Microdissection of tissue sections from formalin-fixed, paraffin-embedded tumor material allows separation of microscopic sites within a sample. DNA can easily be extracted, and polymerase chain reaction (PCR) technology makes it possible to perform different molecular biologic analyses on small cell populations. The presence of normal cells or tumor heterogeneity may cause false negatives in allelic imbalance (AI) studies. Microdissected well-defined cell populations from a tumor section are assumed to increase the sensitivity of AI analyses. The present study has evaluated this in colorectal liver metastases by comparing genotypes in frozen biopsies with genotypes in microdissected archival samples from the same patients. Constitutional genotypes were obtained from corresponding peripheral blood leukocytes as well as normal liver tissue. Archival samples (n = 43) from 16 patients were analyzed after microdissection with 2-5 of 10 selected microsatellite markers. Frozen biopsies from one metastasis of each patient had previously been investigated at numerous microsatellite loci. From those results we selected, for the comparable analysis of archival samples, 41 tumor genotypes at 10 loci representing 11 heterozygotes, 13 AI, 7 losses of heterozygosity (LOHs), 8 homozygotes, and 2 microsatellite unstable cases. The microdissected samples revealed AI or a complete loss of one allele (LOH) in 5 of 11 (45%) genotypes that were previously evaluated as unchanged (retained heterozygosity) in the frozen biopsies, and LOH in 8 of 13 (62%) genotypes at loci known to exhibit AI in the frozen biopsies. Microsatellite instability, LOH, and homozygosity found in the frozen samples were all confirmed by analyses of the archival material. Intertumoral genetic heterogeneity was found in samples from two patients. The same allelic intensities were seen in DNA from tumor-close liver tissue as in blood DNA from the same patient except in one sample. The present study shows a 54% increase in sensitivity of genetic alterations if pure tumor cell components are used (five "new" AIs and LOHs and eight "new" LOHs among previously scored heterozygotes [n = 11] and AI [n = 13], respectively). In total, a 93% success rate (108/115 analyses) was obtained using standard PCR conditions for the 10 selected markers. The fact that standard PCR conditions and 5-micron tumor sections are used shows how easy these analyses are to perform, and that only minor amounts of valuable archival material is used.