1. The Choice atrial fibrillation patients increased risk of strokeatrial fibrillation patients increased risk of stroke –can reduce with warfarin, but increased bleeding risk without treatment 100 patients will suffer:without treatment 100 patients will suffer: –12 strokes (6 major, six minor), 3 serious gi bleeds in 1 year warfarin would increase bleeds in 100 patients to 5 per year (2 additional bleeds)warfarin would increase bleeds in 100 patients to 5 per year (2 additional bleeds) how many strokes must we prevent to make it worth taking warfarin with increased risk of bleeding?how many strokes must we prevent to make it worth taking warfarin with increased risk of bleeding?

2. PHYSICIAN AND PATIENT STROKE THRESHOLDS FOR WARFARIN

3. Physician and patient mean stroke threshold for warfarin Baseline risk of 12 strokes and 3 major bleeds in 100 patients over 2 yearsBaseline risk of 12 strokes and 3 major bleeds in 100 patients over 2 years Given warfarin would increase the risk of major bleeds to 5 in 100 patients, we then determined the minimum number of strokes that needed to be prevented for a participant to feel warfarin was justifiedGiven warfarin would increase the risk of major bleeds to 5 in 100 patients, we then determined the minimum number of strokes that needed to be prevented for a participant to feel warfarin was justified

4. The Choice without treatment 100 patients will suffer:without treatment 100 patients will suffer: –12 strokes (six major, six minor), 3 serious gi bleeds in 1 year warfarin would decrease strokes in 100 patients to 4 per year (8 fewer strokes, 4 major, minor)warfarin would decrease strokes in 100 patients to 4 per year (8 fewer strokes, 4 major, minor) how many bleeds would you accept in 100 patients over a year, and still be willing to administer/take warfarin?how many bleeds would you accept in 100 patients over a year, and still be willing to administer/take warfarin?

6. Physician and patient mean bleeding threshold for warfarin Baseline risk of 12 strokes and 3 major bleeds in 100 patients over 2 yearsBaseline risk of 12 strokes and 3 major bleeds in 100 patients over 2 years Given warfarin would decrease the risk of stroke to 4 in 100 patients, we then determined the maximum number of excess bleeds that participants were willing to acceptGiven warfarin would decrease the risk of stroke to 4 in 100 patients, we then determined the maximum number of excess bleeds that participants were willing to accept

7. Values and Preferences every intervention has benefits, risks, inconvenience, costsevery intervention has benefits, risks, inconvenience, costs decision a trade-offdecision a trade-off values and preferences differvalues and preferences differ Cochrane reviews particularly vulnerable because world-wideCochrane reviews particularly vulnerable because world-wide Cochrane reviews shouldn’t make recommendationsCochrane reviews shouldn’t make recommendations

8. Issues for this Workshop should Cochrane reviews structure discussion?should Cochrane reviews structure discussion? –highlight tradeoffs and potential impact of values –highlight implementation, applicability issues guideline developers using Cochrane reviewsguideline developers using Cochrane reviews –should they grade recommendations? –should they use a uniform system (and if so, what should it look like)

9. Osteoporosis Common, serious morbidityCommon, serious morbidity –vertebral and non-vertebral fractures Many agents availableMany agents available –what should we offer women Evidence versus recommendationsEvidence versus recommendations

10. Relative Risk with 95%CI of Vertebral Fracture After Treatment with Calcium Favours Calcium Favours Control & & & & & & & Chevalley 0.45 (0.11 to 1.88) Recker (w/fractures) 0.58 (0.35 to 0.97) Recker (w/o fractures) 1.36 (0.70 to 2.62) Reid 0.45 (0.11 to 1.94) Riggs 0.90 (0.38 to 2.18) Hansson 0.87 ( 0.10 to 7.71) Pooled Estimate 0.77 (0.54 to 1.09) 00.511.522.53 Relative Risk, 95% CI Prevention Trials (n = 45) (n = 92) (n = 99) (n = 122) (n = 177) (n = 41) (n = 576)

11. Relative Risk with 95% CI of Non-Vertebral Fracture after Treatment with Calcium Favours Calcium Favours Control ' ' ' Chevally 0.48 ( 0.07 to 3.38) Riggs 0.93 ( 0.44 to 1.96) Pooled Estimate 0.86 (0.43 to 1.72) 00.511.522.533.5 Prevention Trials Relative Risk, 95% CI (n = 45) (n = 177) (n = 222)

12. Relative Risk with 95% CI for Vertebral Fractures after Treatment with Vitamin D Favours Vitamin D Favours Control ' ' ' ' ' ' ' ' ' ' Baeksgaard(1998) 0.33(0.01 to 8.06) Gallagher (1990) 0.90 (0.42 to 1.89) Orimo (1994) 0.37 (0.09 to 1.44) Ott (1989) 1.46 ( 0.59 to 3.62) Tilyard (1992) 0.43 ( 0.31 to 0.61) Guesens (1986) 0.88 (0.43 to 1.80) Orimo (1987) 0.46 (0.31 to 0.69) Caniggia (1984) 0.20 (0.01 to 3.54) Pooled Hydroxylated Vitamin D Estimate 0.61 ( 0.42 to 0.87) Pooled Estimate 0.60 (0.42 to 0.84) 00.511.522.5 Standard Vitamin D (IU) Hydroxylated Vitamin D (ug) (N =160) (N =50) (N = 80) (N = 86) (N = 622) (N =32) (N = 86) (N = 14) (N = 970) (N =1130)

13. Relative Risk with 95% CI for Non-Vertebral Fractures after Treatment with Vitamin D ' ' ' ' ' ' ' ' ' Chapuy (1992) 0.75 ( 0.61 to 0.91) Lips (1996) 1.04 (0.77 to 1.41) Dawson-Hughes* (1997) 0.45 (0.22 to 0.91) Pooled Standard Vitamin D Estimate 0.78 (0.55 TO 1.09) Ott (1989) 2.20 ( 0.52 to 9.24) Tilyard (1992) 0.50 ( 0.25 to 1.00) Orimo (1994) 1.10 (0.02 to 2.0) Pooled Hydroxylated Vitamin D Estimate 0.87 (0.29 to 2.59) Pooled Estimate 0.77 (0.57 to 1.04) 00.511.522.533.5 Standard Vitamin D (IU) Hydroxylated Vitamin D (ug) * Prevention Trial Favours Vitamin D Favours Control (N =3270) (N =1916) (N =213) (N = 5399) (N = 86) (N =622) (N = 80) (N =788) (N = 6187)

14. RR of Vertebral Fracture after Treatment with HRT ' ' ' ' ' ' ' Lufkin 1992 0.63 (0.28, 1.43) Greenspan 1998 (0.70 (0.06, 7.55) Wilalawansa 1998 0.40 (0.09, 1.77) Hulley 1998 0.74 (0.37, 1.47) Alexandersen 1999 2.78 ( 0.12, 65.09) WHI 2002 0.65 (0.44, 0.97) Pooled Estimate 0.66 (0.49, 0.90) 0.010.1110100 Relative Risk (95% CI) Favours HRT Favours Control (N = 75) (N = 193) (N = 32) (N = 2763) (N = 52) (N = 16608) (N = 19723)

15. RR of Non-Vertebral Fracture after Treatment with HRT ' ' ' ' ' ' ' ' Greenspan 1998 (0.70 (0.22, 2.22) Komulainen 1997 0.40 (0.16, 0.99) Wilalawansa 1998 1.00 (0.07, 14.79) Hulley 1998 0.90 (0.69, 1.19) Hosking 1998 0.98 ( 0.29, 3.34)) Alexandersen 1999 0.31 ( 0.03, 2.76) WHI 2002 0.68 ( 0.46, 0.99) Pooled Estimate 0.78 (0.64, 0.96) 0.010.1110100 Relative Risk (95% CI) Favours HRT Favours Control (N =2763) (N =193) (N =612) (N =36) (N =232) (N =50) (N =16608) (N =20494)

16. Relative Risk with 95% CI for Vertebral & Non-Vertebral Fractures After Treatment with Raloxifene ' ' ' ' ' ' Ettinger 0.59 (0.50 to 0.70) Lufkin 1.15 (0.75 to 1.75) Pooled Vertebral Fracture Estimate 0.64 ( 0.55 to 0.75) Ettinger 0.91 (0.79 to 1.06) Lufkin 0.51 ( 0.12 to 2.16) Pooled Non Vertebral Fracture Estimate 0.91 ( 0.78 to 1.06) 0.1110 * All Trials Secondary Treatment (N = 7705) ( N = 143) (N = 7848) ( N = 7705) (N= 143) (N = 7848) Vertebral Fractures Non-Vertebral Fractures Fixed Effects Model Vertebral fracture results from Lufkin trial based on 15% cutoff in reduction of vertebrae ( baseline to 1 year) Favours Raloxifene Favours Control

17. Weighted Relative Risk for Vertebral Fractures after Treatment with Etidronate Favours Etidronate Favours Control Osteoporotic and Non-Osteoporotic Populations (Primary Prevention Trials: Herd, Meunier, and Pouilles [n = 315] not included due to low incidence of fractures) * Treatment and Control Groups Received Phosphate & & & & & & & & & & Watts 0.52 (0.19 to 1.40) Watts* 0.47 (0.14 to 1.61) Pooled Prevention Estimate 0.62 (0.30 to 1.27) Montessori 0.14 (0.01 to 2.67) Pacifici 1.10 (0.35 to 3.44) Storm 0.64 (0.35 to 1.17) Wimalawansa 1998 0.67 (0.21 to 2.18) Lyritis 0.47 ( 0.17 to 1.36) Pooled Treatment Estimate 0.68 (0.42 to 1.10) Pooled Estimate 0.63 ( 0.44 to 0.99) 0.0010.010.1110 Relative Risk, 95% CI (N = 80) (N = 57) (N = 66) (N = 209 ) (N =214) (N = 35) (N = 1076) Prevention Trials Treatment Trials (N = 738) (N = 338) (N = 100)

18. Weighted Relative Risk for Non-Vertebral Fractures after Treatment with Etidronate Favours Etidronate Favours Control Osteoporotic and Non-Osteoporotic Populations * Montessori Trial (N=80) not included in figure due to zero Non-Vertebral Fractures occuring. ** Treatment and Control Groups Received phosphate & & & & & & & & & & Watts 1.23 (0.68 to 2.22) Watts** 1.16 (0.57 to 2.35) Meunier 0.71 (0.15 to 3.32) Pouilles 0.55 (0.16 to 1.9) Pooled Prevention Trial Estimate: 1.06 (0.71 to 1.60) Storm 0.85 (0.31 to 2.37) Wimalawansa 1998 1.06 (0.12 to 9.24) Lyritis 0.64 (0.18 to 2.30) Pooled Treatment Trial Estimate: 0.79 (0.38 to 1.67) Pooled Estimate 0.99 (0.69 to 1.42) 0.1110 Relative Risk, 95% CI Prevention Trials Treatment Trials (N = 54 ) (N = 109 ) (N = 586 ) (N = 66 ) (N = 209) (N = 214 ) (N = 35 ) (N = 281) (N = 867) (N = 100)

19. Relative Risk with 95% CI for Vertebral Fractures for Doses of 5mg or Greater of Alendronate Adami and Hoskings trials not included in figure due to low vertebral fracture incidence. ' ' ' ' ' ' ' ' ' ' McClung 0.34 (0.04 to 3.25) Pooled Prevention Estimate 0.45(0.06 to 3.15) Bone 0.68 ( 0.21 to 2.18) Chesnut 0.25 (0.03 to 2.34) Liberman (USA) 0.52 ( 0.24 to 1.15) Liberman (Int) 0.52 ( 0.20 to 1.34) Black 0.53 (0.41 to 0.69) Cummings 0.51 ( 0.31 to 0.84) Pooled Treatment Estimate 0.53 (0.43 to 0.65) Pooled Estimate 0.52 (0.43 to 0.65) 0.010.1110 Prevention Trials Favours Alendronate Favours Control (n = 355) (n = 1355) (n = 184) (n = 157) (n = 478) (n = 516) (n = 2027) (n = 4432 ) (n = 8005 ) (n = 9360) Treatment Trials

20. Risk Ratios and Summary Estimates with 95% CI for Non-Vertebral Fractures for Dose of 10mg or Greater of Alendronate ' ' ' ' ' ' ' ' ' McClung 0.79 (0.28 to 2.24) Adami 0.36 (0.07 to 1.80) Chesnut 0.43 (0.11 to 1.65) Liberman (USA) 0.55 (0.31 to 0.97) Liberman (Int) 0.65 (0.32 to 1.34) Pols 0.47 (0.26 to 0.83) Rosen 0.35 (0.15 to 0.77) Pooled Treatment Estimate 0.49 (0.36 to 0.67) Pooled Estimate 0.51 (0.38 to 0.69) 0.010.1110 Prevention Trials Treatment Trials Favours Alendronate Favours Control (n =267) (n = 211) (n = 125) (n = 380) (n =412) (n = 1908) (n =419) (n = 3455) (n = 3722)

21. Relative Risk with 95% CI for Non-Vertebral Fractures after Treatment with Risedronate (Final Year, All Doses) ' ' ' ' ' ' ' Mortensen (1998) 0.49 (0.12 to 2.03) Harris (1999) 0.64 (0.42 to 0.98) Clemensen (1997) 0.70 (0.45 to 1.09) McClung (Abstract) 0.71 (0.36 to 1.40) Reginster (2000) 0.71 (0.47 to 1.06) Pooled Treatment Estimate 0.69 (0.55 to 0.86) Pooled Estimate 0.68 (0.54 to 0.85) 00.511.522.5 Prevention Trials Treatment Trials Favours Risedronate Favours Control (N = 111) (N = 1627 ) (N =132) (N = 648) (N =812) (N =3219 ) (N =3330 )

22. Relative Risk with 95% CI for Vertebral Fractures after Treatment with Risedronate (Final Year, All Doses) Favours Risedronate Favours Control ' ' ' ' ' ' ' ' ' Mortensen (1998) 2.44 (0.12 to 49.43) Harris 1- year (1999) 0.59 (0.36 to 0.97) Harris - 3 year (1999) 0.64 (0.47 to 0.87) Clemensen (1997) 1.52 (0.56 to 4.15) Fogelman (Abstract) 0.72 (0.45 to1.15) Reginster 1 - year (2000) 0.55 (0.34 to 0.87) Reginster 3 - year (2000) 0.60 (0.44 to 0.81) Pooled Treatment Estimate 0.63 (0.54 to 0.75) Pooled Estimate 0.64 (0.54 to 0.85) 0.1110 Prevention Trials Treatment Trials (N = 111) (N = 1278) (N =1374) (N = 132) (N = 541) (N = 663) (N = 690) (N = 4687) (N =4789)

23. Early treatment may be appropriate Baseline risk of fracture from alendronate RCTs over 2 year periodBaseline risk of fracture from alendronate RCTs over 2 year period non-osteoporoticNNTsnon-osteoporoticNNTs –vertebral 0.12%1,790 –non-vertebral 2.54% 80 osteoporoticosteoporotic –vertebral 2.88% 72 –non-vertebral 6.85% 24

24. Benefits Drugs that reduce vertebral fracturesDrugs that reduce vertebral fractures –vitamin D, HRT, raloxifene, risedronate, alendronate Drugs that reduce non-vertebral fracturesDrugs that reduce non-vertebral fractures –risedronate (1/3 RRR), alendronate (1/2 RRR)

25. Values and Preferences high value: reducing fractures, no uncertaintyhigh value: reducing fractures, no uncertainty –choose alendronate high value: reducing fractures, no inconveniencehigh value: reducing fractures, no inconvenience –alendronate upright 30 minutes before meal –choose residronate high value on “natural” treatment, low costhigh value on “natural” treatment, low cost –calcium and vitamin D high value on fracture reduction – early treatmenthigh value on fracture reduction – early treatment high value living without medication – late treatmenthigh value living without medication – late treatment

26. Grading Recommendations methodologic strengthmethodologic strength –High (RCT), intermediate (quasi-RCTs), low (observational), insufficient (other) –implementation, consistency, directness decisiondecision –do it, don’t do, toss-up strength of decisionstrength of decision –strong (across range of values, most would choose –weak (different choices across range of values)