THANK YOU FOR JOINING ISMPP U TODAY! The program will begin promptly at 11:00 am eastern

Transcription

1 THANK YOU FOR JOINING ISMPP U TODAY! The program will begin promptly at 11:00 am eastern

2 ISMPP WOULD LIKE TO THANK.. the following Corporate Platinum Sponsors for their ongoing support of the society

3 ISMPP ANNOUNCEMENTS Registration for the 8 th Annual Meeting, April 23-25, 2012, will open shortly. Watch your inbox Exhibit and Sponsorship opportunities for the 8 th Annual Meeting will be available before the end of the year. Remember to renew your 2012 membership. If not done by January 31 st, you will not receive the dial in information for the February ISMPP U Applications are now being accepted for the March 2012 seating of the CMPP exam. Application deadline: February 15, 2012 This session qualifies for 1 CMPP Recertification Credit hour

4 FACULTY Presenter Leslie Citrome, MD, MPH Adjunct Professor of Psychiatry & Behavioral Sciences, New York Medical College, Valhalla, New York, USA Moderator Tom Drake, MA, CMPP Senior Director, Business Development, The JB Ashtin Group

5 DISCLOSURE STATEMENT Leslie Citrome is the Section Editor for Psychiatry for the International Journal of Clinical Practice. In the last 12 months, Leslie Citrome, is a consultant for, has received honoraria from, or has conducted clinical research supported by the following: AstraZeneca, Alexza, Alkermes, Avanir, Bristol-Myers Squibb, Eli Lilly, Janssen, Lundbeck, Merck, Novartis, Noven, Otsuka, Pfizer, Shire, Sunovion and Valeant.

6 What editors look for in a clinically meaningful presentation of trial results in a manuscript Leslie Citrome, MD, MPH Adjunct Professor of Psychiatry & Behavioral Sciences, New York Medical College, Valhalla, New York, USA

7

8 Percent of Patients in Remission at 6 Weeks The difference in remission for a major depressive episode at 6 weeks for Drug A versus Drug B is highly statistically significant 35% 34% 33% 32% 31% P < % Drug A Drug B Citrome L. Acta Psychiatr Scand 2008;117(6):

9 Percent of Patients in Remission at 6 Weeks The difference in remission for a major depressive episode at 6 weeks for Drug A versus Drug B is highly statistically significant, but clinically irrelevant How irrelevant is this? Can we quantify this? 35% 34% 33% P < % 31% 30% Drug A Drug B Citrome L. Acta Psychiatr Scand 2008;117(6):

10 What Is Evidence-Based Medicine? Clinical Judgment EBM Relevant Scientific Evidence Patients Values and Preferences Sackett DL, et al. BMJ. 1996;312(7023): Citrome L, Ketter TA. Int J Clin Pract. 2009;63(3):

11 What Is Treatment Effectiveness? Efficacy Does Rx reduce Sx? Treatment Effectiveness Combines all measures Adherence/ Persistence Will Pt take Rx? Tolerability and Safety Does Rx cause SE? Lehman AF, et al. Am J Psychiatry. 2004;161(2 suppl):1-56. Swartz MS, et al. Schizophr Bull. 2003;29(1): Lieberman JA, et al. N Engl J Med. 2005;353(12):

12 Citrome L, Ketter TA. Int J Clin Pract. 2009;63(3): EBM: 5 Steps to Success

13 Concepts Related to Benefit/Risk: P Value The smaller the P value, the more convinced you are that something is going on that is not just random This does not state anything about the size or the importance of the nonrandom effect P value is not the same as effect size

14 Concepts Related To Benefit / Risk: Effect Size This gives an indication of how big the treatment effect is in terms of reduction in symptoms, or other outcome of interest The greater the effect size, the more convinced you are that the intervention will have a clinically important impact This does not state anything about statistical significance of the observed outcome in a clinical trial Effect size is not the same as P value

15

16 Concepts Related to Benefit/Risk: Effect Size - Number Needed to Treat NNT is one measure of effect size It is independent of P value and does not say anything about the likelihood of the difference between treatments being due to chance alone Helps you judge the clinical significance of a statistically significant result

17 Number Needed to Treat How many patients would you need to treat with drug A instead of drug B before you would encounter one additional responder, or one additional adverse outcome? The smaller the NNT, the larger the differences between the two drugs

18 Calculating NNT Is Easy What is the NNT for an outcome for drug A versus drug B? f A = frequency of outcome for drug A f B = frequency of outcome for drug B Attributable risk (AR) = f A f B NNT = 1/AR By convention, and to avoid exaggerating differences, we round up the NNT to the next higher whole number

19 Calculating NNT Is Easy What is the NNT for an outcome for drug A versus drug B? f A = frequency of outcome for drug A f B = frequency of outcome for drug B Attributable risk (AR) = f A f B NNT = 1/AR For example, drug A results in remission 50% of the time, but drug B By convention, results and in to remission avoid exaggerating 20% of the time. differences, we round up the NNT to the next higher whole number NNT = 1/[ ] = 1/0.30 = 3.33 Round up to 4

20 % Likelihood of Being Relapse-Free QUESTION What is the NNT? * Relapse in Schizophrenia: Medication versus No Medication Maintenance medication group Medication withdrawal group Months of Follow-Up 75% 23% A. 1 B. 2 C. 3 D..53 Adapted from DeQuardo JR et al. Journal of Psychiatry Research 1998;32:

21 % Likelihood of Being Relapse-Free QUESTION What is the NNT? * Relapse in Schizophrenia: Medication versus No Medication Maintenance medication group Medication withdrawal group Months of Follow-Up 75% 23% A. 1 B. 2 C. 3 D..53 NNT = 1/( )=1/.53=1.92, round up to 2 Adapted from DeQuardo JR et al. Journal of Psychiatry Research 1998;32:

22 Percent of Patients in Remission at 6 Weeks The difference in remission for a major depressive episode at 6 weeks for Drug A versus Drug B is highly statistically significant, but clinically irrelevant 35% 34% 33% 32% 31% 30% NNT = 100 P < Drug A Drug B NNT= 1/( )=1/0.01=100 Citrome L. Acta Psychiatr Scand 2008;117(6):

23 What Is A Clinically Important NNT? A small NNT of 2 would be a hugely important difference Single-digit NNTs are important enough to notice in day-to-day clinical practice A large NNT of 100 or more means that there is little difference between choosing drug A or drug B for the outcome measured Some NNTs may be clinically important, even if they are relatively large, for example when the outcome is death Some NNTs may be clinically irrelevant, even if they are relatively small, for example when the outcome is a mild dry mouth

24 Clinical Relevance is Key Scales need to measure symptoms that clinicians and patients care about Definitions of response and remission need to reflect what is considered clinically important If thresholds are set too low or too high, differences found between treatments can be exaggerated or absent Clinical decision-making is based on effectiveness in the real world rather than efficacy in the laboratory

25 NUMBER NEEDED TO TREAT 1000 An NNT of occurs when both interventions have the same rate for the outcome measured NNT values of this magnitude are irrelevant when comparing interventions except when evaluating the utility of immunizations or when examining lethal outcomes 100 Double and triple digit NNT values are usually not important when comparing routine efficacy measures, but may become important regarding adverse outcomes that have long-term consequences 10 Single digit NNT values are usually important enough to see differences in routine clinical practice A NNT of 9 is a small effect size; NNT of 8.96 equals Cohen s d of 0.2 A NNT of 4 is a medium effect size; NNT of 3.6 equals Cohen s d of 0.5 A NNT of 3 is a large effect size; NNT of 2.3 equals Cohen s d of 0.8 A NNT of 1 can only occur if one intervention has a rate of 100% for the outcome measured and the other intervention has a rate of 0% Citrome L. Acta Psychiatr Scand 2008;117(6):

26 Examples of NNT for Psychiatric Conditions Disorder Treatment Comparison Outcome Measure NNT Major depression Acute mania Antidepressant vs placebo Valproate or lithium vs placebo 50% Reduction in HAM-D 3 50% Reduction in SADS-M 5 Bipolar disorder Lithium vs placebo Relapse 3 Schizophrenia Antipsychotic vs placebo 40% Reduction in BPRS or much improved CGI scale Panic disorder SSRI vs placebo Panic free 3-6 Social phobia Paroxetine vs placebo Much improved CGI scale 3 Obsessive-compulsive disorder Bulimia nervosa SSRI vs placebo 35% Reduction in Y-BOCS 4-5 Antidepressants vs placebo 2-5 Remission 9 Pinson L et al. Psychiatric Services 2003;54:

27 Examples of NNT for Medical Conditions Condition Intervention Prevented Event NNT Diabetes 1 Insulin Neuropathy 15 Acute myocardial infarction (MI) 2 Prematurely born baby 3 Streptokinase and aspirin Prenatal corticoid Diastolic blood pressure Antihypertensive drugs for 5 years Diastolic blood pressure Antihypertensive drugs for 5 years Death in 5 weeks Respiratory distress syndrome or prematurity Death, stroke, or MI NNT also depends on individual baseline risk Death, stroke, or MI Centre for Evidence-Based Medicine. Available at: Accessed Dec 17, Second International Study of Infarct Survival Collaborative Group. Lancet. 1988;2(8607): Crowley PA. Am J Obstet Gynecol. 1995;173(1): A'Court C. BMJ. 2002;324(7350):1375.

28 What Is NNH? NNH is number needed to harm We would use NNH when referring to an outcome we are trying to avoid It is calculated the same way as NNT

29 NNT NNH

30 P Values vs NNT P VALUE Indicates statistical significance Independent of effect size NNT Indicates clinical significance Independent of P value Statistical and clinical significance can be expressed together by calculating the confidence interval.

31 Limitations Of Using NNT / NNH GIGO Results are only calculable for binary or dichotomous events that are either present or absent, and do not apply to continuous variables such as the value of a blood test or the amount of weight change in kilograms However, we can transform continuous variables that have clinically significant thresholds, such as weight gain > 7%, and then calculate the NNT for that outcome

32 There Are Other Measures of Effect Size Effect size Range of possible values (weakest, i.e. no difference, to strongest) Typical example of a small effect Typical example of a large effect Relative measures Relative risk 1 to 2 4 Odds ratio 1 to 2 4 Hazard ratio 1 to 2 4 Relative risk increase 1 to <100% 300% Absolute measures Attributable risk 0 to 100% <10% 33%-50% Number needed to treat to Cohen s d 0 to Area under the curve 0.50 to 1.00 or 0.50 to Success rate difference 0 to Citrome L. Acta Psychiatr Scand 2010;121: See also: Kraemer HC, Kupfer DJ. Biol Psychiatry 2005;59:

33 Percent of Patients in Remission at 6 Weeks Relative versus absolute differences: Is Drug A (30% remission) is 50% better than Drug B (20% remission)? * 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% NNT = 10 P< NNT= Drug A1/( )=1/0. 1=10 Drug B Citrome L Acta Psych Scand. 2008;117:

34 Contrasting Absolute and Relative Risk Prospective Results from the Women s Health Initiative

35 Contrasting Absolute and Relative Risk Prospective Results from the Women s Health Initiative In each story, the media highlighted the change in risk associated with aspirin -- noting prominently something to the effect that aspirin users had a "20 percent lower risk" compared with nonusers. The implied message in many of the stories was that women should consider taking aspirin to avoid breast cancer. Schwartz LM et al. The Washington Post Tuesday, May 10, 2005.

36 Contrasting Absolute and Relative Risk Prospective Results from the Women s Health Initiative Absolute risk The risk of developing breast cancer for postmenopausal women who do not take aspirin on a regular basis is 955/194, 884 person-years, or 0.49% Relative risk Taking an aspirin a day for at least 5 years reduces risk by 20% to 99/24,398 person-years, or 0.41%; this is a relative risk reduction of 20% The absolute risk reduction is only 0.08% versus a relative risk reduction of 20% Harris RE et al. Cancer Research 2003;63:

37 Contrasting Absolute and Relative Risk Prospective Results from the Women s Health Initiative Another way to present these results would be to say that a woman's chance of being free from breast cancer over the next five years was 98.4 percent if she used aspirin and 98 percent if she did not. Seeing the actual risks leaves a very different impression than a statement like aspirin lowers breast cancer risk by 20 percent. Schwartz LM et al. The Washington Post Tuesday, May 10, 2005.

38 Relatively Speaking, How Odd Can Hazards Be? Strength of the association may be high, but statistical significance may not be reached (if confidence intervals includes 1.0) HIGH OR, NS LOW OR, NS HIGH OR, p<.05 SIGNIFICANT RISK (OR 3.0) LOW OR p< POSSIBLY SIGNIFICANT RISK (OR 2.0 but < 3.0) IFFY RISK (OR<2.0) Citrome L. PsychEdUp 2005;1(10):2-3.

39 Summary P values tell us only about the likelihood that a result is not due to chance; P values are irrelevant when it comes to clinically meaningfulness, whether the P value is < 0.05 or < Effect size calculations based on clinically valid and meaningful outcomes are a necessary component of any appraisal of any clinical trial or test Relative and absolute effect size measures provide different perspectives Ratios can be misleading and exaggerate clinical differences NNT can appear to trivialize the risk of rare but potentially important adverse events

40 QUESTIONS... To ask a question, please type your query into the Q&A chat box at the bottom left of your screen. Every attempt will be made to answer all questions.

41 NEXT ISMPP U Date: January 11, 2012 Topic: Obtaining and Maintaining Your CMPP Credential Speakers: Angela Cairns, CMPP Chair, Certification Board and Jason McDonough, PhD, CMPP Chair, Recertification Committee

42 THANK YOU FOR ATTENDING! We hope you enjoyed today s presentation Please take a moment to fill out the survey sent to you after today s program so you can provide valuable feedback, as it will help us to develop future educational offerings