ࡱ> wyv` _+bjbj .rUG)))8)tf*?++:P+P+P+///c?e?e?e?e?e?e?$@h6Cn?0./00?P+P+?f4f4f40 P+P+c?f40c?f4f4V;@hc<P+ + 0Tv;)20'< =?0?3<0C83Cc<Cc<</"0f450I0///??4^///?0000$6 6 Date: 9/8/05Dev. Date:Duration: Topic: Evidence based practiceID:Title: Interpreting confidence IntervalsAuthor(s): Julia LaceyDeveloper(s): Learning objective: Interpret the meaning of confidence intervals around measures of treatment effects in clinical trialsKeywords: confidence intervals, significance, sample sizeSummary: This RLO defines the term confidence intervals and demonstrates how they can be used to determine the significance and range of possible sizes of a treatment effect.Presentation: Introduction In a clinical trial, there is always a possibility that the effect seen may have occurred by chance. Confidence intervals are a useful approach to assessing the role of chance. Consider the following scenario. A trial is being carried out to investigate the effectiveness of drug A in preventing heart attacks. 10 patients receive drug A, and 2 of these patients have heart attack within the next 12 months. 10 patients receive placebo and 3 of these patients have a heart attack within the next 12 months. Drug A has reduced the risk of a heart attack by one third, in other words, a relative risk reduction of 33%. Its looking good. On the basis on this trial, would you be happy to recommend this treatment for your patients? The role of chance If this trial was repeated in another group of 10 similar patients, would you expect to see exactly the same results again? Would it be surprising if, for example, 1 patient in each group had a heart attack? Would it be surprising if 2 patients in the placebo group and 3 patients in the treatment group had a heart attack? Try changing the values in the table and look at the effect on the relative risk reduction. (pause) Note how values for the relative risk reduction changes. The true effect on relative risk reduction is somewhere within this range of values and may be larger or smaller than the effect observed in the original trial. This range, within which the true value could plausibly lie given the size of the difference actually observed, is called the confidence interval. 95% Confidence Intervals Confidence intervals should be reported in a clinical trial, but see the resources at the end if you want to know how to calculate them yourself. The actual calculated confidence interval for the relative risk reduction in the trial described above is -4.4% - 70%. Confidence intervals are created at the 95% level this means that there is a 5% chance that the true value lies outside this range. Assessing statistical significance from a confidence interval Confidence intervals can be calculated for any effect measure, not just relative risk reduction, and can be used to determine if the size of the effect is significant. If the confidence interval includes the value reflecting no effect we cannot rule out the possibility that the intervention has no effect and the result is statistically non-significant. In the table opposite, drag and drop the values of no effect into the appropriate place.(Pause) Now let us go back to our trial where the relative risk reduction was 33% (display white box). It looked good, but was it statistically significant, or might it have occurred by chance? Click on yes or no to indicate your opinion. What about continuous data? Some trials dont measure effect in terms of did it happen or not, so called dichotomous data. Instead they measure the effect on a continuous measure such as blood pressure, or months of survival. Confidence Intervals can also be calculated for these measures, and used to assess significance of the result. Consider a trial comparing the effect of drug X and placebo on FEV1 in patients with chronic obstructive airways disease. At the end of the trial, the FEV1 of patients who received drug X has, on average, declined 50ml less than in those who received placebo. The 95% confidence interval is 5mL-100mL (display white box). Is this result statistically significant? Click on yes or no to indicate your opinion. Is this result clinically significant? Click on yes or no to indicate your opinion. Images etc: Tossing a coin or rolling a die. Images of 2 rows of hearts 8 beating &2 not and 7 beating & 3 not. Build up a table with this information on. (see table 1) Give a yes and no button to click on. Answer revealed when they click yes Answer revealed when they click No  Images of hearts as above. Present table one again, but let the student manipulate the figure in column 3 (drop down list, can select 1, 3 or 5 only for placebo, and 1,2, or 3 for drug A). Make the figures in column 4 change to the appropriate figures as in table 2). When student manipulates figures, numbers of hearts beating changes.  Drag and drop the figures to the appropriate place in table (see table 3).  Give a yes and no button to click on. Answer revealed when they click  Illustration of dichotomous vs continuous data (e.g. no of patients with FEV1 decline > 50mL after 1 year vs decline in FEV1 at 1 year. Image of patient using inhaler labelled placebo and other using inhaler labelled drug X. Graph of results (see figure 1)  Give a yes and no button to click on. Answer revealed when they click  Give a yes and no button to click on. Answer revealed when they click   The width of the confidence interval In our heart attack example, the true effect of the intervention lies within a wide range and the trial is not very informative. In the table on the right, look at what happens when our trial is repeated in a larger number of patients. The relative risk reduction has stayed the same, but try manipulating the figures as before and see what happens. (Pause) Display white box. You can see that not only has the trial now reached statistical significance, but the confidence intervals become narrower with a bigger trial. The narrower the confidence interval, the more sure you can be of the true size of the effect of the intervention. Show table 4 and let students manipulate figures in column 4 as before as shown in column 3 of table 5. Figures in column 5 of table 4 change as shown in column 4 of table 5!Assessment: Please answer the questions below to test your understanding of confidence intervals. Confidence intervals can be used to assess both clinical and statistical significance. TRUE/FALSE (True statistical significance is shown if the CI does not include the value of no effect. Clinical significance is shown if the smallest likely effect is still considered to be clinically worthwhile) In a trial comparing a new treatment for breast cancer with established treatment, the relative risk of recurrence within 5 years was 0.88 (95% CI 0.73 1.09). Has the new treatment shown a statistically significant benefit? YES/NO (answer NO as the range includes one - the value of no effect for relative risk, but no evidence of effect is not the same as evidence of no effect) In a trial comparing the effect of an osteoporosis treatment with placebo on prevention of hip fracture, the number needed to treat (NNT) was 38 (95% CI 23 560). Has the treatment shown a statistically significant benefit? YES/NO(YES as does not include the value of no effect) In the osteoporosis trial above, has the treatment shown a clinically significant effect? YES/NO (NO the confidence interval is very wide, meaning that as many as 1 in 23 patients, or as few as 1 in 560 patients may benefit. The trial was too small to be fully informative and should be repeated in a larger number of patients). Links (1): Straus S et al (2005) Evidence-Based Medicine:How to practice and teach EBM 3rd Ed Elsevier Appendix 1 Confidence IntervalsLinks (2): Montori et al (2004) Tips for leaners of EBM: Measures of precision (confidence intervals) Canadian Medical Association Journal 171(6):611-615. Available at www.cmaj.caGlossary:Related concepts:Authors notes:Devs notes:Time: Total:Total:     UCEL Reusable Learning Object Specification V2.0 4/11/02 Last update:  DATE \@ "dd/MM/yyyy" 02/08/2006 Page  PAGE 1 of  NUMPAGES 2 2003 Dawn Leeder for UCEL No, the trial is too small and the results cannot be trusted RRR = 33% (95% CI -4.4% - 70%) This would be reported in a paper as RRR = 33% (95% CI -4.4% - 70%) (no or correct)The relative risk reduction is not statistically significant as the confidence interval encompasses zero - the value of no effect. Reduction in FEV1 decline = 50mL (95% CI 5mL-100mL) 10pts RRR = 33% (95% CI-4.4% - 70%) 100pts RRR = 33% (95% CI 17%-50%) 1000pts RRR = 33% (95% CI 23%-43% (correct or no)The result is statistically significant as the confidence interval does not include zero - the value of no effect. Correct, you probably feel the trial is too small and the results cannot be trusted Whether the result is clinically significant depends on whether you consider the smallest possible effect to still be clinically worthwhile.  *BHO]p}  M X  ~   K ²yj^jRhrCJOJQJaJh [CJOJQJaJh[h.VCJOJQJaJhdH(CJOJQJaJh[h[CJOJQJaJh[hoN@CJOJQJaJh[htCJOJQJaJh[h.V5CJOJQJaJhbCJOJQJhb5CJOJQJhtCJOJQJhWbCJOJQJhbCJOJQJ ht5 hb5  $CG$IfU'(^+GHkd$$Ifl4r; u%2  ``20'4 laf4p2Hqr|  @$IfskdZ$$Ifl4 ;%' ` 0'4 laf4p $If nh$Ifkd$$Ifl40;*% `0'4 laf4p  N $Ifskd$$Ifl4;%' ` 0'4 laf4p N O $Ifskd$$Ifl4 ;%' ` 0'4 laf4p     T'(A $IfgdZ-$Ifpkd.$$Ifl ;%' ` 0'4 lap    ! . 4 > I Y ST;EZǷuuffufuffh[htCJOJQJaJh[hg7CJOJQJaJh6CJOJQJaJhuCJOJQJaJh[h1 CJOJQJaJhrCJOJQJaJh[hoN@5CJOJQJaJh[h1 5CJOJQJaJhdT5CJOJQJh[hoN@CJOJQJaJh[h.VCJOJQJaJ$Z[z&'(AV\ILMN微埓wgwXLX?LhZ-5CJOJQJaJh)CJOJQJaJhZ-hZ-CJOJQJaJh)hWb6CJOJQJaJhWbCJOJQJaJh'hZ-5CJOJQJaJhoN@CJOJQJaJh[hoN@CJOJQJaJhrh'5CJOJQJaJh'CJOJQJaJhrCJOJQJaJh[hg7CJOJQJaJh[htCJOJQJaJh;CJOJQJaJAM+,M$If $Ifgd& $Ifgdnw $IfgdZ- $Ifgd[N>$8'ؼǹČٌthݳ55Cϴ5䴳ϴ<CJOJQJaJh):CJOJQJaJh%CJOJQJaJhMfCJOJQJaJh\$CJOJQJaJh)CJOJQJaJhxAhxACJOJQJaJhMf5CJOJQJaJh'hZ-5CJOJQJaJhZ-CJOJQJaJhxACJOJQJaJ'(D`c Neg+öêÞtgWHhnYihbCJ OJQJaJ h'hC^5CJOJQJaJh):5CJOJQJaJhxAhxACJOJQJaJhh&WCJOJQJaJh&CJOJQJaJhnwCJOJQJaJh?CJOJQJaJhCJH*OJQJaJhCJOJQJaJh&WCJH*OJQJaJh5CJOJQJaJh&WCJOJQJaJhdTCJOJQJaJ+,KLMY=?@AB׹树sf\P:\*jhCJ OJQJUaJ mHnHuhnYiCJ OJQJaJ htCJOJQJh[htCJOJQJhnYihtCJ OJQJaJ hnYihnYiCJ OJQJaJ hnYih[CJ OJQJaJ 0jhnYihCJ OJQJUaJ mHnHuhnYihWbCJ OJQJaJ hnYihbCJ OJQJaJ hnYiho<CJ OJQJaJ hnYih.VCJ OJQJaJ hnYihbCJ aJ @BCDEa=?@ $IfgdJv $Ifgd $Ifgdo< $IfgdnYi$IfBCEO`aQ|fZNfBh5CJOJQJaJhJvCJOJQJaJhxACJOJQJaJ*jhnwCJOJQJUaJmHnHuh[h CJOJQJaJh CJOJQJaJhZ-CJOJQJaJh[hg7CJOJQJaJh[htCJOJQJaJh[ho<CJOJQJaJhWbCJOJQJaJho<CJOJQJaJhnYiCJOJQJaJhnwCJOJQJaJ=>?CD`c&FGHK÷{oo`J>h#l+CJOJQJaJ*jh&CJOJQJUaJmHnHuh[hnwCJOJQJaJh5CJOJQJaJhCJOJQJaJh CJOJQJaJhFuCJOJQJaJh+CJOJQJaJh&WCJOJQJaJhnwCJOJQJaJh]FCJOJQJaJhJvCJOJQJaJ*jhnwCJOJQJUaJmHnHuh[hJvCJOJQJaJ@ABCDGIJK $Ifgd+ $Ifgdnw $Ifgd]F$If#op> ? @ L ~h~bQ@b5hb5CJOJQJ h?hbCJOJQJ^JaJ h?h?CJOJQJ^JaJ hbCJ*jh?CJOJQJUaJmHnHuh'h5CJOJQJaJh!CJOJQJaJh5CJOJQJaJh55CJOJQJaJh5hbCJOJQJh[h#l+CJOJQJaJh#l+CJOJQJaJ*jh+CJOJQJUaJmHnHuh[h+CJOJQJaJo? nlcccc]]$If $Ifgd5kd$$Ifl4J0;B%  0'4 laf4p? @ L !!O#nhhh[NA & F$Ifgd?" $If^gdy_ & F$Ifgd!$Ifkd$$Ifl40;]%"` 0'4 laf4pL !(!!!!!""#"H"J""""""" #M#N#O#P###$$0$1$7$g$h$$$$$ܽȒzzlzhuh@5CJOJQJh@CJOJQJhuh?"5CJOJQJh;yCJOJQJh%CJOJQJh?"CJOJQJh?"h?"CJOJQJh!5CJOJQJh?CJOJQJhy_CJOJQJhy_h!CJOJQJh~5CJOJQJh?"5CJOJQJ'O#P#h$i$%%%%?&Ikd$$Ifl40;]%"  0'4 laf4p$If $Ifgd;y & F$Ifgd?" $Ifgd?"$%%%%%%&&>&I&&&&T'U'V'X'Y'['\'^'_'a'c'''''''ӾӾӣɟscNc(jht0JB*CJOJQJUphht0JB*CJOJQJphhtB*CJOJQJphht5B*OJQJphhthpbjhpbUhbhNB-5CJOJQJhUHhUH5CJH*OJQJhUH5CJOJQJhbCJOJQJhb5CJOJQJh~5CJOJQJh?"h;yCJOJQJhuCJOJQJ?&&&&&kkdl$$Ifl0;*% 0'4 lap$If&&'$Ifpkd<$$Ifl;%' ` 0'4 lap ''"'/'$Ifpkd$$Ifl;%' ` 0'4 lap /'0'7'8'9':';'B'C'D'E'F'G'qkkkkkkkkkkk$Ifkd $$Ifl0;*% `0'4 lap G'H'I'J'K'R'S'T'U'W'X'Z'[']'^'`'a'b'c'''''(((((Ff $If''''''''''''''''((((9(:(T(U(t(u(v((((((yuyugYhJvCJOJQJ^JaJhnYiCJOJQJ^JaJhJv hJvhJvhJvCJOJQJaJh[hh[CJ aJ hh[CJ OJQJ^JaJ hhnYiCJ OJQJ^JaJ hpbhtht0JB*CJOJQJph*h)0JB*CJOJQJmHnHphu(jht0JB*CJOJQJUph((((T(U(t(u(v((((O)P)))))))))x*y***gdnYigd&gd5gdgdJv((/)M)N)O)P)`)a)j)m)r)z))))))))*w*x*y*~zl^MIh& h[h&CJOJQJ^JaJh&CJOJQJ^JaJhCJOJQJ^JaJh5 hJvh5h5CJOJQJaJhC^ hJvhhCJOJQJaJhdTCJH*OJQJaJhdTCJOJQJaJhJv h[hJvCJOJQJ^JaJhJvCJOJQJ^JaJhCJOJQJ^JaJhP-CJOJQJ^JaJy********[+\+]+_+ijүhb h[h+CJOJQJ^JaJh+CJOJQJ^JaJh+hnYihnYiCJ aJ hnYihnYiCJ OJQJ^JaJ hnYiCJ OJQJ^JaJ *\+]+^+_+gd+,/ =!"#$%7 X$$If!vh55255 5#v#v2#v#v #v:V l4 ``20'55255 54af4p2$$If!vh5'#v':V l4  ` 0'5'4af4p $$If!vh55#v#v:V l4 `0'554af4p$$If!vh5'#v':V l4 ` 0'5'4af4p $$If!vh5'#v':V l4  ` 0'5'4af4p $$If!vh5'#v':V l  ` 0'5'4ap $$If!vh55 #v#v :V l4J 0'55 / 4af4p$$If!vh5"5 #v"#v :V l4 0'+5"5 4af4p$$If!vh5"5 #v"#v :V l4 0'+5"5 4af4p$$If!vh55#v#v:V l 0'554ap$$If!vh5'#v':V l ` 0'5'4ap $$If!vh5'#v':V l ` 0'5'4ap $$If!vh55#v#v:V l `0'554apL$$If!vh5555555q55 5 5 5 5 5 555#v#v#v#vq#v#v #v #v #v #v #v #v#v#v:V l4 "```````֪0'5555q55 5 5 5 5 5 555/ 4af4p֪kdT $$Ifl4z;yI *D_f"%q "```````֪0'DDDD4 laf4p֪<@< nwNormalCJ_HmH sH tH N@N Heading 1$7$8$@&5CJOJPJQJDA@D Default Paragraph FontVi@V  Table Normal :V 44 la (k(No List 4@4 Header  !4 @4 Footer  !.)@. Page Number>_9qbF_#   >_9qbFI_# r $CGHqrNO T'(AM  +,M@BCDEa=?@ABCDGIJKo?@LOPhi?"/0789:;BCDEFGHIJKRSTUWXZ[]^`abc        T U t u v O!P!!!!!!!!!x"y"""\#]#`#0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x 0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x0x 0| 000000 0 0 000 00 00 00 00 0 0 0 00 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 @0y00@0y00@0y00@0y00@0@0@0@0@0@0@0@0@0@0@0@0@0y00p000000000000000000000000Pw $CGHqrNO T'(AM  +,M@Ba=?@ABCIJKo?@LO`#0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 00000000000000000000@0 @0@0j0(04)K@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0@0j0:0;h0?0h0?0@0@0@0@0@0@0j0C0Dh0C0h0C0@0@0@0@0@0 @0 @0@0@0@00  0 0 000 0ۊ0 0f(0ۊ00,  IK ZN'+BL $'(y*_+ "#$&'),.467GH N A@? O#?&&'/'G'(*_+!%(*+-/012358^+Xoz!8@P (  H  #  B    B    H   #   H   #   B     H  #  H  #   H  #  B S  ?@=Go_# tg *t} tZ t F t  1t6 t rt =' tet#f\.`# `#8*urn:schemas-microsoft-com:office:smarttagsdate  x 1120022005489DayMonthYear`bJQho"'UUWWXXZ[]^`a    ]#`#AL ]f'4UUWWXXZ[]^`a  b!l!!!P"]"""]#`#3333333333333 J ,UUWWXXZ[]^`a  U t !!`#UUWWXXZ[]^`a  `#9}[E8FES_r| /2zP^`o(. ^`hH. pLp^p`LhH. @ @ ^@ `hH. ^`hH. L^`LhH. ^`hH. ^`hH. PLP^P`LhH. 7^7`o(hH. ^`hH. pLp^p`LhH. @ @ ^@ `hH. ^`hH. L^`LhH. ^`hH. ^`hH. PLP^P`LhH. 71e^7`1o(hH. ^`hH. pLp^p`LhH. @ @ ^@ `hH. ^`hH. L^`LhH. ^`hH. ^`hH. PLP^P`LhH.ES_9}[/2z(        IH>AEnw1 !):dT;Kz?"\$%dH(#l+NB-@oN@xA]FUH> V.V [C^y_aWb.eMfnYiPmFuJv;y<?u&W5g7((@' trt+~&2)j#,Z-bP-[H6pbo<)Vf pZXN $CGHqrNO?@?"/0789:;BCDEFGHIJKRST`#""@ `0_#@UnknownGz Times New Roman5Symbol3& z Arial3z Times 1hfff88!4dEE2qHX?((2Date: Dawn LeederSchool of Nursing   Oh+'0|   , 8 D P\dltDate:  Dawn LeederNormalSchool of Nursing2Microsoft Office Word@@:@b;@b;՜.+,0 hp  University of Cambridge8E Date: Title  !"#$%&'()*+,-./0123456789;<=>?@ABCEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdeghijklmopqrstuxRoot Entry F`_v;zData :G1TableDCWordDocument.rSummaryInformation(fDocumentSummaryInformation8nCompObjq  FMicrosoft Office Word Document MSWordDocWord.Document.89q