Effectiveness and safety of Glucosamine, chondroitin, the two in combination, or celecoxib in the treatment of osteoarthritis of the knee

Transcription

1 Effectiveness and safety of Glucosamine, chondroitin, the two in combination, or celecoxib in the treatment of osteoarthritis of the knee Chao Zeng, Jie Wei 2,3, Hui Li, Yi-lun Wang, Dong-xing Xie, Tuo Yang, Shu-guang Gao, Yu-sheng Li, Wei Luo, Guang-hua Lei * Appendix search strategies. PUBMED DATABASE. osteoarthriti*[tiab] or osteoarthriti*[mh] 2. osteoarthro*[tiab] or gonarthriti*[tiab] or gonarthro*[tiab] or coxarthriti*[tiab] or coxarthro*[tiab] or osteo?arthritis[tiab] 3. knee*[tiab] and (pain*[tiab] or discomfort*[tiab]) 4. knee*[tiab] and stiff*[tiab] 5. or/-6 6. "celecoxib" [Supplementary Concept] 7. Celecoxib[tiab] or celebrex[tiab] or Celebra[tiab] or Onsenal[tiab] 8. 7 or 8 9. "Chondroitin"[Mesh]. Chondroitin*[tiab]. 9 or 2. Glucosamine [Mesh] 3. Glucosamine[tiab] 4. Acetylglucosamine[tiab] or n-acetylglucosamine[tiab] or n-acetyl-dglucosamine[tiab] 5. Or/ placebo[tiab] 7. 8 and 8. 8 and and 6 2. and 5 2. and and Or/ randomized[tiab] 25. controlled[tiab] 26. random*[tiab] 27. rct [text word] 28. trial*[tiab] 29. groups[tiab] 3. ((singl*[tiab] or doubl*[tiab] or tripl*[tiab]) and (mask*[tiab] or blind*[tiab])) 3. Or/24-3

2 32. 5 and 23 and 3 Embase. osteoarthritis /exp 2. (osteoarthriti* or osteoarthro* or gonarthriti* or gonarthro* or gonarthro* or coxarthriti* or coxarthro* or arthros* or arthrot* or osteo*arthritis):ti,ab 3. (knee* near/3 (pain* or ach* or discomfort*)):ti,ab 4. (knee* near/3 stiff*):ti,ab 5. Or/-4 6. celecoxib /exp 7. (celecoxib or celebrex or Celebra or Onsenal):ti,ab 8. 6 or 7 9. chondroitin /exp. Chondroitin:ti,ab. 9 or 2. Glucosamine /exp 3. Glucosamine:ti,ab 4. (Acetylglucosamine or n-acetylglucosamine or n-acetyl-d-glucosamine):ti,ab 5. 2 or 3 or 4 6. Placebo:ti,ab 7. 8 and 6 8. and and and 2. 8 and and Or/ (random* or control* or trial*):ti,ab 25. Groups:ti,ab 26. rct :ti,ab 27. ((singl* or doubl*or tripl*) and (mask* or blind*)):ti,ab 28. Or/ and 23 and 28 Cochrane. (osteoarthritis* or osteoarthro* or gonarthriti* or gonarthro* or coxarthriti* or coxarthro* or arthros* or arthrot*):ab,ti 2. MeSH descriptor Osteoarthritis explode all trees 3. or 2 4. MESH descriptor celecoxib explode all trees 5. (celecoxib or celebrex or Celebra or Onsenal):ab,ti 6. 4 or 5 7. MESH descriptor chondroitin explode all trees 8. Chondroitin:ab,ti

3 9. 7 or 8. MESH descriptor Glucosamine explode all trees. (Glucosamine or Acetylglucosamine or n-acetylglucosamine or n-acetyl-dglucosamine):ab,ti 2. or 3. Placebo:ab,ti 4. 6 and and and and and and 3 2. Or/ and 2 Appendix 2 WinBUGs codes for network meta-analysis. a. Random effects model for multi-arm trials (for continuous data, calculating SMD) model { for(i in :ns) { w[i,]<- delta[i,t[i,]]<- ss[i]<- sum(n[i,:na[i]]) nom[i]<- sum(nom[i,:na[i]]) pooled.sd[i]<- sqrt(nom[i]/(ss[i]-na[i])) J[i]<- -(3/((4*(ss[i]-na[i]))-)) #Normal Likelihood# for (k in :na[i]) { y[i,k] ~ dnorm(phi[i,t[i,k]],prec[i,k]) se[i,k]<- sd[i,k]/sqrt(n[i,k]) var[i,k]<- se[i,k]*se[i,k] prec[i,k]<- /var[i,k] nom[i,k]<- (n[i,k]-)*sd[i,k]*sd[i,k] #Parameterization of the model# phi[i,t[i,]]<- u[i]*(pooled.sd[i]/j[i]) for (k in 2:na[i]) { phi[i,t[i,k]]<- (u[i]+delta[i,t[i,k]])*(pooled.sd[i]/j[i]) delta[i,t[i,k]] ~ dnorm(md[i,t[i,k]],taud[i,t[i,k]]) md[i,t[i,k]]<- d[t[i,k]] - d[t[i,]] + sw[i,k] taud[i,t[i,k]]<- tau *2*(k-)/k w[i,k]<- (delta[i,t[i,k]] - d[t[i,k]] + d[t[i,]])

4 sw[i,k]<- sum(w[i,:k-])/(k-) #Priors# SD ~ dnorm(,)i(,) tau<- /pow(sd,2) for(k in :(ref-)) { d[k] ~ dnorm(,.) for(k in (ref+):nt) { d[k] ~ dnorm(,.) for(i in :ns) { u[i] ~ dnorm(,.) #Estimated & Predicted Standardized Mean Differences# d[ref]<- for (c in :(ref-)) { SMD.ref[c]<- d[c] - d[ref] predsmd.ref[c] ~ dnorm( SMD.ref[c],tau) for (c in (ref+):nt) { SMD.ref[c]<- d[c] - d[ref] predsmd.ref[c] ~ dnorm( SMD.ref[c],tau) for (c in :(nt-)) { for (k in (c+):nt) { SMD[c,k]<- d[c] - d[k] predsmd[c,k] ~ dnorm(smd[c,k],tau) #Ranking of treatments# for(k in :nt) { order[k]<- rank(d[],k) # this is when the outcome is positive - omit 'nt+-' when the outcome is negative most.effective[k]<-equals(order[k],) for(j in :nt) { effectiveness[k,j]<- equals(order[k],j) for(k in :nt) { for(j in :nt) { cumeffectiveness[k,j]<- sum(effectiveness[k,:j])

5 #SUCRAS# for(k in :nt) { SUCRA[k]<- sum(cumeffectiveness[k,:(nt-)]) /(nt-) #Fit of the Model# for(i in :ns) { for(k in :na[i]) { Darm[i,k]<-(y[i,k]-phi[i,t[i,k]])*(y[i,k]-phi[i,t[i,k]])/var[i,k] D[i]<- sum(darm[i,:na[i]]) D.bar<- sum(d[]) b. Random effects model for multi-arm trials (for continuous data, calculating MD) model { for (i in :ns) { w[i, ] <-.E+ delta[i, ] < -.E+ mu[i] ~ dnorm(.e+,.e-4) for (k in :na[i]) { var[i, k] < -pow(se[i, k], 2) prec[i, k] <- /var[i, k] y[i, k] ~ dnorm(theta[i, k], prec[i, k]) theta[i, k] < -mu[i] + delta[i, k] dev[i, k] < -(y[i, k] - theta[i, k]) * (y[i, k] - theta[i, k]) * prec[i, k] resdev[i] < -sum(dev[i, :na[i]]) for (k in 2:na[i]) { delta[i, k] ~ dnorm(md[i, k], taud[i, k]) md[i, k] <- d[t[i, k]] - d[t[i, ]] + sw[i, k] taud[i, k] < -tau * 2 * (k - )/k w[i, k] <- (delta[i, k] - d[t[i, k]] + d[t[i, ]]) sw[i, k] < -sum(w[i, :k - ])/(k - ) totresdev < -sum(resdev[]) d[] <-.E+ for (k in 2:nt) { d[k] ~ dnorm(.e+,.e-4)

6 sd ~ dunif(.e+, 5) tau <- pow(sd, -2) for (c in :(nt - )) { for (k in (c + ):nt) { diff[c, k] < -(d[c] - d[k]) for (k in :nt) { rk[k] < -rank(d[], k) best[k] < -equals(rk[k], ) for (j in :nt){ effectiveness[k,j]<-equals(rk[k],j) for(k in :nt){ for (j in :nt){ cumeffectiveness[k,j]<-sum(effectiveness[k,:j]) for(k in :nt){ SUCRA[k]<-sum(cumeffectiveness[k,:(nt-)])/(nt-) c. Random effects model for multi-arm trials (for category data) model{ # *** PROGRAM STARTS for(i in :ns){ # LOOP THROUGH STUDIES w[i,] <- # adjustment for multi-arm trials is zero for control arm delta[i,] < - # treatment effect is zero for control arm mu[i] ~ dnorm(,.) # vague priors for all trial baselines for (k in :na[i]) { # LOOP THROUGH ARMS r[i,k] ~ dbin(p[i,k],n[i,k]) # binomial likelihood logit(p[i,k]) < - mu[i] + delta[i,k] # model for linear predictor rhat[i,k] <- p[i,k] * n[i,k] # expected value of the numerators #Deviance contribution dev[i,k] < - 2 * (r[i,k] * (log(r[i,k]) -log(rhat[i,k])) + (n[i,k]-r[i,k]) * (log(n[i,k]-r[i,k]) - log(n[i,k] - rhat[i,k])))

7 # summed residual deviance contribution for this trial resdev[i] < - sum(dev[i,:na[i]]) for (k in 2:na[i]) { # LOOP THROUGH ARMS # trial-specific LOR distributions delta[i,k] ~ dnorm(md[i,k],taud[i,k]) # mean of LOR di stributions (with multi-arm trial correction) md[i,k] <- d[t[i,k]] - d[t[i,]] + sw[i,k] # precision of LOR distributions (with multi-arm trial correction) taud[i,k] < - tau *2*(k -)/k # adjustment for multi-arm RCTs w[i,k] <- (delta[i,k] - d[t[i,k]] + d[t[i,]]) # cumulative adjustment for multi-arm trials sw[i,k] < - sum(w[i,:k -])/(k-) totresdev < - sum(resdev[]) # Total Residual Deviance d[]<- # treatment effect is zero for reference t reatment # vague priors for treatment effects for (k in 2:nt){ d[k] ~ dnorm(,.) sd ~ dunif(,5) # vague prior for between -trial SD tau <- pow(sd,-2) # between-trial precision = (/between -trial variance) # pairwise ORs and LORs for all possi ble pair-wise comparisons, if nt>2 for (c in :(nt -)) { for (k in (c+):nt) { or[c,k] < - exp(d[k] - d[c]) lor[c,k] < - (d[k] -d[c]) # ranking on relative scale for (k in :nt) { # rk[k] < - nt+ -rank(d[],k) # assumes events are good rk[k] < - rank(d[],k) # assumes events are bad best[k] < - equals(rk[k],) #calculate probability that treat k is best # *** PROGRAM ENDS Appendix 2 WinBUGs codes for network meta-analysis. a. Random effects model for multi-arm trials (for continuous data, calculating SMD) model { for(i in :ns) {

8 w[i,]<- delta[i,t[i,]]<- ss[i]<- sum(n[i,:na[i]]) nom[i]<- sum(nom[i,:na[i]]) pooled.sd[i]<- sqrt(nom[i]/(ss[i]-na[i])) J[i]<- -(3/((4*(ss[i]-na[i]))-)) #Normal Likelihood# for (k in :na[i]) { y[i,k] ~ dnorm(phi[i,t[i,k]],prec[i,k]) se[i,k]<- sd[i,k]/sqrt(n[i,k]) var[i,k]<- se[i,k]*se[i,k] prec[i,k]<- /var[i,k] nom[i,k]<- (n[i,k]-)*sd[i,k]*sd[i,k] #Parameterization of the model# phi[i,t[i,]]<- u[i]*(pooled.sd[i]/j[i]) for (k in 2:na[i]) { phi[i,t[i,k]]<- (u[i]+delta[i,t[i,k]])*(pooled.sd[i]/j[i]) delta[i,t[i,k]] ~ dnorm(md[i,t[i,k]],taud[i,t[i,k]]) md[i,t[i,k]]<- d[t[i,k]] - d[t[i,]] + sw[i,k] taud[i,t[i,k]]<- tau *2*(k-)/k w[i,k]<- (delta[i,t[i,k]] - d[t[i,k]] + d[t[i,]]) sw[i,k]<- sum(w[i,:k-])/(k-) #Priors# SD ~ dnorm(,)i(,) tau<- /pow(sd,2) for(k in :(ref-)) { d[k] ~ dnorm(,.) for(k in (ref+):nt) { d[k] ~ dnorm(,.) for(i in :ns) { u[i] ~ dnorm(,.) #Estimated & Predicted Standardized Mean Differences# d[ref]<- for (c in :(ref-)) { SMD.ref[c]<- d[c] - d[ref] predsmd.ref[c] ~ dnorm( SMD.ref[c],tau)

9 for (c in (ref+):nt) { SMD.ref[c]<- d[c] - d[ref] predsmd.ref[c] ~ dnorm( SMD.ref[c],tau) for (c in :(nt-)) { for (k in (c+):nt) { SMD[c,k]<- d[c] - d[k] predsmd[c,k] ~ dnorm(smd[c,k],tau) #Ranking of treatments# for(k in :nt) { order[k]<- rank(d[],k) # this is when the outcome is positive - omit 'nt+-' when the outcome is negative most.effective[k]<-equals(order[k],) for(j in :nt) { effectiveness[k,j]<- equals(order[k],j) for(k in :nt) { for(j in :nt) { cumeffectiveness[k,j]<- sum(effectiveness[k,:j]) #SUCRAS# for(k in :nt) { SUCRA[k]<- sum(cumeffectiveness[k,:(nt-)]) /(nt-) #Fit of the Model# for(i in :ns) { for(k in :na[i]) { Darm[i,k]<-(y[i,k]-phi[i,t[i,k]])*(y[i,k]-phi[i,t[i,k]])/var[i,k] D[i]<- sum(darm[i,:na[i]]) D.bar<- sum(d[]) b. Random effects model for multi-arm trials (for continuous data, calculating MD) model { for (i in :ns) { w[i, ] <-.E+

10 delta[i, ] < -.E+ mu[i] ~ dnorm(.e+,.e-4) for (k in :na[i]) { var[i, k] < -pow(se[i, k], 2) prec[i, k] <- /var[i, k] y[i, k] ~ dnorm(theta[i, k], prec[i, k]) theta[i, k] < -mu[i] + delta[i, k] dev[i, k] < -(y[i, k] - theta[i, k]) * (y[i, k] - theta[i, k]) * prec[i, k] resdev[i] < -sum(dev[i, :na[i]]) for (k in 2:na[i]) { delta[i, k] ~ dnorm(md[i, k], taud[i, k]) md[i, k] <- d[t[i, k]] - d[t[i, ]] + sw[i, k] taud[i, k] < -tau* 2 *(k- )/k w[i, k] <- (delta[i, k] - d[t[i, k]] + d[t[i, ]]) sw[i, k] < -sum(w[i, :k - ])/(k - ) totresdev < -sum(resdev[]) d[] <-.E+ for (k in 2:nt) { d[k] ~ dnorm(.e+,.e-4) sd ~ dunif(.e+, 5) tau <- pow(sd, -2) for (c in :(nt - )) { for (k in (c + ):nt) { diff[c, k] < -(d[c]- d[k]) for (k in :nt) { rk[k] < -rank(d[], k) best[k] < -equals(rk[k], ) for (j in :nt){ effectiveness[k,j]<-equals(rk[k],j) for(k in :nt){ for (j in :nt){ cumeffectiveness[k,j]<-sum(effectiveness[k,:j])

11 for(k in :nt){ SUCRA[k]<-sum(cumeffectiveness[k,:(nt-)])/(nt-) c. Random effects model for multi-arm trials (for category data) model{ # *** PROGRAM STARTS for(i in :ns){ # LOOP THROUGH STUDIES w[i,] <- # adjustment for multi-arm trials is zero for control arm delta[i,] < - # treatment effect is zero for control arm mu[i] ~ dnorm(,.) # vague priors for all trial baselines for (k in :na[i]) { # LOOP THROUGH ARMS r[i,k] ~ dbin(p[i,k],n[i,k]) # binomial likelihood logit(p[i,k]) < - mu[i] + delta[i,k] # model for linear predictor rhat[i,k] <- p[i,k] * n[i,k] # expected value of the numerators #Deviance contribution dev[i,k] < - 2 * (r[i,k] * (log(r[i,k]) -log(rhat[i,k])) + (n[i,k]-r[i,k]) * (log(n[i,k]-r[i,k]) - log(n[i,k] - rhat[i,k]))) # summed residual deviance contribution for this trial resdev[i] < - sum(dev[i,:na[i]]) for (k in 2:na[i]) { # LOOP THROUGH ARMS # trial-specific LOR distributions delta[i,k] ~ dnorm(md[i,k],taud[i,k]) # mean of LOR di stributions (with multi-arm trial correction) md[i,k] <- d[t[i,k]] - d[t[i,]] + sw[i,k] # precision of LOR distributions (with multi-arm trial correction) taud[i,k] < - tau *2*(k -)/k # adjustment for multi-arm RCTs w[i,k] <- (delta[i,k] - d[t[i,k]] + d[t[i,]]) # cumulative adjustment for multi-arm trials sw[i,k] < - sum(w[i,:k -])/(k-) totresdev < - sum(resdev[]) # Total Residual Deviance d[]<- # treatment effect is zero for reference t reatment # vague priors for treatment effects for (k in 2:nt){ d[k] ~ dnorm(,.)

12 sd ~ dunif(,5) tau <- pow(sd,-2) # vague prior for between -trial SD # between-trial precision = (/between -trial variance) # pairwise ORs and LORs for all possi ble pair-wise comparisons, if nt>2 for (c in :(nt -)) { for (k in (c+):nt) { or[c,k] < - exp(d[k] - d[c]) lor[c,k] < - (d[k] -d[c]) # ranking on relative scale for (k in :nt) { # rk[k] < - nt+ -rank(d[],k) # assumes events are good rk[k] < - rank(d[],k) # assumes events are bad best[k] < - equals(rk[k],) #calculate probability that treat k is best # *** PROGRAM ENDS Appendix 4 Risk of bias of included study.

13

14 Figure Risk of bias summary: review authors' judgements about each risk of bias item for each included study. Figure 2 Risk of bias graph: review authors' judgements about each risk of bias item presented as percentages across all included studies. Appendix 5 Absolute treatment effects (change from baseline) for pain and function. Treatment Pain (SMD) Pian (WOMAC -) Function (SMD) Function (WOMAC -) Celecoxib -.9 (-.23, -.96) (-3.8, (-.6, -.82) -.97 (-2.23,-.72) 2.4) Glucosamine -.85 (-.5, -.65) -2.3 (-2.63,-.63) -.56 (-.78, -.35) -.8 (-.64,-.74) Chondroitin -.2 (-.3, -.74) (-3.28,-.85) -.87 (-.9, -.54) -.83 (-2.5,-.3) glucosamine -. (-.66, -.34) -2.5 (-4.5,-.85) -.66 (-2.45, -.86) (-5.5,-.8) + chondroitin Appendix 6 Rankings for effects of pain relief and function improvement.

15 celecoxib glucosamine chondroitin glucosamine+chondroitin placebo Graphs by Treatment Rank Rankings for effects of pain relief. Graph displays distribution of probabilities for each treatment. X-axis represents the possible rank of each treatment (from the best rank to worse according to the outcomes),y-axis represents the cumulative probability for each treatment to be the best option, among the best two options, among the best three options, and so on.

16 celecoxib glucosamine chondroitin glucosamine+chondroitin placebo Graphs by Treatment Rank Rankings for effects of function improvement. Graph displays distribution of probabilities for each treatment. X-axis represents the possible rank of each treatment (from the best rank to worse according to the outcomes),y-axis represents the cumulative probability for each treatment to be the best option, among the best two options, among the best three options, and so on. Appendix 7 Heterogeneity and publication bias between direct comparisons Pain difference Heterogeneity (P/I 2 ) Publication bias (P value of begg s test) Cel vs pla.2/3%.45 GS vs pla.5/46%.28 CS vs pla./8%.26 Cel vs GS./8%. Cel vs GS+CS.7/7%. Function difference Cel vs pla.3/3%.89 GS vs pla.9/4%.72 CS vs pla.3/8%.73 Cel vs GS.6/7%.

17 Cel vs GS+CS.88/%. Joint width narrowing GS vs pla./75%.73 CS vs pla.2/67%.8 GS+CS vs pla.33/%. GS vs CS.29/9%. GS vs GS+CS.2/8%. CS vs GS+CS.2/38%. Withdrawal due to AE Cel vs pla.2/2%.4 GS vs pla.77/%.28 CS vs pla.97/%.72 GS+CS vs pla.75/%. Cel vs GS.95/%. Cel vs GS+CS.32/%. GS vs CS.44/%. GS vs GS+CS.54/%. CS vs GS+CS.93/%. SAEs Cel vs pla.97/%.44 GS vs pla.37/%. CS vs pla.83/%.3 Cel vs GS+CS.8/%. Number of patients with AE Cel vs pla.9/27%.53 GS vs pla.89/%.35 CS vs pla.4/%.45 GI AE Cel vs pla.88/%.39 GS vs pla.99/%.76 CS vs pla.39/4%.45 CV AE Cel vs pla.63/%.86 GS vs pla.5/%.8

18 CS vs pla.84/%. CNS AE Cel vs pla.8/%.6 GS vs pla.99/%.44 CS vs pla.53/%. Infection Cel vs pla.25/9%.45 GS vs pla.48/%.55 CS vs pla.26/26%. MU AE Cel vs pla.53/%. GS vs pla.48/%.7 CS vs pla.2/59%. Skin AE Cel vs pla.73/%. GS vs pla.52/%.7 CS vs pla.7/62%. GS, glucosamine; CS, chondroitin; pla, placebo; Cel, celecoxib; AE, adverse event; SAE, serious adverse event; GI, gastrointestinal; CV, cardiovascular; CNS, central nervous system; MU, musculoskeletal