Because there are little satisfactory data on change in severity of diabetic polyneuropathy (DP) over time from study of population-based cohorts of diabetic patients in epidemiologic surveys of DP, it is difficult to predict outcome or morbidity or to identify risk factors; it is also difficult to estimate statistical power for use in controlled clinical trials. In this longitudinal study of almost 200 patients from the Rochester Diabetic Neuropathy Study (RDNS) cohort, we assess which symptoms, clinical examinations, tests, or combinations of examinations and tests (composite scores) are best used as minimal criteria for the diagnosis of DP and as a quantitative measure of severity of DP. An abnormality (> or = 97.5th percentile) of a composite score that included the Neuropathy Impairment Score of the lower limbs plus seven tests (NIS(LL)+7 tests), was a better minimal criteria for DP than clinical judgment alone or previously published minimal criteria. First, it provided a more comprehensive assessment of neuropathic impairment. Second, it avoided the overestimated frequency of DP when the minimal criteria for DP was any one or two abnormalities from multiple measurements. Minimal criteria using nerve conduction and reduced heart beat response to deep breathing identified approximately twice as many patients with DP than did clinical examination and vibration detection threshold using CASE IV. This difference could be used to subclassify state 1 DP. Although various individual measures of DP, for example, vibration detection threshold (as evaluated by CASE IV and the 4, 2, and 1 stepping algorithm [see text]), were good measures of worsening, the composite score NIS(LL)+7 tests (assessing neuropathic impairment) was much better at showing monotone worsening. Using this composite score, the average diabetic patient in the RDNS worsened by 0.34 points per year, whereas patients with diabetic polyneuropathy worsened by 0.85 points per year. On the assumption that a therapeutic agent may prevent worsening of DP but not cause improvement, controlled clinical trials of patients with DP would need to be conducted for a period of 3 years to achieve a meaningful change of 2 NIS points (the level of abnormality considered by a Peripheral Nerve Society consensus group to be clinically meaningful).