The principles f evidence-based medicine By the end f this mdule yu shuld be able t: Describe what evidence based medicine is Knw where t find quality evidenced based medicine n the internet Be able t critically appraise a research paper Dctrs are nw faced with an increasing bdy f infrmatin and a rapidly changing evidence base frm which t guide their clinical decisin making prcess. This can make it difficult t maintain up-tdate clinical practice and may result in heavy reliance n ut-f-date undergraduate teaching, textbks and anecdtal evidence frm prir experiences. Evidence-based medicine is therefre a prcess f cntinued medical educatin which invlves a systematic apprach t finding, evaluating and implementing research findings as the basis fr clinical decisins. It is abut asking questins, finding and appraising the relevant data, and harnessing that infrmatin fr everyday clinical practice 1. Of curse a dctr s expertise is still f great imprtance when making clinical decisins and shuld be cmbined with the principles f evidence-based medicine in rder t prvide ptimum patient care and infrmed patient chice. There are fur steps in evidence based medicine 1 : 1. Frmulate a clinical questin 2. Search the literature fr relevant clinical articles 3. Evaluate the evidence fund 4. Implement useful findings int clinical practice 1. Frmulate a clear clinical questin frm a patient's prblem The questin may relate t any aspect f patient care including diagnsis, prgnsis and management. The scenari belw is an example. A 32 year ld lady is seen fr her initial visit in the antenatal clinic at 16 weeks gestatin in her 4 th pregnancy. She has a pr bstetric histry and all her pregnancies have been cmplicated by severe pre-eclampsia. In her first pregnancy she was delivered by cesarean sectin at 34 weeks due t pre-eclampsia and her sn frm that pregnancy is nw 5 years f age and well. In her 2 nd and 3 rd pregnancies she was again delivered prematurely at 28 weeks and 27 weeks gestatin respectively by cesarean sectin fr fulminating pre-eclampsia, but unfrtunately neither baby survived. Her sister als suffered frm pre-eclampsia in her last 2 pregnancies and was given aspirin which she wuld nw like yu t prescribe her. Yu are unfamiliar with this tpic. What kind f questins regarding the management f this lady with aspirin wuld yu like t find ut frm the literature? Jt dwn sme f yur answers nw. Des the use f aspirin reduce the risk f develping pre-eclampsia? Des the effect f aspirin differ if started befre r after 20 weeks gestatin? Des the effect f aspirin differ if high dse aspirin is used versus lw dse? Des the use f aspirin in pregnancy result in increased risk f haemrrhage? Is there an increase in perinatal mrbidity r mrtality assciated with its use? Nw think abut hw yu may g abut finding the answer t these questins. 2. Search the literature fr relevant clinical articles Unfrtunately, althugh textbks are an invaluable resurce fr updating and refreshing knwledge n a particular tpic, they wuld have t be cntinually re-published in rder t
include the mst up-t-date infrmatin. Frtunately, wide access t the internet and imprved cmputer literacy amngst dctrs means that there is a wealth f infrmatin within reach. Databases that can be used include: Medline Cchrane Knwledge finder Pubmed Embase Once the references f relevant papers have been fund, mst f these databases will prvide a link t the abstract f the paper and at times t the full text. Full texts can als be fund as riginal paper cpies in libraries. 3. Evaluate the evidence fr its validity and usefulness Once yu have fund a relevant research paper frm yur search, the next task is t be able t critically appraise the wrk that has been carried ut. It is imprtant t remember that nt all research is valid, relevant r applicable t yur wn patient ppulatin. Therefre with the vast numbers f papers being published in jurnals each mnth, dctrs must be able t pick ut the mst rbust research and have the tls t be able t relate the results back t their wn patients. Here are sme questins that yu may want t ask yurself when appraising a research paper. Methdlgy: What type f study is this? If it is a randmised trial were the participants and/ r researchers blinded? What is the main questin (utcme) that this study is tying t answer? Are there clearly-defined selectin criteria fr which patients were included and excluded? Was the study adequately cntrlled? Are there any bvius biases in the way the patients were selected r in the way that data was cllected? Have they clearly described every aspect f the methdlgy s that the study culd be reprduced? Results: Are all patients that were included int the start f the study accunted fr? Is there any missing data? Were the grups cmparable in all imprtant aspects except fr the variable being studied? If nt has this been accunted fr? Are there any bvius cnfunding influences? Was the study large enugh t make the results valid (See statistical tls- pwer)? Was the study cntinued fr lng enugh and was fllw up cmplete enugh t make the results credible? Discussin Are the results discussed with reference t ther imprtant literature? D the discussin and cnclusins speculate far beynd what has been shwn in the study? Generalisability:
The results f the study may nt be generalisable t yur patients even if the study had a sund methdlgy. Fr example a study shwing benefit f an interventin in Swedish wmen aged 20-30 years may nt be relevant t lder wmen in Sub-Saharan Africa. Nw jt dwn sme ideas f hw yu wuld design a trial t answer the first questin frm the previus example (Des the use f aspirin reduce the risk f develping pre-eclampsia?) using the imprtant pints abve. Here is a link f a paper published in the Lancet in 2007 regarding aspirin use and the risk f pre-eclampsia. CLASP: a randmised trial f lw-dse aspirin fr the preventin and treatment f preeclampsia amng 9364 pregnant wmen. CLASP (Cllabrative Lw-dse Aspirin Study in Pregnancy) Cllabrative Grup. Lancet. 343(8898):619-29, 1994. Hw des yur design cmpare t their methdlgy? What are sme f the gd and bad pints f this paper? The Cchrane library cntains reviews that analyse the results frm many different trials in rder t prvide infrmatin n pssible healthcare interventins. See a review by Duley et al. (2007) 2 fr a review n the use f aspirin in preventing pre-eclampsia.. 4. Statistical tls required fr critical appraisal: Pwer: The pwer f a trial is determined using a mathematical calculatin that tells us if the number f patients enrlled in the trial is sufficient t detect a difference between treatment arms. It estimates the ability f a trial t detect a statistically significant difference f a particular size between the treated and cntrl grups. Number needed t treat The number needed t treat (NNT) is an epidemilgical measure used in assessing the effectiveness f a health-care interventin, typically a new treatment. The NNT is the number f patients wh need t be treated in rder t prevent ne additinal bad utcme. Fr example in the MAGPIE trial the NNT was 91. That is the same as saying 91 wmen need t be treated with magnesium sulphate t prevent ne eclamptic fit. It is defined as the inverse f the abslute risk reductin
Attributable risk Attributable risk is the disease rate in expsed persns minus that in unexpsed persns. Relative risk Relative risk (RR) is the risk f an event (r f develping a disease) relative t expsure. Relative risk is a rati f the prbability f the event ccurring in the expsed grup versus a nn-expsed grup. Relative risk is less relevant t making decisins in risk management than is attributable risk as it des nt take int accunt the abslute risk f the cnditin. Fr example, if given a chice between a dubling in their risk f develping preeclampsia r a dubling in their risk f develping acute fatty liver f pregnancy, mst infrmed peple wuld pt fr the latter. The relative risk is the same (tw), but the crrespnding attributable risk is lwer because acute fatty liver f pregnancy is a rarer disease. Abslute risk Abslute risk is the chance f a persn develping a specific disease ver a specified time-perid. Fr example, let s say in a grup f 20,000 wmen, 1,600 develp endmetrial cancer ver 50 years. The risk f any ne individual getting this disease wuld be 1,600 divided by 20,000 r 0.08. That means that the abslute risk wuld be 8 percent r 8 in 100. Odds rati The dds rati is a way f cmparing whether the prbability f a certain event is the same fr tw grups. It is calculated by dividing the dds in the treated r expsed grup by the dds in the cntrl grup. An dds rati f 1 implies that the event is equally likely in bth grups. An dds rati greater than ne implies that the event is mre likely in the first grup. An dds rati less than ne implies that the event is less likely in the first grup. Fr example, n average 51 bys are brn in every 100 births, s the dds f any randmly chsen delivery being that f a by is: Number f bys 51 / number f girls 49, r abut 1.04. We culd als have calculated the same answer as the rati f the baby being a by (0.51) and it nt being a by (0.49). 5. Implement useful findings in clinical practice Once the paper has been evaluated and has been deemed valid and relevant dctrs can use this infrmatin in a number f ways: Use findings t directly alter a patient s care Develp team prtcls and hspital guidelines.
Use the evidence as a basis fr audit. A gd way t ensure cntinued prfessinal develpment in this way and t maintain evidence based practices in yur department is t set up a jurnal club. A jurnal club is a grup f individuals wh meet regularly t critically evaluate recent articles in scientific literature. Each meeting is usually rganized arund a defined subject in basic r applied research. Summary f learning pints: Evidence based medicine is part f cntinued prfessinal develpment and invlves the systematic appraisal f research in rder t imprve ur clinical practice. It needs t be cmbined with clinical experience and expertise and tailred t yur individual patient. Results f research papers may be invalidated by pr methdlgy r statistical analysis and may be irrelevant t yur patient ppulatin. Identifying gd quality research relies n the ability t apprpriately interpret their results and apply them apprpriately. Jurnal clubs are a gd way f keeping up-t-date and practicing critical appraisal skills. References: 1. Evidence based medicine: an apprach t clinical prblem-slving. W Rsenberg et al. BMJ 1995;310:1122-1126 2. Duley L, Hendersn-Smart DJ, Meher S, King JF. Antiplatelet agents fr preventingpreeclampsia and its cmplicatins. Cchrane Database f Systematic Reviews 2007, Issue 2. Art. N.: CD004659. DOI:10.1002/14651858.CD004659.pub2. Useful website: http://www.cebm.net/ Further Reading: Heneghan C. and Badench D.:(2006) Evidence-Based Medicine Tlkit, 2nd Editin Straus S.E. et al. (2005): Evidence-Based Medicine: Hw t Practice and Teach EBM. 3 rd Editin Evidence based medicine: what it is and what it isn't. D Sackett et al. BMJ 1996;312:71-72 Rse G, Barker DJP. Epidemilgy fr the uninitiated. 3rd ed. Lndn: BMJ Bks, 1993. Hw t read a paper: getting yur bearings (deciding what the paper is abut). T Greenhalgh. BMJ 1997;315:243-246