Total CRQ score formulas: Standard errors and standard deviations for total CRQ scores To calculate the standard deviations and m mXE mXM mXD SEx SE XF 4 m mXE mXM mXD Var XF 4 standard errors for the total CRQ treatment group mean, and the total CRQ control group and the standard deviation by multiplying the mean, and the total CRQ mean difference, we standard error by the squareroot of the assumed a correlation of ρ=0.5. Let mEF, mEE, number of patients in the treatment group, nx mEM and mED denote the means for the Fatigue(F), Emotional(E), Mastery(M) and Dyspnea(D) domains in the experimental group, let mCF, mCE, mCM and mCD denote the means for the four domains in the control m mXE mXM mXD SDx SD XF 4 m mXE mXM mXD nx SD XF 4 group. Let x denote the treatment group (either E or C). We calculate the total variance for the average of treatment means across domains using the formula Now let mF, mE, mM and mD denote the mean differences for the four domains. We obtain the standard error of the average of the mean differences across domains using the formula m mXE mXM mXD Var XF 4 1 Var (mxi ) Cov mxi , mxj 16 i {F , E , M , D} i { F , E , M , D} i j SDE2 SDC2 m mE mM mD SE F 4 nE nC MID units formulas: Pooling MID Where we use that Cov xi , x j Var xi Var x j 0.5 Var xi Var x j standardized mean differences Assume that a trial reports a MD on some disease-specific HRQL instrument X, and assume that the minimally important different for instrument X, MIDX, has been established. We obtain the standard error by taking the squareroot of the variance The estimated MD is a random variable. If we standardize this random variable by dividing it by the MIDX, we get a new random variable, MD/MIDX. We know from basic probability theory that because MIDX is a mi MDi MIDB and Var mi Var MDi MIDB2 constant, the variance of MD/MIDX is given by MD Var 2 MIDX Var MD MIDX2 By defining the trial weights as wi=Var(mi)-1, we can use the fixed-effect model inverse variance method to pool the MIDstandardized mean differences using the That is, the variance of the mean difference formula divided by the square of the MID. Further, the standard error of MD/MIDX is given by Var MD SE MD MD SE 2 MIDX2 MIDX MIDX k k mˆ wi mi / wi i 1 i 1 Where m̂ denotes the pooled MIDstandardized mean difference. The standard error of m̂ can be calculated using the formula Now suppose a meta-analysis included k trials. The first j trials use disease-specific instrument A, and the last k-j trials use disease-specific instrument B. Let MDi denote the mean difference observed in trial i, let MIDA denote the minimally important difference established for instrument A, and let MIDB denote the minimally important difference established for instrument B. Further, let mi denote the MID standardized effect for trial i. To pool results across trials using MIDs we must first estimate the mi and its associated variance for all trials. For i=1, …, j we have mi MDi MIDA and Var mi and for i=j+1, …, k we have Var MDi MIDA2 k se(mˆ ) 1 / wi i 1 and confidence intervals can subsequently be derived. Pooling of MID-standardized mean differences is naturally extended to the random-effects model using weights wi=(Var(mi)+τ2)-1, where τ2 denotes the between-trial variance.