Sir Karl Popper, swans, and the general practitioner

Sir Karl Popper, swans, and the general practitioner
of 23
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  A PATIENT ’ S JOURNEY Sir Karl Popper, swans, and the general practitioner At age 50 years, Ron Berghmans had back pain, which despite visits to his general practitionerincreased in severity. Eventually Ron referred himself to a neurologist and realised things weremore serious than anyone first thought Ron Berghmans lecturer  1 , Harry C Schouten professor  2 1 Department of Health, Ethics and Society, CAPHRI School for Public Health and Primary Care, Maastricht University, PO Box 616, 6200MDMaastricht, Netherlands ; 2 Department of Internal Medicine, University Hospital Maastricht, PO Box 5800, 6202 AZ Maastricht, NetherlandsThis is one of a series of occasional articles by patients about theirexperiences that offer lessons to doctors. The BMJ  welcomescontributions to the series. Please contact Peter Lapsley( for guidance. What is truth? And what is truth in the encounter between apatient and a general physician? Is truth always determinedeither from the “observer” or the “patient’s” perspective andthus subjective? Or does something like objective truth exist?DuringmyjourneyasapatientIexperiencedpainandemotionalsuffering.Ididnotworryaboutphilosophicalquestionsrelatingto truth in general or truth as it applies to medicine and themedical encounter. Questions relevant for the philosophy of science seem to be far removed from the sickbed and thedelivery of professional medical care and treatment. But arethey?The most difficult part of my journey through the medicalsystem was the entry. As a 50 year old I already had a historyof low back pain. Around the time of my 50th birthday the painincreased. This was a reason to consult my general practitioner(GP). Quickly, and without physical examination, she wasconvinced that there was no reason to worry. She knew I wasa busy academic who had a stressful working life with myresearchandteaching.Heradvicewastotrytoreducethestressas much as possible by not working too hard (in my patient fileshe noted: “functional, non-specific low back pain resultingfromworkstress”).Heradvicedidn’thelp.Onthecontrary,theback pain increased and a month later I consulted her again.This time she prescribed pain medication. This was of no availeither. I returned to see her again some weeks later and askedher to refer me for radiography. She was very reluctant to dothis and explained the “dangers” of radiation. Only after muchpressure did she agree. The radiographs of the lower parts of mybackdidn’tshowanymajoraberrations,althoughthereweresome minor abnormalities at the vertebrae. My GP seemedcontent to see these peculiarities as the possible (probable)causes of my pain. Nevertheless, my situation kept worsening.During the long nights I experienced severe pain; I would liein bed in a separate room, howling like a wolf (as my wife said)andhavinginvoluntarycontractionsofmylegs.Itbecamemoreand more difficult to stand up and walk, and I lost the ability tostay upright.In the Dutch healthcare system the GP has a gatekeeping role.This makes it difficult for patients to access specialisedhealthcare services without a referral from their GP. In my casemy GP refused to refer me to a neurologist. As my medicalcondition kept worsening, I decided to make an appointmentforaconsultationwithaneurologistinBelgium(wherepatientshave the freedom to contact specialised healthcare staff independently).WhenIconfrontedmyGPaboutthis,shenotedin my patient file: “It’s the patient’s wish, so be it. . .”Ultimately, this neurologist referred me to the neurosurgerydepartment,whereanassistantdecidedtodomagneticresonanceimaging of the whole of my back (not just the lower parts).Thesescansshowedthatcanceroustissue,whichwasallaroundmy lungs and belly, was already pressing on my spinal cord.This explained the progressive and extreme pain I had beenexperiencing. Chemotherapy and autologous stem celltransplantation were successful.Ten months had passed between my first encounter with myGP and the final diagnosis of non-Hodgkin lymphoma and itwasonlybecauseathoughtfulneurosurgeryassistantwaswillingto look beyond the lower back (on which the GP had focused)that the cancerous tissue became visible and the truth wasrevealed. I crossed the threshold and my journey as a cancerpatient eventually started.So what about the truth? Does it matter? Yes, it does. Lookingback at my entry into the medical system, I came to realise theimportance, indeed the lifesaving, potential of the work of thephilosopher of science Sir Karl Popper. In particular, his viewson truth are very relevant for medicine and GP’s. My GP’s way Correspondence to: R Berghmans For personal use only: See rights and reprints BMJ  2011;343:d5469 doi: 10.1136/bmj.d5469 Page 1 of 2 Practice PRACTICE  of approaching the truth would be abhorrent to Popper. Bysearching for confirmation of her initial diagnosis she foundnothingbutatruththatwasalreadythere.AsPopperwouldsay,if you suppose all swans are white, you will not find any swansthan white swans. Popper’s revolutionary insight was that youshould, in fact, be searching for black swans. Only then will itbe possible to gain better knowledge about the world and whathappensinit.Thissocalledfalsificationprinciplehashadgreatimpact on the philosophy of science and the way in whichempirical scientific knowledge is achieved. Briefly, it involvestheformulationoftestablehypotheses(thatis,predictionsaboutwhat is the matter or what will happen) and the intention tofalsify these expectations. Do not try to look for white swans,but instead look for black swans. Only this way, in Popper’sview, can the truth be approached.Subjectively, I think the course set by my GP (and continuedby several of my specialists who based their actions on theinformation given by her) might have had serious if not fatalconsequences for my health and existence. This is obviously anobservation I cannot confirm or falsify but nevertheless doesnot seem unreasonable.Doctors, andinparticular GP’s,shouldalwaysbearinmind thefallibility of their assumptions. This is the central message KarlPopper has for the daily practice of medicine. When I typed“falsification+principle+medical +diagnosis” into Google itshowed 223 000 hits, but only a few bits of information wereusable. One relevant piece was the Wikipedia entry“Confirmationbias”(,whichstated:“Confirmationbias(alsocalledconfirmatorybias or myside bias) is a tendency for people to favorinformation that confirms their preconceptions or hypothesesregardlessofwhethertheinformationistrue.Asaresult,peoplegatherevidenceandrecallinformationfrommemoryselectively,and interpret it in a biased way. The biases appear in particularfor emotionally significant issues and for established beliefs.”And with regards to physical and mental health Wikipediastated: “Raymond Nickerson, a psychologist, blamesconfirmation bias for the ineffective medical procedures thatwereusedforcenturiesbeforethearrivalofscientificmedicine.Ifapatientrecovered,medicalauthoritiescountedthetreatmentas successful, rather than looking for alternative explanationssuch as that the disease had run its natural course.” I found nolinkstoinformationonconfirmationbiaswithregardstomedicaldiagnosis, treatment, or general physicians.AtnopointduringtheprocesswasmyGPwillingtoreconsiderher initial diagnosis, although my situation was progressivelyworsening and becoming unbearable. She was, and stayed,convinced that her truth was the truth—that I had “aspecificlow back pain,”—and this preconception kept her away fromconsidering any contrary evidence that would refute thispreconception. She didn’t look for black swans, all she sawwere white swans.My hypothesis is that in medical schools and education no, orinsufficient,attentionisbeinggiventothephilosophyofsciencein general, and Popper’s theory in particular. Now it’s time totry to refute this hypothesis, and make real progress in dailymedical practice, for the sake of patients with complaints thatappear to be clear, obvious, undeniable and easy to explain, butare not. This requires a transformation in the mindset of GP’sand a fundamental willingness to try not to look for support fortheir convictions, but to search instead for alternativeexplanations for the complaints of patients. Competing interests: All authors have completed the ICMJE uniformdisclosure form onrequest from the corresponding author) and declare: no support fromany organisation for the submitted work; no financial relationships withany organisations that might have an interest in the submitted work inthe previous three years; no other relationships or activities that couldappear to have influenced the submitted work.Provenance and peer review: Not commissioned; not externally peerreviewed. 1 Groopman J. How doctors think. First Mariner Books, 2008.2 A predictive model for aggressive non-Hodgkin’s lymphoma. The InternationalNon-Hodgkin’s Lymphoma Prognostic Factors Project. N Engl J Med  1993;329:987. Accepted: 6 June 2011Cite this as: BMJ  2011;343:d5469  © BMJ Publishing Group Ltd 2011 For personal use only: See rights and reprints BMJ  2011;343:d5469 doi: 10.1136/bmj.d5469 Page 2 of 2 PRACTICE  A clinician’s perspective When you see a patient with a new complaint, your way of thinking may be influenced by previous encounters with thisspecific patient or, alternatively, by your experience with previous patients with similar complaints. Many times this maybe very helpful for these patients and may spare them potentially dangerous diagnostic procedures. However, sometimesthis biased personal experience may damage a patient, as described by Ron Berghmans.Although we all know in this era of evidence based medicine that personal experience, or even expert opinion, is generallyconsideredtoberelativelyweakevidence,weallruntheriskposedbyflawedfirstorsecondimpressions.Readersdoubtingthis really should read How Doctors Think  by Jerome Groopman. 1 This book teaches us to be very modest about ourprofessional approach to individual patients’ complaints and should stimulate us all to approach patients in an open way,designing a formal listing of differential diagnoses, based on likelihood of occurrence.Being a haematologist working in a university hospital setting, I realise that I am rarely the first physician to see the patient.As in Ron’s experience, it is generally the GP who considers a malignancy and requests a computed tomography scan,or a neurologist who raises the suspicion of a malignancy resulting in a referral to someone like me, who then can hardlymiss the obvious diagnosis.Large cell lymphoma is a disease srcinating from our developing immune system. The malignant tumour is generally verychemosensitive, and the combination of chemotherapy and a monoclonal antibody results in high response rates and asubstantial subset of patients cured. Did the diagnostic delay harm Ron? We will never know. However, a diagnostic delaymayresultinalargertumourburden,moreextensivespreadofthedisease(higherstageormoreextranodallocalisations),ordevelopmentofthetumourtoamoreaggressivesubtype(alsoassociatedwithahigherleveloflactatedehydrogenase),which are all characteristics of a poor prognosis, as shown by the international prognostic index. 2 But this is only part ofthe problem. The fear and anxiety for a patient when their complaints are not taken seriously by their physicians can bedevastating and result in loss of confidence in our healthcare providers and may influence compliance and perseverance.Or even worse, some patients may turn to alternative medicine, resulting in even more damage.Especially in these days of limited financial budgets and high work load we all should try to spend healthcare money asappropriately as possible, without being biased by our opinions, first impressions, and budgetary constraints.Harry C Schouten For personal use only: See rights and reprints BMJ  2011;343:d5469 doi: 10.1136/bmj.d5469 Page 3 of 2 PRACTICE  Estimating treatment effects for individual patientsbased on the results of randomised clinical trials OPEN ACCESS Johannes A N Dorresteijn epidemiologist and medical doctor  1 , Frank L J Visseren professor of vascular medicine, epidemiologist, and internist  1 , Paul M Ridker Eugene Braunwald professor of medicine, epidemiologist, and cardiologist  2 , Annemarie M J Wassink internist and postdoctoral researcher  1 , Nina P Paynter assistant professor of epidemiology  2 , Ewout W Steyerberg professor of medical decision making, and methodologist  3 , Yolanda van der Graaf professor of epidemiology and imaging  4 , Nancy R Cook associate professor of biostatistics and epidemiology  2 1 Department of Vascular Medicine, University Medical Center Utrecht, PO Box 85500, 3508 GA Utrecht, Netherlands; 2 Division of PreventiveMedicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, USA; 3 Department of Public Health, Erasmus Medical Center,Rotterdam, Netherlands; 4 Julius Center for Health Sciences and Primary Care, Utrecht, Netherlands Abstract Objectives To predict treatment effects for individual patients based ondatafromrandomisedtrials,takingrosuvastatintreatmentintheprimaryprevention of cardiovascular disease as an example, and to evaluatethe net benefit of making treatment decisions for individual patientsbased on a predicted absolute treatment effect. Setting As an example, data were used from the Justification for theUse of Statins in Prevention (JUPITER) trial, a randomised controlledtrial evaluating the effect of rosuvastatin 20 mg daily versus placebo onthe occurrence of cardiovascular events (myocardial infarction, stroke,arterial revascularisation, admission to hospital for unstable angina, ordeath from cardiovascular causes). Population 17 802 healthy men and women who had low densitylipoproteincholesterollevelsoflessthan3.4mmol/LandhighsensitivityC reactive protein levels of 2.0 mg/L or more. Methods Data from the Justification for the Use of Statins in Preventiontrial were used to predict rosuvastatin treatment effect for individualpatients based on existing risk scores (Framingham and Reynolds) andon a newly developed prediction model. We compared the net benefitofpredictionbasedrosuvastatintreatment(selectivetreatmentofpatientswhosepredictedtreatmenteffectexceedsadecisionthreshold)withthenet benefit of treating either everyone or no one. Results The median predicted 10 year absolute risk reduction forcardiovascular events was 4.4% (interquartile range 2.6-7.0%) basedon the Framingham risk score, 4.2% (2.5-7.1%) based on the Reynoldsscore, and 3.9% (2.5-6.1%) based on the newly developed model(optimalfitmodel).Predictionbasedtreatmentwasassociatedwithmorenet benefit than treating everyone or no one, provided that the decisionthreshold was between 2% and 7%, and thus that the number willing totreat (NWT) to prevent one cardiovascular event over 10 years wasbetween 15 and 50. Conclusions Data from randomised trials can be used to predicttreatmenteffectintermsofabsoluteriskreductionforindividualpatients,based on a newly developed model or, if available, existing risk scores.The value of such prediction of treatment effect for medical decisionmaking is conditional on the NWT to prevent one outcome event. Trial registration number NCT00239681. Introduction Usually the results of trials are implemented in clinical practicebyeithertreatingallpatients(inthecaseofapositivetrialresult)or treating no one (in the case of a negative trial result),expecting the treatment effect for every patient to be similar tothe average treatment effect in the srcinal trial. Cliniciansintuitively know that this idea is oversimplified because inrealitysomepatientsbenefitmorethanaveragefromtreatment,whereas others do not or may even be harmed. 1-6 The direct translation of trial results to individual patients inclinical practice is, however, complicated by some importantlimitations. The treatment effects of randomised trials are Correspondence to: F L J Visseren F.L.J.Visseren@umcutrecht.nlExtra material supplied by the author (see 1: development of optimal fit modelAppendix 2: example calculation of net benefit assessment methodAppendix 3: application of treatment effect prediction methods to other datasets No commercial reuse: See rights and reprints BMJ  2011;343:d5888 doi: 10.1136/bmj.d5888 Page 1 of 13 Research RESEARCH  typically expressed in terms of relative risks or hazard ratios ata group level—that is, treatment versus control. Yet treatmentthat is associated with a considerable reduction in relative risk will still result in a modest absolute effect when the incidencerate of the disease is low. Absolute risk reduction is usuallymoreinformativebecauseitcombinestherelativeriskreductionand the incidence rate of the disease outcome. 1 2 The absoluterisk reduction is sometimes expressed in trial reports as thenumber needed to treat (NNT). Still, implicit in the use of estimates at group level is that all patients are at average risk and all have the same likelihood of response to treatment.Usually at least one of these two assumptions is untrue becausethe expected absolute risk reduction resulting from treatmentoften depends on the characteristics of individual patients. 1-6 Although prespecified subgroup analyses take a step towardsidentifying those characteristics of patients that modify thetreatment effect, some important limitations are retained. Insubgroupanalysesthestudycohortistypicallydividedaccordingtothepresenceorabsenceofasinglepatientcharacteristicsuchas diabetes, age (below or above a certain limit), or sex, and theeffect of the intervention is presented accordingly. However,these univariable analyses do not fully incorporate all availablepatient characteristics and are less well powered but still returnrelative,ratherthanabsolute,averageeffectmeasuresatagrouplevel. 2-4 Amorecomprehensiveapproachtowardsmakingwellinformeddecisions about treatment is to predict the treatment effect forindividual patients based on all relevant characteristicstogether. 1-5 7 Although not yet widely appreciated, data fromrandomised controlled trials usually provide an opportunity todevelop models for the prediction of a treatment effect on thebasis of individual patient characteristics. 1 5 Such models canenable clinicians to estimate a treatment effect for individualpatients in terms of absolute risk reduction for the disease of interest.Thiscanbedonebeforethestartofintendedtreatment,and therefore decisions about treatment can be based on suchpredictions. 4 Moreover, individualised predictions of treatmenteffect provide an opportunity to determine which implicationstheresultsofrandomisedtrialsshouldhaveinclinicalpractice. 6 Makingtreatmentdecisionsonthebasisofapredictedtreatmenteffect for individual patients may in some situations result inmore net benefit on a group level than treating all patients (inthe case of a positive trial result) or treating no one (in the caseofanegativetrialresult).Althoughthisapproachisoccasionallyused in the research of cancer 8-10 and cardiovascular disease, 11-13 the full potential has yet to be recognised by both researchersand clinicians.We developed and evaluated methods for predicting treatmenteffect using rosuvastatin in individual patients in a primaryprevention setting based on data from the Justification for theUseofStatinsinPreventiontrial. 14 Thisstudywasarandomised,double blind, placebo controlled, multicentre trial that showedonaveragea44%relativeriskreductioninmajorvasculareventsin those treated with rosuvastatin. As the trial was carried outin a primary prevention cohort at moderate absolute risk forcardiovascular disease, the overall treatment effect was modestforaverageabsoluteriskreduction.Thereforethetrialrepresentsa typical situation in which the prediction of treatment effectcan be used to identify those who will benefit from treatment.We predicted treatment effects for individual patients based ondata from randomised trials, taking rosuvastatin treatment inprimary prevention of cardiovascular disease as an example,and evaluated the net benefit of making treatment decisions forindividualpatientsbasedonpredictedabsolutetreatmenteffect. Methods The design, rationale, and outcomes of the Justification for theUse of Statins in Prevention trial are described in detailelsewhere. 14-16 Briefly, the trial evaluated the effect of rosuvastatin 20 mg daily compared with placebo on theoccurrence of myocardial infarction, stroke, arterialrevascularisation, admission to hospital for unstable angina, ordeath from cardiovascular causes among 17 802 apparentlyhealthy men and women who had low density lipoproteincholesterollevelsoflessthan3.4mmol/L(130mg/dL)andhighsensitivity C reactive protein levels of 2.0 mg/L or more. Aftera median follow-up of 1.9 years the hazard ratio for occurrenceof the primary end point was 0.56 (95% confidence interval0.46 to 0.69), favouring rosuvastatin. 14 Univariable subgroupanalyses (for example, for age, sex, smoking status, ethnicity,and Framingham risk score) showed no significant deviationsfrom this effect size. Estimating treatment effects for individualpatients Weestimatedthebaseline10yearriskforcardiovascularevents(myocardial infarction, stroke, arterial revascularisation,admission to hospital for unstable angina, or death fromcardiovascularcauses)forindividualpatientsifuntreated,usingthe existing Framingham risk score 17 and the Reynolds risk score,withoutpriorupdatingorrefittingofthecoefficients(seebox). 18 19 Basedontheassumptionthattreatmenteffectincreaseslinearly with baseline risk (fig 1⇓), we estimated the patient’sresidual risk when given treatment by multiplying baseline risk by the overall relative effect measure (relative risk or hazardratio) from the srcinal trial report. Consequently the estimatedabsoluteriskreductionachievedbytreatmentwithrosuvastatinfor 10 years (10 year treatment effect) is equal to the differencebetween these two [individual treatment effect=(1−overallrelative effect measure from trial)×baseline risk derived froman existing prediction model].Alternatively we developed a new prediction model (optimalfit model) based on trial data only (see web extra appendix 1).A theoretical advantage of this strategy over using existing risk scores is that the model may be better calibrated to thepopulation of interest. Furthermore, such a model is not basedon the assumption that treatment effect increases linearly withbaseline risk: modification of the treatment effect by patientcharacteristics can be tested and, if significant, included in themodel. Importantly, even in the absence of subgroup effectsdefined by univariable characteristics, a multivariable adjustedprediction model may contain such modifications of treatmenteffect. In situations where no existing prediction models areavailable, developing a new prediction model may be the onlyoption. Performance of the prediction models We assessed the calibration of the predictions based on theFraminghamriskscore,theReynoldsriskscore,andtheoptimalfit model. To do this we plotted the observed Kaplan-Meiersurvival for cardiovascular events at two years within 10ths of the predicted survival against the mean predicted two yearsurvival of each 10th and by the P value derived from theHosmer-Lemeshowtest.Basedontheassumptionthatthehazardrate is constant and thus survival is exponential over time wederived two year risk estimates of the Framingham risk scoreand the Reynolds risk score from the 10 year predicted risks.Discrimination was assessed by calculation of the C statistic. No commercial reuse: See rights and reprints BMJ  2011;343:d5888 doi: 10.1136/bmj.d5888 Page 2 of 13 RESEARCH

Soluções Tetra

Mar 23, 2018


Mar 23, 2018
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks