The work of Rosansky et al1 suggesting that early-start hemodialysis may be harmful is an important contribution to the debate and may well prove a seminal article. A potential flaw, though, is the introduction of survivor bias brought about by choosing to calculate patient survival from the start of hemodialysis in each of the estimated glomerular filtration rate (eGFR) strata rather than from a common eGFR starting point. The difference in the proportion of patients who died in the 2 years following initiation of dialysis between a higher (early-start) and lower (late-start) eGFR strata may simply be the number of patients who died prior to initiation of dialysis in the lower strata with an eGFR at time of death between that of the higher and lower strata. Future, preferably prospective, studies need to take this into account.
Pickering JW. Survivor Bias in Early- vs Late-Start Hemodialysis Studies. Arch Intern Med. 2011;171(5):477-478. doi:10.1001/archinternmed.2011.55