Pitfalls of practicing cancer epidemiology in resource-limited settings: the case of survival and loss to follow-up after a diagnosis of Kaposi's sarcoma in five countries across sub-Saharan Africa.

BACKGROUND Survival after diagnosis is a fundamental concern in cancer epidemiology. In resource-rich settings, ambient clinical databases, municipal data and cancer registries make survival estimation in real-world populations relatively straightforward. In resource-poor settings, given the deficiencies in a variety of health-related data systems, it is less clear how well we can determine cancer survival from ambient data. METHODS We addressed this issue in sub-Saharan Africa for Kaposi's sarcoma (KS), a cancer for which incidence has exploded with the HIV epidemic but for which survival in the region may be changing with the recent advent of antiretroviral therapy (ART). From 33 primary care HIV Clinics in Kenya, Uganda, Malawi, Nigeria and Cameroon participating in the International Epidemiologic Databases to Evaluate AIDS (IeDEA) Consortia in 2009-2012, we identified 1328 adults with newly diagnosed KS. Patients were evaluated from KS diagnosis until death, transfer to another facility or database closure. RESULTS Nominally, 22% of patients were estimated to be dead by 2 years, but this estimate was clouded by 45% cumulative lost to follow-up with unknown vital status by 2 years. After adjustment for site and CD4 count, age <30 years and male sex were independently associated with becoming lost. CONCLUSIONS In this community-based sample of patients diagnosed with KS in sub-Saharan Africa, almost half became lost to follow-up by 2 years. This precluded accurate estimation of survival. Until we either generally strengthen data systems or implement cancer-specific enhancements (e.g., tracking of the lost) in the region, insights from cancer epidemiology will be limited.

Survival analysis of longevity in siblings

Do siblings of centenarians tend to have longer life spans? To answer this question, life spans of 184 siblings for 42 centenarians have been evaluated. Two important questions have been addressed in analyzing the sibling data. First, a standard needs to be established, to which the life spans of 184 siblings are compared. In this report, an external reference population is constructed from the U.S. life tables. Its estimated mortality rates are treated as baseline hazards from which the relative mortality of the siblings are estimated. Second, the standard survival models which assume independent observations are invalid when correlation within family exists, underestimating the true variance. Methods that allow correlations are illustrated by three different methods. First, the cumulative relative excess mortality between siblings and their comparison group is calculated and used as an effective graphic tool, along with the Product Limit estimator of the survival function. The variance estimator of the cumulative relative excess mortality is adjusted for the potential within family correlation using Taylor linearization approach. Second, approaches that adjust for the inflated variance are examined. They are adjusted one-sample log-rank test using design effect originally proposed by Rao and Scott in the correlated binomial or Poisson distribution setting and the robust variance estimator derived from the log-likelihood function of a multiplicative model. Nether of these two approaches provide correlation estimate within families, but the comparison with the comparison with the standard remains valid under dependence. Last, using the frailty model concept, the multiplicative model, where the baseline hazards are known, is extended by adding a random frailty term that is based on the positive stable or the gamma distribution. Comparisons between the two frailty distributions are performed by simulation. Based on the results from various approaches, it is concluded that the siblings of centenarians had significant lower mortality rates as compared to their cohorts. The frailty models also indicate significant correlations between the life spans of the siblings. ^

Optimally Combining Censored and Uncensored Datasets

Economists and other social scientists often face situations where they have access to two datasets that they can use but one set of data suffers from censoring or truncation. If the censored sample is much bigger than the uncensored sample, it is common for researchers to use the censored sample alone and attempt to deal with the problem of partial observation in some manner. Alternatively, they simply use only the uncensored sample and ignore the censored one so as to avoid biases. It is rarely the case that researchers use both datasets together, mainly because they lack guidance about how to combine them. In this paper, we develop a tractable semiparametric framework for combining the censored and uncensored datasets so that the resulting estimators are consistent, asymptotically normal, and use all information optimally. When the censored sample, which we refer to as the master sample, is much bigger than the uncensored sample (which we call the refreshment sample), the latter can be thought of as providing identification where it is otherwise absent. In contrast, when the refreshment sample is large and could typically be used alone, our methodology can be interpreted as using information from the censored sample to increase effciency. To illustrate our results in an empirical setting, we show how to estimate the effect of changes in compulsory schooling laws on age at first marriage, a variable that is censored for younger individuals. We also demonstrate how refreshment samples for this application can be created by matching cohort information across census datasets.

Predictors of survival in Alzheimer's disease

Alzheimer's disease (AD) is associated with greater mortality and reduced survival among individuals with Alzheimer's disease as compared to those without dementia. It is uncertain how these survival estimates change when the clinical signs and/or symptoms of comorbid conditions are present in individuals' with Alzheimer's disease. Cardiovascular risk factors such as hypertension, hyperlipidemia, congestive heart failure, coronary artery disease, and diabetes mellitus are common conditions in the aged population. Independently, these factors influence mortality and may have an additive effect on reduced survival in an individual with concomitant Alzheimer's disease. The bulk of the evidence from previous research efforts suggests an association between vascular co-morbidities and Alzheimer's disease incidence, but their role in survival remains to be elucidated. The objective of this proposed study was to examine the effects of cardiovascular comorbidities on the survival experience of individuals with probable Alzheimer's disease in order to identify prognostic factors for life expectancy following onset of disease. This study utilized data from the Baylor College of Medicine Alzheimer's Disease Center (ADC) longitudinal study of Alzheimer's disease and other memory disorders. Individuals between the ages of 55-69, 70-79, and ≥80 had a median survival from date of onset of 9.2 years, 8.0 years, and 7.2 years, respectively (p<0.001) and 5.5 years, 4.3 years, and 3.4 years from diagnosis. Sex was the strongest predictor of death from onset of AD, with females having a 30 percent lower risk compared to males. These findings further support the notion that age (both from onset and from diagnosis) and sex are the strongest predictors of survival among those with AD. ^

Simulation study of joint transition and Weibull survival model with shared parameters

This study investigates a theoretical model where a longitudinal process, that is a stationary Markov-Chain, and a Weibull survival process share a bivariate random effect. Furthermore, a Quality-of-Life adjusted survival is calculated as the weighted sum of survival time. Theoretical values of population mean adjusted survival of the described model are computed numerically. The parameters of the bivariate random effect do significantly affect theoretical values of population mean. Maximum-Likelihood and Bayesian methods are applied on simulated data to estimate the model parameters. Based on the parameter estimates, predicated population mean adjusted survival can then be calculated numerically and compared with the theoretical values. Bayesian method and Maximum-Likelihood method provide parameter estimations and population mean prediction with comparable accuracy; however Bayesian method suffers from poor convergence due to autocorrelation and inter-variable correlation. ^

The role of smoking status on postoperative complications and survival in lung cancer patients presenting to M.D. Anderson Cancer Center

It is estimated that 50% of all lung cancer patients continue to smoke after diagnosis. Many of these lung cancer patients who are current smokers often experience tremendous guilt and responsibility for their disease, and feel it might be too late for them to quit smoking. In addition, many oncologists may be heard to say that it is 'too late', 'it doesn't matter', 'it is too difficult', 'it is too stressful' for their patients to stop smoking, or they never identify the smoking status of the patient. Many oncologists feel unprepared to address smoking cessation as part of their clinical practice. In reality, physicians can have tremendous effects on motivating patients, particularly when patients are initially being diagnosed with cancer. More information is needed to convince patients to quit smoking and to encourage clinicians to assist patients with their smoking cessation. ^ In this current study, smoking status at time of lung cancer diagnosis was assessed to examine its impact on complications and survival, after exploring the reliability of smoking data that is self-reported. Logistic Regression was used to determine the risks of smoking prior to lung resection. In addition, survival analysis was performed to examine the impact of smoking on survival. ^ The reliability of how patients report their smoking status was high, but there was some discordance between current smokers and recent quitters. In addition, we found that cigarette pack-year history and duration of smoking cessation were directly related to the rate of a pulmonary complication. In regards to survival, we found that current smoking at time of lung cancer diagnosis was an independent predictor of early stage lung cancer. This evidence supports the idea that it is "never too late" for patients to quit smoking and health care providers should incorporate smoking status regularly into their clinical practice.^

Modeling survival and the development of second primary tumors among Hodgkin's disease patients

Hodgkin's disease (HD) is a cancer of the lymphatic system. Survivors of HD face varieties of consequent adverse effects, in which secondary primary tumors (SPT) is one of the most serious consequences. This dissertation is aimed to model time-to-SPT in the presence of death and HD relapses during follow-up.^ The model is designed to handle a mixture phenomenon of SPT and the influence of death. Relapses of HD are adjusted as a covariate. Proportional hazards framework is used to define SPT intensity function, which includes an exponential term to estimate explanatory variables. Death as a competing risk is considered according to different scenarios, depending on which terminal event comes first. Newton-Raphson method is used to estimate the parameter estimates in the end.^ The proposed method is applied to a real data set containing a group of HD patients. Several risk factors for the development of SPT are identified and the findings are noteworthy in the development of healthcare guidelines that may lead to the early detection or prevention of SPT.^

Racial disparities in patient survival after liver transplantation in the United States

Background. Racial disparities in healthcare span such areas as access, outcomes after procedures, and patient satisfaction. Previous work suggested that minorities experience less healthcare and worse survival rates. In adult orthotopic liver transplantation (OLT) mixed results have been reported, with some showing African-American recipients having poor survival compared to Caucasians, and others finding no such discrepancy. ^ Purpose. This study’s purpose was to analyze the most recent United Network for Organ Sharing (UNOS) data, both before and after the implementation of the Model for End-Stage Liver Disease (MELD)/Pediatric End-Stage Liver Disease (PELD) scoring system, to determine if minority racial groups still experience poor outcomes after OLT. ^ Methods. The UNOS dataset for 1992-2001 (Era I) and 2002-2007 (Era II) was used. Patient survival rates for each Era and for adult and pediatric recipients were analyzed with adjustment. A separate multivariate analysis was performed on African-American adult patients in Era II in order to identify unique predictors for poor patient survival. ^ Results. The overall study included 66,118 OLT recipients. The majority were Caucasian (78%), followed by Hispanics (13%) and African-Americans (9%). Hispanic and African-American adults were more likely to be female, have Hepatitis C, to be in the intensive care unit (ICU) or ventilated at time of OLT, to have a MELD score ≥23, to have a lower education level, and to have public insurance when compared to Caucasian adults (all p-values < 0.05). Hispanic and African-American pediatric recipients were more likely have public insurance and less likely to receive a living donor OLT than were Caucasian pediatric OLT recipients (p <0.05). There was no difference in the likelihood of having a PELD score ≥21 among racial groups (p >0.40). African-American adults in Era I and Era II had worse patient survival rates than both Caucasians and Hispanic (pair-wise p-values <0.05). This same disparity was seen for pediatric recipients in Era I, but not in Era II. Multivariate analysis of African-American recipients revealed no unique predictors of patient death. ^ Conclusions. African-American race is still a predictor of poor outcome after adult OLT, even after adjustment for multiple clinical, demographic, and liver disease severity variables. Although African-American and Hispanic subgroups share many characteristics previously thought to increase risk of post-OLT death, only African-American patients have poor survival rates when compared to Caucasians. ^

Ethnic differences in tumor characteristics, treatment and survival in glioma patients

Purpose. A descriptive analysis of glioma patients by race was carried out in order to better elucidate potential differences between races in demographics, treatment, characteristics, prognosis and survival. ^ Patients and Methods. Among 1,967 patients ≥ 18 years diagnosed with glioma seen between July 2000 and September 2006 at The University of Texas M.D. Anderson Cancer Center (UTMDACC). Data were collated from the UTMDACC Patient History Database (PHDB) and the UTMDACC Tumor Registry Database (TRDB). Chi-square analysis, uni- /multivariate Cox proportional hazards modeling and survival analysis were used to analyze differences by race. ^ Results. Demographic, treatment and histologic differences exist between races. Though risk differences were seen between races, race was not found to be a significant predictor in multivariate regression analysis after accounting for age, surgery, chemotherapy, radiation, tumor type as stratified by WHO tumor grade. Age was the most consistent predictor in risk for death. Overall survival by race was significantly different (p=0.0049) only in low-grade gliomas after adjustment for age although survival differences were very slight. ^ Conclusion. Among this cohort of glioma patients, age was the strongest predictor for survival. It is likely that survival is more influenced by age, time to treatment, tumor grade and surgical expertise rather than racial differences. However, age at diagnosis, gender ratios, histology and history of cancer differed significantly between race and genetic differences to this effect cannot be excluded. ^

The role of cardiovascular comorbidities in ovarian cancer survival

Objective. One facet of cancer care that often goes ignored is comorbidities, or diseases that exist in concert with cancer. Comorbid conditions may affect survival by influencing treatment decisions and prognosis. The purpose of this secondary data analysis was to identify whether a history of cardiovascular comorbidities among ovarian cancer patients influenced survival time at the University of Texas M. D. Anderson Cancer Center. The parent study, Project Peace, has a longitudinal design with an embedded randomized efficacy study which seeks to improve detection of depressive disorders in ovarian, peritoneal, and fallopian tube cancers. ^ Methods. Survival time was calculated for the 249 ovarian cancer patients abstracted by Project Peace staff. Cardiovascular comorbidities were documented as present, based upon information from medical records in addition to self reported comorbidities in a baseline study questionnaire. Kaplan-Meier survival curves were used to compare survival time among patients with a presence or absence of particular cardiovascular comorbidities. Cox Regression proportional models accounted for multivariable factors such as age, staging, family history of cardiovascular comorbidities, and treatment. ^ Results. Among our patient population, there was a statistically significant relationship between shorter survival time and a history of thrombosis, pericardial disease/tamponade, or COPD/pulmonary hypertension. Ovarian cancer patients with a history of thrombosis lived approximately half as long as patients without thrombosis (58.06 months vs. 121.55 months; p=.001). In addition, patients who suffered from pericardial disease/tamponade had poorer survival than those without a history of pericardial disease/tamponade (48 months vs. 80.07 months; p=.002). Ovarian cancer patients with a history of COPD or pulmonary hypertension had a median survival of 60.2 months, while the median survival for patients without these comorbidities was 80.2 months (p=.014). ^ Conclusion. Especially because of its relatively lower survival rate, greater emphasis needs to be placed on the potential influence of cardiovascular comorbid conditions in ovarian cancer.^

Hormone therapy and physical activity as predictors of lung cancer risk and survival: The California Teachers Study

Cigarette smoking is responsible for the majority of lung cancer cases worldwide; however, a proportion of never smokers still develop lung cancer over their lifetime, prompting investigation into additional factors that may modify lung cancer incidence, as well as mortality. Although hormone therapy (HT), physical activity (PA), and lung cancer have been previously examined, the associations remain unclear. This study investigated exposure to HT and PA that may modulate underlying mechanisms of lung cancer etiology and progression among women by using existing, de-identified data from the California Teachers Study (CTS).^ The CTS cohort, established in 1995–1996, has 133,479 active and retired female teachers and administrators, recruited through the California State Teachers Retirement System, and followed annually for cancer diagnosis, death, and change of address. Each woman enrolled in the CTS returned a questionnaire covering a wide variety of issues related to cancer risk and women's health, including recent and past HT use and physical activity, as well as active and environmental cigarette smoke exposure. Complete data to assess the associations between HT and lung cancer risk and survival were available for 60,592 postmenopausal women. Between 1995 and 2007, 727 of these women were diagnosed with invasive lung cancer; 441 of these died. Complete data to assess the associations between PA and lung cancer risk and survival were available for 118,513 women. Between 1995 and 2007, 853 of these women were diagnosed with invasive lung cancer; 516 of these died.^ After careful adjustment for smoking habits and other potential confounders, no measure of HT use was associated with lung cancer risk; however, any HT use (vs. no use) was associated with a decrease in lung-cancer-specific mortality. Specifically, among women who only used estrogen (E-only), decreases in lung cancer mortality were seen for recent use, but not for former use; no association was observed for estrogen plus progestin (E+P). Furthermore, among former users of HT, a statistically significant decrease in lung cancer mortality was observed for E-only use within 5 years prior to baseline, but not for E-only use >5 years prior to baseline. Neither long-term recreational PA nor recent recreational PA alone were associated with lung cancer risk; however, among women with a BMI<25 and ever smokers, high long-term moderate+strenuous PA was associated with a decrease in lung cancer risk. Women with non-local disease showed a decrease in lung cancer mortality associated with increasing duration of strenuous long-term activity, and 1.50-3.00 h/wk/y of recent moderate or recent strenuous PA. Long-term moderate PA was associated with decreased lung cancer mortality in never smokers, whereas recent moderate PA was associated with increased lung cancer mortality in current smokers. ^ Placing our findings in the context of the current literature, HT does not appear to be associated with lung cancer risk and previous studies reporting a protective effect of HT use on lung cancer risk may be subject to residual confounding by smoking. Looking at our findings regarding PA overall, the evidence still remains inconclusive regarding whether or not physical activity influence lung cancer risk or mortality. Our results suggest that recreational PA may associated with decreased lung cancer risk among women with BMI<25 and ever smoking-women; however, residual confounding by smoking should be strongly considered. To our knowledge, this is the first study to investigate lifetime recreational PA and lung cancer mortality among women. Our results contribute to the growing body of knowledge regarding non-smoking-related risk factors for lung cancer incidence and mortality among women. Given the potential clinical and interventional significance, further study and validation of these findings is warranted.^

Demographic analysis of Certified Industrial Hygienists (CIH) with the use of survival analysis techniques

Objective. The goal of this study is to characterize the current workforce of CIHs, the lengths of professional practice careers of the past and current CIHs.^ Methods. This is a secondary data analysis of data compiled from all of the nearly 50 annual roster listings of the American Board of Industrial Hygiene (ABIH) for Certified Industrial Hygienists active in each year since 1960. Survival analysis was performed as a technique to measure the primary outcome of interest. The technique which was involved in this study was the Kaplan-Meier method for estimating the survival function.^ Study subjects: The population to be studied is all Certified Industrial Hygienists (CIHs). A CIH is defined by the ABIH as an individual who has achieved the minimum requirements for education, working experience and through examination, has demonstrated a minimum level of knowledge and competency in the prevention of occupational illnesses. ^ Results. A Cox-proportional hazards model analysis was performed by different start-time cohorts of CIHs. In this model we chose cohort 1 as the reference cohort. The estimated relative risk of the event (defined as retirement, or absent from 5 consecutive years of listing) occurred for CIHs for cohorts 2,3,4,5 relative to cohort 1 is 0.385, 0.214, 0.234, 0.299 relatively. The result show that cohort 2 (CIHs issued from 1970-1980) has the lowest hazard ratio which indicates the lowest retirement rate.^ Conclusion. The manpower of CIHs (still actively practicing up to the end of 2009) increased tremendously starting in 1980 and grew into a plateau in recent decades. This indicates that the supply and demand of the profession may have reached equilibrium. More demographic information and variables are needed to actually predict the future number of CIHs needed. ^

Modeling the survival of T-cell lymphocytes using compound branching processes

Radiotherapy has been a method of choice in cancer treatment for a number of years. Mathematical modeling is an important tool in studying the survival behavior of any cell as well as its radiosensitivity. One particular cell under investigation is the normal T-cell, the radiosensitivity of which may be indicative to the patient's tolerance to radiation doses.^ The model derived is a compound branching process with a random initial population of T-cells that is assumed to have compound distribution. T-cells in any generation are assumed to double or die at random lengths of time. This population is assumed to undergo a random number of generations within a period of time. The model is then used to obtain an estimate for the survival probability of T-cells for the data under investigation. This estimate is derived iteratively by applying the likelihood principle. Further assessment of the validity of the model is performed by simulating a number of subjects under this model.^ This study shows that there is a great deal of variation in T-cells survival from one individual to another. These variations can be observed under normal conditions as well as under radiotherapy. The findings are in agreement with a recent study and show that genetic diversity plays a role in determining the survival of T-cells. ^

Geographic variation in the utilization and survival associated with oxaliplatin chemotherapy in patients with stage III colon cancer

Background: Overall objectives of this dissertation are to examine the geographic variation and socio-demographic disparities (by age, race and gender) in the utilization and survival of newly FDA-approved chemotherapy agents (Oxaliplatin-containing regimens) as well as to determine the cost-effectiveness of Oxaliplatin in a large nationwide and population-based cohort of Medicare patients with resected stage-III colon cancer. Methods: A retrospective cohort of 7,654 Medicare patients was identified from the Surveillance, Epidemiology and End Results – Medicare linked database. Multiple logistic regression was performed to examine the relationship between receipt of Oxaliplatin-containing chemotherapy and geographic regions while adjusting for other patient characteristics. Cox proportional hazard model was used to estimate the effect of Oxaliplatin-containing chemotherapy on the survival variation across regions using 2004-2005 data. Propensity score adjustments were also made to control for potential bias related to non-random allocation of the treatment group. We used Kaplan-Meier sample average estimator to calculate the cost of disease after cancer-specific surgery to death, loss-to follow-up or censorship. Results: Only 51% of the stage-III patients received adjuvant chemotherapy within three to six months of colon-cancer specific surgery. Patients in the rural regions were approximately 30% less likely to receive Oxaliplatin chemotherapy than those residing in a big metro region (OR=0.69, p=0.033). The hazard ratio for patients residing in metro region was comparable to those residing in big metro region (HR: 1.05, 95% CI: 0.49-2.28). Patients who received Oxalipaltin chemotherapy were 33% less likely to die than those received 5-FU only chemotherapy (adjusted HR=0.67, 95% CI: 0.41-1.11). KMSA-adjusted mean payments were almost 2.5 times higher in the Oxaliplatin-containing group compared to 5-FU only group ($45,378 versus $17,856). When compared to no chemotherapy group, ICER of 5-FU based regimen was $12,767 per LYG, and ICER of Oxaliplatin-chemotherapy was $60,863 per LYG. Oxaliplatin was found economically dominated by 5-FU only chemotherapy in this study population. Conclusion: Chemotherapy use varies across geographic regions. We also observed considerable survival differences across geographic regions; the difference remained even after adjusting for socio-demographic characteristics. The cost-effectiveness of Oxaliplatin in Medicare patients may be over-estimated in the clinical trials. Our study found 5-FU only chemotherapy cost-effective in adjuvant settings in patients with stage-III colon cancer.^

Data quality in trauma transfusion studies and the impact of missing data on predicting massive transfusion

Maximizing data quality may be especially difficult in trauma-related clinical research. Strategies are needed to improve data quality and assess the impact of data quality on clinical predictive models. This study had two objectives. The first was to compare missing data between two multi-center trauma transfusion studies: a retrospective study (RS) using medical chart data with minimal data quality review and the PRospective Observational Multi-center Major Trauma Transfusion (PROMMTT) study with standardized quality assurance. The second objective was to assess the impact of missing data on clinical prediction algorithms by evaluating blood transfusion prediction models using PROMMTT data. RS (2005-06) and PROMMTT (2009-10) investigated trauma patients receiving ≥ 1 unit of red blood cells (RBC) from ten Level I trauma centers. Missing data were compared for 33 variables collected in both studies using mixed effects logistic regression (including random intercepts for study site). Massive transfusion (MT) patients received ≥ 10 RBC units within 24h of admission. Correct classification percentages for three MT prediction models were evaluated using complete case analysis and multiple imputation based on the multivariate normal distribution. A sensitivity analysis for missing data was conducted to estimate the upper and lower bounds of correct classification using assumptions about missing data under best and worst case scenarios. Most variables (17/33=52%) had <1% missing data in RS and PROMMTT. Of the remaining variables, 50% demonstrated less missingness in PROMMTT, 25% had less missingness in RS, and 25% were similar between studies. Missing percentages for MT prediction variables in PROMMTT ranged from 2.2% (heart rate) to 45% (respiratory rate). For variables missing >1%, study site was associated with missingness (all p≤0.021). Survival time predicted missingness for 50% of RS and 60% of PROMMTT variables. MT models complete case proportions ranged from 41% to 88%. Complete case analysis and multiple imputation demonstrated similar correct classification results. Sensitivity analysis upper-lower bound ranges for the three MT models were 59-63%, 36-46%, and 46-58%. Prospective collection of ten-fold more variables with data quality assurance reduced overall missing data. Study site and patient survival were associated with missingness, suggesting that data were not missing completely at random, and complete case analysis may lead to biased results. Evaluating clinical prediction model accuracy may be misleading in the presence of missing data, especially with many predictor variables. The proposed sensitivity analysis estimating correct classification under upper (best case scenario)/lower (worst case scenario) bounds may be more informative than multiple imputation, which provided results similar to complete case analysis.^