The purpose of this review is to update surgeons on finding and applying best evidence in clinical surgery. Large amounts of evidence exist to support clinical decisions, yet it can be difficult to find and apply the best answers to clinical questions. We often start with pubmed, although in 2010 there are better ways of going about searching. A pubmed search tends to produce a feast or a famine of results, with no straightforward way of sorting useful from distracting papers. Articles disagree with one another, yet not everyone can be correct. Advances in surgical research take a long time to influence clinical practice.
Newer tools such as clinical practice guidelines and evidence based reviews are meant to provide clinicians with quick access to high quality, clinically relevant research. Search strategies specifically designed to answer clinical questions can be used when credible practice guidelines or evidence based reviews do not exist. Understanding and applying the tools of evidence based medicine makes it easier to find the best information to support clinical decisions.
Surgeons need to make clinical decisions all the time, and if you count the decisions you make during a single consultation, you will find that there are many of them - what is the diagnosis? does the diagnosis need confirmatory tests? what is the prognosis? does the prognosis apply to this patient? The most important decision we make is how to treat the patient - whether to operate or not, which operation to do, how to do it, when, what else to do in the after care. In order to make the best decisions, we want to use the best information - but what is the best information? Usually, the best information is an unbiased summary of real outcomes experienced by real patients as a result of the decision.
The Pyramid of clinically useful information |
|
The Evidence Based Medicine movement contends that the best way to make decisions about patients is to explicitly and conscientiously apply the results of well designed clinical studies - that is, properly designed and controlled studies looking at the outcomes of real clinical decisions on real clinical patients. Evidence Based Medicine is a widespread movement that had its origins in the 1980’s at McMaster University in Ontario, Canada under David Sackett and his colleagues (1).
Training in clinical epidemiology and applying its tools to different fields in medicine and surgery has become a very common career path for doctors in North America and Europe. There are many unanswered questions, and clinical science is advancing all the time. Many questions or decisions we make in daily practice have newer and better answers than we may be aware of or than we may apply.
The evidence based medicine movement began with considerations of biases in clinical studies, showed us how to critically appraise individual studies, and developed ways of combining studies using systematic reviews, meta-analyses, decision support tools, and practice guidelines. This review will go backwards through time - starting with practice guidelines and decision support tools, then looking at systematic reviews and meta analyses, then looking at critical appraisal of individual studies, sources of bias, and considerations in research design. Searching for best evidence should be efficient, and we should use the work of others (practice guidelines and decision support tools) if it is applicable and of high quality.
A practice guideline is a document which gives recommendations about clinical practice and summarizes the evidence behind those recommendations. An excellent example of clinical practice guidelines in surgery are the guidelines produced by the American Academy of Orthopaedic Surgeons. These can be found online at Clinical Practice Guidelines and are worth looking at. Treatments of different types of fractures, prevention of thromboembolic disease, and other common problems are addressed via a very rigorous, structured review of the literature performed by a group of experts. The guidelines give a recommendation, a grade indicating whether the recommendation is based on strong or on weak clinical evidence, and a referenced summary of the clinical studies which were used to support the recommendation. Excluded studies are also referenced, and the reasons for exclusion listed. Advantages of these guidelines are that they are produced by and for surgeons through a professional association, and that they are produced using a pre-defined, transparent process to maximize objectivity and minimize bias.
Disadvantages of these guidelines are that there is a limited number of them (specify) and therefore not every question relevant to clinical practice is addressed. If a guideline is available, it is often the best starting point in answering a clinical question. Guidelines are likely to be better if they are made by and for surgeons, and with a transparent, objective, and well documented process for identifying and retrieving information in the literature.
These services are commercially produced clinical reviews which summarize clinical evidence in a manner focused on decisions made in clinical practice. They are comprehensive, in the manner of a textbook, and organized in a similar way. Internet based and CD/DVD based products are available (Uptodate and Dynamed) {links for Ptolemy users}. They are expensive to subscribe to, but are available free via the University of Toronto library for Ptolemy users (www.ptolemy.ca). In a manner similar to a practice guideline, they give recommendations about clinical decisions and support these recommendations by referencing high quality, carefully evaluated clinical studies. The editorial board and authors are responsible for the standards of the review and their incentive is to keep standards high in order to keep selling subscriptions. UptoDate, in particular, has a very large and dedicated following among physicians. Surgical questions are addressed in these reviews also, although they are not as widely read by surgeons.
UptoDate and Dynamed vary considerably in the depth and sophistication of treatment of surgical topics. Ptolemy users should log on and use the search box to explore some of their favourite topics. A search on partial mastectomy leads to a very complete writeup of the indications, outcomes, surgical techniques and anatomical considerations related to this strategy. In addition the topics of axillary node dissection and chemotherapy, and all other related aspects of care, come up in search bars. The writeup is, as advertised, updated within the past few months and the type and quality of the underlying literature is evaluated. A similar extensive and well referenced writeup is found when looking for surgical deep venous thrombosis, for example. On the other hand, looking for the specifics of care around paediatric femoral fractures, or around avascular necrosis of the hip, yields an organized outline similar to that found in most textbooks – not bad, but nothing special. The reason for this is that high quality clinical evidence does not exist for these latter topics – very few randomized trials or prospective studies.
Comparing the American Academy of Orthopaedic Surgeons practice guidelines with UptoDate, UptoDate gives a lot more structure and context about epidemiology, mechanism of injury, anatomy, clinical assessment, and radiographic assessment but glossed fairly lightly over the treatment. The AAOS guideline gives a series of 12 to 15 very precise statements about treatment recommendations, and in the full document summarizes all of the actual patient outcomes that support these recommendations. The AAOS guidelines are better evidence, UptoDate is easier reading, and as time passes and the literature gets better the products like UptoDate and Dynamed will be more and more widely used.
Disadvantages of these reviews are that they can only be as good as the underlying literature that they are based on. Surgeons have lagged behind physicians in producing high quality, relevant clinical studies but this is changing. Most major surgical journals are increasingly publishing randomized trials and high quality clinical evidence, which is displacing case reports and case series and other low quality clinical evidence. In 2010, most clinical decisions made in surgery are not supported by high quality clinical evidence. In addition, the scope of surgical practice in Africa means that some of these guidelines and reviews may not be applicable – while others may be entirely applicable.
The ‘evidence based orthopaedics’ section in the journal of bone and joint surgery behaves like a ‘journal club’ providing summaries and commentaries of high quality studies (generally randomized trials) which appear in other journals but are relevant to practicing orthopaedic surgeons (2). It allows practitioners to keep up with developments in their field, by selecting from the small portion of the literature most likely to influence clinical practice. Physicians have an entire journal (American College of Physicians Journal Club) which provides a similar service. In general surgery, the British Journal of Surgery has a section called ‘Scientific Surgery’ which links to every randomized trial in surgery published in over 50 journals since 1998. The British Journal of Surgery has been strongly influenced by the move towards randomized trials, with sometimes half the published research being randomized trials! It is impossible to argue that randomized trials do not apply in surgery in 2010, and in fact there are still trials of placebo surgery being published in high quality journals. For example a trial compared arthroscopic debridement of the osteoarthritic knee with sham surgery and showed no difference, it was published in the New England Journal of Medicine in 2002 (3) along with an editorial discussing the ethics of placebo surgery, which the journal obviously agreed with in this case (4).
A systematic review means that all of the literature relevant to a topic has been identified, screened for inclusion, abstracted, and analyzed in the review. A systematic review should include, in the methods section, a detailed description of the search strategy and the inclusion and exclusion criteria for studies as well as details of who did what to go from the source articles to the final review. A systematic review is the building block on which a recommendation in a practice guideline or a decision support tool should be based. If a systematic review is done well, it should meet the standards of a scientific experiment, in that it should produce the same results and conclusions no matter who does it. This means that objective, usually quantifiable clinical information is going to be sought. It also means that the highest quality or level of evidence (see below) will determine the recommendation made.
The Cochrane Collaboration has set out to standardize methods for performing systematic reviews, and to collect the results of such reviews in the online, updated Cochrane Library. The collaboration is academically based and is not commercially funded. Cochrane reviews are often published in summary form in a commercial or open access journal, as well as in complete form in the online Cochrane library. If there is a Cochrane review of a particular topic, that is often an excellent single source for obtaining an objective and systematic summary of the evidence supporting a clinical decision. The Cochrane library is available online, as a full-text resources, for users of the Ptolemy library (www.ptolemy.ca).
A meta-analysis is an optional component of a systematic review in which data from multiple small studies is quantitatively combined. Sometimes a systematic review is done solely for the purpose of performing a metasanalysis, ans sometimes a systematic review stands alone without a meta analysis, so although the terms are related they are not interchangeable. The best use of a meta-analysis is to pool the results of randomized trials to come up with an overall conclusion. Meta-analyses can statistically ‘prove’ things that the underlying trials could not, by creating an effective larger sample size. Meta-analyses can also provide the best estimate of the effect of a treatment, by combining trials which have different results.
An early problem identified with meta-analysis is that if the results of a trial are ‘negative’ - ie the new treatment is no better than, or perhaps worse than, the treatment to which it is being compared, then the trial is less likely to be completed, less likely to be submitted, and less likely to be published if submitted. The best example of this affecting clinical medicine is that the initial trials and metaanalyses evaluating the effectiveness of selective serotonin reuptake inhibitors in treating childhood depression were strongly biased (5). Had all the trials performed been published, a balanced picture would have emerged suggesting the drugs have smaller, or even no, benefits. Some of the trial data were owned by the companies which make the drugs, and the negative results were deliberately suppressed. This is why, at present, randomized trials are required to be registered, so that meta-analyses can identify unpublished, ‘negative’ trials and avoid publication bias.
Many Cochrane reviews and other systematic reviews contain meta-analyses. In general, the presence of a meta-analysis means that the information supporting the review is high quality and objective. A common means of displaying a meta-analysis is to use a summary plot, in which the point estimates of effect size and the confidence intervals around these estimates are shown for each trial and the overall estimate of effect, with a narrower confidence interval because of higher numbers of subjects, is shown as a diamond at the bottom of the graph.
A narrative review includes any review which does not meet the methodological standards of a systematic review or a meta-analysis. All the reviews in Surgery in Africa are narrative reviews. Are they useful? They are ‘evidence based’ to varying degrees. The sophisticated reader will seek out the quality of evidence referenced in the review, and the sophisticated writer will make the nature, level, and quality of evidence readily apparent.
Narrative reviews have a long history in clinical medicine and there are entire clinical journals in every field devoted to review articles (eg. Surgical Clinics of North America, Orthopaedic Clinics of North America to name but two of dozens.) Narrative reviews are popular with readers because they help to give a logical framework for thinking about a particular subject. A good narrative review can be current and correct about a lot of clinical questions. However, whenever the literature is treated selectively there is a strong tendency for intended or unintended biases to come out in the conclusions, usually well referenced and ‘supported’. Readers of narrative reviews need to be more aware of the quality of the underlying papers, and more questioning of the conclusions and biases of the author. We hope that Surgery in Africa reviews are useful because they view the world literature through the eyes of surgeons experienced with surgery in resource constrained settings - in a way we are trying to introduce a particular, useful, bias to the review.
A well prepared practice guideline, systematic review, or meta-analysis would generally supersede the conclusions of a narrative review if they differed.
A set of narrative reviews of particular relevance is the themed issue on evidence based surgery, in Surgical Clinics of North America (6).
This phrase usually describes case report with just enough reference to establish how a single, very specific case relates to what is previously published. It is not really meant to be a review that answers clinical questions in a systematic or complete way. Ignore it if you are looking for answers. Read it only for questions, puzzles, or interest. Case reports are more likely to describe something unique about a particular patient than to describe typical results of treatment – so they do not, in general, support clinical decision making about the common things that we treat. When we do come across unique or interesting cases in our own practice, we may find it helpful to read and reflect upon the thoughts or experiences of others.
Pubmed is a common starting point for finding answers to clinical questions.
Suppose we have decided to repair an umbilical hernia (see August review to decide whether this is a good idea or not!) and we would like to know whether a repair using mesh will reduce the recurrence rate compared with a repair using sutures alone. We are going to use a pubmed search. Open another browser window or tab on your computer as you read this, and follow along. If you have a Ptolemy account, log in first so you can get full text. First, we try a ‘usual’ pubmed search. I typed ‘mesh in hernia repair’ into pubmed and am shown the first 20 of 3315 results - a mishmash of case reports, clinical studies, and other types of articles arranged in reverse chronological order. A feast of information - but where is the good stuff? Reverse chronological order hardly puts the most relevant information first – and 3315 results are too many to browse through and evaluate if there is a simpler way. Now, try a clinical query. On the PubMed home page, under PubMed Tools, there is a link for Clinical Queries. Click it and another search page opens up. Again type in ‘mesh in hernia repair’. There are now three columns of results. The left column lists the first five of 1282 clinical trials about the question - the first five happen to be randomized trials, but the list includes other clinical studies. The middle column lists 5 of 82 systematic reviews. The first one (when I accessed 2010 August 1st) was an article entitled “Does mesh offer an advantage over tissue in the open repair of umbilical hernias? A systematic review and meta - analysis” this paper counts as ‘the good stuff’ for two reasons. First, it directly addresses the comparison I am interested in and the decision I am trying to make. Second, it is fairly high up on the ‘pyramid’ of clinically useful evidence. It is a systematic review with a quantitative meta-analysis, so it will reference and summarize all of the randomized clinical trials addressing the question. If this paper is well written, then I will have a good answer to this question after reading it alone.
The most important step in improving what one finds in pubmed is to get the question right. Writing out a full question before searching, and thinking about different ways of asking the question, should yield a list of words relating to important concepts in the search. The pubmed search box is designed to accept multiple terms at once. The software will map each word to the keywords used by the US National Library of Medicine to index articles. These keywords are a systematic and carefully designed vocabulary also referred to as MeSH headings, which stands for medical subject headings. You can look up MeSH headings, or see how pubmed has translated your terms into MeSH headings, but in general this is not necessary today because the algorithms for translating terms used by clinicians into MeSH headings used by librarians are very good.
There are two ways of thinking about what happens next. One way of thinking about it is that pubmed will find the articles which contain all or most of the keywords in the search. The other way of thinking about it is that pubmed will do a separate search for each keyword, and then combine the sets using ‘AND’ (see diagram). I prefer to think of it in the latter fashion, because it helps me know what to do if I have too many results or too few results.
If I have between twenty and two hundred results I will simply scan down them to make sure that most of the article titles are relevant to what I want to find out. If they are, then I usually scan the titles, open the abstracts of the promising titles in separate tabs, and open the full text links of the abstracts I like to download the articles (refer to Ptolemy user guide). If most of the results are not relevant, I have done a bad search. I rewrite my question and rearrange it using different terms.
If I have several thousand results I worry. Looking at the titles will help me to decide whether the results are all relevant. If they are, then I need to either narrow the search or look the topic up in a textbook. Narrowing the search is a simple matter of working out which concept I have missed in my search, or being more specific about what types of papers I want to see. Adding a word for an extra concept in the search box will limit and focus my search.
For example, suppose I am performing a total knee replacement tomorrow and I want the latest results on heparin prophylaxis. I can type ‘heparin prophylaxis’ and see 12035 results. Too many to read through. Add the term ‘knee’ and we are down to 510 results. Still too many – but better and more relevant. Now add the term ‘randomized’. Now I am down to 150 results. I can scan that. About 20 down (now) is a paper from the Cochrane database of systematic reviews entitled “Direct thrombin inhibitors versus vitamin K antagonists or low molecular weight heparins for prevention of venous thromboembolism following total hip or knee replacement.” It reports on 14 randomized trials involving over 20,000 patients and shows that low molecular weight heparin is equally effective at preventing DVTs, and the newer drugs (direct thrombin inhibitors) have a higher risk of bleeding and a higher all cause mortality. I am up to date on the high quality evidence and can comfortably prescribe low molecular weight heparin for this indication.
Just above the pubmed search box is a box for limits. Using the same example above – I have heparin prophylaxis knee randomized in the search box – I click limits and see a set of checkboxes. These allow me to restrict the search according to several criteria – date, language, gender, patient age, etc. Try the limits under type of article. Click on meta-analysis and press search – the fifteen metaanalyses are all listed. Go back into limits and unclick metaaanlysis, reclick randomized trial. Now I have eighty six actual randomized trials. Using the limits in this way is very useful if I have lots of articles and want the highest quality. The other limits to try are practice guideline, and clinical trial. A practice guideline search does not work well for this example, it brings up an article about trying to implement a guideline rather than the guidelines themselves. A clinical trial search will be a little broader than a randomized trial search – it will find all prospective studies with a control group.
Finding information
Pubmed searching
Clinical Decision Support Tools (eg. UptoDate, Dynamed)
- designed to find individual studies
- useful for rare conditions or very specific questions
- less useful for day to day decisions in practice – too much information
- clinical queries will organize pubmed results around practice decisions
- designed to organize the literature around clinically relevant questions
- useful for reviewing best evidence about common conditions
- will refer to practice guidelines, systematic reviews, and individual papers
- should comment on both the level of evidence and the quality of the study
Levels of evidence give a broad way of classifying clinical studies in a hierarchy. (see text box).
Level one studies are randomized trials and they give the highest level of clinical evidence. This is because they are true experiments. If the treatment is chosen by randomization, then the patients who received the treatment and those who received the comparator cannot differ on any known, or unknown, factor influencing the outcome. Differences in outcome are the result of the treatment randomly assigned, and this is the ‘gold standard’ level of evidence. Just because a trial is described as a randomized trial does not mean it is one (how were the patients randomized? Did the randomization work ie. Are the patients in each group similar?). Randomized trials can also produce biased results, even if the randomization worked. Two common causes of bias are loss to followup and lack of blinding. If all the patients who did poorly received the same treatment, but they never come back to be evaluated, the results will be biased. Any trial with more than 20 % loss of followup should be discounted. Blinding means the people evaluating the results should not know which treatment was applied, it can also mean the patients do not know (double blind). If the trial is not, or cannot, be blinded then the prejudice of the patients or evaluators about which treatment is better can show up in the results. This is not such a problem with objective endpoints, like survival, but can be a considerable difficulty if the outcome is more subjective, like quality of life. A randomized trial with near complete followup, blinded evaluators, and clinically important outcomes is an important unit of clinical evidence.
If more than one randomized trial is available, look for a meta-analysis or look for consistency of trial results. Then, in general, look no further.
Level two studies are prospective studies with a comparison group. Some patients received the treatment, and some received the alternative (a different treatment, or no treatment). The more similar the two groups of patients are, the more likely it is that differences in outcomes is related to the treatment. Look carefully for differences – a surgical trial in which the patients selected for surgery were all younger and healthier, and those selected to wait were older and sicker – is obviously biased. We expect younger and healthier people to do better. A prospective study with a control group is still better than a retrospective study – because the investigator is selecting the patient according to the treatment, rather than according to the results. If the follow-up is not complete, again that can lead to a biased conclusion.
If only level two prospective studies are available, they are useful in making clinical decisions. Some types of questions in surgery are much more difficult to test in a randomized trial, and some people claim there are surgical questions that can never be answered in a randomized trial.
Level three studies have a control group but the study is performed after the fact – when the outcome is known. There is still a comparison made explicitly in the study between patients who received the treatment in question and those who received an alternative.
Level four studies are case series. They report on outcomes of patients, but all of the patients in the study had the outcome in question. There is no explicit comparison, and that makes case series difficult to use in clinical medicine. Everyone’s patient population is different, and everyone’s selection of patients is different. Implicitly, the author of a case series wants you to compare his results with what you believe would happen to similar patients. For certain questions a case series might be sufficient to change practice – the so-called ‘penicillins or parachutes’ interventions where a near universally fatal outcome is made survivable. Interventions with less than total success, for conditions with less than total fatality, are much more difficult to evaluate by studying case series.
One of the reasons surgeons like case series is because they are often accompanied by very detailed technique sections – line drawings, descriptive paragraphs, even videos showing how to do the surgery. It is important to separate the usefulness of a technical description, from the scientific validity of the results. Surgery is difficult because we have to pick the right operation, and we have to do it the right way. Lower quality evidence articles may be useful in getting the technical details right, if they are well prepared.
Levels of Evidence (Individual Studies)
Level one evidence – randomized controlled trial
Level two evidence – prospective study with control group
Level three evidence – retrospective study with control or comparison group
Level four evidence – case series without control group
Level five evidence – expert opinion without
Cannot assign a level of evidence to: Basic science studies, animal studies, case reports, editorials, review articles
After categorizing an article by broad level of evidence (and rejecting it quickly if a higher level of evidence is available), the first thing I look for is what the conclusion of the study is. I usually do this by looking back and forth between the question and the answer in the abstract. If the question is of interest to me, and the answer might be useful in my practice, then it is worth reading further. If not, on to the next paper.
If the conclusion is relevant to practice, we need to look at the results and then the methods. If the results do not relate to or support the conclusion, stop. A surprisingly large number of patients is still published where the conclusions represent the opinion of the author, reviewer, or editor but are not actually reflected in the results. IF the results do support the conclusion, then look carefully at the results themselves. Were the most important outcomes measured? Were the outcomes measured at the right time? Were the differences in outcomes large enough to be clinically important? Were the differences statistically significant? A clinically important difference in an outcome that is well measured is potentially important. If the study is of high quality, then you may decide to apply the results.
Quality of a study is a different concept from level of evidence. A randomized trial can be of higher or lower quality, and evaluation of the quality of randomized trials is done with the CONSORT statement (7) (see attached checklist for details). Many journals require structured reporting of trials according to the checklist, and application of the checklist has resulted in improved quality of reporting of randomized trials (8). Quality tools can be applied to nonrandomized studies, many of the aspects in the CONSORT statement are not about randomization but do make a study more valid.
Quality of Studies – simple considerations
- At least ten patients per group (minimum criterion)
- Inclusion and exclusion criteria well defined
- All eligible patients accounted for, followup on 80% or more of patients
- Clinically important outcomes measured at an appropriate time frame
- Compliance with treatment, cointerventions recieved, and possible confounding factors discussed
- Method of randomization appropriate.
- Analysis according to randomization
- Method of blinding subjects and assessors of outcomes described
- Evaluation of blinding
The final step in critical appraisal of an individual study is to look closely at the patients who are being reported on. Consider the inclusion criteria, the exclusion criteria, and the patient characteristics (usually summarized in ‘table 1’ of the trial). If the patient we are deciding about fits, we can use the results of the study. If the patient we are considering does not fit the trial, we must make a judgement about how much the results of the study can be generalized. We could also go back to searching, and use more limits.
Bias is a systematic deviation of the results of an investigation from the truth. I have driven cars with considerable bias in the steering, and it is possible to correct for bias if you know which way to adjust. The problem with bias in clinical studies is that it is unpredictable. It can be difficult to prevent bias in studies which are not randomized, controlled, and blinded. When randomized trial results are compared with the results of previous, nonrandomized studies of the same clinical question the most common finding is that the result of the randomized trial is similar in direction, but smaller in magnitude, than the result of a nonrandomized study. So nonrandomized studies overstate treatment effects compared with the truth. Unfortunately, we cannot simply ‘correct’ for this bias because there are plenty of examples where nonrandomized studies have shown the opposite effect from a randomized trial, or have shown large effects where subsequent randomized studies have shown none at all. For example, hormone replacement therapy after menopause was believed to protect against cardiovascular disease until a large randomized trial established the opposite (9). Beta Carotene and Alpha tocopherol were also believed to be protective against coronary disease based on epidemiological evidence, but their effects were neutral or actually harmful when studied in randomized trials (10,11). Dietary fibre was similarly believed to have a protective effect against colon cancer, but supplementation with fibre did not prevent cancer when studied prospectively in a randomized trial (12). Bias arises at many steps in the evidence process, from design of the study through publication – the tendency to publish positive results and suppress neutral or negative information. Although randomized studies are the gold standard of evidence, and there are widely quoted examples where nonrandomized studies have been wrong, there is also a reassuring agreement between randomized and nonrandomized study designs in many areas of clinical medicine (13).
Critical appraisal, done properly, is a lot of work. It is a good exercise in critical thinking for trainees, and it is useful to run a journal club where the best articles about topics of interest can be carefully reviewed. It is, however, too much work to sit down and critically appraise multiple original research papers every time we have a clinical decision to make. A balance needs to be struck between doing your own thinking, versus having ready access to useful information. This is why the author favours practice guidelines produced by professional societies. A lot of work goes into them. Done properly, these guidelines represent conscientious efforts of our own colleagues to sort and filter the information which is actually of value for our patients. It can be disappointing to find that we are not as certain about some things as we think we should be. Most surgical teachers can be dogmatic about things which are not founded in fact. Moving from bias to science is a slow process and can be done explicitly. Remember that unproven does not mean untrue.
In an ideal world, we would get a single answer – the right answer – to any question about clinical medicine. In the surgical literature, which grows day by day, it seems we can find any answer we look for. In 2010 we do not want to start with a pubmed search for individual studies. We want to look for practice guidelines, systematic reviews, meta-analyses and clinical decision support tools such as dynamed and uptodate. We want to understand the strengths and limitations of each of these sources of information, and apply them conscientiously to our surgical practices.
Credit to: Altman DG, Schulz KF, Moher D, Egger M, Davidoff F, Elbourne D, et al. The revised CONSORT statement for reporting randomized trials: explanation and elaboration. Ann.Intern.Med. 2001 Apr 17;134(8):663-694.
(1) Sackett DL. Evidence based medicine: what it is and what it isn't. BMJ.British medical journal 1996;312(7023):71. http://simplelink.library.utoronto.ca/url.cfm/121137
(2) Wright JG, Swiontkowski MF. Introducing a New Journal Section: Evidence-Based Orthopaedics. J Bone Joint Surg Am 2000;82:759. http://simplelink.library.utoronto.ca/url.cfm/121136
(3) Moseley JB. A controlled trial of arthroscopic surgery for osteoarthritis of the knee. N.Engl.J.Med. 2002;347(2):81. http://simplelink.library.utoronto.ca.myaccess.library.utoronto.ca/url.cfm/121124
(4) Horng S. Is placebo surgery unethical? N.Engl.J.Med. 2002;347(2):137. http://simplelink.library.utoronto.ca/url.cfm/121125
(5) Whittington CJ. Selective serotonin reuptake inhibitors in childhood depression: systematic review of published versus unpublished data. Lancet 2004;363(9418):1341. http://simplelink.library.utoronto.ca/url.cfm/121126
(6) Meakins J, Gray M. Evidence-Based Surgery. Surgical Clinics of North America 2006;86(1):1-226. http://simplelink.library.utoronto.ca/url.cfm/121127
(7) Altman DG, Schulz KF, Moher D, Egger M, Davidoff F, Elbourne D, et al. The revised CONSORT statement for reporting randomized trials: explanation and elaboration. Ann.Intern.Med. 2001 Apr 17;134(8):663-694. http://simplelink.library.utoronto.ca/url.cfm/121128
(8) Moher D. Use of the CONSORT statement and quality of reports of randomized trials: a comparative before-and-after evaluation. JAMA (Chicago, Ill.) 2001;285(15):1992. http://simplelink.library.utoronto.ca/url.cfm/121129
(9) Hulley S, Grady D, Bush T, Furberg C, Herrington D, Riggs B, et al. Randomized trial of estrogen plus progestin for secondary prevention of coronary heart disease in postmenopausal women. Heart and Estrogen/progestin Replacement Study (HERS) Research Group. JAMA 1998 Aug 19;280(7):605-613. http://simplelink.library.utoronto.ca/url.cfm/121130
(10) Hennekens CH, Buring JE, Manson JE, Stampfer M, Rosner B, Cook NR, et al. Lack of effect of long-term supplementation with beta carotene on the incidence of malignant neoplasms and cardiovascular disease. N.Engl.J.Med. 1996 May 2;334(18):1145-1149. http://simplelink.library.utoronto.ca/url.cfm/121131
(11) Yusuf S, Dagenais G, Pogue J, Bosch J, Sleight P. Vitamin E supplementation and cardiovascular events in high-risk patients. The Heart Outcomes Prevention Evaluation Study Investigators. N.Engl.J.Med. 2000;342(3):154-160. http://simplelink.library.utoronto.ca/url.cfm/121132
(12) Schatzkin A, Lanza E, Corle D, Lance P, Iber F, Caan B, et al. Lack of effect of a low-fat, high-fiber diet on the recurrence of colorectal adenomas. Polyp Prevention Trial Study Group. N.Engl.J.Med. 2000 Apr 20;342(16):1149-1155. http://simplelink.library.utoronto.ca/url.cfm/121133
(13) Concato J. Randomized, controlled trials, observational studies, and the hierarchy of research designs. N.Engl.J.Med. 2000;342(25):1887. http://simplelink.library.utoronto.ca/url.cfm/121135
