"Paradoxical" shortening of REM latency on first recording night in major depressive disorder: Clinical and polysomnographic correlates

Among 92 inpatients with major depressive disorders, variability in REM latency (RL) during the first two recording nights was assessed by means of an "adaptation coefficient" (AC = night 1 RL - night 2 RL × 100 ÷ mean RL for nights 1 and 2). Although mean RL was very similar for both nights (48.1 and 50.7 min), individual ACs showed a gaussian distribution [range: -176.5-171.4; mean: -1.2 (67.3); median: -4.4]. Forty-two patients (45.7%) exhibited shortening of RL on night 2 compared with night 1 (positive AC, corresponding to an "expected" evolution), 48 patients (52.2%) displayed the opposite pattern (negative AC, corresponding to a "paradoxical" evolution), and two patients had identical values on both nights. Extent of increase in RL from first to second night (i.e., extent of "paradoxical" evolution) correlated significantly with increasing duration of current episode, earlier age of onset, and poorer clinical response to tricyclic antidepressants. A cutoff AC of -7 correctly classified 62% of patients according to treatment response. The research diagnostic criteria-based retarded subtype of depression was associated with a more negative AC (i.e., a more "paradoxical" evolution), and the situational subtype was associated with a more positive AC (i.e., a more "expected" evolution) than the remainder of the sample. The subgroup with the most negative ACs (i.e., largest increases in RL from first to second night) also had longer sleep latencies, whereas the subgroup with the most positive ACs (i.e., largest decreases in RL) had higher REM activity and REM density as compared with the remainder of the sample. These results suggest that RL in major depression is not a static parameter and that the study of its within-subject variability can be helpful for diagnostic confirmation and prediction of treatment response. © 1985.