ABSTRACT. We have tested the hypothesis that autonomic instability, reflected in increased variability of heart rate and respiratory frequency, characterized high risk babies who died of sudden infant death syndrome. Using computer-based methods, we compared the power spectra of instantaneous heart rate and respiration on coded tape recordings from seven asymptomatic siblings and 10 babies with symptomatic apnea who died of sudden infant death syndrome to 34 age-and sex-matched controls. We confirmed our previous observation of increased respiratory bandwidth, an index of variability in respiratory frequency (p = 0.009) but failed to confirm our finding of increased low frequency fluctuations in heart rate (p = 0.18). In addition, we found an increase in mean respiratory frequency during quiet breathing (p = 0.001) and a significant relationship between respiratory bandwidth and mean respiratory frequency (r = 0.604,p = 0.0002). These variables along with those from a previous analysis of the same data base yield a discriminant function with 82% sensitivity and 100% specificity. These results confirm previous suggestions that high risk babies who die of sudden infant death syndrome exhibit autonomic instability. (Pediatr Res 22: 158-162,1987) Abbreviations RBW, respiratory band width SIDS, sudden infant death syndrome ROC, receiver operating curve HR, heart rate PB, periodic breathing Hz, Hertz (cycles/s)We have hypothesized that the at-risk state for sudden and unexplained death in infants (SIDS) is characterized by abnormal patterns of fluctuations in heart rate and respiratory activity (1, 2). In our first test of this hypothesis, we compared fluctuations in instantaneous heart rate and respirations using power spectrum analysis and identified two abnormalities that were statistically different in high risk babies who died of SIDS compared to controls (2). SIDS babies exhibited increased fluctuations in heart rate at frequencies of 0.02-0.10 Hz and increased variability of respiratory frequency. However, in a blinded trial of this hypothesis, using recordings obtained prospectively by Gordon et al. (3) on English babies, we were unable to identify SIDS Received June 10,1986; accepted February 26, 1987. Address correspondence to Dan iel G. Shannon, M.D., Pediatric Pulmonary Unit, Massachusetts General Hospital, Boston, MA 02114. babies among control s by either increased low frequency HR fluctuations or increased variability of respiratory frequency. We concluded that the differences between these two studies might be explained by differences in methods or most probably in the populations. In order to test the initial hypothesis, we have repeated the analysis of data from our high risk population using a larger, blinded and coded data set in order to avoid any possibility of bias. The coding was performed by a technician who was unaware of the past history of the subjects and of the results ofa pneumogram analysis of recordings obtained with an independent software program (4). Using the latter program ,...