Advertisement
Mayo Clinic Proceedings Home

To Screen or Not to Screen

Reconciling Individual and Population Perspectives on Screening
  • L. Rachid Salmi
    Correspondence
    Correspondence: Address to L. Rachid Salmi, MD, PhD, Institut de Santé Publique, d'Épidémiologie et de Développement, Université de Bordeaux, 146 rue Léo-Saignat, F-33076 Bordeaux cedex, France.
    Affiliations
    University of Bordeaux, ISPED, Centre INSERM U-1219 Bordeaux Population Health, Bordeaux, France

    INSERM, ISPED, Centre INSERM U-1219 Bordeaux Population Health, Bordeaux, France

    CHU de Bordeaux, Pole de Sante Publique, Service d'Information Medicale, Bordeaux, France
    Search for articles by this author
  • Gaëlle Coureau
    Affiliations
    University of Bordeaux, ISPED, Centre INSERM U-1219 Bordeaux Population Health, Bordeaux, France

    INSERM, ISPED, Centre INSERM U-1219 Bordeaux Population Health, Bordeaux, France

    CHU de Bordeaux, Pole de Sante Publique, Service d'Information Medicale, Bordeaux, France
    Search for articles by this author
  • Marion Bailhache
    Affiliations
    University of Bordeaux, ISPED, Centre INSERM U-1219 Bordeaux Population Health, Bordeaux, France

    INSERM, ISPED, Centre INSERM U-1219 Bordeaux Population Health, Bordeaux, France

    CHU de Bordeaux, Pole de Pediatrie, Place Amélie Raba Léon, Bordeaux, France
    Search for articles by this author
  • Simone Mathoulin-Pélissier
    Affiliations
    University of Bordeaux, ISPED, Centre INSERM U-1219 Bordeaux Population Health, Bordeaux, France

    INSERM, ISPED, Centre INSERM U-1219 Bordeaux Population Health, Bordeaux, France

    Institut Bergonié, Unité de Recherche et d'Épidémiologie Cliniques, Inserm CIC1401, Bordeaux, France
    Search for articles by this author

      Abstract

      Screening is the early detection of a latent disorder by a test to allow early intervention with the aim of improving prognosis. Individual and population perspectives on screening are perceived as opposing interests of patients and the population. In this article, we try to reconcile these perspectives. The individual perspective is based on the clinical experience of a better prognosis at early stages and patients with missed opportunities. In the population perspective, screening is based on a population-oriented, evidence-based model and addresses the acceptability and possible negative effects, including for people without the disorder. Known possible obstacles to a positive effect of screening include a short latent stage, lead time, overdiagnosis, lack of acceptability, poor performance of tests, and misclassification of outcome. Randomized trials of screening are challenging and need an adaptation of standards such as the Consolidated Standards of Reporting Trials (CONSORT). Simulating the effects of screening can allow the consideration of complex screening strategies and other options to help avoid biases related to treatment improvement and prevention success. Reconciling both perspectives is possible by considering that hypotheses underlying the former are prerequisites for the latter. From an evidence-based medicine and policy perspective, we suggest that recommending screening or prescribing a test is unethical if all possible obstacles are not documented by providing the best available evidence.

      Abbreviations and Acronynms:

      CONSORT (Consolidated Standards of Reporting Trials), CT (computed tomographic), DET (detection bias), ERSPC (European Randomized Study of Screening for Prostate Cancer), FN (false-negative), FP (false-positive), LTB (lead-time bias), NELSON (European NEderlands-Leuvens Screening ONderzoek), NLST (National Lung Screening Trial), OVD (overdiagnosis), PSA (prostate-specific antigen), PSB (prevention success bias), SDB (sticky diagnosis bias), SEL (selection bias), SLB (slippery linkage bias), TIB (treatment improvement bias), TN (true-negative), TP (true-positive)
      Screening is the early detection of a latent disorder with the goal of starting treatment early to improve prognosis or intervening to avoid unwanted consequences of the disorder.
      • Wilson J.M.G.
      • Jungner G.
      The Principles and Practice of Screening for Disease.
      For example, screening for breast cancer with mammography is proposed in many developed countries to initiate treatment of localized tumors and hopefully achieve cure.
      • Nelson H.D.
      • Tyne K.
      • Naik A.
      • Bougatsos C.
      • Chan B.K.
      • Humphrey L.
      Screening for breast cancer: an update for the U.S. Preventive Services Task Force.
      Screening of blood donors for infections is performed to avoid contamination of transfusion recipients.
      • Salunkhe V.
      • van der Meer P.F.
      • de Korte D.
      • Seghatchian J.
      • Gutiérrez L.
      Development of blood transfusion product pathogen reduction treatments: a review of methods, current applications and demands.
      Screening is fully described as a test proposed for a target population, and an organized process including a frequency of screening, a confirmed diagnostic procedure for those who tested positive, an early intervention for those confirmed, and an outcome to improve.
      When targeted at a complete population, screening is called systematic screening.
      • Morrison A.S.
      Screening in Chronic Disease.
      Screening can be limited to an explicitly defined subgroup based on criteria such as age or other markers of high risk of developing the disorder or a poor prognosis; it is then called targeted or selective screening.
      • Morrison A.S.
      Screening in Chronic Disease.
      Systematic and targeted screening are 2 forms of what is also called mass screening.
      • Morrison A.S.
      Screening in Chronic Disease.
      Screening can also be proposed outside of any program to patients consulting a physician for an unrelated problem; this opportunistic approach is called case finding.
      • Fletcher R.H.
      • Fletcher S.W.
      • Wagner E.H.
      Clinical Epidemiology: The Essentials.
      Whereas systematic and selective screening are usually performed within programs developed with a population-based, public health perspective, case finding depends on the physician's clinical, individual perspective on what is best for his or her patient.
      The individual and population perspectives are perceived as opposing interests of patients and populations. Briefly, screening may seem like an obviously good idea from a clinical experience perspective because a perfect test would allow either the chance for reassurance or the possibility of improving prognosis with early intervention. Nevertheless, screening is a complex intervention from the population perspective,
      • Campbell M.
      • Fitzpatrick R.
      • Haines A.
      • et al.
      Framework for design and evaluation of complex interventions to improve health.
      therefore implying specific evaluations of the advantages and disadvantages for all individuals to whom the test will be proposed, including people with and without the disorder.
      Herein, using examples of screening for chronic diseases (eg, cancer, human immunodeficiency virus infection, and psychosis) and in specific populations (eg, children and the elderly), we illustrate how screening is even more complex than usually perceived, and we try to reconcile the individual and population perspectives. We define and illustrate issues and biases that are potential obstacles to the implementation or positive effects of screening and that need to be addressed in clinical trials and modeling of the effects of screening; we expect to promote adequate methods to document the effects of screening and to improve evidence-based decisions regarding screening.

      Individual Perspective on Screening

      The Obvious Seduction of Early Detection

      The individual perspective considers that screening anticipates an intervention for patients with a poor prognosis: (1) for many disorders, patients have a better prognosis or response to treatment at early compared with later stages
      • Rosenberg J.
      • Chia Y.L.
      • Plevritis S.
      The effect of age, race, tumor size, tumor grade, and disease stage on invasive ductal breast cancer survival in the U.S. SEER database.
      ; (2) some patients seen late in the disorder process have a history of previous contacts with health services for manifestations possibly linked to early stages of the disorder.
      • Lyratzopoulos G.
      • Vedsted P.
      • Singh H.
      Understanding missed opportunities for more timely diagnosis of cancer in symptomatic patients after presentation.
      For example, proposals of early detection of child abuse have been triggered by the observation of children seen at the stage of serious or lethal injuries but who had been previously seen by social services or in emergency departments.
      • Tursz A.
      • Crost M.
      • Gerbouin-Rérolle P.
      • Cook J.
      Underascertainment of child abuse fatalities in France: retrospective analysis of judicial data to assess underreporting of infant homicides in mortality statistics.
      • Thorpe E.L.
      • Zuckerbraun N.S.
      • Wolford J.E.
      • Berger R.P.
      Missed opportunities to diagnose child physical abuse.
      In such contexts, it is logical to hypothesize that an active search for early stages of the disorder should improve the prognosis (Figure 1A).
      Figure 1
      Figure 1Hypotheses underlying screening according to the individual (A) and population (B) perspectives. FN = false-negative; FP = false-positive; TN = true-negative; TP = true-positive.
      This logical hypothesis explains why screening is so “popular” among physicians dealing in their practice with patients with a late diagnosis or a poor prognosis. This popularity is reflected by the US Preventive Services Task Force recommendations: as of July 2015, of 150 active recommendations, 106 (70.7%) are about screening. Still, the difficulty of screening is reflected by the many instances in which screening is contraindicated (23 grade D recommendations) or the evidence is insufficient (39 grade I recommendations) (for details, see Supplemental Table 1, available online at http://www.mayoclinicproceedings.org).

      Difficulties in Judging the Clinical Effects of Screening

      The hypothesis underlying the individual perspective is logical, but it can be flawed in three ways. First, early detection is possible only if there is a preclinical stage (obstacle 1) and if this stage is long enough for patients with a poor prognosis, if untreated, to have time to benefit from earlier treatment.
      • Wilson J.M.G.
      • Jungner G.
      The Principles and Practice of Screening for Disease.
      The preclinical stage is the period when the disorder is present but without any of the manifestations that usually trigger diagnosis. Many cancers have a potentially long preclinical phase, making screening relevant and implying repeated testing. In other instances, the preclinical phase is always short (bacterial contamination by blood transfusion)
      • Perez P.
      • Salmi L.R.
      • Folléa G.
      • et al.
      Determinants of transfusion-associated bacterial contamination: results of the French BACTHEM case-control study.
      or too poorly defined or documented to guarantee its existence (some forms of child abuse).
      • Lauterbach D.
      • Armour C.
      Symptom trajectories among child survivors of maltreatment: findings from the Longitudinal Studies of Child Abuse and Neglect (LONGSCAN).
      Second, the anticipation of diagnosis related to the application of a test before any symptom occurrence will seem to improve prognosis even if treatment is not better early compared with later. For example, a patient with a severe form of lung cancer who would have died rapidly after diagnosis
      • Wao H.
      • Mhaskar R.
      • Kumar A.
      • Miladinovic B.
      • Djulbegovic B.
      Survival of patients with non-small cell lung cancer without treatment: a systematic review and meta-analysis.
      would seem to survive longer if diagnosis is performed a few months earlier as part of case finding. Because this lead time (obstacle 2)
      • Schwartz M.
      Estimates of lead time and length bias in a breast cancer screening program.
      is usually imperceptible at the individual level,
      • Schwartz M.
      Estimates of lead time and length bias in a breast cancer screening program.
      it is the main justification for conducting randomized trials of early vs delayed treatment.
      • Bell K.J.
      • Bossuyt P.
      • Glasziou P.
      • Irwig L.
      Assessment of changes to screening programmes: why randomisation is important.
      Third, the screening test might detect preclinical forms that would never have been diagnosed had the test not been performed. This is the phenomenon of overdiagnosis (obstacle 3),
      • Carter J.L.
      • Coletti R.J.
      • Harris R.P.
      Quantifying and monitoring overdiagnosis in cancer screening: a systematic review of methods.
      • Carter S.M.
      • Rogers W.
      • Heath I.
      • Degeling C.
      • Doust J.
      • Barratt A.
      The challenge of overdiagnosis begins with its definition.
      that is, the identification of forms of the disorder with a spontaneous good prognosis. Overdiagnosis is likely to be very important, for instance, when screening for prostate cancer using prostate-specific antigen because many of the detected lesions, although meeting the pathologic criteria for invasive cancer, would never progress to severe or lethal stages.
      • Carter S.M.
      • Williams J.
      • Parker L.
      • et al.
      Screening for cervical, prostate, and breast cancer: interpreting the evidence.
      Overdiagnosis, the extreme of what was called length bias,
      • Schwartz M.
      Estimates of lead time and length bias in a breast cancer screening program.
      is a major clinical problem because persons detected with these forms will have no benefit from an early treatment that they do not need (what is called overtreatment) and will experience the adverse effects of the test, the treatment, and of being labeled with the diagnosis.
      • Morgan D.J.
      • Brownlee S.
      • Leppin A.L.
      • et al.
      Setting a research agenda for medical overuse.
      Length bias is also accentuated by the fact that cases with a short preclinical phase, corresponding to forms of the disorder with the poorest prognosis, are most likely to be missed by a single test and, therefore, are more likely to be detected through clinical presentation than after screening.
      • Schwartz M.
      Estimates of lead time and length bias in a breast cancer screening program.
      Both lead time and overdiagnosis are major issues when evaluating screening at the population level because they tend to make screening seem better than it actually is.

      Population Perspective on Screening

      The Obvious Complexity of Screening a Target Population

      The population perspective considers that screening not only is an issue of early intervention for patients with a poor prognosis but also is proposed to a population that includes people with all forms of the disorder and a majority of people without the disorder.
      • Morrison A.S.
      Screening in Chronic Disease.
      A screening recommendation, systematic or case finding, should consider the acceptability and possible negative effects of the process and the perspective of people without the disorder (Figure 1B).
      • Pinsky P.F.
      Principles of cancer screening.
      Acceptability (obstacle 4) is a key issue in the individual and population perspectives because most people in the target population either do not need screening (they do not have the disorder) or do not feel that they need screening (they have the disorder but no manifestations). The latter individuals, who will be detected early (the true-positives), might not feel the need for a potentially aggressive treatment (screening for cancer)
      • Pinsky P.F.
      Principles of cancer screening.
      or stigmatizing intervention (screening for decreased ability to drive an automobile in the elderly).
      • Leproust S.
      • Lagarde E.
      • Salmi L.R.
      Systematic screening for unsafe driving due to medical conditions: still debatable.
      In the former group of people without the disorder, adverse effects of tests and, potentially, being false-positive are likely to be unacceptable.
      • Bond M.
      • Pavey T.
      • Welch K.
      • et al.
      Systematic review of the psychological consequences of false-positive screening mammograms.
      Conversely, being falsely reassured as “in good health” after a first test (false-negative) can also negatively affect the acceptability of a later-round test and its results.
      • Petticrew M.
      • Sowden A.
      • Lister-Sharp D.
      False-negative results in screening programs: medical, psychological, and other implications.

      Further Difficulties in Judging the Population Effects of Screening

      From the population perspective, the recommendation to develop a screening program or policy or to introduce screening in clinical practice should occur only after demonstrating that the advantages for the population are greater than the disadvantages.
      • Morrison A.S.
      Screening in Chronic Disease.
      Lead time and overdiagnosis, already discussed from the individual perspective, are major issues in the evaluation of the population effects of screening and have to be considered in any comparison of consequences in populations with and without the program, policy, or practice.
      • Duffy S.W.
      • Nagtegaal I.D.
      • Wallis M.
      • et al.
      Correcting for lead time and length bias in estimating the effect of screen detection on cancer survival.
      Furthermore, overdiagnosis also occurs because persons without the disorder may be diagnosed as having incidental findings apart from the target condition; for example, screening for lung cancer by chest radiography or computed tomography also identifies asymptomatic nonevolutive lesions.
      • Demirci I.
      • Herold S.
      • Kopp A.
      • Flaßhove M.
      • Klosterhalfen B.
      • Janßen H.
      Overdiagnosis of a typical carcinoid tumor as an adenocarcinoma of the lung: a case report and review of the literature.
      Selection (obstacle 5), another issue, can occur at several steps of the process. The ability of screening programs to reach the target population has been documented to vary across socioeconomic strata.
      • Wools A.
      • Dapper E.A.
      • Leeuw J.R.
      Colorectal cancer screening participation: a systematic review.
      For example, the populations included in randomized trials of screening for lung cancer by low-dose computed tomography in heavy smokers were of higher socioeconomic levels and smoked less than the general population in the United States,
      • Aberle D.R.
      • Adams A.M.
      • Berg C.D.
      • et al.
      Baseline characteristics of participants in the randomized National Lung Screening Trial.
      Denmark,
      • Hestbech M.S.
      • Siersma V.
      • Dirksen A.
      • Pedersen J.H.
      • Brodersen J.
      Participation bias in a randomised trial of screening for lung cancer.
      and the United Kingdom.
      • McRonald F.E.
      • Yadegarfar G.
      • Baldwin D.R.
      • et al.
      The UK Lung Screen (UKLS): Demographic profile of first 88,897 approaches provides recommendations for population screening.
      In established screening programs, acceptability depends on the type of test (higher for mammography
      • Lagerlund M.
      • Sontrop J.M.
      • Zackrisson S.
      Psychosocial factors and attendance at a population-based mammography screening program in a cohort of Swedish women.
      than for fecal occult blood tests
      • Hobbs F.D.
      • Cherry R.C.
      • Fielding J.W.
      • Pike L.
      • Holder R.
      Acceptability of opportunistic screening for occult gastrointestinal blood loss.
      ), and it is lower in high- than in low-risk groups (screening for human immunodeficiency virus infection).
      • Irwin K.L.
      • Valdiserri R.O.
      • Holmberg S.D.
      The acceptability of voluntary HIV antibody testing in the United States: a decade of lessons learned.
      Detection issues (obstacle 6) are related to the performance of tests, reflected by their accuracy
      • Pinsky P.F.
      Principles of cancer screening.
      and reliability.
      • Streiner D.L.
      • Norman G.R.
      Health Measurement Scales. A Practical Guide to Their Development and Use.
      A test developed and demonstrated to be accurate (ie, correctly classifying both people with and without the disorder) and reliable (ie, providing stable results when repeated in the same individuals or across observers) for individual diagnosis in the clinical setting might not be appropriate for screening. For a fixed sensitivity and specificity, the positive predictive value will be much lower and the negative predictive value higher in a screening context than in clinical diagnostic context. Indeed, the prevalence of the disorder is lower in populations targeted by screening programs or case finding than in a diagnostic context.
      • Leeflang M.M.
      • Rutjes A.W.
      • Reitsma J.B.
      • Hooft L.
      • Bossuyt P.M.
      Variation of a test's sensitivity and specificity with disease prevalence.
      Thus, tests will usually be better to rule out the disorder than to identify early stages. Even with a high specificity, false-positive results and their negative consequences are likely to be considerably frequent. Favoring the specificity to reduce false-positive results is likely to decrease the sensitivity and, thus, the effectiveness of screening by increasing false-negative results. Moreover, achieving the same level of reliability and accuracy across all structures (eg, laboratories) and professionals (eg, radiologists) involved in a screening program is hard to achieve.
      • Soh B.P.
      • Lee W.
      • Kench P.L.
      • et al.
      Assessing reader performance in radiology, an imperfect science: lessons from breast screening.
      Reliability in real-life conditions needs to be assessed before implementation of a screening program and must be monitored in a quality assurance program.
      • Elfström K.M.
      • Arnheim-Dahlström L.
      • von Karsa L.
      • Dillner J.
      Cervical cancer screening in Europe: quality assurance and organisation of programmes.
      Other issues, such as selecting a population with an adequate spectrum of preclinical disease and patients with competing disorders that mimic the target condition,
      • Montori V.M.
      • Wyer P.
      • Newman T.B.
      • Keitz S.
      • Guyatt G.
      for the Evidence-Based Medicine Teaching Tips Working Group
      Tips for learners of evidence-based medicine, 5: the effect of spectrum of disease on the performance of diagnostic tests.
      and considering the possible variation of sensitivity and specificity with prevalence,
      • Leeflang M.M.
      • Rutjes A.W.
      • Reitsma J.B.
      • Hooft L.
      • Bossuyt P.M.
      Variation of a test's sensitivity and specificity with disease prevalence.
      are also important to consider when assessing the performance of tests in a screening context.

      Research Implications

      The Challenging Randomized Screening Trial

      The previously mentioned issues justify the need to adequately assess both the positive and negative effects of screening on populations, not only on individuals with the disorder. Ideally, this assessment must be based on a properly conducted randomized community trial: usually, screening is proposed to half the target population, and the other half is allocated to what is considered usual care in the absence of screening.
      • Juffs H.G.
      • Tannock I.F.
      Screening trials are even more difficult than we thought they were.
      Such trials must consider all issues expected in reports of randomized trials, as defined by the Consolidated Standards of Reporting Trials (CONSORT).
      • Schulz K.F.
      • Altman D.G.
      • Moher D.
      CONSORT 2010 Statement: updated guidelines for reporting parallel group randomised trials.
      Many of the CONSORT items, however, are considered difficult to address
      • Cals J.W.
      • van Amelsvoort L.G.
      • Kotz D.
      • Spigt M.G.
      CONSORT 2010 Statement: unfinished update?.
      and, in our opinion, need to be adapted: these items concern the objective of the trial; the choice of an appropriate control arm; the choice of relevant yet measurable outcomes, including all aspects of the benefit to harm balance; and sample size calculation (Supplemental Table 2, available online at http://www.mayoclinicproceedings.org).
      The first issue concerns the objective of the trial (obstacle 7). The challenge is to demonstrate that screening, considered in all its dimensions, is superior to usual care.
      • Juffs H.G.
      • Tannock I.F.
      Screening trials are even more difficult than we thought they were.
      This implies that the compared strategies are defined and operationally described in (Table, items 5a and 5b) (1) the screening group (initial test, confirmation strategy, early intervention for those with the disorder, process of dealing with false-positive and false-negative results) and (2) the control group of usual care in the health care system where the study is conducted (accessibility to diagnosis and treatment). Demonstrating a favorable advantages to disadvantages ratio also implies that all relevant positive and negative effects be documented, in the methods and results, similarly for all participants, whether they have the disorder or not (Table, items 6a-6e) and in the screening and control groups.
      TableMain Specific Methodological Issues to Be Addressed in Trials Assessing the Effects of Screening
      CT = computed tomographic; ERSPC = European Randomized Study of Screening for Prostate Cancer; NELSON = European NEderlands-Leuvens Screening ONderzoek; NLST = National Lung Screening Trial; PSA = prostate-specific antigen.
      Adapted from the Consolidated Standards of Reporting Trials (CONSORT) Statement (see Schulz et al39 for original CONSORT items and the full list of items adapted to screening trials in Supplemental Table 2).
      Section/topicItem No.
      Items of the original CONSORT or proposed numbers; see the complete original list in Supplemental Table 2.
      Checklist itemExamples of observed limitations or variations
      Introduction
       Objective2bIdentify the study as assessing whether the proposed screening program is superior to usual care in the absence of screening in terms of mortalityNone of the 2 major trials of screening for prostate cancer with PSA expressed the notion of superiority in the objective
      • Andriole G.L.
      • Crawford E.D.
      • Grubb R.L.
      • et al.
      Mortality results from a randomized prostate-cancer screening trial.
      • Schröder F.H.
      • Hugosson J.
      • Roobol M.J.
      • et al.
      Screening and prostate-cancer mortality in a randomized European study.
      Methods
       Interventions5aAll steps of screening are defined and operationally described: initial test, confirmation strategy, early intervention, process of dealing with false-positive and false-negative resultsIn the NLST assessing CT screening for lung cancer in heavy smokers, the description of screening only provided information on the test; modalities of confirmation and treatment were left to the referring physician's initiative
      • Aberle D.R.
      • Adams A.M.
      • Berg C.D.
      • et al.
      Baseline characteristics of participants in the randomized National Lung Screening Trial.
      5bThe comparison strategy of usual care in the absence of screening and specificities of the health care system are definedIn the NLST,
      • Aberle D.R.
      • Adams A.M.
      • Berg C.D.
      • et al.
      Baseline characteristics of participants in the randomized National Lung Screening Trial.
      the control strategy included a chest radiograph, whereas the NELSON trial, the largest European trial, compared CT screening with usual care in the Netherlands and Belgium
      • van Iersel C.A.
      • de Koning H.J.
      • Draisma G.
      • et al.
      Risk-based selection from the general population in a screening trial: selection criteria, recruitment and power for the Dutch-Belgian randomised lung cancer multi-slice CT screening trial (NELSON).
      5cAny changes in screening or comparison interventions after the trial commenced, and reasonsIn prostate cancer screening, the threshold for considering a PSA positive has been lowered over time
      • Andriole G.L.
      • Crawford E.D.
      • Grubb R.L.
      • et al.
      Mortality results from a randomized prostate-cancer screening trial.
      • Schröder F.H.
      • Hugosson J.
      • Roobol M.J.
      • et al.
      Screening and prostate-cancer mortality in a randomized European study.
       Outcomes6aThe prespecified main outcome is all-cause mortality or, if not possible, disorder-specific or incidence-based mortality
      See the text for a discussion of relevance and the limits of all-cause, disorder-specific, and incidence-based mortality.
      The use of suicide ideation as the main outcome in a trial of screening for risk of suicide precludes concluding on the actual effects on risk
      • Crawford M.J.
      • Thana L.
      • Methuen C.
      • et al.
      Impact of screening for risk of suicide: randomised controlled trial.
      6bDeaths and their causes, and all other relevant outcomes, are assessed similarly in all groupsIn the ERSPC,
      • Schröder F.H.
      • Hugosson J.
      • Roobol M.J.
      • et al.
      Screening and prostate-cancer mortality in a randomized European study.
      verification of causes of death in patient records was recommended but varied across countries
      6cThe prespecified length of follow-up is long enoughIn screening for many cancers, it is estimated that the full benefit of a program would be observed only after 20 y of follow-up
      • Hanley J.A.
      • McGregor M.
      • Liu Z.
      • Strumpf E.C.
      • Dendukuri N.
      Measuring the mortality impact of breast cancer screening.
      ; the use of surrogate end points might help shorten the follow-up needed
      • Cuzick J.
      • Cafferty F.H.
      • Edwards R.
      • Møller H.
      • Duffy S.W.
      Surrogate endpoints for cancer screening trials: general principles and an illustration using the UK Flexible Sigmoidoscopy Screening Trial.
      6dPractical modalities of follow-up are similar in all groupsNo standardization within or across studies of screening for asymptomatic carotid artery stenosis
      • Jonas D.E.
      • Feltner C.
      • Amick H.R.
      • et al.
      Screening for asymptomatic carotid artery stenosis: a systematic review and meta-analysis for the U.S. Preventive Services Task Force.
      6eKey secondary outcomes (disorder-related morbidity, compliance with tests and intervention, contamination, adverse effects of tests and intervention, false-positives, false-negatives; overdiagnosis, psychological impact) are defined and monitored similarly in all groupsThe 2 major trials of PSA screening for prostate cancer have not documented most secondary outcomes for all individuals in all groups
      • Andriole G.L.
      • Crawford E.D.
      • Grubb R.L.
      • et al.
      Mortality results from a randomized prostate-cancer screening trial.
      • Schröder F.H.
      • Hugosson J.
      • Roobol M.J.
      • et al.
      Screening and prostate-cancer mortality in a randomized European study.
       Sample size7aHypotheses underlying sample size calculation are reasonableIn the Dante trial of lung cancer screening, the expected 50% reduction of mortality is not compatible with current hypotheses
      • Infante M.
      • Lutman F.R.
      • Cavuto S.
      • et al.
      Lung cancer screening with spiral CT: baseline results of the randomized DANTE trial.
       Statistical methods12aAnalysis is intent to screen, and accounting for actual length of follow-upIn the ERSPC, the authors considered excluding from the analysis centers with a large proportion of contamination in the control group and performed their final analysis in a subgroup of men, therefore breaking the randomization
      • Schröder F.H.
      • Hugosson J.
      • Roobol M.J.
      • et al.
      Screening and prostate-cancer mortality in a randomized European study.
      Discussion
       Interpretation22Interpretation consistent with results, balancing actual benefits and all aspects of harm and feasibility issuesIn screening for cervical, prostate, or breast cancer, there is a gap between evidence regarding efficacy and harm
      • Carter J.L.
      • Coletti R.J.
      • Harris R.P.
      Quantifying and monitoring overdiagnosis in cancer screening: a systematic review of methods.
      a CT = computed tomographic; ERSPC = European Randomized Study of Screening for Prostate Cancer; NELSON = European NEderlands-Leuvens Screening ONderzoek; NLST = National Lung Screening Trial; PSA = prostate-specific antigen.
      b Adapted from the Consolidated Standards of Reporting Trials (CONSORT) Statement (see Schulz et al
      • Schulz K.F.
      • Altman D.G.
      • Moher D.
      CONSORT 2010 Statement: updated guidelines for reporting parallel group randomised trials.
      for original CONSORT items and the full list of items adapted to screening trials in Supplemental Table 2).
      c Items of the original CONSORT or proposed numbers; see the complete original list in Supplemental Table 2.
      d See the text for a discussion of relevance and the limits of all-cause, disorder-specific, and incidence-based mortality.

      The Controversial Definition of Relevant Outcomes

      One major issue is what main outcome (obstacle 8) should be used to document the possible advantages of screening.
      • Hanley J.A.
      Measuring mortality reductions in cancer screening trials.
      From both the individual and population perspectives (Figure 1), the main outcome is the most severe consequence for which it is hypothesized that screening can improve prognosis. For example, when screening for inability to drive an automobile in the elderly, the main outcome must include all occurrences of collisions with at least a severe or lethal injury to either the driver or other road users.
      • Leproust S.
      • Lagarde E.
      • Salmi L.R.
      Systematic screening for unsafe driving due to medical conditions: still debatable.
      For most chronic diseases for which screening has been proposed, the main outcome is death,
      • Pinsky P.F.
      Principles of cancer screening.
      but it is debated what is the most relevant parameter
      • Hanley J.A.
      Measuring mortality reductions in cancer screening trials.
      (Table, items 2b and 6a).
      Case fatality is relevant from the individual perspective (Figure 1A) in trials comparing early and delayed treatment, but it is not as relevant from the population perspective (Figure 1B) because it considers only people with the disorder. Furthermore, in the presence of overdiagnosis, case fatality will seem lower in screened than in unscreened populations, and the effect of lead-time bias will be increased. Disorder-specific mortality is the parameter estimated in most cancer screening trials.
      • Hanley J.A.
      Measuring mortality reductions in cancer screening trials.
      Proponents of disorder-specific mortality argue that, first, it is the parameter most directly related to an improvement in prognosis and, second, it is unrealistic to hope for a positive effect to be reflected by a change in all-cause mortality, unless follow-up is very long
      • Hanley J.A.
      Measuring mortality reductions in cancer screening trials.
      • Tabar L.
      • Duffy S.W.
      • Yen M.F.
      • et al.
      All-cause mortality among breast cancer patients in a screening trial: support for breast cancer mortality as an end point.
      ; the huge sample size needed to show an improvement in all-cause mortality, a major obstacle in trials based on all-cause mortality, is another issue discussed by proponents of disorder-specific mortality.
      • Tabar L.
      • Duffy S.W.
      • Yen M.F.
      • et al.
      All-cause mortality among breast cancer patients in a screening trial: support for breast cancer mortality as an end point.
      • Steele R.
      • Brewster D.
      Should we use total mortality rather than cancer specific mortality to judge cancer screening programmes? no.
      It can also be argued that whenever screening can reduce the incidence of the disease, for instance when the diagnostic procedure allows removal of precancerous lesions (eg, removal of polyps during colonoscopy in colonic cancer screening),
      • Schoen R.E.
      • Pinsky P.F.
      • Weissfeld J.L.
      • et al.
      Colorectal-cancer incidence and mortality with screening flexible sigmoidoscopy.
      disorder-specific mortality can be an appropriate outcome parameter.
      Proponents of all-cause mortality,
      • Black W.C.
      • Haggstrom D.A.
      • Welch H.G.
      All-cause mortality in randomized trials of cancer screening.
      including the authors, argue that the follow-up needs, anyway, to be very long because screening cannot have a positive effect before a substantial number of patients in the preclinical phase are identified and the resulting improved prognosis results in a change in disorder-specific mortality.
      • Hanley J.A.
      • McGregor M.
      • Liu Z.
      • Strumpf E.C.
      • Dendukuri N.
      Measuring the mortality impact of breast cancer screening.
      Furthermore, it can be argued, from the population perspective, that only all-cause mortality can reflect the most severe consequences occurring in people with and without the disorder, resulting from disorder- and screening-related morbidity.
      • Black W.C.
      • Haggstrom D.A.
      • Welch H.G.
      All-cause mortality in randomized trials of cancer screening.
      It is also relevant because a decrease in disorder-specific mortality implies that survivors remain subject to competitive causes of deaths, especially in aged populations.
      • Seidman H.
      • Silverberg E.
      • Bodden A.
      Probabilities of eventually developing and of dying of cancer (risk among persons previously undiagnosed with the cancer).
      • Jemal A.
      • Simard E.P.
      • Dorell C.
      • et al.
      Annual Report to the Nation on the Status of Cancer, 1975–2009, featuring the burden and trends in human papillomavirus (HPV)–associated cancers and HPV vaccination coverage levels.
      Another reason to prefer all-cause to disorder-specific mortality is that the latter is easily subject to misclassification.
      • Black W.C.
      • Haggstrom D.A.
      • Welch H.G.
      All-cause mortality in randomized trials of cancer screening.
      Indeed, the validity of disorder-specific mortality assumes that cause of death is accurately determined and that screening has negligible effects on other causes of death. Not meeting these assumptions results in 2 biases.
      • Juffs H.G.
      • Tannock I.F.
      Screening trials are even more difficult than we thought they were.
      • Black W.C.
      • Haggstrom D.A.
      • Welch H.G.
      All-cause mortality in randomized trials of cancer screening.
      (1) Sticky diagnosis bias can occur in a randomized trial if validation of outcomes is not adequately standardized, resulting in the screening group in some deaths from other causes being unduly attributed to the disorder; this bias underestimates the positive effect of screening. (2) Slippery linkage bias occurs because deaths due to adverse effects of screening-related procedures are not counted in disorder-related mortality; this bias overestimates the positive impact of screening. This lack of consideration of screening-related morbidity is another possible argument in favor of all-cause over disorder-specific mortality.
      • Black W.C.
      • Haggstrom D.A.
      • Welch H.G.
      All-cause mortality in randomized trials of cancer screening.
      To resolve the controversy regarding all-cause and disorder-specific mortality, some authors have suggested assessing the real-life impact of screening based on excess mortality in those with the disorder.
      • Brenner H.
      • Hoffmeister M.
      • Jansen L.
      Comparisons of colorectal cancer mortality between screening participants and the general population are strongly biased unless an incidence-based mortality approach is used.
      This approach consists of comparing the observed number of deaths in screening participants with the expected number derived from disease-specific mortality in the general population. However, this approach is also subject to strong overestimation of the number of expected deaths and of the benefit of screening if it is not based on an incidence-based mortality method, which standardizes for time of diagnosis and considers that screening participants do not have a date of previous diagnosis.
      • Brenner H.
      • Hoffmeister M.
      • Jansen L.
      Comparisons of colorectal cancer mortality between screening participants and the general population are strongly biased unless an incidence-based mortality approach is used.
      Still, because the method can be derived from data available in cancer and death registries, it probably gives a good reflection of the actual effect of a screening program in circumstances in which a trial is not feasible or not completed.
      • Brenner H.
      • Hoffmeister M.
      • Jansen L.
      Comparisons of colorectal cancer mortality between screening participants and the general population are strongly biased unless an incidence-based mortality approach is used.
      Whatever the option, the population perspective implies that investigators should thoroughly document all relevant positive and negative outcomes of the disorder and screening. This is a huge challenge, as reflected by the fact that systematic reviews on screening usually identify limitations of published trials: focus on the initial test rather than on the program,
      • Zakher B.
      • Cantor A.G.
      • Pappas M.
      • Daeges M.
      • Nelson H.D.
      Screening for gonorrhea and chlamydia: a systematic review for the U.S. Preventive Services Task Force.
      insufficient statistical power,
      • Cantor A.G.
      • Bougatsos C.
      • Dana T.
      • Blazina I.
      • McDonagh M.
      Routine iron supplementation and screening for iron deficiency anemia in pregnancy: a systematic review for the U.S. Preventive Services Task Force.
      poorly compensated by attempts to mix studies or centers with different populations or screening,
      • Schröder F.H.
      • Hugosson J.
      • Roobol M.J.
      • et al.
      Screening and prostate-cancer mortality in a randomized European study.
      • Oudkerk M.
      • Heuvelmans M.A.
      Screening for lung cancer by imaging: the Nelson study.
      insufficient follow-up,
      • LeBlanc E.S.
      • Zakher B.
      • Daeges M.
      • Pappas M.
      • Chou R.
      Screening for vitamin D deficiency: a systematic review for the U.S. Preventive Services Task Force.
      lack of standardization of outcome assessment
      • Lin J.S.
      • O'Connor E.
      • Rossom R.C.
      • Perdue L.A.
      • Eckstrom E.
      Screening for cognitive impairment in older adults: a systematic review for the U.S. Preventive Services Task Force.
      and follow-up procedures,
      • Jonas D.E.
      • Feltner C.
      • Amick H.R.
      • et al.
      Screening for asymptomatic carotid artery stenosis: a systematic review and meta-analysis for the U.S. Preventive Services Task Force.
      unclear definition and poor documentation of secondary outcomes,
      • Carter S.M.
      • Rogers W.
      • Heath I.
      • Degeling C.
      • Doust J.
      • Barratt A.
      The challenge of overdiagnosis begins with its definition.
      • Zakher B.
      • Cantor A.G.
      • Pappas M.
      • Daeges M.
      • Nelson H.D.
      Screening for gonorrhea and chlamydia: a systematic review for the U.S. Preventive Services Task Force.
      • Lin J.S.
      • O'Connor E.
      • Rossom R.C.
      • Perdue L.A.
      • Eckstrom E.
      Screening for cognitive impairment in older adults: a systematic review for the U.S. Preventive Services Task Force.
      and major changes in the interventions or interventions becoming obsolete.
      • Bell K.J.
      • Bossuyt P.
      • Glasziou P.
      • Irwig L.
      Assessment of changes to screening programmes: why randomisation is important.
      • Jonas D.E.
      • Feltner C.
      • Amick H.R.
      • et al.
      Screening for asymptomatic carotid artery stenosis: a systematic review and meta-analysis for the U.S. Preventive Services Task Force.
      It is therefore argued that good screening trials are difficult to conduct,
      • Juffs H.G.
      • Tannock I.F.
      Screening trials are even more difficult than we thought they were.
      and our proposed specifications of CONSORT (Table) make the challenge even bigger. Nevertheless, trials are important because they are the gold standard to document the effect of any intervention.

      The Possibility of Simulating the Effects of Screening

      Given the difficulties of conducting adequate trials and the delays before relevant results are available, some authors have suggested that simulations using, for example, Markov models
      • Chiu S.Y.
      • Malila N.
      • Yen A.M.
      • Anttila A.
      • Hakama M.
      • Chen H.H.
      Analytical decision model for sample size and effectiveness projections for use in planning a population-based randomized controlled trial of colorectal cancer screening.
      could be used to assess the effects of screening.
      • Etzioni R.
      • Gulati R.
      • Cooperberg M.R.
      • Penson D.M.
      • Weiss N.S.
      • Thompson I.M.
      Limitations of basing screening policies on screening trials: the US Preventive Services Task Force and prostate cancer screening.
      Such simulations can consider all steps of the natural history of the disorder and can compare the values of case fatality and of disorder-specific and all-cause mortality in the presence and absence of screening (Figure 2) for various lag times, screening intervals, starting ages, stopping ages, and ascertainment algorithms. The proposed model, however, must consider that many methodological choices are specific to a given disorder/testing procedure.
      Figure 2
      Figure 2The course of a disorder and biases to consider in randomized trials, modeling, or cohort studies of the effects of screening programs. The solid lines indicate the possible transitions in the absence of screening; dashed lines, the additional possible transitions in the presence of screening. Nonspecific mortality also includes mortality related to competitive causes, which are neither disorder specific nor care related. DET = detection bias; LTB = lead-time bias; OVD = overdiagnosis; PSB = prevention success bias; SDB = sticky diagnosis bias; SEL = selection bias; SLB = slippery linkage bias; TIB = treatment improvement bias (see definitions in text).
      One strength of models is that they can be fed by data from many sources (randomized trials; observational studies such as cohort, accuracy, and repeatability studies; and expert judgments using formal consensus techniques) and can simulate the effect of uncertainties in estimates of relevant parameters.
      • Hamra G.
      • MacLehose R.
      • Richardson D.
      Markov Chain Monte Carlo: an introduction for epidemiologists.
      For example, a Markov model could be used to simulate the impact of a hypothetical program screening for pediatric abusive head trauma despite an extremely scarce literature on the performance of possible tests and the definition and length of the preclinical phase.
      • Bailhache M.
      • Bénard A.
      • Salmi L.R.
      Simulation of the impact of programs for prevention and screening of Pediatric Abusive Head Trauma.
      Similar models have been developed, for example, for screening for psychosis,
      • Cougnard A.
      • Salmi L.R.
      • Salamon R.
      • Verdoux H.
      A decision analysis model to assess the feasibility of the early detection of psychosis in the general population.
      age-related macular degeneration,
      • Tamura H.
      • Goto R.
      • Akune Y.
      • Hiratsuka Y.
      • Hiragi S.
      • Yamada M.
      The clinical effectiveness and cost-effectiveness of screening for age-related macular degeneration in Japan: a Markov modeling study.
      and prostate cancer.
      • Mühlberger N.
      • Kurzthaler C.
      • Iskandar R.
      • et al.
      The ONCOTYROL prostate cancer outcome and policy model: effect of prevalence assumptions on the benefit-harm balance of screening.
      Models are also potentially appropriate to consider the potential effects of issues such as selection, detection, lead time, overdiagnosis, sticky diagnosis, and slippery linkage.
      • Seigneurin A.
      • Labarère J.
      • Duffy S.W.
      • Colonna M.
      Overdiagnosis associated with breast cancer screening: a simulation study to compare lead-time adjustment methods.
      Another advantage is that models can complement trials by simulating more complex questions than can be addressed in trials.
      • Karnon J.
      • Brennan A.
      • Akehurst R.
      Decision modeling to inform decision making: seeing the wood for the trees.
      The typical trial compares one screening strategy with one control group; there are a few instances of trials comparing several screening strategies with a control group. For example, the Multicenter Italian Lung Detection trial compared 2 rhythms of screening for lung cancer by low-dose computed tomography, annually or every other year, with the absence of screening.
      • Pastorino U.
      • Rossi M.
      • Rosato V.
      • et al.
      Annual or biennial CT screening versus observation in heavy smokers: 5-year results of the MILD trial.
      Such trials are even more difficult to conduct than 2-group trials and are less likely to meet all the expectations of CONSORT.
      • Schulz K.F.
      • Altman D.G.
      • Moher D.
      CONSORT 2010 Statement: updated guidelines for reporting parallel group randomised trials.
      It is, therefore, difficult to imagine larger trials comparing more screening strategies or screening with other interventions (prevention or treatment).

      The Need to Compare Screening With Other Options

      Prevention and treatment are interventions that should be considered alternatives to screening. Failure to consider these alternatives when assessing screening can result in 2 issues, similar to the classic history bias in evaluation research
      • Campbell D.T.
      • Stanley J.C.
      Experimental and Quasi-experimental Designs for Research.
      : treatment improvement bias (obstacle 9) and prevention success bias (obstacle 10). Treatment improvement bias occurs when progress has occurred between the start and the final analysis of a trial. For example, a dramatic improvement has been observed in the prognosis of breast cancer
      • Gøtzsche P.C.
      • Jørgensen K.J.
      The Breast Screening Programme and misinforming the public.
      owing to the availability of targeted treatments.
      • Vera-Badillo F.E.
      • Templeton A.J.
      • de Gouveia P.
      • et al.
      Androgen receptor expression and outcomes in early breast cancer: a systematic review and meta-analysis.
      Were trials of mammography screening to be started in the present decade, rather than 40 years ago, it is likely that the observed positive effect of screening, if any, would be small.
      • Bell K.J.
      • Bossuyt P.
      • Glasziou P.
      • Irwig L.
      Assessment of changes to screening programmes: why randomisation is important.
      Treatment improvement bias decreases the disorder-specific mortality in the usual care group; it also dramatically decreases the power of trials, making it even less likely to demonstrate a statistically and clinically significant difference between screening and usual care.
      • Hanley J.A.
      • McGregor M.
      • Liu Z.
      • Strumpf E.C.
      • Dendukuri N.
      Measuring the mortality impact of breast cancer screening.
      Prevention success bias is a consequence of a decrease in the incidence of the disorder owing to effective preventive interventions. For example, the positive effect of screening for driving under the influence of alcohol is lowered by the effectiveness of laws.
      • Voas R.B.
      • Fell J.C.
      Preventing impaired driving opportunities and problems.
      In a trial of prevention vs screening, the result will favor the prevention arm. If a trial compares screening with no intervention in a population in which prevention has been disseminated, then the trial will lack power. Treatment improvement bias and prevention success bias are more likely to occur when the length of follow-up needed to expect a positive effect of screening is long.
      • Campbell D.T.
      • Stanley J.C.
      Experimental and Quasi-experimental Designs for Research.
      Another effect of the decreased incidence of the disorder is to worsen detection and selection issues: fewer people with the disorder increases the likelihood of false-positive results and their consequences.
      • Leeflang M.M.
      • Rutjes A.W.
      • Reitsma J.B.
      • Hooft L.
      • Bossuyt P.M.
      Variation of a test's sensitivity and specificity with disease prevalence.
      For example, the model comparing prevention, screening, and the combination of prevention and screening with no intervention to deal with pediatric abusive head trauma showed that prevention would be effective in decreasing mortality but that both screening and the combination of screening and prevention could actually increase mortality, mostly because stigmatization of parents—by increasing parental poor self-esteem, depression, and social isolation—might increase the risk of abuse.
      • Bailhache M.
      • Bénard A.
      • Salmi L.R.
      Simulation of the impact of programs for prevention and screening of Pediatric Abusive Head Trauma.

      Reconciling Perspectives and Meeting Ethical Challenges

      Reconciling individual and population health perspectives is possible by considering that verification of hypotheses underlying the former is a prerequisite for the latter (Figure 3). The only subtlety is that a screening program or policy is justified only if the disorder is a public health issue (Figure 3, box 1), that is, that its consequences are important in terms of mortality, severe morbidity, or costs to society.
      • Wilson J.M.G.
      • Jungner G.
      The Principles and Practice of Screening for Disease.
      • Morrison A.S.
      Screening in Chronic Disease.
      • Ahern J.
      • Hubbard A.
      • Galea S.
      Estimating the effects of potential public health interventions on population disease burden: a step-by-step illustration of causal inference methods.
      Mass screening is a public health option, but issues common to the individual and population perspectives (a well-defined and long preclinical phase, accurate and reliable tests in the preclinical phase, and early treatment demonstrated to be better than later treatment) have to be addressed whether screening is systematic, targeted, or case finding.
      • Etzioni R.
      • Gulati R.
      • Cooperberg M.R.
      • Penson D.M.
      • Weiss N.S.
      • Thompson I.M.
      Limitations of basing screening policies on screening trials: the US Preventive Services Task Force and prostate cancer screening.
      Overdiagnosis, acceptability, selection, and detection issues are also relevant to both perspectives.
      Figure 3
      Figure 3Seven key issues reconciling the individual and population perspectives on screening and their consequences for decisions (ovals) and implications for research (dotted-line boxes). Dark orange boxes indicate the only issues traditionally considered in the individual model; light orange boxes are suggested to be part of an evidence-based model (see the text for details).
      Furthermore, we argue that consideration of acceptability (of tests, early treatment, and errors) is a foundation of evidence-based medicine and evidence-based policy.
      • Sackett D.L.
      • Rosenberg W.M.
      • Gray J.A.
      • Haynes R.B.
      • Richardson W.S.
      Evidence based medicine: what it is and what it isn't.
      • Charles C.
      • Gafni A.
      • Freeman E.
      The evidence-based medicine model of clinical practice: scientific teaching or belief-based preaching?.
      • Evidence-Based Medicine Working Group
      Evidence-based medicine: a new approach to teaching the practice of medicine.
      The challenge of evidence-based medicine is to integrate scientific evidence, the values and perspectives of the patient/person on his or her health and the intervention, the experience of the physician, and the availability of health care resources.
      • Charles C.
      • Gafni A.
      • Freeman E.
      The evidence-based medicine model of clinical practice: scientific teaching or belief-based preaching?.
      Addressing the patient's values and perspectives is central to acceptability (Figure 3, box 6); individuals with different expectations from life and different coping may, indeed, give different values to benefits and harms. Nevertheless, the explanation of options (to screen or not to screen) and their consequences requires a previous demonstration that the advantages of screening the target population are larger than the disadvantages (Figure 3, box 5). Also, it should be the experience of physicians that most tests are negatives, most positive tests are false-positives, and screening-related morbidity can be frequent for some disorders.
      • Etzioni R.
      • Gulati R.
      • Cooperberg M.R.
      • Penson D.M.
      • Weiss N.S.
      • Thompson I.M.
      Limitations of basing screening policies on screening trials: the US Preventive Services Task Force and prostate cancer screening.
      Thus, issues of effectiveness and acceptability belong to both perspectives. Finally, screening is feasible only if costs are acceptable for the individual and the health system (Figure 3, box 7). However, any evaluation of costs should not be limited to the cost of testing but should include costs directly or indirectly related to treatment and errors
      • Campbell M.
      • Fitzpatrick R.
      • Haines A.
      • et al.
      Framework for design and evaluation of complex interventions to improve health.
      ; costs should also be reevaluated during the program.
      We further suggest that recommending screening or prescribing a test is unethical when all issues have not been properly addressed.
      • Carter S.M.
      • Williams J.
      • Parker L.
      • et al.
      Screening for cervical, prostate, and breast cancer: interpreting the evidence.
      • Scott P.A.
      Unsupervised self-testing as part public health screening for HIV in resource-poor environments: some ethical considerations.
      • van Dam L.
      • Bretthauer M.
      Ethical issues in colorectal cancer screening.
      • Petrova D.
      • Garcia-Retamero R.
      • Cokely E.T.
      Understanding the harms and benefits of cancer screening: a model of factors that shape informed decision making.
      Addressing all issues by adequate research, however, is a huge challenge.
      • Rossi P.G.
      • Ferroni E.
      • Barca A.
      • et al.
      Quality appraisal of documents producing recommendations for breast, colorectal and cervical cancer screening.
      Nevertheless, we submit that a positive benefit-to-harm balance should be better documented, whether through better-conducted randomized controlled trials, modeling, or both, before any recommendation of screening. One has to accept that although screening is the responsibility of health care professionals, relevant research documenting the benefits and harms of screening is not limited to the clinical setting; integrating epidemiologic, clinical, and behavioral and social sciences, mathematical modeling and economics can properly be addressed by multidisciplinary research groups familiar with the evaluation of complex public health interventions.
      • Campbell M.
      • Fitzpatrick R.
      • Haines A.
      • et al.
      Framework for design and evaluation of complex interventions to improve health.
      In that context, one should question whether it is ethical to continue trials that will never provide the information needed to guide recommendations.
      • Juffs H.G.
      • Tannock I.F.
      Screening trials are even more difficult than we thought they were.
      We believe that modeling is a promising approach because models can be used to estimate the potential effects of proposed screening and, using data from cohorts of populations where screening has been implemented, estimate the real-life effects of existing programs.
      • Karnon J.
      • Brennan A.
      • Akehurst R.
      Decision modeling to inform decision making: seeing the wood for the trees.
      Nevertheless, we suggest that trials, if better designed and monitored to appropriately consider key issues discussed herein, remain important sources of evidence to inform benefit and harm. Modeling and trials are not in opposition; both should be used whenever screening is considered. Finally, we suggest that better formalization of reflections on screening, as proposed in this article, will make it easier for clinicians and other decision makers, as well as researchers, to define best-evidence practices and policies. One has to accept, however, that not prescribing screening is often ethically the best option and that research, prevention, and improved treatment should always be considered as alternatives to screening.

      Conclusion

      Screening—the early detection of a latent disorder to improve prognosis—is an important public health intervention for which the individual and population perspectives can be reconciled. Many obstacles and biases are well documented, and most apply equally to the individual and population perspectives and can be addressed in randomized controlled trials or modeling of the effects of screening, provided these studies are adapted to the specific disorder and testing procedures. The individual and population perspectives can both benefit from stronger emphasis on an evidence-based model of screening, in which the interests and values of all individuals, whether they have or do not have the targeted disorder, are considered.

      Supplemental Online Material

      Supplemental Online Material

      Supplemental material can be found online at: http://www.mayoclinicproceedings.org. Supplemental material attached to journal articles has not been edited, and the authors take responsibility for the accuracy of all data.

      References

        • Wilson J.M.G.
        • Jungner G.
        The Principles and Practice of Screening for Disease.
        World Health Organization, Geneva, Switzerland1968
        • Nelson H.D.
        • Tyne K.
        • Naik A.
        • Bougatsos C.
        • Chan B.K.
        • Humphrey L.
        Screening for breast cancer: an update for the U.S. Preventive Services Task Force.
        Ann Intern Med. 2009; 151: 727-737
        • Salunkhe V.
        • van der Meer P.F.
        • de Korte D.
        • Seghatchian J.
        • Gutiérrez L.
        Development of blood transfusion product pathogen reduction treatments: a review of methods, current applications and demands.
        Transfus Apher Sci. 2015; 52: 19-34
        • Morrison A.S.
        Screening in Chronic Disease.
        Oxford University Press, New York, NY1985
        • Fletcher R.H.
        • Fletcher S.W.
        • Wagner E.H.
        Clinical Epidemiology: The Essentials.
        Williams & Wilkins, Baltimore, MD1982
        • Campbell M.
        • Fitzpatrick R.
        • Haines A.
        • et al.
        Framework for design and evaluation of complex interventions to improve health.
        BMJ. 2000; 321: 694-696
        • Rosenberg J.
        • Chia Y.L.
        • Plevritis S.
        The effect of age, race, tumor size, tumor grade, and disease stage on invasive ductal breast cancer survival in the U.S. SEER database.
        Breast Cancer Res Treat. 2005; 89: 47-54
        • Lyratzopoulos G.
        • Vedsted P.
        • Singh H.
        Understanding missed opportunities for more timely diagnosis of cancer in symptomatic patients after presentation.
        Br J Cancer. 2015; 112: S84-S91
        • Tursz A.
        • Crost M.
        • Gerbouin-Rérolle P.
        • Cook J.
        Underascertainment of child abuse fatalities in France: retrospective analysis of judicial data to assess underreporting of infant homicides in mortality statistics.
        Child Abuse Negl. 2010; 34: 534-544
        • Thorpe E.L.
        • Zuckerbraun N.S.
        • Wolford J.E.
        • Berger R.P.
        Missed opportunities to diagnose child physical abuse.
        Pediatr Emerg Care. 2014; 30: 771-776
        • Perez P.
        • Salmi L.R.
        • Folléa G.
        • et al.
        Determinants of transfusion-associated bacterial contamination: results of the French BACTHEM case-control study.
        Transfusion. 2001; 41: 862-872
        • Lauterbach D.
        • Armour C.
        Symptom trajectories among child survivors of maltreatment: findings from the Longitudinal Studies of Child Abuse and Neglect (LONGSCAN).
        J Abnorm Child Psychol. 2015; 44: 369-379
        • Wao H.
        • Mhaskar R.
        • Kumar A.
        • Miladinovic B.
        • Djulbegovic B.
        Survival of patients with non-small cell lung cancer without treatment: a systematic review and meta-analysis.
        Syst Rev. 2013; 2: 10
        • Schwartz M.
        Estimates of lead time and length bias in a breast cancer screening program.
        Cancer. 1980; 46: 844-851
        • Bell K.J.
        • Bossuyt P.
        • Glasziou P.
        • Irwig L.
        Assessment of changes to screening programmes: why randomisation is important.
        BMJ. 2015; 350: h1566
        • Carter J.L.
        • Coletti R.J.
        • Harris R.P.
        Quantifying and monitoring overdiagnosis in cancer screening: a systematic review of methods.
        BMJ. 2015; 350: g7773
        • Carter S.M.
        • Rogers W.
        • Heath I.
        • Degeling C.
        • Doust J.
        • Barratt A.
        The challenge of overdiagnosis begins with its definition.
        BMJ. 2015; 350: h869
        • Carter S.M.
        • Williams J.
        • Parker L.
        • et al.
        Screening for cervical, prostate, and breast cancer: interpreting the evidence.
        Am J Prev Med. 2015; 49: 274-285
        • Morgan D.J.
        • Brownlee S.
        • Leppin A.L.
        • et al.
        Setting a research agenda for medical overuse.
        BMJ. 2015; 351: h4534
        • Pinsky P.F.
        Principles of cancer screening.
        Surg Clin North Am. 2015; 95: 953-966
        • Leproust S.
        • Lagarde E.
        • Salmi L.R.
        Systematic screening for unsafe driving due to medical conditions: still debatable.
        BMC Public Health. 2008; 8: 27
        • Bond M.
        • Pavey T.
        • Welch K.
        • et al.
        Systematic review of the psychological consequences of false-positive screening mammograms.
        Health Technol Assess. 2013; 17: 1-170
        • Petticrew M.
        • Sowden A.
        • Lister-Sharp D.
        False-negative results in screening programs: medical, psychological, and other implications.
        Int J Technol Assess Health Care. 2001; 17: 164-170
        • Duffy S.W.
        • Nagtegaal I.D.
        • Wallis M.
        • et al.
        Correcting for lead time and length bias in estimating the effect of screen detection on cancer survival.
        Am J Epidemiol. 2008; 168: 98-104
        • Demirci I.
        • Herold S.
        • Kopp A.
        • Flaßhove M.
        • Klosterhalfen B.
        • Janßen H.
        Overdiagnosis of a typical carcinoid tumor as an adenocarcinoma of the lung: a case report and review of the literature.
        World J Surg Oncol. 2012; 10: 19
        • Wools A.
        • Dapper E.A.
        • Leeuw J.R.
        Colorectal cancer screening participation: a systematic review.
        Eur J Public Health. 2015; 26: 158-168
        • Aberle D.R.
        • Adams A.M.
        • Berg C.D.
        • et al.
        Baseline characteristics of participants in the randomized National Lung Screening Trial.
        J Natl Cancer Inst. 2010; 102: 1771-1779
        • Hestbech M.S.
        • Siersma V.
        • Dirksen A.
        • Pedersen J.H.
        • Brodersen J.
        Participation bias in a randomised trial of screening for lung cancer.
        Lung Cancer. 2011; 73: 325-331
        • McRonald F.E.
        • Yadegarfar G.
        • Baldwin D.R.
        • et al.
        The UK Lung Screen (UKLS): Demographic profile of first 88,897 approaches provides recommendations for population screening.
        Cancer Prev Res. 2014; 7: 362-371
        • Lagerlund M.
        • Sontrop J.M.
        • Zackrisson S.
        Psychosocial factors and attendance at a population-based mammography screening program in a cohort of Swedish women.
        BMC Womens Health. 2014; 14: 33
        • Hobbs F.D.
        • Cherry R.C.
        • Fielding J.W.
        • Pike L.
        • Holder R.
        Acceptability of opportunistic screening for occult gastrointestinal blood loss.
        BMJ. 1992; 304: 483-486
        • Irwin K.L.
        • Valdiserri R.O.
        • Holmberg S.D.
        The acceptability of voluntary HIV antibody testing in the United States: a decade of lessons learned.
        AIDS. 1996; 10: 1707-1717
        • Streiner D.L.
        • Norman G.R.
        Health Measurement Scales. A Practical Guide to Their Development and Use.
        3rd ed. Oxford University Press, Oxford, England2003
        • Leeflang M.M.
        • Rutjes A.W.
        • Reitsma J.B.
        • Hooft L.
        • Bossuyt P.M.
        Variation of a test's sensitivity and specificity with disease prevalence.
        CMAJ. 2013; 185: E537-E544
        • Soh B.P.
        • Lee W.
        • Kench P.L.
        • et al.
        Assessing reader performance in radiology, an imperfect science: lessons from breast screening.
        Clin Radiol. 2012; 67: 623-628
        • Elfström K.M.
        • Arnheim-Dahlström L.
        • von Karsa L.
        • Dillner J.
        Cervical cancer screening in Europe: quality assurance and organisation of programmes.
        Eur J Cancer. 2015; 51: 950-968
        • Montori V.M.
        • Wyer P.
        • Newman T.B.
        • Keitz S.
        • Guyatt G.
        • for the Evidence-Based Medicine Teaching Tips Working Group
        Tips for learners of evidence-based medicine, 5: the effect of spectrum of disease on the performance of diagnostic tests.
        CMAJ. 2005; 173: 385-390
        • Juffs H.G.
        • Tannock I.F.
        Screening trials are even more difficult than we thought they were.
        J Natl Cancer Inst. 2002; 94: 156-157
        • Schulz K.F.
        • Altman D.G.
        • Moher D.
        CONSORT 2010 Statement: updated guidelines for reporting parallel group randomised trials.
        BMJ. 2010; 340: c332
        • Cals J.W.
        • van Amelsvoort L.G.
        • Kotz D.
        • Spigt M.G.
        CONSORT 2010 Statement: unfinished update?.
        J Clin Epidemiol. 2011; 64: 579-582
        • Andriole G.L.
        • Crawford E.D.
        • Grubb R.L.
        • et al.
        Mortality results from a randomized prostate-cancer screening trial.
        N Engl J Med. 2009; 360: 1310-1319
        • Schröder F.H.
        • Hugosson J.
        • Roobol M.J.
        • et al.
        Screening and prostate-cancer mortality in a randomized European study.
        N Engl J Med. 2009; 360: 1320-1328
        • van Iersel C.A.
        • de Koning H.J.
        • Draisma G.
        • et al.
        Risk-based selection from the general population in a screening trial: selection criteria, recruitment and power for the Dutch-Belgian randomised lung cancer multi-slice CT screening trial (NELSON).
        Int J Cancer. 2007; 120: 868-874
        • Crawford M.J.
        • Thana L.
        • Methuen C.
        • et al.
        Impact of screening for risk of suicide: randomised controlled trial.
        Br J Psychiatry. 2011; 198: 379-384
        • Hanley J.A.
        • McGregor M.
        • Liu Z.
        • Strumpf E.C.
        • Dendukuri N.
        Measuring the mortality impact of breast cancer screening.
        Can J Public Health. 2013; 104: e437-e442
        • Cuzick J.
        • Cafferty F.H.
        • Edwards R.
        • Møller H.
        • Duffy S.W.
        Surrogate endpoints for cancer screening trials: general principles and an illustration using the UK Flexible Sigmoidoscopy Screening Trial.
        J Med Screen. 2007; 14: 178-185
        • Jonas D.E.
        • Feltner C.
        • Amick H.R.
        • et al.
        Screening for asymptomatic carotid artery stenosis: a systematic review and meta-analysis for the U.S. Preventive Services Task Force.
        Ann Intern Med. 2014; 161: 336-346
        • Infante M.
        • Lutman F.R.
        • Cavuto S.
        • et al.
        Lung cancer screening with spiral CT: baseline results of the randomized DANTE trial.
        Lung Cancer. 2008; 59: 355-363
        • Hanley J.A.
        Measuring mortality reductions in cancer screening trials.
        Epidemiol Rev. 2011; 33: 36-45
        • Tabar L.
        • Duffy S.W.
        • Yen M.F.
        • et al.
        All-cause mortality among breast cancer patients in a screening trial: support for breast cancer mortality as an end point.
        J Med Screen. 2002; 9: 159-162
        • Steele R.
        • Brewster D.
        Should we use total mortality rather than cancer specific mortality to judge cancer screening programmes? no.
        BMJ. 2011; 343: d6397
        • Schoen R.E.
        • Pinsky P.F.
        • Weissfeld J.L.
        • et al.
        Colorectal-cancer incidence and mortality with screening flexible sigmoidoscopy.
        N Engl J Med. 2012; 366: 2345-2357
        • Black W.C.
        • Haggstrom D.A.
        • Welch H.G.
        All-cause mortality in randomized trials of cancer screening.
        J Natl Cancer Inst. 2002; 94: 167-173
        • Seidman H.
        • Silverberg E.
        • Bodden A.
        Probabilities of eventually developing and of dying of cancer (risk among persons previously undiagnosed with the cancer).
        Cancer. 1978; 28: 33-46
        • Jemal A.
        • Simard E.P.
        • Dorell C.
        • et al.
        Annual Report to the Nation on the Status of Cancer, 1975–2009, featuring the burden and trends in human papillomavirus (HPV)–associated cancers and HPV vaccination coverage levels.
        J Natl Cancer Inst. 2013; 105: 175-201
        • Brenner H.
        • Hoffmeister M.
        • Jansen L.
        Comparisons of colorectal cancer mortality between screening participants and the general population are strongly biased unless an incidence-based mortality approach is used.
        J Clin Epidemiol. 2014; 67: 184-189
        • Zakher B.
        • Cantor A.G.
        • Pappas M.
        • Daeges M.
        • Nelson H.D.
        Screening for gonorrhea and chlamydia: a systematic review for the U.S. Preventive Services Task Force.
        Ann Intern Med. 2014; 161: 884-894
        • Cantor A.G.
        • Bougatsos C.
        • Dana T.
        • Blazina I.
        • McDonagh M.
        Routine iron supplementation and screening for iron deficiency anemia in pregnancy: a systematic review for the U.S. Preventive Services Task Force.
        Ann Intern Med. 2015; 162: 566-576
        • Oudkerk M.
        • Heuvelmans M.A.
        Screening for lung cancer by imaging: the Nelson study.
        JBR-BTR. 2013; 96: 163-166
        • LeBlanc E.S.
        • Zakher B.
        • Daeges M.
        • Pappas M.
        • Chou R.
        Screening for vitamin D deficiency: a systematic review for the U.S. Preventive Services Task Force.
        Ann Intern Med. 2015; 162: 109-122
        • Lin J.S.
        • O'Connor E.
        • Rossom R.C.
        • Perdue L.A.
        • Eckstrom E.
        Screening for cognitive impairment in older adults: a systematic review for the U.S. Preventive Services Task Force.
        Ann Intern Med. 2013; 159: 601-612
        • Chiu S.Y.
        • Malila N.
        • Yen A.M.
        • Anttila A.
        • Hakama M.
        • Chen H.H.
        Analytical decision model for sample size and effectiveness projections for use in planning a population-based randomized controlled trial of colorectal cancer screening.
        J Eval Clin Pract. 2011; 17: 123-129
        • Etzioni R.
        • Gulati R.
        • Cooperberg M.R.
        • Penson D.M.
        • Weiss N.S.
        • Thompson I.M.
        Limitations of basing screening policies on screening trials: the US Preventive Services Task Force and prostate cancer screening.
        Med Care. 2013; 51: 295-300
        • Hamra G.
        • MacLehose R.
        • Richardson D.
        Markov Chain Monte Carlo: an introduction for epidemiologists.
        Int J Epidemiol. 2013; 42: 627-634
        • Bailhache M.
        • Bénard A.
        • Salmi L.R.
        Simulation of the impact of programs for prevention and screening of Pediatric Abusive Head Trauma.
        J Neurotrauma. 2016; 33: 1397-1403
        • Cougnard A.
        • Salmi L.R.
        • Salamon R.
        • Verdoux H.
        A decision analysis model to assess the feasibility of the early detection of psychosis in the general population.
        Schizophr Res. 2005; 74: 27-36
        • Tamura H.
        • Goto R.
        • Akune Y.
        • Hiratsuka Y.
        • Hiragi S.
        • Yamada M.
        The clinical effectiveness and cost-effectiveness of screening for age-related macular degeneration in Japan: a Markov modeling study.
        PLoS One. 2015; 10: e133628
        • Mühlberger N.
        • Kurzthaler C.
        • Iskandar R.
        • et al.
        The ONCOTYROL prostate cancer outcome and policy model: effect of prevalence assumptions on the benefit-harm balance of screening.
        Med Decis Making. 2015; 14: 758-772
        • Seigneurin A.
        • Labarère J.
        • Duffy S.W.
        • Colonna M.
        Overdiagnosis associated with breast cancer screening: a simulation study to compare lead-time adjustment methods.
        Cancer Epidemiol. 2015; 39: 1128-1135
        • Karnon J.
        • Brennan A.
        • Akehurst R.
        Decision modeling to inform decision making: seeing the wood for the trees.
        Med Decis Making. 2010; 10: E20-E22
        • Pastorino U.
        • Rossi M.
        • Rosato V.
        • et al.
        Annual or biennial CT screening versus observation in heavy smokers: 5-year results of the MILD trial.
        Eur J Cancer Prev. 2012; 21: 308-315
        • Campbell D.T.
        • Stanley J.C.
        Experimental and Quasi-experimental Designs for Research.
        Houghton Mifflin Co, Boston, MA1963
        • Gøtzsche P.C.
        • Jørgensen K.J.
        The Breast Screening Programme and misinforming the public.
        J R Soc Med. 2011; 104: 361-369
        • Vera-Badillo F.E.
        • Templeton A.J.
        • de Gouveia P.
        • et al.
        Androgen receptor expression and outcomes in early breast cancer: a systematic review and meta-analysis.
        J Natl Cancer Inst. 2014; 106: djt319
        • Voas R.B.
        • Fell J.C.
        Preventing impaired driving opportunities and problems.
        Alcohol Res Health. 2011; 34: 225-235
        • Ahern J.
        • Hubbard A.
        • Galea S.
        Estimating the effects of potential public health interventions on population disease burden: a step-by-step illustration of causal inference methods.
        Am J Epidemiol. 2009; 169: 1140-1147
        • Sackett D.L.
        • Rosenberg W.M.
        • Gray J.A.
        • Haynes R.B.
        • Richardson W.S.
        Evidence based medicine: what it is and what it isn't.
        BMJ. 1996; 312: 71-72
        • Charles C.
        • Gafni A.
        • Freeman E.
        The evidence-based medicine model of clinical practice: scientific teaching or belief-based preaching?.
        J Eval Clin Pract. 2011; 17: 597-605
        • Evidence-Based Medicine Working Group
        Evidence-based medicine: a new approach to teaching the practice of medicine.
        JAMA. 1992; 268: 2420-2425
        • Scott P.A.
        Unsupervised self-testing as part public health screening for HIV in resource-poor environments: some ethical considerations.
        AIDS Behav. 2014; 18: S428-S444
        • van Dam L.
        • Bretthauer M.
        Ethical issues in colorectal cancer screening.
        Best Pract Res Clin Gastroenterol. 2014; 28: 315-326
        • Petrova D.
        • Garcia-Retamero R.
        • Cokely E.T.
        Understanding the harms and benefits of cancer screening: a model of factors that shape informed decision making.
        Med Decis Making. 2015; 35: 847-858
        • Rossi P.G.
        • Ferroni E.
        • Barca A.
        • et al.
        Quality appraisal of documents producing recommendations for breast, colorectal and cervical cancer screening.
        Epidemiol Biostat Public Health. 2014; 11: e9062