Bayesian bivariate meta‐analysis of diagnostic test studies using integrated nested Laplace approximations

For bivariate meta‐analysis of diagnostic studies, likelihood approaches are very popular. However, they often run into numerical problems with possible non‐convergence. In addition, the construction of confidence intervals is controversial. Bayesian methods based on Markov chain Monte Carlo (MCMC) sampling could be used, but are often difficult to implement, and require long running times and diagnostic convergence checks. Recently, a new Bayesian deterministic inference approach for latent Gaussian models using integrated nested Laplace approximations (INLA) has been proposed. With this approach MCMC sampling becomes redundant as the posterior marginal distributions are directly and accurately approximated. By means of a real data set we investigate the influence of the prior information provided and compare the results obtained by INLA, MCMC, and the maximum likelihood procedure SAS PROC NLMIXED . Using a simulation study we further extend the comparison of INLA and SAS PROC NLMIXED by assessing their performance in terms of bias, mean‐squared error, coverage probability, and convergence rate. The results indicate that INLA is more stable and gives generally better coverage probabilities for the pooled estimates and less biased estimates of variance parameters. The user‐friendliness of INLA is demonstrated by documented R‐code. Copyright © 2010 John Wiley & Sons, Ltd.     

Determining joint carrier probabilities of cancer-causing genes using Markov chain Monte Carlo methods

In genetic counseling for cancer risk, the probability of carrying a mutation of a cancer-causing gene plays an important role. Family history of various cancers is important in calculating this probability. BRCAPRO is a widely used software for calculating the probability of carrying mutations in BRCA1 and BRCA2 genes given the family history of breast and ovarian cancer in first- and second-degree relatives. BRCAPRO uses an analytical (exact) calculational procedure. Using Markov chain Monte Carlo (MCMC) methods, we extend BRCAPRO to handle, in principle, any type of cancer, family history, any number of genes and alleles that each gene may have. When the information used in this MCMC approach is the same as for BRCAPRO (two genes: BRCA1 and BRCA2; two cancers: breast and ovarian; first- and second-degree relatives only), the two approaches give essentially the same answer. Extending the model to include (1) prostate cancer, (2) two mutated alleles of BRCA2, namely, mutations in Ovarian Cancer Cluster Region (OCCR) and non-OCCR region, and (3) relatives of degree greater than second-degree, leads to different carrier probabilities. The MCMC approach is a useful tool in building a comprehensive model to give accurate estimates of carrier probabilities. Such an approach will be even more important as additional information about the genetics of various cancers becomes available.     

A survey regarding the implementation of cancer screening among municipalities in Japan

OBJECTIVE: The objective of this study was to conduct a survey regarding implementation of cancer screening in all municipalities in Japan. METHODS: In February 2003, we sent out a questionnaire regarding cancer screening to all 3,242 municipalities in Japan. RESULTS: A total of 2,342 municipalities (72.2%) completed and returned the questionnaire. In the year 2002, the percentages of municipalities which implemented each cancer screening were 49.4% for breast cancer using mammography, 33.7% for prostate cancer, 20.3% for cervical cancer for women younger than 30 years old, 5.7% for lung cancer using helical CT, 5.0% for gastric cancer using the PG method and 4.8% for ovarian cancer. CONCLUSIONS: Screening for breast cancer using mammography, as recommended by the 4-th plan for Health Services for the Aged, was implemented by less than half of the municipalities. Parameters, including eligible age group, target group, and method, widely differed among the municipalities. A standardized system for cancer screening programs should be recommended.     

Geographical variation in cancer patient survival in Finland: chance,confounding, or effect of treatment?

STUDY OBJECTIVE--The aim was to determine whether survival of cancer patients in Finland varies with their place of residence, and if so, what proportion of the variation might be due to health services rather than to confounding variables. DESIGN--Patients with breast and prostatic cancer diagnosed in Finland between 1970 and 1981 were classified by place of residence (from 21 hospital districts), and area specific 5 year relative survival rates were estimated. SETTING--This was a population based survey of the whole of Finland. PATIENTS--16,754 cases of breast cancer and 9483 cases of prostatic cancer were identified. Of these, 0.5% of breast cancers and 4.1% of prostatic cancers were excluded because diagnosis was based only on necropsy findings or because the only information available was from the death certificate. MEASUREMENTS AND MAIN RESULTS--There was a large variation in rates, ranging from 59% to 76% for breast cancer, and from 30% to 65% for prostatic cancer. However, after accounting for age of patient and extent of disease, the standardised differences for prostatic cancer closely followed random distribution, indicating equal results of treatment in different areas. For breast cancer there was more variation than expected by chance and patients resident in any of the university central hospital districts with modern radiotherapy equipment survived better than other patients. CONCLUSIONS--There is little indication that large variations in crude mortality rates from these two cancers in different parts of Finland are due to inequalities of medical care, though a small effect on breast cancer survival which might be care related was shown.     

Family history of reproductive cancers and ovarian cancer risk: an Italian case-control study.

The relation between family history of ovarian, breast, and endometrial cancer and risk of epithelial ovarian carcinoma was analyzed within the framework of a case-control study conducted from 1983 to 1989. The study included 755 cases of ovarian cancer and 2,023 controls in hospital for a spectrum of acute nongynecologic, hormonal, or neoplastic conditions in the Greater Milan area, Italy. Eighteen cases (2%) and 24 controls (1%) reported a history of ovarian cancer in a first-degree relative: The corresponding multivariate adjusted odds ratio (OR) was 1.9 (95% confidence interval (CI) 1.1-3.6). The risk of ovarian cancer was elevated in women reporting a family history of breast cancer (OR = 1.6, 95% CI 1.1-2.3), but no significant association emerged with a family history of endometrial cancer (OR = 1.3, 95% CI 0.8-1.7). When the data were stratified by family history of breast cancer, a family history of ovarian cancer was over 10 times more frequent in both cases and controls who reported a family history of breast cancer than in cases and controls reporting no family history of breast cancer. The estimated odds ratio for ovarian cancer associated with a family history of the disease was 2.3 (95% CI 1.1-4.5) in women not reporting a family history of breast cancer, but no association emerged in the subgroup of women reporting a family history of breast cancer. These results confirm that a family history of ovarian cancer increases the risk of the disease, but the percentage of ovarian cancer cases explained by a family history of the disease is small: Less than 1% of observed cases in this study could be attributed to this "family risk factor."     

Cure fraction model with random effects for regional variation in cancer survival

Assessing regional differences in the survival of cancer patients is important but difficult when separate regions are small or sparsely populated. In this paper, we apply a mixture cure fraction model with random effects to cause‐specific survival data of female breast cancer patients collected by the population‐based Finnish Cancer Registry. Two sets of random effects were used to capture the regional variation in the cure fraction and in the survival of the non‐cured patients, respectively. This hierarchical model was implemented in a Bayesian framework using a Metropolis‐within‐Gibbs algorithm. To avoid poor mixing of the Markov chain, when the variance of either set of random effects was close to zero, posterior simulations were based on a parameter‐expanded model with tailor‐made proposal distributions in Metropolis steps. The random effects allowed the fitting of the cure fraction model to the sparse regional data and the estimation of the regional variation in 10‐year cause‐specific breast cancer survival with a parsimonious number of parameters. Before 1986, the capital of Finland clearly stood out from the rest, but since then all the 21 hospital districts have achieved approximately the same level of survival. Copyright © 2010 John Wiley & Sons, Ltd.     

Exploring heterogeneity in tumour data using Markov chain Monte Carlo

We describe a Bayesian approach to incorporate between-individual heterogeneity associated with parameters of complicated biological models. We emphasize the use of the Markov chain Monte Carlo (MCMC) method in this context and demonstrate the implementation and use of MCMC by analysis of simulated overdispersed Poisson counts and by analysis of an experimental data set on preneoplastic liver lesions (their number and sizes) in the presence of heterogeneity. These examples show that MCMC-based estimates, derived from the posterior distribution with uniform priors, may agree well with maximum likelihood estimates (if available). However, with heterogeneous parameters, maximum likelihood estimates can be difficult to obtain, involving many integrations. In this case, the MCMC method offers substantial computational advantages.     

A correlated frailty model with long-term survivors for estimating the heritability of breast cancer

The aim of this study is to investigate the role of genetics and environment in susceptibility to breast cancer (frailty). An interdisciplinary approach was adopted, combining a correlated frailty-mixture model with genetic equations, allowing for decomposition of the frailty variance into genetic and environmental components. In addition, the possibility that a fraction of the population under study is 'immune' to the disease is evaluated, and changes in heritability estimates introducing a fraction of non-susceptible individuals are determined. The methodology is applied to breast cancer data from the Swedish Twin Registry, including information about all female monozygotic and dizygotic twin pairs born in Sweden between 1886 and 1967. The inferential problem is solved in a Bayesian framework and the numerical work is carried out using Markov chain Monte Carlo (MCMC) methods.     

From Multiple Quality Indicators of Breast Cancer Care Toward Hospital Variation of a Summary Measure

ObjectivesTo improve quality in breast cancer care, large numbers of quality indicators are collected per hospital, but benchmarking remains complex. We aimed to assess the validity of indicators, develop a textbook outcome summary measure, and compare case-mix adjusted hospital performance.MethodsFrom a nationwide population-based registry, all 79 690 nonmetastatic breast cancer patients surgically treated between 2011 and 2016 in 91 hospitals in The Netherlands were included. Twenty-one indicators were calculated and their construct validity tested by Spearman’s rho. Between-hospital variation was expressed by interquartile range (IQR), and all valid indicators were included in the summary measure. Standardized scores (observed/expected based on case mix) were calculated as above (>100) or below (<100) expected. The textbook outcome was presented as a continuous and all-or-none score.ResultsThe size of between-hospital variation varied between indicators. Sixteen (76%) of 21 quality indicators showed construct validity, and 13 were included in the summary measure after excluding redundant indicators that showed collinearity with others owing to strong construct validity. The median all-or-none textbook outcome score was 49% (IQR 42%-54%) before and 49% (IQR 48%-51%) after case-mix adjustment. From the total of 91 hospitals, 3 hospitals were positive (3%) and 9 (10%) were negative outliers.ConclusionsThe textbook outcome summary measure showed discriminative ability when hospital performance was presented as an all-or-none score. Although indicator scores and outlier hospitals should always be interpreted cautiously, the summary measure presented here has the potential to improve Dutch breast cancer quality indicator efforts and could be implemented to further test its validity, feasibility, and usefulness.     

Familial breast and ovarian cancer: a Swedish population-based register study

A cohort of offspring of mothers with breast or ovarian cancer diagnosed in 1958-1993 was established using Swedish population-based registers. The children (n = 158,041) were born between 1941 and 1993, and their cancer incidence was followed between 1961 and 1993. A total of 3,257 tumors in 3,102 children were found. Observed numbers of cases were compared with expected numbers based on national calendar year-, age-, and sex-specific incidences. For daughters of women with breast cancer, the standardized morbidity ratios for being diagnosed with breast cancer and ovarian cancer before age 50 years were 1.99 (95% confidence interval (CI): 1.86, 2.14) and 1.28 (95% CI: 1.05, 1.54), respectively. The corresponding figures for daughters of women with ovarian cancer were 1.79 (95% CI: 1.55, 2.07) and 2.38 (95% CI: 1.77, 3.12). The risks were raised if the mother's cancer was diagnosed at a young age, the mother had multiple breast/ovarian diagnoses, or there was a sister with breast/ovarian cancer. Among all offspring, increased risks were found for thyroid cancer, testicular cancer, and malignant melanoma, while lung cancer risk was decreased if the mother had had breast cancer. The authors developed a variance estimator for the standardized morbidity ratio to cope with overdispersion due to dependency within families.     

2014-2018年兰州市妇女病患病情况分析

目的 了解2014-2018年兰州市妇女病患病情况,为兰州市妇女保健工作的开展和防治措施的制定提供科学依据。方法 收集2014-2018年兰州市3县5区妇幼保健院上报的卫生年报,使用SPSS 20.0经卡方和卡方趋势检验以及动态数列描述分析数据。结果 2014-2018年兰州市接受妇女病筛查总人数为1384375,患病人数为509997,患病率为36.84%,5年来妇女病患病呈上升趋势(χ2趋势=159.20,P＜0.05),且城区和县区之间患病率的差异具有统计学意义(χ2=30672.01,P＜0.05)。其中宫颈炎和阴道炎为最常见的妇女病,患病率分别为16.46%、16.44%,均呈上升趋势;子宫肌瘤患病率为2.21%,尖锐湿疣患病率为0.05%,二者均呈下降趋势;肿瘤的患病率宫颈癌为38.07/10万、乳腺癌为26.93/10万、卵巢癌为8.24/10万,患病率均呈下降趋势。结论 2014-2018年兰州市妇女妇女病患病率较高且呈上升趋势;城乡差距较大。妇科炎症居妇女病首位且有上升趋势,妇科肿瘤发病呈下降趋势。应加强妇女病的筛查和诊疗,提高妇女保健工作质量。     

2001至2014年我国孕产期保健服务现状及效果分析

目的 分析2001至2014年我国孕产期保健服务的现状和效果,探讨各项保健服务指标与孕产妇和围生儿死亡率的相关性.方法 分别采用动态数列和多样本秩和检验分析各指标的时间趋势和地区差异,运用秩相关分析探讨各保健服务指标与孕产妇和围生儿死亡率的相关性.结果 2001至2014年,我国各孕产期保健服务指标均呈上升趋势,其中住院分娩率上升最快,平均上升速度达2.10%;2014年,不同地区在住院分娩率和新法接生率的差异仍具有统计学意义(χ2值分别为22.587、21.357,均P<0.05).同时,孕产妇和围生儿死亡率呈下降趋势,年平均下降速度分别为6.25%和6.73%;2014年,两指标的地区差异仍具有统计学意义(χ2值分别为17.856、14.455,均P<0.05).孕产妇和围生儿死亡率与各保健服务指标之间均存在强的负相关(|rs|值为0.776~0.996,均P≤0.001),其中两项死亡率均与住院分娩率和新法接生率的相关性最大.结论 2014年,我国住院分娩率、新法接生率以及孕产妇和围生儿死亡率仍存在地区差异,尚需进一步改善孕产期保健服务的公平性和可及性,缩小孕产妇和围生儿死亡率的地区差异.     

Estimating potential savings in cancer deaths by eliminating regional and social class variation in cancer survival in the Nordic countries.

STUDY OBJECTIVES: To examine equity in the health care system with regard to cancer patient care by estimating the level of systematic regional variation in cancer survival in the Nordic countries. Specifically, those cancer sites which exhibit high levels of systematic regional variation in survival and hence inequity were identified. Estimating the reduction in cancer deaths which could be achieved by eliminating this variation so that everyone receives effective care will provide a readily interpretable measure of the amount of systematic regional variation. A comprehensive analysis of regional variation in survival has not previously been conducted so appropriate statistical methodology must be developed. SETTING AND PARTICIPANTS: All those aged 0-90 years who had been diagnosed with at least one of 12 common malignant neoplasms between 1977 and 1992 in Denmark, Finland, Norway, and Sweden. DESIGN: A separate analysis was conducted for each country. Regression models for the relative survival ratio were used to estimate the relative risk of excess mortality attributable to cancer in each region after correcting for age and sex. An estimate of the amount of systematic regional variation in survival was obtained by subtracting the estimated expected random variation from the observed regional variation. An estimate was then made of the potential reduction in the number of cancer deaths for 2008-12 if regional variation in survival were eliminated so that everyone received the same level of effective care. MAIN RESULTS: Between 2008 and 2012, an estimated 2.5% of deaths from cancers in the 12 sites studied could be prevented by eliminating regional variation in survival. The percentage of potentially avoidable deaths did not depend on country or sex but it did depend on cancer site. There was no relationship between the level of regional variation in a given country and the level of survival. The cancer sites for which the greatest percentage savings could be achieved were melanoma (11%) and cervix uteri (6%). The sites for which the highest number of deaths could be prevented were prostate, colon, melanoma, and breast. CONCLUSIONS: This methodology showed a small amount of systematic regional variation in cancer survival in the Nordic countries. The cancer sites with high levels of regional variation identified are potential targets for cancer control programmes.     

Distribución municipal de la incidencia de los tumores más frecuentes en un área de elevada mortalidad por cáncer

Objective

To describe the geographic distribution patterns of the municipal incidence of the most common tumours in the Huelva province (Spain) as compared to the estimated incidence for all of Spain.

Prevention in familial breast cancer: counseling and prophylactic mastectomy

In five families with an apparent excess of breast cancer, four women elected prophylactic mastectomy to prevent breast cancer, and nine others sought counseling to explore possible means of control. Counseling addressed the women's risk factors (the largest being family history), the treatment and prognosis of the tumor, and a comparison of medical surveillance and prophylactic surgery. Personalized relative risk estimates, defined for individual factors by epidemiologic studies, were summarized as the probability of developing breast cancer within five years and ranged from 0.2% to 24%. In 1- to 12-year follow-up, none developed breast cancer. Of those counseled, the five who elected subcutaneous mastectomy with implantation and the four who chose surveillance had similar magnitude and recall of risk estimates and knowledge of control options. None of those choosing surveillance had, in fact, followed a regular program of examination. Controlling breast cancer may be possible through individualized counseling of high risk women.     

Noncontraceptive estrogen use and epithelial ovarian cancer

The relation of noncontraceptive estrogen use to epithelial ovarian cancer was evaluated in a case-control study conducted in hospitals mainly in the northeastern United States. There were 377 cases diagnosed within the year before hospital admission and 2,030 hospital controls; data were collected by interview in the hospital. Compared with women who never took noncontraceptive estrogens, the overall relative risk estimate for women whose estrogen use lasted at least one year and was not combined with progestogens or testosterone was 1.2 (95% confidence interval (CI) 0.8-1.9), after taking into account risk factors for ovarian cancer. There were 55 cases of the endometrioid, clear cell, or malignant mixed mesodermal cell type; the corresponding relative risk estimate was 0.9 (95% CI 0.3-3.0). There were 26 cases of undifferentiated cell type, with a relative risk estimate of 3.6 (95% CI 1.2-11). Relative risk estimates were similar in a subset of the cases (57%) for which pathology slides were reviewed. For estrogen use of long duration, use of high-dose preparations, or use in the distant past, the relative risk estimates were not significantly different from 1.0. The estimates were elevated for some categories of use, but not consistently--for example, for an interval of 5-9 years since estrogen use began (relative risk (RR) = 2.7), but not after shorter or longer intervals, and for use of conjugated estrogens with a dose of 0.3 mg (RR = 3.2) or 1.25 mg (RR = 2.4), but not for doses of 0.625 mg or 2.5 mg. The relative risk estimate was also elevated for use by nulliparous women (RR = 2.4). The results suggest that, overall, noncontraceptive estrogen use is not associated with the risk of epithelial ovarian cancer. Furthermore, our data do not support the hypothesis that estrogens increase the risk of endometrioid ovarian cancer. The elevated estimates could be due to multiple stratification of the data, but they should be explored in further studies, given the lethality of ovarian cancer and the common use of estrogens by postmenopausal women.     

Analysis of familial breast cancer in genetic analysis workshop 9: Summary of findings

As part of the 9th Genetic Analysis Workshop held in Val Morin, Quebec, October 16-18, 1994, four workshop participants analyzed a large breast cancer data set. This data set consisted of phenotype and genetic marker data on 3884 individuals in 214 families with at least four cases of breast cancer contributed by members of an international breast cancer consortium. Two of the four papers [Barrett and Rigby; Commenges; this issue] utilized variants of affected pair methods to assess linkage of breast/ovarian cancer to the 17q markers in the data set. The third paper by Bansal et al. used a Monte-Carlo approach to examine the question of intrafamilial clustering of breast and ovarian cancer. The last paper in the series [Leal and Ott] described a method of computing support intervals for risk estimates when there are uncertainties associated with the parameter estimates used to compute these risks. © 1995 Wiley-Liss, Inc.     

Cancer of the breast and reproductive tract in relation to use of oral contraceptives

Effects of oral contraception on cancers of the female breast and reproductive tract are critically reviewed from human studies reported since 1980. The cumulative risk of breast cancer through 59 years of age appears to bear no relationship to oral contraceptive (OC) use whatsoever. Studies restricted to women under age 45, however, raise concern about a possible adverse effect from OC use before a first-term pregnancy. A duration-related protective effect against endometrial cancer occurs from use of combined OCs. The risk is reduced by about 40% with 2 years of use, and by about 60% with 4 or more years of oral contraception. Oral contraception in excess of 3 years protects against ovarian cancer. Four years of use confers a 50% reduction in risk and 7 or more years of use confers a 60%-80% reduction in ovarian cancer risk. Studies of cervical dysplasia and carcinoma in situ suggest elevated risks with 2 or more years of OC use, although results are difficult to interpret in view of numerous factors that might distort the findings. The risk of invasive cervical cancer appear to be unaffected by up to 5 years of oral contraception. Beyond this, there is evidence suggesting an elevated risk which approaches a 2-fold increase at 10 years of use. Cancers of the vagina and fallopian tube are extremely rare. Their risks have yet to be characterized in relation to oral contraception.     

Effect of Case-Mix and Random Variation on Breast Cancer Care Quality Indicators and Their Rankability

ObjectivesHospital comparisons to improve quality of care require valid and reliable quality indicators. We aimed to test the validity and reliability of 6 breast cancer indicators by quantifying the influence of case-mix and random variation.MethodsThe nationwide population-based database included 79 690 patients with breast cancer from 91 Dutch hospitals between 2011 and 2016. The indicator-scores calculated were: (1) irradical breast-conserving surgery (BCS) for invasive disease, (2) irradical BCS for ductal carcinoma-in-situ, (3) breast contour–preserving treatment, (4) magnetic resonance imaging (MRI) before neo-adjuvant chemotherapy, (5) radiotherapy for locally advanced disease, and (6) surgery within 5 weeks from diagnosis. Case-mix and random variation adjustments were performed by multivariable fixed and random effect logistic regression models. Rankability quantified the between-hospital variation, representing unexplained differences that might be the result of the level of quality of care, as low (<50%), moderate (50%-75%), or high (>75%).ResultsAll of the indicators showed between-hospital variation with wide (interquartile) ranges. Case-mix adjustment reduced variation in indicators 1 and 3 to 5. Random variation adjustment (further) reduced the variation for all indicators. Case-mix and random variation adjustments influenced the indicator-scores of individual hospitals and their ranking. Rankability was poor for indicator 1, 2, and 5, and moderate for 3, 4, and 6.ConclusionsThe 6 indicators lacked validity and/or reliability to a certain extent. Although measuring quality indicators may stimulate quality improvement in general, comparisons and judgments of individual hospital performance should be made with caution if based on indicators that have not been tested or adjusted for validity and reliability, especially in benchmarking.     

Lung cancer: district active treatment rates affect survival

STUDY OBJECTIVE: This study investigates variation in management and treatment of lung cancer patients and determines the impact of any variation in treatment on survival. DESIGN: A retrospective study of population based data held by the Northern & Yorkshire Cancer Registry and Information Service (NYCRIS), comparing active treatment rates for lung cancer with survival by districts. SETTING The then 17 districts in Yorkshire and South Humber, England. PATIENTS: 22 654 patients registered with lung cancer between 1986 and 1994 and followed up until end of 1996. RESULTS: The overall rates of active treatment (surgery, radiotherapy, and chemotherapy) varied between districts from 37% to 56%. One year survival (with 95% CI) was significantly better in the districts with highest rates of active treatment 23% (22% to 24%) compared with 19% (17% to 20%) for those with lowest treatment rates. Non-small cell lung cancer patients (55%) in the districts with highest active treatment rates had an age adjusted relative risk of death during the follow up period, relative to risk of death in the districts with the lower treatment rates of 0.88 (0.83 to 0.92). Clinically diagnosed patients (34%) had an age adjusted RR of 0.92 (0.86 to 0.96). RR in small cell cancer (11%) was not significant. CONCLUSION: This study has shown wide variations in the rates of active treatment for lung cancer patients within districts across one large region of England. Active treatment was strongly associated with improved survival, especially in non-small cell lung cancer.