Mucinous differentiation is associated with both CpG island methylator phenotype and microsatellite instability in colorectal cancer. The mucinous phenotype derives from abundant expression of the colonic goblet cell mucin, MUC2, and de novo expression of gastric foveolar mucin, MUC5AC. We, therefore, investigated the protein expression levels of MUC2 and MUC5AC, as well as MUC5B and MUC6, in molecular subtypes of colorectal cancer. Seven-hundred and twenty-two incident colorectal carcinomas occurring in 702 participants of the Melbourne Collaborative Cohort Study were characterized for methylator status, MLH1 methylation, somatic BRAF and KRAS mutations, microsatellite-instability status, MLH1, MSH2, MSH6, and PMS2 mismatch repair, and p53 protein expression, and their histopathology was reviewed. Protein expression levels of MUC2, MUC5AC, MUC5B, MUC6, and the putative mucin regulator CDX2 were compared with molecular and clinicopathological features of colorectal cancers using odds ratios and corresponding 95% confidence intervals. MUC2 overexpression (>25% positive tumor cells) was observed in 33% colorectal cancers, MUC5B expression in 53%, and de novo MUC5AC and MUC6 expression in 50% and 39%, respectively. Co-expression of two or more of the mucins was commonly observed. Expression of MUC2, MUC5AC and MUC6 was strongly associated with features associated with tumorigenesis via the serrated neoplasia pathway, including methylator positivity, somatic BRAF p.V600E mutation, and mismatch repair deficiency, as well as proximal location, poor differentiation, lymphocytic response, and increased T stage (all…