Is there a way to combine stratified survival curves? Suppose in publications we see stratified parametric survival curves

Is there a way to combine the survival estimates of the left ventricle and right ventricle into a combined survival estimate (just 1 survival curve)? Suppose I know how the survival functions for each group is specified but do not have the raw data, is there a way to combine the two?
 A: For non-parametric Kaplan Meier curves: you need to know the denominator and numerator of failures at each failure time, which is equal to just having the complete survival information anyway, making that a moot point.
For a general survival curve, you could take a harmonic mean based on the sample size. So if arm one has n=100 participants with 80% survival at time t and arm two has n=200 participants with 60% survival, the combined survival is 300/(100*1/80 + 200*1/60) = 65% survival. This approach only works when the proportional hazards assumption is met (the key assumption of the Cox model). This is also the approach that gives the predicted survival conditional upon the failure times in the logrank test, aka the stratified Mantel Haenszel which @MichaelChernick refers to.
The smoothness of the graphic you present suggests that they used a parametric survival model, something like a Weibull where the log hazard decreases linearly. If they report the parameters from the stratified survival model (like a scale and shape), you can write down the mixture likelihood for the two arms and marginalize that likelihood to obtain the marginal survival (which you call combined survival). This is, of course, just as easy to do computationally with numerical integration.
