Re: "Adjusting survival curves for confounders: a review and a new method".

Nieto and Coresh (1) are reviewing alternative methods for presenting adjusted survival curves. The proposed new method is a special case of the Ederer II method currently used at cancer registries worldwide. In addition, there are other frequently used methods for adjusting for deaths and censorings that are not mentioned. It is important to state the context in which any adjusted survival curves are to be used. Age-adjusted survival curves are analogous to directly standardized incidence rates (2). These rates may be expressed as the weighted average of the age-specific survival curve for each study group, where the weights are chosen to be proportional to the age distribution of some external standard population. A term that helps to distinguish this type of adjustment from others and emphasizes that covariates other than age may be adjusted is externally covariate-adjusted survival curves. Adjusted survival curves may also be used to provide a graphic representation of survival obtained from a fitted regression model of survival such as the Cox model or Aalen's additive risk model (3). This curve describes survival for individuals with a particular set of covariates. A term that describes this type of adjustment is internally covariate-adjusted survival curve. Another type of adjusted survival curves is one that describes the survival probability of a randomly selected individual in the study group. It is obtained as the average of the individual survival curves where the individual survival curves are obtained from the fitted regression model. These survival curves are usually called directly adjusted survival curves. This adjusted survival curve is, in mathematical terms, similar to the method used in cancer survival analysis to evaluate expected survival curves (4). It is evaluated as the average of the individual survival curves where the individual survival curves are obtained from population mortality tables. Ederer et al. (5) noted that it might be reasonable to adjust the expected survival curve for deaths in the study group. The Ederer n definition of expected survival is