Poisson Error Structure Research Articles

Spatial Risk Factors for Pillar 1 COVID-19 Excess Cases and Mortality in Rural Eastern England, UK.

Understanding is still developing about spatial risk factors for COVID‐19 infection or mortality. This is a secondary analysis of patient records in a confined area of eastern England, covering persons who tested positive for SARS‐CoV‐2 through end May 2020, including dates of death and residence area. We obtained residence area data on air quality, deprivation levels, care home bed capacity, age distribution, rurality, access to employment centers, and population density. We considered these covariates as risk factors for excess cases and excess deaths in the 28 days after confirmation of positive Covid status relative to the overall case load and death recorded for the study area as a whole. We used the conditional autoregressive Besag—York–Mollie model to investigate the spatial dependency of cases and deaths allowing for a Poisson error structure. Structural equation models were applied to clarify relationships between predictors and outcomes. Excess case counts or excess deaths were both predicted by the percentage of population age 65 years, care home bed capacity and less rurality: older population and more urban areas saw excess cases. Greater deprivation did not correlate with excess case counts but was significantly linked to higher mortality rates after infection. Neither excess cases nor excess deaths were predicted by population density, travel time to local employment centers, or air quality indicators. Only 66% of mortality was explained by locally high case counts. Higher deprivation clearly linked to higher COVID‐19 mortality separate from wider community prevalence and other spatial risk factors.

Comparative assessment of parameter estimation methods in the presence of overdispersion: a simulation study

The Poisson distribution is commonly assumed as the error structure for count data; however, empirical data may exhibit greater variability than expected based on a given statistical model. Greater variability could point to model misspecification, such as missing crucial information about the epidemiology of the disease or changes in population behavior. When the mechanism producing the apparent overdispersion is unknown, it is typically assumed that the variance in the data exceeds the mean (by some scaling factor). Thus, a probability distribution that allows for overdispersion (negative binomial, for example) may better represent the data. Here, we utilize simulation studies to assess how misspecifying the error structure affects parameter estimation results, specifically bias and uncertainty, as a function of the level of random noise in the data. We compare results for two parameter estimation methods: nonlinear least squares and maximum likelihood estimation with Poisson error structure. We analyze two phenomenological models the generalized growth model and generalized logistic growth model to assess how results of parameter estimation are affected by the level of overdispersion underlying in the data. We use simulation to obtain confidence intervals and mean squared error of parameter estimates. We also analyze the impact of the amount of data, or ascending phase length, on the results of the generalized growth model for increasing levels of overdispersion. The results show a clear pattern of increasing uncertainty, or confidence interval width, as the overdispersion in the data increases. While maximum likelihood estimation consistently yields narrower confidence intervals and smaller mean squared error, differences between the two methods were minimal and not practically significant. At moderate levels of overdispersion, both estimation methods yielded similar performance. Importantly, it is shown that issues of parameter uncertainty and bias in the presence of overdispersion can be mitigated with the inclusion of more data.

Sociodemographic variation in the use of chemotherapy and radiotherapy in patients with stage IV lung, oesophageal, stomach and pancreatic cancer: evidence from population-based data in England during 2013\u20132014

BackgroundSociodemographic inequalities in cancer treatment have been generally described, but there is little evidence regarding patients with advanced cancer. Understanding variation in the management of these patients may provide insights into likely mechanisms leading to inequalities in survival.MethodsWe identified 50,232 patients with stage IV lung, oesophageal, pancreatic and stomach cancer from the English national cancer registry. A generalised linear model with a Poisson error structure was used to explore variation in radiotherapy and chemotherapy within 6 months from diagnosis by age, sex, deprivation, ethnicity, cancer site, comorbidity and, additionally, performance status.ResultsThere was substantial variation by cancer site, large gradients by age, and non-trivial associations with comorbidity and deprivation. After full adjustment, more deprived patients were consistently least likely to be treated with chemotherapy alone or chemotherapy and radiotherapy combined compared with less deprived patients with equally advanced disease stage (treatment rate ratio: 0.82 95% CI (0.78, 0.87) for CT, 0.78 95% CI (0.71, 0.85) for CTRT p < 0.0001).ConclusionsThere was marked variation in the management of patients with stage IV cancer. Routinely collected data could be used for surveillance across all cancers to help reduce treatment variation and optimise outcomes among patients with advanced cancer.

The effect of elevation and time on mountain spider diversity: a view of two aspects in the Cederberg mountains of South Africa

AbstractAimOur aim is to test if long‐term patterns in α alpha and β beta diversity along an elevational transect on two aspects of a mountain are consistent through time using spiders as model organisms, quantify the role of elevation and time (seasonal and inter‐annual) in explaining these patterns and partition the relative contribution of nestedness, species turnover and species loss in explaining these diversity patterns.LocationThe transect is across the Cederberg mountains in the Cape Floristic Kingdom, Western Cape, South Africa and is constituted by 17 sites with an elevational range of 1900 m on two aspects of the mountain (east and west).MethodsSpider assemblages were sampled biannually (wet and dry seasons) over 6 years. Four replicates per site, each consisting of a 5 × 2 pitfall grid, were sampled for 5 days sessions. Generalized linear mixed models with Poisson error structure for species richness (alpha diversity), binomial error structure for beta diversity (Jaccard dissimilarity βccand its partitioned components, β‐3and βrich), and Gaussian error structure for beta diversity values standardized by a null model (SES) were used to model the effects of elevation and season on these two indices respectively.ResultsAlthough varying considerably between years, spider alpha diversity had a hump‐shaped pattern on the western aspect and U‐shaped on the eastern aspect. However, season interacted with elevation to produce more complex patterns during the dry season. There was no significant nestedness except for two instances on the western aspect. Replacement accounted for 60–70% of beta diversity between sites, and elevational distance decay in beta diversity was the result of increased turnover on the western aspect and increased species loss on the eastern aspect. Standardized patterns suggest that there are no effects of season on beta diversity except for a decreased distance decay during the dry season on the western aspect.Main conclusionsLarge‐scale predictors of spider alpha diversity explained small amounts variation in spider diversity, pointing to the importance of local and stochastic processes. Regional turnover of spider diversity is mainly the result of niche processes, suggesting localized adaptation of taxa, further supported by the lack of nestedness in assemblages.

Using opportunistic records from a recreational fishing magazine to assess population trends of sharks

Detecting and determining changes in the occurrence and abundance of species is a priority for effective management of resources and for conservation of biodiversity. In the absence of long-term monitoring data, potential population declines may be very difficult to establish. Therefore, alternative information on occurrence of species to infer population trends is highly valued. Records of sharks (i.e., Notorynchus cepedianus, Carcharias taurus, Galeorhinus galeus, and Carcharhinus brachyurus) off northern Argentina from a recreational fishing magazine (Weekend), between 1973 and 2008, were reviewed with the aim of evaluating population trends with opportunistic data sources. For each shark species, the number of occurrences per year in the magazine was registered. Our analyses were based on a nonprobabilistic method (McPherson and Myers’ approach) designed to determine population trends with opportunistic sighting records. In this approach we included the number of classified ads offering fishing guide services published per year in the magazine as a measure of observation effort. For each species, we fitted generalized linear models with a Poisson error structure and a log link, where the response variable was the number of occurrences per year, the explanatory variable was the year, and the logarithm of the number of fishing guide ads was specified as an offset in each model. For both approaches, the models estimated that populations of the four shark species have suffered declines. Estimates produced by these models will help to determine the magnitude of population changes where a paucity of data prevents more precise analysis.

Is melanoma survival influenced by month of diagnosis?

BackgroundDespite being a well-documented phenomenon, seasonal variation in the incidence of cutaneous melanoma is poorly understood, and data on the seasonality of melanoma survival are scarce. We sought to explore the seasonal variation in melanoma incidence and survival in Belgium and to assess the characteristics and outcomes of cases diagnosed during the seasonal peak. MethodsAll cases of invasive cutaneous melanoma—patients over 15 years of age and registered by the Belgian Cancer Registry (BCR) from 2004 to 2009—were included (n=9782). Monthly variations in incidence were estimated by the ratio of the number of cases diagnosed each month to that diagnosed in January (Rmonth/January) using Nam’s method. The observed and relative 5-year survival rates were adjusted on patient and tumour characteristics using Cox proportional hazards regression models and generalised linear models with a Poisson error structure, respectively. ResultsA peak in melanoma incidence was observed in June (RJune/January=1.64, 95% confidence interval (CI)=1.54–1.73). The 5-year observed survival (OS) and relative survival (RS) rates were significantly higher for patients diagnosed in June compared with other months (OSJune=84%, 95%CI=81–86 versus OSOthermonths=79%, 95%CI=78–80; RSJune=93%, 95%CI=90–95 versus RSothermonths=87%, 95%CI=86–88). After adjustment, the 5-year OS remained significantly higher for patients diagnosed in June (hazard ratioJune=0.78, 95%CI=0.62–0.98); however, the 5-year RS was no longer significantly different for patients diagnosed in June compared with other months (relative excess riskJune=1.16, 95%CI=0.73–1.84). ConclusionsThis study demonstrated a seasonal variation in melanoma incidence in Belgium with a peak in June for the period 2004–2009. When adjusted for patient and tumour characteristics, patients diagnosed in June had higher observed survival rates, but relative survival rates did not differ. Our findings do not support an influence of season of diagnosis on melanoma prognosis.

Time Dependent Relative Risks in Life Insurance Medical Underwriting.

Introduction .- Life insurance medicine focuses on mortality hazards in specified periods. People are free to insure their lives for shorter or longer terms. Because the chosen terms range from 1 year to a life time, life insurers have to take into account the fact that the predictive value of risk indicators can and will change over time. The time a risk indicator keeps its predictive value, will be dependent on its biological effects, volatility, and treatability. For a given applicant this implies that the relative hazard (RH) calculated for his/her medical condition should be dependent on the term of the insurance. The main objective of this study is to determine if some commonly used risk indicators - previously used to study age dependency of relative risks - have a predictive value that increases with the observation period. (1) Methods .- This population-based cohort study uses NHANES-data files from the Third National Health and Nutrition Examination Survey (NHANES III) and the NHANES Linked Mortality Files 2010. Only participants aged 20 to 69 that were examined in mobile examination centers, without a history of some prevalent high risk diseases were included. The observed mortality was compared to the expected mortality in a Generalized Linear Model (GLM) with Poisson error structure with two reference populations, which both can serve as preferred reference for life insurers: The United States Life Tables 2008 (USLT) and the 2008 Valuation Basic Tables (VBT) based on the insured population of 35 US Life insurers. The time dependency of the RHs of the systolic blood pressure (SBP), aspartate aminotransferase (ASAT), lactate dehydrogenase (LDH), serum albumin and albuminuria, was assessed, with correction for ethnicity, household income, history of diabetes mellitus, BMI and serum cholesterol. To be able to compare the results with the results of the Age Dependency Study (ADS), the same data, risk indicators, statistical analysis method, and the Generalized Linear Model (GLM) were used. Results .- Contrary to the results in the univariate analyses, the multivariate analyses show no difference between the USLT and VBT models. In both models the RHs of SBP and albuminuria increase over time, while the RH of LDH decreases over time. Only the slopes are different, reflecting the rate of mortality increase between USLT and VBT. The RHs of ASAT and serum albumin are independent of time in both models. Discussion .- Time dependency of RHs can be assumed to exist with many risk indicators. Time dependency may take two forms: decreasing the RH and increasing the RH of a variable as time passes. Medical underwriting guidelines should therefore differentiate between short term and long term life insurances. Medical directors should realize that the presence of time dependency diminishes the value of short term clinical studies for life insurance medical underwriting. Life insurance mortality studies should always contain a discussion of the time dependency of the predictive value of the variables under study.

An Empirical Study of Generalized Linear Model for Count Data

This paper deals with an empirical study of generalized linear model (GLM) for count data. In particular, Poisson regression model which is also known as generalized linear model for Poisson error structure has been widely used in recent years; it is also used in modeling of count and frequency data. Quasi Poisson model was employ for handling over and under dispersion which the data was found to be over dispersed and another way of handling over dispersion is negative binomial regression model. In this study, the two regression model were compare using the Akaike information criterion (AIC), the model with minimum AIC shows the best which implies the Poisson regression model.

Overall and cause-specific mortality in HIV-positive subjects compared to the general population.

IntroductionEmerging non-AIDS related causes of death have been observed in HIV-positive subjects in industrialized countries. We aimed to analyze overall and cause-specific excess of mortality of HIV-positive patients compared to the general population and to assess the effect of prognostic factors.Material and MethodsWe used generalized linear models with Poisson error structure to estimate overall and cause-specific excess of mortality in HIV-positive patients from 2004 to 2012 in the cohort of the Spanish Network of HIV Research (CoRIS), compared to Spanish general population and to assess the impact of multiple risk factors. We investigated differences between short-term and long-term risk factors effects on excess of mortality. Multiple Imputation by Chained Equations was used to deal with missing data.ResultsIn 9162 patients there were 363 deaths, 16.0% were non-AIDS malignancies, 10.5% liver and 0.3% cardiovascular related. Excess mortality was 1.20 deaths per 100 person years (py) for all-cause mortality, 0.16 for liver, 0.10 for non-AIDS malignancies and 0.03 for cardiovascular. Short-term (first-year follow-up) excess Hazard Ratio (eHR) for global mortality for baseline AIDS was 4.27 (95% CI 3.06–6.01) and 1.47 (95% CI 0.95–2.27) for HCV coinfection; long-term (subsequent follow-up) eHR for baseline AIDS was 0.88 (95% CI 0.58–1.35) and 4.48 (95% CI 2.71–7.42) for HCV coinfection. Lower CD4 count and higher viral load at entry, lower education, being male and over 50 years were predictors for overall excess mortality. Excess of liver mortality was higher in patients with CD4 counts at entry below 200 cells compared to those above 350 (eHR: 6.49, 95% CI 1.21–34.84) and in HCV-coinfected patients (eHR: 3.85, 95% CI 0.85– 17.37), although it was borderline significant. Patients over 50 years old (eHR: 5.55, 95%CI 2.4–12.85) and HCV coinfected (eHR: 5.81, 95% CI 2.6–13) showed a higher risk of non-AIDS malignancies mortality excess. Excess of cardiovascular mortality was related with HCV coinfection (eHR: 6.68, 95% CI 1.25–35.73).ConclusionsOur results show overall, liver, non-AIDS malignancies and cardiovascular excess of mortality associated with being HIV-positive, despite improvements in HIV disease management and antiretroviral therapies. Differential short-term and long-term effect of AIDS before entry and HCV coinfection was found for overall mortality.

No socioeconomic inequalities in ovarian cancer survival within two randomised clinical trials.

Background:Ovarian cancer is the leading cause of death among cancers of the female genital tract, with poor outcomes despite chemotherapy. There was a persistent socioeconomic gradient in 1-year survival in England and Wales for more than 3 decades (1971–2001). Inequalities in 5-year survival persisted for more than 20 years but have been smaller for women diagnosed around 2000. We explored one possible explanation.Methods:We analysed data on 1406 women diagnosed with ovarian cancer during 1991–1998 and recruited to one of two randomised clinical trials. In the second International Collaborative Ovarian Neoplasm (ICON2) trial, women diagnosed between 1991 and 1996 were randomised to receive either the three-drug combination cyclophosphamide, doxorubicin and cisplatin (CAP) or single-agent carboplatin given at optimal dose. In the ICON3 trial, women diagnosed during 1995–1998 were randomised to receive either the same treatments as ICON2, or paclitaxel plus carboplatin.Relative survival at 1, 5 and 10 years was estimated for women in five categories of socioeconomic deprivation. The excess hazard of death over and above background mortality was estimated by fitting multivariable regression models with Poisson error structure and a dedicated link function in a generalised linear model framework, adjusting for the duration of follow-up and the confounding effects of age, Federation of Gynecology and Obstetrics (FIGO) stage and calendar period.Results:Unlike women with ovarian cancer in the general population, no statistically significant socioeconomic gradient was seen for women with ovarian cancer treated in the two randomised controlled trials. The deprivation gap in 1-year relative survival in the general population was statistically significant at −6.7% (95% CI (−8.1, −5.3)), compared with −3.6% (95% CI (−10.4, +3.2)) in the trial population.Conclusions:Although ovarian cancer survival is significantly lower among poor women than rich women in England and Wales, there was no evidence of an association between socioeconomic deprivation and survival among women with ovarian cancer who were treated and followed up consistently in two well-conducted randomised controlled trials. We conclude that the persistent socioeconomic gradient in survival among women with ovarian cancer, at least for 1-year survival, may be due to differences in access to treatment and standards of care.

Increasing incidence of cataract surgery: Population-based study

To estimate the incidence of cataract surgery in a defined population and to determine longitudinal cataract surgery patterns. Mayo Clinic, Rochester, Minnesota, USA. Cohort study. Rochester Epidemiology Project (REP) databases were used to identify all incident cataract surgeries in Olmsted County, Minnesota, between January 1, 2005, and December 31, 2011. Age-specific and sex-specific incidence rates were calculated and adjusted to the 2010 United States white population. Data were merged with previous REP data (1980 to 2004) to assess temporal trends in cataract surgery. Change in the incidence over time was assessed by fitting generalized linear models assuming a Poisson error structure. The probability of second-eye cataract surgery was calculated using the Kaplan-Meier method. Included were 8012 cataract surgeries from 2005 through 2011. During this time, incident cataract surgery significantly increased (P<.001), peaking in 2011 with a rate of 1100 per 100 000 (95% confidence interval, 1050-1160). The probability of second-eye surgery 3, 12, and 24 months after first-eye surgery was 60%, 76%, and 86%, respectively, a significant increase compared with the same intervals in the previous 7 years (1998 to 2004) (P<.001). When merged with 1980 to 2004 REP data, incident cataract surgery steadily increased over the past 3 decades (P<.001). Incident cataract surgery steadily increased over the past 32 years and has not leveled off, as reported in Swedish population-based series. Second-eye surgery was performed sooner and more frequently, with 60% of residents having second-eye surgery within 3 months of first-eye surgery.

Trends in the Incidence of Venous Thromboembolism Adjusted for Body Mass Index (BMI).

AbstractAbstract 2256 Background:The incidence of venous thromboembolism (VTE) may be increasing due to the obesity epidemic. Objective:To estimate the incidence of VTE and describe trends in incidence adjusted for the population's increase in BMI. Methods:Using the resources of the Rochester Epidemiology Project, we identified all Olmsted County, MN residents with an incident deep vein thrombosis (DVT) or pulmonary embolism (PE) over the 13-year period, 1988–2000 (n=1,400). Age and BMI at VTE diagnosis and gender were collected for all cases. Age was categorized into 10 year intervals. Using a model for the population distribution of BMI derived from several (n=6) non-VTE Olmsted County cohorts (n=8,011 individuals), we calculated the probability of falling into each of five BMI categories (<18.5, 18.5–24, 25–29, 30–34, 35+) for each age, gender and calendar year. In Olmsted County in the median age group, the prevalence of BMI >30 kg/m2increased from approximately 15% in 1980 to 36% in 2000. We applied these probabilities to the decennial census data for 1980, 1990, 2000 using linearly interpolated census values for intra-census years to obtain Olmsted County age/sex/calendar year/BMI category denominators. Generalized linear modeling assuming a Poisson error structure, and using a log link function, and a log (population) offset was used to assess the relationship of crude incidence rates to gender, year of diagnosis, age and BMI. Results:The overall average age- and sex-adjusted annual VTE incidence was 120 per 100,000 person-years (DVT: 66 per 100,000; PE: 54 per 100,000), with higher age-adjusted rates among men than women (129 versus 113 per 100,000, respectively). VTE incidence rates increased exponentially with age for both genders, ranging from 3 to 1,079 per 100,000 for age groups 0–19 to 90–110 years. Age and sex-adjusted VTE incidence increased by increasing BMI category (88, 112, 106, 150, and 195 per 100,000 person-years, respectively, by increasing BMI category). Unadjusted for BMI, VTE incidence was unchanged from 1988 through 2000 (p=0.70). After adjusting for age, gender and calendar year, VTE incidence increased with increasing BMI (p<0.0001). The calendar year incidence slope estimate decreased numerically from 3% per decade to minus 4% per decade after adjusting for increasing BMI, remaining nonsignificant (p=0.57). Conclusions:VTE remains a major national health problem, especially among the obese. Despite the dramatically increasing population BMI, and the significant association of BMI category with VTE incidence, the overall incidence of VTE remained unchanged over the timeframe, 1988–2000. Disclosures:No relevant conflicts of interest to declare.

Population-based study of breast cancer in older women: prognostic factors of relative survival and predictors of treatment

BackgroundA large proportion of women with breast cancer (BC) are elderly. However, there is a lack of information regarding BC prognostic factors and care in this population. The aims of this study were to assess the prognostic factors of relative survival (RS) among women with BC aged ≥ 75 years old and to identify the predictive factors of treatments administered to this population.MethodsA population-based study was performed using data from the Cote d’Or breast and gynaecological cancer registry. Women aged 75 years and older with primary invasive BC and resident in Cote d’Or at the time of diagnosis made between January 1998 and December 2008 were retrospectively selected. Prognostic factors of RS were estimated in a generalized linear model with a Poisson error structure. RS rate for the whole population was given at 5 years. Logistic regression models were used to identify the predictors of the treatments administered.ResultsSix hundred and eighty-one women were included. Median age at diagnosis was 80. Comorbidities (p=0.02), pT stage (p=0.04), metastases (p=<0.001), having a family doctor (p=0.03) and hormone-receptor status (p=0.006) were independent prognostic factors of RS. The RS rate at 5 years for the whole population was 78.2%, 95%CI = [72.2-83.0]. Age, pT stage, metastases, histoprognostic SBR grade, hormone receptor status and comorbidities were frequently found to be predictors of treatment with surgery alone, hormone therapy alone, breast conserving surgery plus adjuvant therapy and mastectomy plus adjuvant therapy.ConclusionsComorbid conditions adversely affect survival in older women with breast cancer. Moreover the results of this study showed that there are numerous predictors of the type of treatment administered, and that the most important were age and comorbidities.

What Is the Impact of Impatiens parviflora on Diversity and Composition of Herbal Layer Communities of Temperate Forests?

The aim was to estimate the impacts of invasive Impatiens parviflora on forests’ herbal layer communities. A replicated Before-After-Control-Impact field experiment and comparisons with adjacent uninvaded plots were used. The alien’s impact on species richness was tested using hierarchical generalized mixed effect models with Poisson error structure. Impact on species composition was tested using multivariate models (DCA, CCA, RDA) and Monte-Carlo permutation tests. Removal plots did not differ in native species richness from neither invaded nor adjacent uninvaded plots, both when the treatment’s main effect or its interaction with sampling time was tested (Chi2 = 0.4757, DF = 2, p = 0.7883; Chi2 = 7.229, DF = 8, p = 0.5121 respectively). On the contrary, ordination models revealed differences in the development of plots following the treatments (p = 0.034) with the invaded plots differing from the adjacent uninvaded (p = 0.002). Impatiens parviflora is highly unlikely to impact native species richness of invaded communities, which may be associated with its limited ability to create a dense canopy, a modest root system or the fact the I. parviflora does not represent a novel and distinctive dominant to the invaded communities. Concerning its potential impacts on species composition, the presence of native clonal species (Athyrium filix-femina, Dryopteris filix-mas, Fragaria moschata, Luzula luzuloides, Poa nemoralis) on the adjacent uninvaded plots likely makes them different from the invaded plots. However, these competitive and strong species are more likely to prevent the invasion of I. parviflora on the adjacent uninvaded plots rather than being themselves eliminated from the invaded communities.

Random parameter models for accident prediction on two-lane undivided highways in India

Introduction Generalized linear modeling (GLM), with the assumption of Poisson or negative binomial error structure, has been widely employed in road accident modeling. A number of explanatory variables related to traffic, road geometry, and environment that contribute to accident occurrence have been identified and accident prediction models have been proposed. The accident prediction models reported in literature largely employ the fixed parameter modeling approach, where the magnitude of influence of an explanatory variable is considered to be fixed for any observation in the population. Similar models have been proposed for Indian highways too, which include additional variables representing traffic composition. The mixed traffic on Indian highways comes with a lot of variability within, ranging from difference in vehicle types to variability in driver behavior. This could result in variability in the effect of explanatory variables on accidents across locations. Random parameter models, which can capture some of such variability, are expected to be more appropriate for the Indian situation. Method The present study is an attempt to employ random parameter modeling for accident prediction on two-lane undivided rural highways in India. Three years of accident history, from nearly 200 km of highway segments, is used to calibrate and validate the models. Results The results of the analysis suggest that the model coefficients for traffic volume, proportion of cars, motorized two-wheelers and trucks in traffic, and driveway density and horizontal and vertical curvatures are randomly distributed across locations. Conclusions The paper is concluded with a discussion on modeling results and the limitations of the present study.

Population-based study of ovarian cancer in Côte d'Or: prognostic factors and trends in relative survival rates over the last 20 years

BackgroundThe aim of this population-based study was to assess independent prognostic factors in ovarian cancer using relative survival (RS) and to investigate changes in RS rates from 1982 to 2005.MethodsData on 748 patients with ovarian cancer were provided by the Côte d'Or gynaecologic cancer registry. The RS was estimated using a generalized linear model with a Poisson error structure. Relative survival and its 95% confidence interval (CI) were described at the following specific time points 1, 3 and 5 years. The effect of prognostic factors on survival was assessed with multivariate analyses of RS.ResultsThe median follow-up was 12 years. The RS rates at 1, 3 and 5 years were 81%, 55% and 44%, respectively. As compared with the period 1982-1989, an improvement in survival was found for the period 1998-2005: HR = 0.52[0.40-0.67]. Women who lived in urban areas had better RS: HR = 0.82[0.67-0.99]. Patients with epithelial types of ovarian cancer other than mucinous or endometrioid cancer had worse RS than those with serous histology. Age ≥ 70 years was associated with lower survival.ConclusionsPeriod of diagnosis, stage at diagnosis, histology, place of residence and age were independent prognostic factors for survival in ovarian cancer. An improvement in the survival rate was observed after 1998 but a significant improvement was limited to advanced stage cancers.

Modelling relative survival in the presence of incomplete data: a tutorial

Missing data frequently create problems in the analysis of population-based data sets, such as those collected by cancer registries. Restriction of analysis to records with complete data may yield inferences that are substantially different from those that would have been obtained had no data been missing. 'Naive' methods for handling missing data, such as restriction of the analysis to complete records or creation of a 'missing' category, have drawbacks that can invalidate the conclusions from the analysis. We offer a tutorial on modern methods for handling missing data in relative survival analysis. We estimated relative survival for 29 563 colorectal cancer patients who were diagnosed between 1997 and 2004 and registered in the North West Cancer Intelligence Service. The method of multiple imputation (MI) was applied to account for the common example of incomplete stage at diagnosis, under the missing at random (MAR) assumption. Multivariable regression with a generalized linear model and Poisson error structure was then used to estimate the excess hazard of death of the colorectal cancer patients, over and above the background mortality, adjusting for significant predictors of mortality. Incomplete information on stage, morphology and grade meant that only 55% of the data could be included in the 'complete-case' analysis. All cases could be included after indicator method (IM) or MI method. Handling missing data by MI produced a significantly lower estimate of the excess mortality for stage, morphology and grade, with the largest reductions occurring for late-stage and high-grade tumours, when compared with the results of complete-case analysis. In complete-case analysis, almost 50% of the information could not be included, and with the IM, all records with missing values for stage were combined into a single 'missing' category. We show that MI methods greatly improved the results by exploiting all the information in the incomplete records. This method also helped to ensure efficient inferences about survival were made from the multivariate regression analyses.

Relative survival after hospitalisation for hip fracture in older people in New South Wales, Australia

Survival after hospitalisation for hip fracture by age group and sex relative to survival in the general population was assessed in people aged 65+. Men had double the risk of death compared with women to 1 year, but age effects lasted only to 3 months. Clinical outcomes need to be improved. We assessed the relative survival of hospitalised fall-related hip fracture patients aged 65+ years leaving hospital in New South Wales, Australia, between July 2000 and December 2003. We carried out a population-based study of all hospital separations for NSW residents with a principal diagnosis of hip fracture (ICD-10-AM S72.0 to S72.2) and first external cause of fall (ICD-10-AM codes W00 to W19), linked to NSW death data. A total of 16,836 cases were included. Relative survival 3 to 36 months post-admission by 10-year age groups and sex was calculated, using NSW life tables for 2002-2004. Relative excess risk was modelled using a generalised linear model with Poisson error structure, using the life table data. One-year cumulative relative survival in 65- to 74-year-olds was 82% (men), 90% (women); in 85+-year-olds 65% (men), 80% (women). Men have a relative excess risk of death of 2.2 (95% CI 2.03-2.38) times that of women. Only 21% of deaths mention the hip fracture as contributing to death. There is a need to reduce the number of hip fractures and improve clinical outcomes for older people hospitalised with hip fractures.

Application of the Tweedie distribution to zero-catch data in CPUE analysis

We focus on the zero-catch problem of CPUE (catch per unit effort) standardization. Because the traditional CPUE model with a log-normal error structure cannot be applied in this case, three methods have often been utilized as follows: (1) Ad hoc method adds a small constant value to all response variables. (2) Catch model with a Poisson or negative-binomial (NB) error structure. (3) Delta-type two-step method such as the delta-normal model (after estimating the ratio of zero-catch using a logit or probit model, a model such as CPUE log-normal or Catch-Poisson is applied to CPUE without zero-data). However, there are some statistical problems with each of these methods. In this paper, we carried out the CPUE standardization mainly using the Tweedie distribution model based on the actual by-catch data (silky shark, Carcharhimus falciformis, in the North Pacific Ocean caught by Japanese training vessels) including many observations with zero-catch (>2/3rd) and tuna fishery data as a target (yellowfin tuna, Thunnus albacares, in the Indian Ocean caught by Japanese commercial vessels) where the ratio of zero-catch is not so high (<1/3rd). The Tweedie model is an extension of compound Poisson model derived from the stochastic process where the weight of the counted objects (i.e., number of fish) has a gamma distribution and has an advantage of handling the zero-catch data in a unified way. We also compared four candidate models, the Catch-NB model, ad hoc method, Delta-lognormal model (delta-type two-step method) and Tweedie distribution, through CPUE analyses of actual fishery data in terms of the statistical performance. Square error and Pearson's correlation coefficient were calculated based on the observed CPUE and the corresponding predicted CPUE using the n-fold cross-validation. As a result, the differences in the trend of CPUE between years and model performance between the ad hoc method and Tweedie model were found to be not so large in the example of yellowfin tuna (target species). However, the statistical performance of Tweedie distribution is rather better than Delta-lognormal model, the Catch-NB distribution and ad hoc method in the example of silky shark (by-catch species). Standardized CPUE year trend of ad hoc method was found to be quite different from that of the Tweedie distribution and other two models. Model performance of the Tweedie distribution is good judging from the 5-fold cross-validation using the fishery data if including many zero-catch data such as by-catch species.

Trends in the Incidence of Deep Vein Thrombosis and Pulmonary Embolism: A 35-Year Population-Based Study.

Background: Recent trends in the incidence of venous thromboembolism (VTE), including idiopathic vs. non-idiopathic VTE, have not been well described.Objective: To estimate the incidence of deep vein thrombosis (DVT) and pulmonary embolism with or without DVT (PE), and describe trends in incidence.Methods: Using the resources of the Rochester Epidemiology Project, we identified all Olmsted County, MN residents with an incident DVT and PE over the 35-year period, 1966–2000 (n=3342). For all cases, the complete medical records in the community were reviewed for demographic and baseline characteristics previously identified as risk factors for VTE. Generalized linear models assuming a Poisson error structure, and using a log link function, and a log (population) offset will be used to assess the relationship of crude incidence rates to gender, year of diagnosis and age at diagnosis.Results: The overall average age- and sex-adjusted annual VTE incidence was 122 per 100,000 person-years (DVT, 56 per 100,000; PE, 66 per 100,000), with higher age-adjusted rates among men than women (134 versus 115 per 100,000, respectively). VTE incidence rates increased exponentially with age for both genders, ranging from 4 to 1110 per 100,000 for age groups 0–19 to 90–110 years. Compared to the 5-year period, 1981–85 (when non-invasive diagnostic testing became routinely available), the overall VTE incidence through 2000 remains unchanged. However, the DVT incidence and the PE incidence significantly increased and decreased, respectively, adjusting for age and gender (p<0.001 for both). The overall age- and sex-adjusted annual incidence of idiopathic VTE was 11.7 per 100,000 person-years (DVT, 6.6 per 100,000; PE, 5.1 per 100,000), with age-adjusted rates also higher among men than women (15.1 vs. 9.1 per 100,000). Interestingly, again compared to 1981–85, idiopathic VTE incidence decreased for 1991–95 (p=0.001) and 1996–2000 (p=0.32), adjusting for age and gender. Idiopathic DVT incidence decreased for 1991–95 (p=0.09), and idiopathic PE incidence decreased for both 1991–95 (p=0.004) and 1996–2000 (p=0.03). The overall age- and sex-adjusted annual incidence of non-idiopathic VTE was 109.4 per 100,000 (DVT, 48.4 per 100,000; PE, 60.7 per 100,000), again, with age-adjusted rates higher in men than women (115.1 vs. 106.8 per 100,000). Non-idiopathic DVT incidence increased steadily since 1981–85 (p=0.006, p<0.001, and p<0.001 for increasing DVT incidence for 1986–1990–1991–1995–1996–2000, respectively, adjusting for age and gender). Non-idiopathic PE incidence, however, remained unchanged for 1986–2000.Conclusions: VTE remains a major national health problem, especially among the elderly. Despite improved VTE prophylaxis efficacy and utilization, the overall incidence of VTE remains unchanged. However, the decreasing incidence of idiopathic DVT, and particularly idiopathic PE (with its associated poor survival) raises the possibility that the total number of VTE(PE)-related deaths may also be decreasing, albeit slightly. This hypothesis requires formal testing. The increasing or steady incidence of non-idiopathic DVT and PE, respectively, suggests the need for more widespread, effective VTE prophylaxis.