Can we predict who hurts most after pediatric adenotonsillectomy?

Adenotonsillectomy is one of the most common operations we perform in children. It is also one whose pain we manage least well once the child goes home, where it can run into the second postoperative week and is easy to underestimate. In this editorial, we review the recent article by Frigon et al. “Association between obstructive sleep-disordered breathing severity and pain trajectories after adenotonsillectomy in children: a prospective cohort study” (1). Their study uses data generated from an initial sample intending to validate the Parents’ Postoperative Pain Measure (PPPM) as sampled twice daily using at-home pain diary assessments in patients ages 2–12 between postoperative days (POD) 0–14. The current analysis explores whether obstructive sleep-disordered breathing (oSDB) severity is associated with pain severity and analgesic use after extracapsular electrocautery adenotonsillectomy. Using data from 124 patients with completed PPPM data, the authors performed multiple imputation by chained equations (MICE) followed by group-based trajectory modeling (GBTM) to describe postoperative pain experience trajectories. The authors settled on three group-based trajectory cohorts by PPPM score including mild pain (scores of ≤6), moderate pain (scores of 6 and ≤10), and severe/prolonged pain (scores of 10–15). Their study reports two main conclusions. First, a higher preoperative modified Snoring, Trouble Breathing, Un-Refreshed (mSTBUR) score predicted the severe trajectory. Second, the severe group was more likely to require an unscheduled revisit/return to hospital.

The question of post-discharge pain prediction is important, particularly for the otolaryngologist who is tasked with prescribing analgesics and managing post-discharge pain once the recovery from anesthesia is complete. Most of what we know surrounding the at-home pain course comes from group averages, usually from a single mean curve that describes no actual patient. Trajectory modeling does something different. It pulls the population apart into the patterns and variations we recognize in clinical practice: most children recover predictably and uneventfully; some have limited pain while a minority experiences pain disproportionately and often requires additional care. However, to explore how covariates influence membership in a group, the variables impacting the outcome of interest (in this case, at home pain) should be similar.

Severe obstructive sleep apnea (OSA) is the signature concern for fatal opioid-induced respiratory depression following pediatric adenotonsillectomy (2). This concern was largely due to codeine-induced fatalities in children with oSDB with genetic polymorphisms of CYP2D6 activity resulting in ultra-rapid metabolism (3). However, recurrent exposure to intermittent hypoxia has been shown to alter opioid-receptor expression in the brainstem (4,5). This begs the question; does this change in receptor expression result in heightened opioid sensitivity (6,7)?

Interestingly, much of the initial concern surrounding the potential sensitivity to opiates stems from the authors’ previous publication suggesting recurrent hypoxia in children to be associated with increased analgesic sensitivity to opiates (6,7). That study assessed the number of opioid doses administered in the post anesthesia care unit (PACU) based on pre-operative SpO₂ nadir on sleep study. The study included a small sample size of 22 patients under a non-standardized anesthetic regiment with grouping drawn from another small, non-uniform prior study of theirs (6). They observed that the group with more severe intermittent hypoxia by pre-op sleep study required fewer opioid doses in the PACU, concluding that intermittent hypoxia resulted in increased sensitivity to opiates (6). However, this observation did not account for intra-operative anesthetic differences in analgesics and attempts to directly correlate preoperative hypoxia severity with pain and sensitivity, a practice they repeated in the present study. Though the authors note that physiologically, repeated drops in oxygen may increase opioid-receptor expression in the brain and heighten opioid sensitivity, multiple subsequent studies in both adults and pediatrics directly measuring respiratory parameters have not demonstrated increased respiratory sensitivity in response to opiates based on OSA severity (8-11). Certainly, opioids impact respiratory rate and tidal volume, however the magnitude of which does not appear to be influenced by the severity of OSA (8).

With regard to pain influence, long-term exposure to intermittent hypoxia in the brain and fragmented sleep as occurs in OSA have been observed to be pro-inflammatory and may even result in hyperalgesia (4,12-14). The authors also refer to concerns based on their previous reporting of OSA severity being a risk factor for respiratory depression due to opioids (1). In the study, Frigon et al. take the approach that oSDB serves as a possible marker for pain severity in addition to a risk for opioid mediated respiratory depression (1). This dichotomy presents a therapeutic paradox; how to approach patients if severe oSDB imparts a greater risk for pain while simultaneously placing these same children at greater risk for respiratory depression from opioid administration. For anyone who manages these children after discharge, that reframing is worth exploration. Can we tell in advance which child will need more, and give it without the drugs that oSDB makes most dangerous? The trouble is that these data answer that question only tentatively. Several features of the analysis argue against changing practice on the basis of this publication. The two major issues is the use of the mSTBUR questionnaire to stratify oSDB severity and secondly, the early postoperative pain scores used to derive the trajectory cohorts.

Start with the explanatory variable: obstructive sleep-disordered breathing. The authors attempt to determine whether the severity of oSDB is associated with the pain trajectory for the initial POD 10. In Frigon et al. study, the majority of patients did not undergo a formal preoperative polysomnographic study (PSG), rather, the authors conducted the analysis based on parental responses to the mSTBUR questionnaire (15,16). While the STBUR is a validated tool, its purpose is to screen for those at risk of perioperative respiratory adverse events (PRAEs) based on 5 yes/no questions related to symptoms suggestive of sleep disordered breathing (15,16). However, the STBUR is designed to assess risk of PRAEs based on symptoms of oSDB and does not correlate with the degree of severity of OSA (15,16). Using a yes/no screen as a graded severity scale and a pain predictor, asks much more of it than its validation supports. The authors’ own data show this strain. The mSTBUR did not correlate with the oximetry-based severity metric (ρ=–0.045; P=0.75), which is hard to reconcile with using it as a proxy for physiologic disease. Further, on univariable analysis it did not separate the mild trajectory from the severe one [odds ratio (OR) 0.90; 95% CI: 0.72–1.14; P=0.38] (1). Significance emerged only after the model was adjusted with an age-by-mSTBUR interaction term (OR 0.52; 95% CI: 0.27–0.97; P=0.04) (1). Though this is defensible as the effect does seem concentrated in the youngest children, it demonstrates how weak the signal is. For the average child in the study, the effect was minimal, depending on a handful of toddlers in the severe group. A signal that appears only under adjustment is a hypothesis, not a triage tool. Perhaps the most important fact is the minimal difference between trajectory groups with 73% of patients in Trajectory 1 falling into mSTBUR scores of 3–5 compared with 74.5% in trajectory 3 (1). Group membership based on mSTBUR score is highly limited by sample size even after adding simulated patients for the analysis. Therefore, whatever the STBUR is worth as an anesthetic risk screening assessment tool, it is not a substitute for polysomnography as a measure for OSA severity, and certainly not a modified unvalidated version. Given the small sample and the use of mSTBUR rather than a measure of actual severity, this paper’s title and conclusion that OSA severity is associated with pain is an overreach.

There is also the question of what the trajectories really are. The three groups are separated almost entirely by where they start with respect to analgesia management and how they were managed. In tracing out these trajectories, the groups started with vastly different constants in mean PPPM scores on postoperative day 0 (4.52, 8.60, and 10.54) in trajectories 1–3 respectively (1). This was followed by an observed maximum mean PPPM scores on postoperative day 1 in all trajectories (4.58, 8.64, and 10.96, for trajectories 1–3 respectively) (1). In other words, a child’s trajectory is mostly a restatement of how much pain they experienced within the first 24 hours postoperatively. Further, because the POD 1 score is one of the inputs used to build the trajectory models, using it to “predict” the trajectory is nearly circular. That children in severe pain early on (POD 1) tend to stay in pain is intuitive but does not represent a finding about sleep-disordered breathing. Their supplemental Tab. 1 highlights the vast difference in both opioid and non-opioid administration for patients on the day of surgery between groups with those in trajectory 1 receiving significantly greater doses of nearly all opioid and non-opioid analgesics when compared with patients in trajectory 2 and especially in trajectory 3 (1). While this may represent deliberate caution in children with known severe oSDB, the difference in non-opioid analgesics would suggest otherwise. Similarly, treatment and disposition were different across trajectory groups; and moved with the outcome. The severe group (trajectory 3), by contrast, received the least amount of opioid intraoperatively [102.2 µg/kg intravenous morphine equivalents (IVME)] compared with 113.3 and 124.5 µg/kg IVME for trajectory groups 1 and 2 respectively. This same group (trajectory 3) required the most opioid in the recovery room and were admitted overnight far more often (57.4% admitted in trajectory 3 compared with 31.5% and 28.9% and in trajectories 1 and 2 respectively). At home, among children given oral morphine, about half the severe group got ≤0.05 mg/kg. The authors fairly and appropriately note this, referring to the analgesic analysis hypothesis-generating, yet the findings were carried through to the paper title and conclusion. And from the surgical side, this cohort’s analgesic approach (where 2/3 of patients received a home morphine prescription) is inconsistent with current North American practice of opioid avoidance post-discharge (17). Thus, the severe group’s “under-treatment” may be an artifact of how these children are managed in this system.

In total, a more pointed hypothesis resulting for these data may suggest that early or perhaps preemptive treatment of pain is most predictive of trajectory? This concept of preemptive multimodal analgesia for improving subsequent pain outcomes in adenotonsillectomy is not new (18).

Pain after any procedure varies widely between patients, for reasons that are unclear and likely multifactorial: genetic differences in analgesic response and inflammation, surgical technique, activity, and coexisting diagnoses such as anxiety, etc. This raises the question of whether imputation is appropriate for a pain analysis (14). There is a small yet decent-sized cohort that actually completed data – then a simulated data set that imputes patient experience from a non-uniform sample of patients with uncontrolled anesthetic and post-discharge regimens. Their supplemental Tab. 6 shows enormous variability between patients in analgesic use (1). In 13.2% of patients, no analgesics were used, while 19.0% used morphine, acetaminophen and ibuprofen in 35.6%, acetaminophen alone in 17.8%, acetaminophen and morphine in 7.5%. The data itself were derived from only 124 of 245 diaries (51%) completed initially. The model relies on a simulated dataset in which 941 of 4,900 PPPM scores (~19%) were filled in by MICE, on top of a sample managed with non-standardized anesthetic and discharge regimens. MICE is certainly a sophisticated response to survey-based data. Multiple imputation assumes that values are missing at random. However, if parents stopped recording precisely when pain was at an extreme (either gone and no longer worth noting, or so severe it was overwhelming), that assumption fails. Imputation then pulls missing values toward the average, blunting the extremes the trajectories exist to capture (19). Moreover, our own prior work, shows that parental education and socioeconomic status strongly affect the assessment and treatment of pain after discharge, which in this case may represent another source of non-random missingness (20,21). With half the diaries incomplete, these assumptions are limiting. In this case, the study derives the groups from the pain scores, which are also the outcomes. So, making predictions of associations and downstream odds ratios based on this invite circular inference (22). This is noted in the revisit rates. The severe group shows a nearly 11-fold odds of a revisit (OR 10.5; 95% CI: 3.1–35.7). However, this comparison relies on only 32 visits from the entire study population. More limiting is the comparison of readmissions. Only 11 out of 245 (4.4%) patients studied were readmitted, limiting reasonable estimate of a difference based on any explanatory variable. Further, the severe group is defined by the outcome, in this case, the pain they had on POD 1, making these variables not independent. The direction of association is believable – children in more pain come back more often. But an odds ratio with a wide interval of plausible values suggests a highly imprecise relationship.

This is an important question to address, predicting which patients will have a greater pain experience. Future studies would benefit from standardized post-discharge analgesic protocols to inform practice based on specific treatment strategies with less variability. While the question of pain and sleep apnea severity is important, a study stratifying patients using formal polysomnography results would be helpful. However, given that the vast majority of patients do not undergo formal polysomnography testing nor is there a suitable alternative to determining OSA severity without polysomnography, presents an inherent limitation. At best, it would be beneficial to know if patients with OSA experience pain after adenotonsillectomy (AT) differently. However, use of non- polysomnographic determinants as surrogates are likely of limited clinical and research potential. That said, where polysomnography is impractical, a standardized and at least partially validated instrument, such as the Pediatric Sleep Questionnaire (PSQ) or the OSA-18, would still provide a more consistent and better-supported baseline estimate of sleep-disordered breathing severity than the mSTBUR, though it would remain an imperfect surrogate.

In conclusion, while the authors’ conclusion that “children with severe oSDB (higher mSTBUR scores) were more likely to have severe prolonged pain and to return more often to hospital” may not be reflected within the data, Frigon et al. do attempt to address an important question that belongs to both of our specialties of pediatric anesthesiology and otolaryngology: how do you anticipate and safely treat the pain of a child whose sleep-disordered breathing may signal more pain and more vulnerability to its treatment?

Oddly, the group that raised the initial concern surrounding opioid sensitivity and concern for respiratory depression in those with OSA continued to have post-discharge morphine prescribed. Fortuitous or not, the study highlights two things worth carrying forward: the importance of early treatment of pain at time of surgery and the burden of undertreated pain after discharge.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.