Statistical issues in randomized trials of cancer screening

Micro-elimination of chronic hepatitis C virus in mental health settings: A prospective multicentre pragmatic trial

INTRODUCTION

Hepatitis C virus (HCV) prevalence is high in the mental health population. We sought to evaluate testing and treatment uptake for HCV following the implementation of a universal nurse led study in inpatient and outpatient mental health populations.

METHODS

From January 2018 to December 2020, we screened mental health inpatients (n = 322) and community mental health patients (n = 615) for HCV with either specialist hepatology nurses or mental health nurses (mental health nurse).

RESULTS

75.5% (464/615) of community patients and 100% (322/322) of inpatients consented to screening, with an HCV antibody-positive prevalence of 12.7% (59/464) in community patients and 19.6% (63/322) in inpatients. RNA detectable prevalence was 4.0% (22/464) and 7.5% (24/322), respectively. Community patients who were screened by specialist hepatology nurses were more likely to consent to screening (94.4% vs. 45.7%, p < 0.001) but had lower proportion of HCV antibody (10.5% vs. 20.3%, p < 0.001) and RNA detectable (4.0% vs. 7.5%, p = 0.018) when compared to mental health nurse screening. Engagement with treatment was 27.0% of community mental health patients and 45.8% of mental health inpatients undergoing treatment. All patients undergoing treatment and underwent sustained viral response (SVR) testing achieved SVR.

DISCUSSION AND CONCLUSIONS

Universal screening of HCV using a nurse-led model has high rates of success in mental health patients with high proportions undergoing screening, with no reduction in the rates of SVR achieved with DAA therapy compared to the general population. Further work is needed to bridge the gap between identification of HCV and treatment among mental health patients.

Epidemiology of Cancer in Kidney Transplant Recipients

Kidney transplantation is the ideal treatment modality for patients with end-stage kidney disease, with excellent outcomes post-transplant compared with dialysis. However, kidney transplant recipients are at increased risk of infections and cancer because of the need for immunosuppression. Kidney transplant recipients have approximately two to three times greater risk of developing cancer than the general population, and cancer is a major contributor to morbidity and mortality. Most of the increased risk is driven by viral-mediated cancers such as post-transplant lymphoproliferative disorder, anogenital cancers, and Kaposi sarcoma. Nonmelanoma skin cancer is the most frequent type of cancer in kidney transplant recipients, likely due to an interaction between ultraviolet radiation exposure and decreased immune surveillance. Occurrence of the more common types of solid organ cancers seen in the general population, such as breast, prostate, lung, and colorectal cancers, is not, or is only mildly, increased post-transplant. Clinical care and future research should focus on prevention and on improving outcomes for important immunosuppression-related malignancies, and treatment options for other cancers occurring in the transplant setting.

Evaluating cancer screening programs using survival analysis

The main purpose of cancer screening programs is to provide early treatment to patients that are diagnosed with cancer on a screening test, thus increasing their chances of survival. To test this hypothesis directly, one should compare the survival of screen-detected cases to the survival of their counterparts not included to the program. In this study, we develop a general notation and use it to formally define the comparison of interest. We explain why the naive comparison between screen-detected and interval cases is biased and show that the total bias that arises in this case can be decomposed as a sum of lead time bias, length time bias, and bias due to overdetection. With respect to the estimation, we show what can be estimated using existing methods. To fill in the missing gap, we develop a new nonparametric estimator that allows us to estimate the survival of the control group, that is, the survival of cancer cases that would be screen-detected among those not included to the program. By joining the proposed estimator with existing methods, we show that the contrast of interest can be estimated without neglecting any of the biases. Our approach is illustrated using simulations and empirical data.

Evidence of cancer progression as the cause of death in men with prostate cancer in Sweden

OBJECTIVE

To assess the strength of the evidence indicative of prostate cancer (PCa) progression as the adjudicated cause of death, according to age at death and PCa risk category.

PATIENTS AND METHODS

Using data from the Prostate Cancer data Base Sweden, we identified a study frame of 5543 men with PCa registered as the cause of death according to the Cause of Death Register. We assessed the evidence of PCa progression through a review of healthcare records for a stratified sample of 495/5543. We extracted data on prostate-specific antigen levels, presence of metastases on imaging, and PCa treatments, and quantified the evidence of disease progression using a points system.

RESULTS

Both no evidence and moderate evidence for PCa progression was more common in men aged >85 years at death than those aged <85 years (29% vs 14%). Among the latter, the proportion with no evidence or moderate evidence for PCa progression was 21% for low-risk, 14% for intermediate-risk, 8% for high-risk, and 0% for metastatic PCa. In contrast, in men aged >85 years, there was little difference in the proportion with no evidence or moderate evidence of PCa progression between PCa risk categories; 31% for low-risk, 29% for intermediate-risk, 29% for high-risk, and 21% for metastatic PCa. Of the 5543 men who died from PCa, 13% (95% confidence interval 5-19%) were estimated to have either no evidence or moderate evidence of PCa progression.

CONCLUSIONS

Weak evidence for PCa progression as cause of death was more common in older men with PCa and in those with low-risk PCa. This has implications for interpretation of mortality statistics especially when assessing screening and early treatment of PCa because the beneficial effect of earlier diagnosis could be masked by erroneous adjudication of PCa as cause of death in older men, particular those with localised disease at diagnosis.

Causal Proportional Hazards Estimation with a Binary Instrumental Variable

Instrumental variables (IV) are a useful tool for estimating causal effects in the presence of unmeasured confounding. IV methods are well developed for uncensored outcomes, particularly for structural linear equation models, where simple two-stage estimation schemes are available. The extension of these methods to survival settings is challenging, partly because of the nonlinearity of the popular survival regression models and partly because of the complications associated with right censoring or other survival features. Motivated by the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer screening trial, we develop a simple causal hazard ratio estimator in a proportional hazards model with right censored data. The method exploits a special characterization of IV which enables the use of an intuitive inverse weighting scheme that is generally applicable to more complex survival settings with left truncation, competing risks, or recurrent events. We rigorously establish the asymptotic properties of the estimators, and provide plug-in variance estimators. The proposed method can be implemented in standard software, and is evaluated through extensive simulation studies. We apply the proposed IV method to a data set from the Prostate, Lung, Colorectal and Ovarian cancer screening trial to delineate the causal effect of flexible sigmoidoscopy screening on colorectal cancer survival which may be confounded by informative noncompliance with the assigned screening regimen.

Long-term deaths from melanoma according to tumor thickness at diagnosis

There is little long-term follow-up information about how the number of melanoma deaths and case fatality vary over time according to the measured thickness of melanoma at diagnosis. This population-based longitudinal cohort study examines patterns and trends in case fatality among 44,531 people in Queensland (Australia) diagnosed with a single invasive melanoma (International Classification of Diseases for Oncology, third revision [ICD-O-3], C44, Morphology 872-879) between 1987 and 2011, including 11,883 diagnosed between 1987 and 1996, with up to 20 years follow-up (to December 2016). The 20-year case fatality increased by thickness, with the percentage of melanoma deaths within 20 years of diagnosis being up to 4.8% for melanomas with measured thickness <0.80 mm, 10.6% for tumors 0.8 to <1.0 mm and generally more than 30% for melanomas measuring 3 mm and more. For melanomas <1.0 mm, most deaths occurred between 5 and 20 years after diagnosis, whereas for thicker melanomas the reverse was true with most deaths occurring within the first 5 years. Five-year case fatality decreased over successive calendar time periods for melanomas <1.0 mm, but not for melanomas ≥1.0 mm. These findings demonstrate that the time course for fatal melanomas varies markedly according to tumor thickness at diagnosis. Improved understanding of the patient factors and characteristics of melanomas, in addition to tumor thickness, which increase the likelihood of progression, is needed to guide clinical diagnosis, communication with patients and ongoing surveillance pathways of patients with potentially fatal lesions.

The WISDOM Personalized Breast Cancer Screening Trial: Simulation Study to Assess Potential Bias and Analytic Approaches

Background

WISDOM (Women Informed to Screen Depending on Measures of Risk) is a randomized trial to assess whether personalized breast cancer screening—where women are screened biannually, annually, biennially, or not at all depending on risk and age—can prevent as many advanced (stage IIB or higher) cancers as annual screening in women ages 40–74 years across 5 years of trial time. The short study time in combination with design choices of not requiring study entry and exit mammograms for all participants may introduce different sources of bias in favor of either the personalized or the annual arm.

Methods

We designed a simulation model and performed 5000 virtual WISDOM trials to assess potential biases. Each virtual trial simulated 65 000 randomly assigned participants who were each assigned a risk stratum and a time to stage of at least IIB cancer sampled from an exponential distribution with the hazard rate based on the risk stratum. Results from the virtual trials were used to evaluate two candidate analysis strategies with respect to susceptibility for introducing bias: 1) difference between arms in total number of events over total trial time, and 2) difference in number of events within complete screening cycles.

Results

Based on the simulations, about 86 stage IIB or higher cancers will be detected within the trial and the total exposure time will be about 74 000 years in each arm. Potential ascertainment bias is introduced at study entry and exit. Analysis strategy 1 works better for the nonscreened stratum, whereas method 2 is considerably more unbiased for the strata of women screened biennially or every 6 months.

Conclusion

Combining the two candidate analysis approaches gives a reasonably unbiased analysis based on the simulations and is the method we will use for the primary analysis in WISDOM. Publishing the WISDOM analysis approach provides transparency and can aid the design and analysis of other individualized screening trials.

Limitations in the Effect of Screening on Breast Cancer Mortality

Purpose

Randomized, controlled trials showed that screening reduces breast cancer mortality rates, but some recent observational studies have concluded that programmatic screening has had minor effect on breast cancer mortality rates. This apparent contradiction might be explained by the use of aggregated data in observational studies. We assessed the long-term effect of screening using individual-level data.

Materials and Methods

Using data from mammography screening in the Copenhagen and Danish national registers, we compared the observed breast cancer mortality rate in women invited to screening with the expected rate in absence of screening. The effect was examined using the “naïve model,” which included all breast cancer deaths; the “follow-up model,” which counted only breast cancer deaths in women diagnosed after their first invitation to screening; and the “evaluation model,” which is similar to the follow-up model during screening age, but after screening age, which counted only breast cancer deaths and person-years in women diagnosed during screening age.

Results

We included 18,781,292 person-years, 976,743 of which were from women invited to screening. The naïve and follow-up models showed, respectively, 10% and 11% reduction in breast cancer mortality after invitation to screening. However, many breast cancer deaths occurred in women whose cancer was diagnosed when they were no longer eligible for screening. Accounting for this dilution, the evaluation model showed a 20% (95% CI, 10% to 29%) reduction in breast cancer mortality after invitation to screening.

Conclusion

Screening had a clear long-term beneficial effect with a 20% reduction in breast cancer–associated mortality in the invited population. However, this effect was, by nature, restricted to breast cancer deaths in women who could potentially benefit from screening. Our study highlights the complexity in evaluating the long-term effect of breast cancer screening from observational data.

Meta-analysis of breast cancer mortality benefit and overdiagnosis adjusted for adherence: improving information on the effects of attending screening mammography

BACKGROUND

Women require information about the impact of regularly attending screening mammography on breast cancer mortality and overdiagnosis to make informed decisions. To provide this information we aimed to meta-analyse randomised controlled trials adjusted for adherence to the trial protocol.

METHODS

Nine screening mammography trials used in the Independent UK Breast Screening Report were selected. Extending an existing approach to adjust intention-to-treat (ITT) estimates for less than 100% adherence rates, we conducted a random-effects meta-analysis. This produced a combined deattenuated prevented fraction and a combined deattenuated percentage risk of overdiagnosis.

RESULTS

In women aged 39-75 years invited to screen, the prevented fraction of breast cancer mortality at 13-year follow-up was 0.22 (95% CI 0.15-0.28) and it increased to 0.30 (95% CI 0.18-0.42) with deattenuation. In women aged 40-69 years invited to screen, the ITT percentage risk of overdiagnosis during the screening period was 19.0% (95% CI 15.2-22.7%), deattenuation increased this to 29.7% (95% CI 17.8-41.5%).

CONCLUSIONS

Adjustment for nonadherence increased the size of the mortality benefit and risk of overdiagnosis by up to 50%. These estimates are more appropriate when developing quantitative information to support individual decisions about attending screening mammography.

Correction factors for self-selection when evaluating screening programmes

Objective

In screening programmes there is recognized bias introduced through participant self-selection (the healthy screenee bias). Methods used to evaluate screening programmes include Intention-to-screen, per-protocol, and the “post hoc” approach in which, after introducing screening for everyone, the only evaluation option is participants versus non-participants. All methods are prone to bias through self-selection. We present an overview of approaches to correct for this bias.

Methods

We considered four methods to quantify and correct for self-selection bias. Simple calculations revealed that these corrections are actually all identical, and can be converted into each other. Based on this, correction factors for further situations and measures were derived. The application of these correction factors requires a number of assumptions.

Results

Using as an example the German Neuroblastoma Screening Study, no relevant reduction in mortality or stage 4 incidence due to screening was observed. The largest bias (in favour of screening) was observed when comparing participants with non-participants.

Conclusions

Correcting for bias is particularly necessary when using the post hoc evaluation approach, however, in this situation not all required data are available. External data or further assumptions may be required for estimation.

Prostate cancer mortality in areas with high and low prostate cancer incidence

Background

The effect of prostate-specific antigen (PSA) screening on prostate cancer mortality remains debated, despite evidence from randomized trials. We investigated the association between prostate cancer incidence, reflecting uptake of PSA testing, and prostate cancer mortality.

Methods

The study population consisted of all men aged 50 to 74 years residing in eight counties in Sweden with an early increase in prostate cancer incidence and six counties with a late increase during two time periods. Incidence of metastatic prostate cancer was investigated in the period from 2000 to 2009, and prostate cancer–specific mortality and excess mortality were investigated in the period from 1990 to 1999 and the period from 2000 to 2009 by calculating rate ratios for high- vs low-incidence counties and rate ratios for the period from 2000 to 2009 vs the period from 1990 to 1999 within these two groups. All statistical tests were two-sided.

Results

There were 4528134 person-years at risk, 1577 deaths from prostate cancer, and 1210 excess deaths in men with prostate cancer in high-incidence counties and 2471373 person-years at risk, 985 prostate cancer deaths, and 878 excess deaths in low-incidence counties in the period from 2000 to 2009. Rate ratios in counties with high vs low incidence adjusted for time period were 0.81 (95% confidence interval [CI] = 0.73 to 0.90) for prostate cancer– specific mortality and 0.74 (95% CI = 0.64 to 0.86) for excess mortality, and the rate ratio of metastatic prostate cancer was 0.85 (95% CI = 0.79 to 0.92).

Conclusions

The lower prostate cancer mortality in high-incidence counties reflecting a high PSA uptake suggests that more-intense as compared with less-intense opportunistic PSA screening reduces prostate cancer mortality.

Breast cancer mortality in participants of the Norwegian Breast Cancer Screening Program

BACKGROUND

The Norwegian Breast Cancer Screening Program started in 1996. To the authors' knowledge, this is the first report using individual-based data on invitation and participation to analyze breast cancer mortality among screened and nonscreened women in the program.

METHODS

Information on dates of invitation, attendance, breast cancer diagnosis, emigration, death, and cause of death was linked by using unique 11-digit personal identification numbers assigned all inhabitants of Norway at birth or immigration. In total, 699,628 women ages 50 to 69 years without prior a diagnosis of breast cancer were invited to the program from 1996 to 2009 and were followed for breast cancer through 2009 and death through 2010. Incidence-based breast cancer mortality rate ratios (MRRs) were compared between the screened and nonscreened cohorts using a Poisson regression model. The MRRs were adjusted for calendar period, attained age, years since inclusion in the cohorts, and self-selection bias.

RESULTS

The crude breast cancer mortality rate was 20.7 per 100,000 women-years for the screened cohort compared with 39.7 per 100,000 women-years for the nonscreened cohort, resulting in an MRR of 0.52 (95% confidence interval, 0.47-0.59). The mortality reduction associated with attendance in the program was 43% (MRR, 0.57; 95% confidence interval, 0.51-0.64) after adjusting for calendar period, attained age, years after inclusion in the cohort, and self-selection bias.

CONCLUSIONS

After 15 years of follow-up, a 43% reduction in mortality was observed among women who attended the national mammographic screening program in Norway.

Most breast cancer screening trials have a flawed design

In the present article, we discuss that why most breast cancer screening trials have a flawed origin. We suggest some solutions to correct these flaws so that more valid and reliable screening trials can be conducted in the future.

Cancer screening trials: nuts and bolts

The most rigorous and valid approach to evaluating cancer screening modalities is the randomized controlled trial (RCT). RCTs are major undertakings and the intricacies of trial design, operations, and management are generally underappreciated by the typical researcher. The purpose of this article is to inform the reader of the "nuts and bolts" of designing and conducting cancer screening RCTs. Following a brief introduction as to why RCTs are critical in evaluating screening modalities, we discuss design considerations, including the choice of design type and duration of follow-up. We next present an approach to sample-size calculations. We then discuss aspects of trial implementation, including recruitment, randomization, and data management. A discussion of commonly employed data analyses comes next, and includes methods for the primary analysis (comparison of cause-specific mortality rates between the screened and control arms for the cancer of interest), as well as for secondary endpoints such as sensitivity. We follow with a discussion of sequential monitoring and interim analysis techniques, which are used to examine the primary outcome while the trial is ongoing. We close with thoughts on lessons learned from past cancer screening RCTs and provide recommendations for future trials. Throughout the presentation we illustrate topics with examples from completed or ongoing RCTs, including the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial and the National Lung Screening Trial (NLST).

Prostate cancer mortality reduction by prostate-specific antigen-based screening adjusted for nonattendance and contamination in the European Randomised Study of Screening for Prostate Cancer (ERSPC)

BACKGROUND

Prostate-specific antigen (PSA) based screening for prostate cancer (PCa) has been shown to reduce prostate specific mortality by 20% in an intention to screen (ITS) analysis in a randomised trial (European Randomised Study of Screening for Prostate Cancer [ERSPC]). This effect may be diluted by nonattendance in men randomised to the screening arm and contamination in men randomised to the control arm.

OBJECTIVE

To assess the magnitude of the PCa-specific mortality reduction after adjustment for nonattendance and contamination.

DESIGN, SETTING, AND PARTICIPANTS

We analysed the occurrence of PCa deaths during an average follow-up of 9 yr in 162,243 men 55-69 yr of age randomised in seven participating centres of the ERSPC. Centres were also grouped according to the type of randomisation (ie, before or after informed written consent).

INTERVENTION

Nonattendance was defined as nonattending the initial screening round in ERSPC. The estimate of contamination was based on PSA use in controls in ERSPC Rotterdam.

MEASUREMENTS

Relative risks (RRs) with 95% confidence intervals (CIs) were compared between an ITS analysis and analyses adjusting for nonattendance and contamination using a statistical method developed for this purpose.

RESULTS AND LIMITATIONS

In the ITS analysis, the RR of PCa death in men allocated to the intervention arm relative to the control arm was 0.80 (95% CI, 0.68-0.96). Adjustment for nonattendance resulted in a RR of 0.73 (95% CI, 0.58-0.93), and additional adjustment for contamination using two different estimates led to estimated reductions of 0.69 (95% CI, 0.51-0.92) to 0.71 (95% CI, 0.55-0.93), respectively. Contamination data were obtained through extrapolation of single-centre data. No heterogeneity was found between the groups of centres.

CONCLUSIONS

PSA screening reduces the risk of dying of PCa by up to 31% in men actually screened. This benefit should be weighed against a degree of overdiagnosis and overtreatment inherent in PCa screening.

Improving the biomarker pipeline to develop and evaluate cancer screening tests

The biomarker pipeline to develop and evaluate cancer screening tests has three stages: identification of promising biomarkers for the early detection of cancer, initial evaluation of biomarkers for cancer screening, and definitive evaluation of biomarkers for cancer screening. Statistical and biological issues to improve this pipeline are discussed. Although various recommendations, such as identifying cases based on clinical symptoms, keeping biomarker tests simple, and adjusting for postscreening noise, have been made previously, they are not widely known. New recommendations include more frequent specimen collection to help identify promising biomarkers and the use of the paired availability design with interval cases (symptomatic cancers detected in the interval after screening) for initial evaluation of biomarkers for cancer screening.

Early reporting for cancer screening trials

OBJECTIVE

Many cancer screening trials involve a screening programme of one or more screenings with follow-up after the last screening. Usually a maximum follow-up time is selected in advance. However, during the follow-up period there is an opportunity to report the results of the trial sooner than planned. Early reporting of results from a randomized screening trial is important because obtaining a valid result sooner translates into health benefits reaching the general population sooner. The health benefits are reduction in cancer deaths if screening is found to be beneficial and more screening is recommended, or avoidance of unnecessary biopsies, work-ups and morbidity if screening is not found to be beneficial and the rate of screening drops.

METHODS

Our proposed method for deciding if results from a cancer screening trial should be reported earlier in the follow-up period is based on considerations involving postscreening noise. Postscreening noise (sometimes called dilution) refers to cancer deaths in the follow-up period that could not have been prevented by screening: (1) cancer deaths in the screened group that occurred after the last screening in subjects whose cancers were not detected during the screening program and (2) cancer deaths in the control group that occurred after the time of the last screening and whose cancers would not have been detected during the screening programme had they been randomized to screening (the number of which is unobserved). Because postscreening noise increases with follow-up after the last screening, we propose early reporting at the time during the follow-up period when postscreening noise first starts to overwhelm the estimated effect of screening as measured by a z-statistic. This leads to a confidence interval, adjusted for postscreening noise, that would not change substantially with additional follow-up. Details of the early reporting rule were refined by simulation, which also accounts for multiple looks.

RESULTS

For the re-analysis of the Health Insurance Plan trial for breast cancer screening and the Mayo Lung Project for lung cancer screening, estimates and confidence intervals for the effect of screening on cancer mortality were similar on early reporting and later.

CONCLUSION

The proposed early reporting rule for a cancer screening trial with post-screening follow-up is a promising method for making results from the trial available sooner, which translates into health benefits (reduction in cancer deaths or avoidance of unnecessary morbidity) reaching the population sooner.

CT screening for lung cancer: update 2007

Screening is the pursuit of the early diagnosis of cancer before symptoms occur. The purpose of early diagnosis is to provide early treatment, which potentially prevents death from the cancer. The usefulness of screening depends on how early the cancer can be diagnosed and how many deaths can be prevented by early treatment as compared with later symptom-prompted diagnosis and treatment. The goal of the Early Lung Cancer Action Project investigators was to develop an efficient methodology that would provide an ever-accumulating, continually updated body of evidence for evaluation of emerging new technologies for screening for cancer. This methodology recognizes that screening is a sequential process that starts with the pursuit of the early diagnosis of cancer followed by early treatment. It also recognizes that diagnostic research is fundamentally different from treatment research. To fully understand the current discussions on the evidence for lung cancer screening, key definitions are provided, including the differentiation between the first, baseline round of screening and all subsequent rounds of repeat screening and baseline and repeat cancers and their distribution by cell type. These definitions are critical in analyzing the results of various screening reports as they are not used by all. To provide optimal screening, a regimen for the diagnostic workup must be specified starting with the definition of the initial test, its positive result, and the workup for a positive result leading to a diagnosis of cancer. Assessment of diagnostic performance does not require a control group, but does require confirmation of the diagnosis. For assessment of the effectiveness of early treatment, a comparison group is needed. The comparison group may be formed by randomly assigning people with screen-diagnosed lung cancer to immediate or delayed treatment, as has been done for prostate cancer. This provides a direct assessment of any potential overdiagnosis of the cancer resulting from screening. Alternatively, a quasiexperimental control group can be used consisting of participants diagnosed with the cancer who have refused or delayed their treatment even though they are candidates for it.

Population screening for lung cancer using computed tomography, is there evidence of clinical effectiveness? A systematic review of the literature

Lung cancer is the leading cause of death among all cancer types in the UK, killing approximately 34 000 people per year. By the time symptoms develop, the tumour is often at an advanced stage and the prognosis is bleak. Treatment at a less advanced stage of disease by surgical resection has been shown to substantially reduce mortality. Screening would be attractive if it could detect presymptomatic lung cancer at a stage when surgical intervention is feasible but has been the subject of scientific debate for the past three decades. The aim of this review was to examine the current evidence on the clinical effectiveness of screening for lung cancer using computed tomography. A systematic literature review searching 15 electronic databases and Internet resources from 1994 until December 2004/January 2005 was carried out. Information was summarised narratively. A total of 12 studies of computed tomography screening for lung cancer were identified including two RCTs and 10 studies of screening without comparator groups. The two RCTs were of short duration (1 year). None examined the effect of screening on mortality compared with no screening. The proportion of people with abnormal computed tomography findings varied widely between studies (5-51%). The prevalence of lung cancer detected was between 0.4% and 3.2% (number needed to screen to detect one lung cancer = 31 to 249). Incidence rates of lung cancer were lower (0.1-1%). Among the detected tumours, a high proportion were stage I or resectable tumours, 100% in some studies. Currently, there is insufficient evidence that computed tomography screening is clinically effective in reducing mortality from lung cancer.

Evaluating markers for the early detection of cancer: overview of study designs and methods

BACKGROUND

The field of cancer biomarker development has been evolving rapidly. New developments both in the biologic and statistical realms are providing increasing opportunities for evaluation of markers for both early detection and diagnosis of cancer.

PURPOSE

To review the major conceptual and methodological issues in cancer biomarker evaluation, with an emphasis on recent developments in statistical methods together with practical recommendations.

METHODS

We organized this review by type of study: preliminary performance, retrospective performance, prospective performance and cancer screening evaluation.

RESULTS

For each type of study, we discuss methodologic issues, provide examples and discuss strengths and limitations.

CONCLUSION

Preliminary performance studies are useful for quickly winnowing down the number of candidate markers; however their results may not apply to the ultimate target population, asymptomatic subjects. If stored specimens from cohort studies with clinical cancer endpoints are available, retrospective studies provide a quick and valid way to evaluate performance of the markers or changes in the markers prior to the onset of clinical symptoms. Prospective studies have a restricted role because they require large sample sizes, and, if the endpoint is cancer on biopsy, there may be bias due to overdiagnosis. Cancer screening studies require very large sample sizes and long follow-up, but are necessary for evaluating the marker as a trigger of early intervention.

Joint National Cancer Institute-Food and Drug Administration Workshop on Research Strategies, Study Designs, and Statistical Approaches to Biomarker Validation for Cancer Diagnosis and Detection

Cancer remains the second leading cause of death in the United States, in spite of tremendous advances made in therapeutic and diagnostic strategies. Successful cancer treatment depends on improved methods to detect cancers at early stages when they can be treated more effectively. Biomarkers for early detection of cancer enable screening of asymptomatic populations and thus play a critical role in cancer diagnosis. However, the approaches for validating biomarkers have yet to be addressed clearly. In an effort to delineate the ambiguities related to biomarker validation and related statistical considerations, the National Cancer Institute, in collaboration with the Food and Drug Administration, conducted a workshop in July 2004 entitled "Research Strategies, Study Designs, and Statistical Approaches to Biomarker Validation for Cancer Diagnosis and Detection." The main objective of this workshop was to review basic considerations underpinning the study designs, statistical methodologies, and novel approaches necessary to rapidly advance the clinical application of cancer biomarkers. The current commentary describes various aspects of statistical considerations and study designs for cancer biomarker validation discussed in this workshop.

Reduction in breast cancer mortality from the organised service screening with mammography: 2. Validation with alternative analytic methods

BACKGROUND

In our companion article, incidence-based mortality analysis of data from breast cancer screening programs in 13 areas in Sweden indicated a 40% to 45% reduction in incidence-based breast cancer mortality among women actually screened. In this article, we apply new analytic methods for the evaluation of breast cancer mortality, using all breast cancer deaths in the period under study.

METHODS

Data were available from 13 areas on breast cancer mortality by year of diagnosis, year of death, and screening exposure. The period of study varied by area, the overall range of year of diagnosis being 1968 to 2001. We had data on 6,231 deaths and an average population of 555,676 women ages 40 to 69 years. Analysis of the effect of being screened was conducted using an alternative statistical analysis applied to all breast cancer deaths in the period of study, in addition to the incidence-based mortality analysis in our companion article. Data were analyzed using Poisson regression and adjusted for self-selection bias, contemporaneous changes in incidence, and changes in mortality independent of screening.

RESULTS

Using all deaths in the period of observation, a significant 42% reduction in breast cancer mortality was observed, adjusting for contemporaneous changes independent of screening [relative risk (RR), 0.58; 95% confidence interval (95% CI), 0.53-0.62]. After further adjustment for self-selection bias, the mortality reduction was 39% (RR, 0.61; 95% CI, 0.55-0.68), also highly significant.

CONCLUSIONS

These results indicate a reduction in breast cancer mortality of 39% in association with screening, after adjustment for contemporaneous changes and self-selection bias. These results confirm previous conclusions arrived at using incidence-based mortality analyses.

Reduction in Breast Cancer Mortality from Organized Service Screening with Mammography: 1. Further Confirmation with Extended Data

BACKGROUND

In an earlier publication, our evaluation of data from breast cancer screening programs in seven Swedish counties suggested a 40% reduction in incidence-based breast cancer mortality among women actually screened. In the current study, we expand the previous analysis from seven counties to 13 large areas within nine counties, including six of the original counties and seven additional areas, examine a longer period of follow-up (20-44 years), apply new analytic methods for the evaluation of incidence-based breast cancer mortality, and estimate the number needed to screen to save one life.

METHODS

Data from six of the original counties (one being excluded as it does not yet have 10 years of follow-up after the initiation of screening), with increased follow-up, and seven additional large areas, within three counties, representing approximately 45% of Swedish women, provide information about age at diagnosis, age at death, and screening history for 542,187 women in the prescreening and 566,423 women in the screening epochs. Regardless of year of diagnosis, there were a total of 6,231 deaths due to breast cancer in the period of study as a whole. Of these, 4,778 were incidence-based deaths in the two epochs, i.e., death among cases diagnosed within either the prescreening or screening period. Data were analyzed using Poisson regression and adjusted, when necessary, for self-selection bias, contemporaneous changes in incidence, and changes in mortality independent of screening.

RESULTS

Attendance was uniformly high, averaging 75% in the screening epochs. Recall rates for assessment varied from 4% to 5% at the first round of screening and approximately 3% at later rounds. Detection rates averaged five breast cancers per 1,000 women screened in the first round, and four breast cancers per 1,000 women screened in subsequent rounds. There was a significant 45% reduction in incidence-based breast cancer mortality among screened women in the screening epoch relative to incidence-based breast cancer mortality in the prescreening epoch (relative risk, 0.55; 95% confidence intervals, 0.51-0.59). After adjusting for self-selection bias, there still was a significant 43% reduction in incidence-based breast cancer mortality associated with screening (relative risk, 0.57; 95% confidence intervals, 0.53-0.62).

CONCLUSIONS

These results indicate a reduction in breast cancer mortality of between 40% and 45% in association with screening, after adjustment for self-selection bias. These results were obtained with modest human costs: the number needed to screen to save one life was estimated as 472.

Computational methods in medical decision making: to screen or not to screen?

Screening for a disease such as cancer is often regarded as a beneficial and successful strategy for reducing mortality. However, as with any clinical treatment or intervention, benefit cannot be assumed, and screening can entail both costs and harms, so the screening as a 'treatment' must undergo evaluation. An evaluation requires a definition of the treatment 'benefit', design of studies to measure that benefit with as little bias and variance as possible, and the development of methods for estimating the potential benefit. In screening studies, the factors most central to the evaluation are unobservable (e.g. earliest point in time at which disease becomes detectable, or 'preclinical'; time at which disease might have been detected in the absence of screening; test sensitivity). Thus, screening programs should be evaluated on scenarios in which these factors are varied, to ensure the robustness of the estimated benefit under a variety of circumstances. This article describes the importance of computational methods and simulations to assess the benefit of screening programs, particularly for cancer, based on randomized screening trials, with special attention to benefit time, lead time, and bias due to length-biased sampling.

Comparing breast cancer mortality rates before-and-after a change in availability of screening in different regions: extension of the paired availability design

BACKGROUND

In recent years there has been increased interest in evaluating breast cancer screening using data from before-and-after studies in multiple geographic regions. One approach, not previously mentioned, is the paired availability design. The paired availability design was developed to evaluate the effect of medical interventions by comparing changes in outcomes before and after a change in the availability of an intervention in various locations. A simple potential outcomes model yields estimates of efficacy, the effect of receiving the intervention, as opposed to effectiveness, the effect of changing the availability of the intervention. By combining estimates of efficacy rather than effectiveness, the paired availability design avoids confounding due to different fractions of subjects receiving the interventions at different locations. The original formulation involved short-term outcomes; the challenge here is accommodating long-term outcomes.

METHODS

The outcome is incident breast cancer deaths in a time period, which are breast cancer deaths that were diagnosed in the same time period. We considered the plausibility of the basic five assumptions of the paired availability design and propose a novel analysis to accommodate likely violations of the assumption of stable screening effects.

RESULTS

We applied the paired availability design to data on breast cancer screening from six counties in Sweden. The estimated yearly change in incident breast cancer deaths per 100,000 persons ages 40-69 (in most counties) due to receipt of screening (among the relevant type of subject in the potential outcomes model) was -9 with 95% confidence interval (-14, -4) or (-14, -5), depending on the sensitivity analysis.

CONCLUSION

In a realistic application, the extended paired availability design yielded reasonably precise confidence intervals for the effect of receiving screening on the rate of incident breast cancer death. Although the assumption of stable preferences may be questionable, its impact will be small if there is little screening in the first time period. However, estimates may be substantially confounded by improvements in systemic therapy over time. Therefore the results should be interpreted with care.

Estimating the cumulative risk of false positive cancer screenings

BACKGROUND

When evaluating cancer screening it is important to estimate the cumulative risk of false positives from periodic screening. Because the data typically come from studies in which the number of screenings varies by subject, estimation must take into account dropouts. A previous approach to estimate the probability of at least one false positive in n screenings unrealistically assumed that the probability of dropout does not depend on prior false positives.

METHOD

By redefining the random variables, we obviate the unrealistic dropout assumption. We also propose a relatively simple logistic regression and extend estimation to the expected number of false positives in n screenings.

RESULTS

We illustrate our methodology using data from women ages 40 to 64 who received up to four annual breast cancer screenings in the Health Insurance Program of Greater New York study, which began in 1963. Covariates were age, time since previous screening, screening number, and whether or not a previous false positive occurred. Defining a false positive as an unnecessary biopsy, the only statistically significant covariate was whether or not a previous false positive occurred. Because the effect of screening number was not statistically significant, extrapolation beyond 4 screenings was reasonable. The estimated mean number of unnecessary biopsies in 10 years per woman screened is.11 with 95% confidence interval of (.10,.12). Defining a false positive as an unnecessary work-up, all the covariates were statistically significant and the estimated mean number of unnecessary work-ups in 4 years per woman screened is.34 with 95% confidence interval (.32,.36).

CONCLUSION

Using data from multiple cancer screenings with dropouts, and allowing dropout to depend on previous history of false positives, we propose a logistic regression model to estimate both the probability of at least one false positive and the expected number of false positives associated with n cancer screenings. The methodology can be used for both informed decision making at the individual level, as well as planning of health services.

Using observational data to estimate an upper bound on the reduction in cancer mortality due to periodic screening

BACKGROUND

Because randomized cancer screening trials are very expensive, observational cancer screening studies can play an important role in the early phases of screening evaluation. Periodic screening evaluation (PSE) is a methodology for estimating the reduction in population cancer mortality from data on subjects who receive regularly scheduled screens. Although PSE does not require assumptions about natural history of cancer it requires other assumptions, particularly progressive detection - the assumption that once a cancer is detected by a screening test, it will always be detected by the screening test.

METHODS

We formulate a simple version of PSE and show that it leads to an upper bound on screening efficacy if the progressive detection assumption does not hold (and any effect of birth cohort is minimal) To determine if the upper bound is reasonable, for three randomized screening trials, we compared PSE estimates based only on screened subjects with PSE estimates based on all subjects.

RESULTS

In the three randomized screening trials, PSE estimates based on screened subjects gave fairly close results to PSE estimates based on all subjects.

CONCLUSION

PSE has promise for obtaining an upper bound on the reduction in population cancer mortality rates based on observational screening data. If the upper bound estimate is found to be small and any birth cohort effects are likely minimal, then a definitive randomized trial would not be warranted.