Добавил:
kiopkiopkiop18@yandex.ru t.me/Prokururor I Вовсе не секретарь, но почту проверяю Опубликованный материал нарушает ваши авторские права? Сообщите нам.
Вуз: Предмет: Файл:
Ординатура / Хирургия / Библиотека им академика М.И. Перельмана / Книга_5537_Библиотеки_им_академика_М_И_Перельмана.pdf
Скачиваний:
0
Добавлен:
31.08.2026
Размер:
22 Мб
Скачать
This page intentionally left blankThis page intentionally left blank
Evidence-based Medicine Primer
Sarah Kharkhanechi*, Amir Abadir†, Abhinav Sharma‡,
Andrew Burke
, George Farag‡, Thanu Nadarajah‡and
Anne Holbrook
‡,§

Introduction

Clinicians who specialize in hospital medicine face numerous information challenges in the course of every working day; there are likely to be at least two important but unanswered clinical questions per patient seen per day. Most clinicians are experienced information gatherers, but many find it challenging to sort high quality “wheat” from the lower quality “chaff” of literature found. We have two choices — either we must limit ourselves to information sources already critically appraised, of which at least one outstanding example exists (MacPlus, available via http://plus.mcmas­ter.ca/macplusfs), or learn how to critically appraise relevant literature ourselves. Fortunately, a systematic and well-tested approach to the latter is available. This chapter describes such an approach to four key types of questions that are routinely encountered: Diagnosis, therapy, prognosis and economics/resource utilization. Each is presented using an example relevant to hospital medicine practice.
15
2
Chapter
*Department of Radiology, McMaster University;
Department of Medicine, University of Toronto, Canada;
Department of Medicine, McMaster University;
§
Division of Clinical Pharmacology & Therapeutics, McMaster University, Canada

Diagnosis

Assessing the value of a diagnostic test is a crucial part of every clini­cian’s practice. It is important to consider the limits and applicability of these tests and understand their role in clarifying the diagnosis.
Clinical Scenario
Mrs. Green is a healthy 45-year-old woman who presents to your clinic with a swollen left calf. There are no significant risk factors for venous thromboembolism (VTE). Are there any tests that can aid you in ruling out a deep venous thrombosis (DVT)?
Diagnosis Study
A study evaluating the use of a latex agglutination quantitative D-Dimer assay in outpatient clinics of four hospitals retrospectively examined 595 unselected patients with a follow-up period of three months.
1
The sensi­tivities, specificities, predictive values, and likelihood ratios of the assay were calculated for all patients using objective testing and eventual clini­cal diagnosis as the gold standard.
Diagnosis Criteria
2–3
a) Availability: Is this test available in your institution? Do you have a
well-trained person to perform the test? Are you sufficiently knowl­edgeable or do you have an expert available to interpret the results? Can the test be completed and reported in the necessary timeframe? If your answer is NO to any of the above practical questions, then there is likely not much point in ordering the test no matter how valid it may be.
b) Validity: This term refers to the accuracy and the ability of the test to
rule in or rule out the diagnosis in question, and thus help you make your clinical decision.
c) Conclusiveness: How much do the results of this test help you in making
a diagnosis? Will you need another test to confirm the diagnosis? This depends on the answers to the questions presented below in Table 1.
16
S. Kharkhanechi et al.
17
Evidence-based Medicine Primer
Table 1. Criteria for Evaluating a Study on a Diagnostic Test
Was there truly uncertainty about the A diagnostic test is useful if it is able to diagnosis in the sample used in distinguish between disorders when they the study? have overlapping clinical presentations. For
example, pulmonary embolism is difficult to diagnose clinically due to variability in presentation and overlap of its symptoms with other diseases.
Was the test being evaluated compared Many diseases do not have a high quality to a gold (reference) standard? Was the diagnostic standard (gold standard) test. gold standard applied to ALL of However, to avoid bias, the new test and the patients? the diagnostic standard test should be
performed on everyone in the study.
Were the results of the gold standard Blinding of the test interpreters is crucial to interpreted without knowledge of the ensure no bias is introduced in the analysis new test result and vice versa? of the results.
What is the “sensitivity” (% patients A highly sensitive test helps to rule out the with condition who have positive condition (SNout — sensitive rules out). results — true positives) of the test? Sensitivity = true positive/(true positive +
false negative).
What is the “specificity” (% patients A highly specific test helps to rule in the who do not have the condition who condition (SPin — Specific rules in). have negative results — true negatives) Specificity = true negative/(true negative+ of the test? false positive).
What is the “likelihood ratio” (LR) The probability of getting this test result in associated with the test? those with the condition divided by those
without the condition. Positive likelihood ratio (LR+) = Sensitivity/(1Specificity), while negative likelihood ratio (LR-) = (1Sensitivity/Specificity). A test with a LR+ more than 10 or a LR- less than
0.1 will provide a considerable change from pre-test to post-test probability.
Have the results of the diagnostic Results of retrospective studies determining accuracy study been confirmed in a Sn, Sp, and LR should be confirmed with a prospective management study? prospective management study where the
results of the diagnostic test are incorporated into patient management decisions.
d) Applicability: Generally, diagnostic tests are only worth considering
if they will change your management decisions. In addition to good diagnostic properties, a test must also be safe to use and provide acceptable cost-effectiveness.
Discussion
The study mentioned above included patients with suspected VTE pre­senting to an outpatient hospital setting.
1
Patients were categorized accord­ing to their pretest probability of VTE as “low,” “intermediate” or “high.” All patients underwent further testing, for example with compression ultra­sound or V/Q scan, except for a portion of patients who had both a low pretest probability and a negative SimpliRED D-Dimer. Patients with no objective evidence of VTE were followed for three months. This differen­tial application of the “gold standard” is a minor methodologic flaw in the study, but is a well-validated clinical approach to VTE. The sensitivity and specificity of the latex agglutination quantitative D-dimer test using a cut­off point of 0.5 µg FEU/mL, were found to be 96 and 45%, respectively. The PPV and NPV of the test were 29 and 98%, with LR+ and LRat 1.4 and 0.09, respectively. These results indicate that for patients with low or moderate pre-test probability of VTE, a negative quantitative latex D­Dimer assay can rule out VTE due to its high sensitivity and good negative likelihood ratio. However, the test is not very helpful when it is positive. A subsequent prospective cohort management study confirmed that a nega­tive D-dimer result safely eliminated the need for further testing in patients with low or moderate pre-test probability of DVT.
4
Mrs. Green, meanwhile, has a low pre-test probability of having a DVT given her history, so a quantitative latex D-Dimer was conducted and its result was negative. She was reassured and sent home.

Prognosis

An estimate of prognosis is essential to guide treatment and inform patients about clinically relevant outcomes.
18
S. Kharkhanechi et al.
Clinical Scenario
Mr. Bing, 65 years of age, is brought to the ER with right-sided facial droop and mild hemiparesis. His symptoms resolve within 60 min of onset. After blood tests and a CT scan, you explain that he has likely suf­fered a transient ischemic attack (TIA). He asks you what his chances are of having a major debilitating stroke.
Prognosis Study
A recent trial looked at the ABCD2 score for determining risk of recurrent TIA, minor stroke or major stroke after presentation with a TIA.
5
This was a prospective sub-study of all consecutive individuals with identified TIA or stroke in the region of Oxfordshire, England. A consecutive sample of 500 patients diagnosed with a TIA were identified and followed-up in person at 30 days.
Prognosis Criteria
6
19
Evidence-based Medicine Primer
Table 2. Criteria for Evaluating a Study on Prognosis
a) Are the results valid (worth reading)?
Does the population in the study include a An inception cohort (everyone entered spectrum of patients with the disease, and at the same stage of their disease, were they entered in the study at a similar preferably early) has the potential to stage of disease? provide the best prognosis information.
Was the duration of follow-up sufficient to The study’s validity may be compromised capture development of the outcome? if there are a large number of patients lost Was follow-up complete? to follow-up relative to those who have
suffered an event or if the follow-up period is too short.
Were the outcomes of interest defined and There should be a clear definition of determined in an objective and unbiased outcomes of interest before study manner? commences. Detailed adjudication rules
may be required for subjective outcomes such as disability.
(Continued )
Discussion
The ABCD2 score (age, blood pressure, clinical findings, duration of symptoms, diabetes) is a tool developed to predict individual risk of stroke at seven days and thus meant to affect the triaging of patients with TIA symptoms.
5
The ABCD2 score in these patients presenting initially
with TIA, was highly predictive of the 7-day risk of major stroke —
12.8% in patients with ABCD2scores of 5 or more versus 1.3% in patients with a score 4 or less (P < 0.001).
Going back to Mr. Bing, he scores 5 on the ABCD2 scoring system (age ≥ 60 = 1 point, unilateral weakness = 2 and duration of symptoms 60 min = 2), giving a 7-day risk of stroke of 12.8%. This ABCD2 score would suggest that the patient should be counseled on the high risk of
20
S. Kharkhanechi et al.
Table 2. (Continued )
Were there adjustments for other Prognostic characteristics, including important prognostic factors? treatments provided, other than those
being studied, will need to be considered.
b) What are the results?
How large was the likelihood of the The quantitative results from studies of outcome event in a specified period of prognosis are the number of outcome time? events that occur over time.
How precise was the estimate of likelihood? The precision of this outcome can be
evaluated based on the confidence interval (CI). The smaller the CI, the greater the precision.
c) How can I apply the results to my patient care?
Were the study patients similar to Consider whether your patient would meet my patient? the inclusion and exclusion criteria of the
study.
Will the results lead directly to selecting Useful prognosis markers will help guide or avoiding therapy? treatment selection.
Are the results useful for reassuring or Giving patients an estimate of outcomes counseling patients? will allow them to make informed
decisions regarding treatment options.
early major stroke and should potentially be hospitalized to facilitate further diagnostic testing and to optimize preventive treatment quickly.

Therapy

As the number of therapeutic options increase, clinicians must be able to identify treatments that will minimize patient morbidity and mortality without undue harm of the treatment itself.
Clinical Scenario
Mrs. Geller is a 65-year-old woman with hypertension who presents to the ER with a non-ST elevation myocardial infarction (NSTEMI). Her tro­ponin T is 0.8 µg/L and her EKG shows inferior ST depression. She has already received 160 mg of aspirin, 25 mg of metoprolol, 5000 units of subcutaneous heparin and 40 mg of atorvastatin. The Emergency physi­cian asks you to decide about clopidogrel.
Therapy Trial
The CURE trial was a large, randomized, placebo-controlled trial that studied the effect of clopidogrel added to aspirin versus placebo plus aspirin, on the composite outcome of cardiovascular (CV) death, non-fatal myocardial infarction (MI) or stroke in patients with acute coronary syndromes without ST-segment elevation.
7
Therapy Criteria
8
21
Evidence-based Medicine Primer
Table 3A. Criteria for Evaluating a Study on Therapy
a) Are the results valid (worth reading)?
Were the patients randomized? Randomization is the key to minimizing
bias — it balances factors that predict outcome, both those known and unknown, between the groups.
(Continued )
22
S. Kharkhanechi et al.
Table 3A. (Continued )
Was randomization allocation concealed? Concealment ensures that those enrolling
patients in the study cannot influence the arm to which patients are allocated (and thus bias the results).
Were all participants blinded? Patients, caregivers, data collectors,
adjudicators, and data analysts should be blinded to group allocation throughout the study.
How complete was the follow-up to the Ideally, every patient’s study outcome end of the study? would be known. If many patients drop
out and are lost to follow-up, the study results may be unreliable even with imputation techniques.
Was intention to treat analysis utilized? Intention to treat analysis (participant
outcomes are assigned to the group that the participant was randomized, no matter whether they stayed in that group or not) is the most conservative method of analyzing a typical randomized trial (non-inferiority trials are different, but are beyond the discussion in this chapter)
Did all of the groups receive similar Similar co-intervention helps ensure that co-interventions? the treatment versus control comparison
is the only cause of any difference in outcome.
b) What are the results of the study?
How large was the treatment effect? The treatment effect is determined by a
number of variables, including the absolute risk, relative risk, relative risk reduction, number needed to treat, and number needed to harm. These are defined below in Table 3B.
How precise was the estimate of the The precision of the effect of a treatment treatment effect? is based on the confidence interval. A
95% confidence interval tells you how widely the true result might vary from the actual result found — the smaller the confidence interval, the greater the precision of the treatment.
(Continued )
23
Evidence-based Medicine Primer
Table 3A. (Continued )
c) Can I apply the results of this study to my patient?
Is there reasonable similarity between my Consider whether your patient would patient and the study patients? meet the inclusion and exclusion criteria
of the study. If not, consider whether there is a compelling reason why your patient would not be similar.
Were all patient-important outcomes Ensure that the investigators have chosen considered? outcomes that would be considered
important by patients. Also ensure that all important adverse effects of the intervention have been documented.
Is treating my patient worth the potential The NNT must be weighed against the harms and costs? adverse effects (see NNH below) as well
as the cost of treatment. (see Economics)
Table 3B. Definitions of Terms to Measure Treatment Effect
Treatment Effect Calculation
Control event rate (CER) Number of events in control (placebo, for
example) arm/Total patients in control arm
Experimental event rate (EER) Number of events treatment arm/Total patients
in treatment arm
Absolute risk reduction (ARR): The CER — EER decrease in risk of a experimental treatment in relation to a control treatment.
Relative risk (RR): Probability of the EER/CER. Relative risk reduction event occurring in the treated group (RRR) = 1-RR versus the non-treated group.
Number needed to treat (NNT): The 1/ARR. The same calculation can be number of people needed to be treated applied to harmful outcomes; this is in order to prevent one bad outcome. termed number needed to harm (NNH).
https://avxhm.se/blogs/hill0