Nocturnal oxygen saturation and painful sickle cell crises in children

Painful crisis is the most common event in children with sickle cell disease (SCD), but the pathogenesis remains poorly understood. Acutely, precipitating factors may include cold, dehydration, infection, and hypoxia, and there are epidemiologic data suggesting that genotype, early dactylitis, leucocytosis, hemoglobin, and hemoglobin F are risk factors for frequent pain.1-4 

There is increasing evidence that pulse oximetry measurements in sickle cell disease reflect arterial oxygen saturation.5Nocturnal desaturation, measured using pulse oximetry, is common in SCD, possibly secondary to upper airway obstruction (UAO).6-10 We have previously published the results of a cohort study investigating the relationship between nocturnal hypoxemia and central nervous morbidity9; we present here the effect on pain as a secondary end point.

The cohort of 95 patients attending the sickle cell clinic at the Queen Elizabeth Children's Hospital in London was studied from January 1, 1991, to April 30, 2000. The hospital ethics committee approved the study, and written informed consent was obtained. Painful crises requiring hospital treatment (emergency department and/or admission) were recorded. A pain rate (hospital days for pain per years of follow-up) was calculated for the whole follow-up period or until death (2 patients) or the initiation of regular transfusion (14 patients) or hydroxyurea (1 patient). Data were recorded on risk factors previously reported in the literature,3 4including dactylitis in the first year of life, mean hemoglobin (early Hb) and white blood cell count (early WBC) from the second year, baseline hemoglobin and hematocrit (Hb, Hct, at sleep study), fetal hemoglobin levels (HbF, only homozygous sickle cell disease [HbSS]), and timing of adenotonsillectomy if performed.

A pulse oximeter (Ohmeda Biox 3700, Hatfield, United Kingdom) was used to record oxygen saturation (SaO₂) continuously during sleep, and the data were analyzed blind to the pain data. Sixty-three sleep studies took place in the patients' home; the remainder took place in the Sleep Laboratory at Great Ormond Street Hospital. Studies lasted from 4.3 to 8.2 hours (median, 7.5 hours). Movement artifact was excluded manually. We examined the data for mean and minimum SaO₂ and percentage of sleep spent at SaO₂ less than 90% and less than 80%. We also classified the sleep studies as with or without significant transient or prolonged dips (more than 4% from baseline) in oxygenation associated with acute pulse rate rises, suggestive of obstructive sleep apnea.

The mean follow-up period from the time of the sleep study was 4.6 years (SD, 2.85 years). The mean pain rate was 2.94 days (SD, 3.8 days) per year; 9 of the 95 patients had no pain episodes recorded. There were 2 sickle-related deaths, and 6 patients had a stroke. Chest syndrome was recorded in 12 patients, with a range of episodes per individual of 0 to 8 events. Table 1shows the baseline characteristics and univariate analysis of the relationship between the risk factors and the average number of pain days per year in the study cohort. Lower baseline Hct and Hb were significantly associated with increased pain rate, as was HbSS genotype. There were significant associations between a decreased pain rate and higher mean SaO₂ (average fall, 0.75 [95% confidence interval (CI), 0.58, 0.92] per unit increase in SaO₂), higher minimum SaO₂, and decreased percentage of sleep study with SaO₂ less than 80% or less than 90%. There were no significant associations with the other variables. Multiple linear regression was used to quantify the association between overnight SaO₂ measurements and pain, both before and after adjustment for potential confounding variables. Of the 5 overnight measurements, mean saturation was most strongly associated with average pain. After accounting for variability in mean saturation, only minimum saturation remained of borderline significance (P = .051) when modeling the variation in pain rates. After adjustment for genotype (SC, SS, Sβthal), dactylitis before 1 year of age, baseline Hb, HbF, early Hb, and WBC, mean saturation remained significantly associated with pain rate (P < .000 05). The presence of dips was additionally significant within this model (P = .007). An increase of one unit in mean SaO₂ was associated with an average fall of 0.83 (95% CI, 0.58, 1.07) in the number of days of pain per year (P < .000 05). The presence of dips was associated with 2.97 (95% CI, 0.85, 5.10) more days of pain on average per year. The results were similar when adjusted for hematocrit in the 79 patients for whom it was available.

For patients who underwent adenotonsillectomy (n = 28; 30%), there was a decrease in the number of days in pain from the year before to the year after surgery (3.75 vs 2.64, a difference of −1.11 [95% CI, −2.68, 0.46] days per year), which did not reach statistical significance (P = .16, paired t test). Nine of 27 patients who underwent adenotonsillectomy had sleep study data collected postoperatively as well as preoperatively because of persistent symptoms, at a median time postoperatively of 4.5 years (range, 0.5-7 years). Six still had significant dips in SaO₂ as defined above, and mean saturation fell from an average of 94.8% to 92.9% (difference, −1.9% [95% CI, −5.9%, 2.0%], P = .29). Interestingly, if the results are split into patients undergoing adenotonsillectomy for obstructive sleep apnea (OSA; n = 11) or for repeated infections (n = 17), the OSA patients do not show any correlation with frequency of painful episodes, but there was an association for those for whom the indication was recurrent infection (P = .02).

In our study, low nocturnal oxygen saturation appears to be highly significantly associated with frequent painful vasoocclusive crisis in SCD, perhaps because it is one of the factors involved in precipitating hypoxic-ischemic pathology in bone marrow. Interestingly, lower hemoglobin was associated with more frequent painful crises in these children, in contrast to the previous literature but in line with our finding that low oxygen saturation is associated with anemia.11 It is possible that the paradoxical association of higher hemoglobin/hematocrit3 and oxygen saturation4 seen in adults with frequent painful crisis is not manifest in early childhood, perhaps because it depends on the degree of adaptive acclimatization to chronic hypoxemia; future longitudinal studies may elucidate this. Increased viscosity as measured by hematocrit was significantly associated with an increased pain rate.

Our data require confirmation in other populations, because there are sources of bias. This is a hospital-based rather than a birth cohort, which may explain the high proportion of children with at least one painful episode requiring hospital attendance. We did not attempt to measure the experience of pain at home, which has considerable methodologic difficulties. At the time of the study, most families in the local community used Queen Elizabeth Hospital for primary care, but with changes in the configuration of health services children are now more likely to be managed at home, and there are now validated methods of assessing pain outside the hospital. We did not have the resources to undertake rigorous reproducibility testing of our home sleep studies. Because the mean saturation is a better predictor of frequent pain than frequent dips, it may be possible to screen the population using a short period of daytime pulse oximetry, and we are at present investigating the reproducibility of clinic measurements and the correlation with overnight saturation monitoring.

Although UAO is relatively common in the sickle population it may not be the major cause of nocturnal hypoxemia, and in this study adenotonsillectomy did not significantly reduce the number of painful crises. Indeed, the main effect of adenotonsillectomy was seen in patients where the indication for operation was recurrent tonsillitis, suggesting that it is the absence of infection that may be the benefit and that this is the real risk factor in this group.12Hypoxemia may persist around the clock,9 perhaps secondary to mechanisms other than UAO and chronic anemia. The size of the airway may be determined in early life, perhaps explaining why adenotonsillectomy does not always improve the oxygen saturation.13 Nocturnal hypoxemia might precipitate vasoocclusion in the lung as well as the bone marrow, with progressive parenchymal lung injury and poor gas exchange14; the investigation of this type of vicious cycle would require a longitudinal study with repeated lung function testing. However, it has long been postulated that hypoxia is a physiologic precipitating factor in the formation of the sickled cell, and it is the basis for avoidance of hypoxic environments, for example, in anesthesia and high-altitude travel.15 The possibility that appropriate management of chronic hypoxemia reduces the incidence of stroke, acute painful crises, and death in children with SCD will be examined in the planned controlled trial, funded by the Stroke Association, of overnight oxygen supplementation.