Purpose: In a randomised study (GM301; dacarbazine with/without oblimersen), patients with advanced melanoma were stratified based on performance status, metastatic site and lactate dehydrogenase (LDH). Progression-free survival and response and durable response rates showed a highly significant difference favouring dacarbazine-oblimersen and a nearly significant survival difference. All efficacy parameters significantly favoured dacarbazine-oblimersen in patients with normal baseline LDH [<= 1.1 x upper limit of normal (ULN)]. Each stratification factor was assessed for an interaction with treatment on survival and an interaction was detected only for LDH. Experimental design: Baseline LDH values in Study GM301 treatment groups were combined and analysed using cutoffs above and below 1 x ULN. Baseline LDH in EORTC study 18951 (dacarbazine, cisplatin, interferon-alfa-2b with/without interleukin-2 in advanced melanoma) was independently analysed using the same approach. In Study GM301, the relation between treatment effect and LDH, treatment effect and tumour size, LDH and tumour size and LDH and disease site were determined. Results: In Study GM301 (N = 760) and Study 18951 (N = 325), LDH was within the upper range of normal for a large number of patients. This was not exhibited in the general population, suggesting such values may be elevated rather than normal in melanoma. A highly ordered and monotonic relationship was apparent between LDH and survival: survival worsened as LDH became more elevated, even when LDH remained within normal range. LDH and tumour size were poorly correlated; elevated LDH was not associated with any one disease site. LDH was highly predictive of oblimersen effect. Conclusion: In designing studies, LDH should be considered, regardless of tumour size or disease site. (C) 2009 Elsevier Ltd. All rights reserved.