Vaso-occlusive crisis pain intensity, frequency, and duration: which best correlates with health-related quality of life in adolescents and adults with sickle cell disease?

Ours was a cross-sectional analysis of HRQOL measures collected on older adolescents and adults with SCD at baseline of a randomized clinical trial investigating the use of community health workers called patient navigators to increase hydroxyurea compliance. Details of the trial have been published[ 48 ]. All patients received care at one of 6 sickle cell clinical specialty sites (5 in Virginia, one in North Carolina). Subjects were enrolled from March 2014-January 2018, had to have SCD, be at least 15 years of age, of HB SS or Sbeta0 genotype, not pregnant, not being treated on a formal chronic transfusion protocol, and be eligible to be on hydroxyurea (HU). Sample size was dictated by power needed to achieve aims for the clinical trial. The study was approved by the Institutional Review Board at each of the study sites. HRQOL was measured using two systems of measures – the Patient-Reported Outcomes Measurement Information System (PROMIS)[ 27 ] and the Adult Sickle Cell Quality of Life Measurement Information System (ASCQ-Me)[ 28 ]. PROMIS is a generic set of measures that has been used to evaluate patient reported HRQOL among children and adults with various chronic conditions, such as SCD, and disabilities as well as among the general population. ASCQ-Me is a SCD disease specific measure that assesses HRQOL; however, the items and domains covered are tailored to experiences unique to the SCD population. All HRQOL measures were converted from raw scores to standardized scores (T-score) which ranged from 0 to 100, with a score of 50 (and SD=10) representing the average of the general US population (PROMIS)[ 27 ] or the average SCD adult (ASCQ-Me) [ 42 ]. The PROMIS scales used in the current study have exhibited good psychometric properties for pediatric and adult populations [ 11 , 51 ], including in individuals with SCD [ 7 , 27 ]. The ASCQ-Me was developed to assess HRQOL in individuals with SCD, and all scales used in the current study have exhibited good psychometric properties in adults with SCD [ 27 , 28 ]. Cronbach alphas for PROMIS and ASCQ-Me in the current study ranged from 0.79 to 0.96. Of note, the current study is the first to use ASCQ-Me in a sample below age 18. According to the ASCQ-Me developers, “ASCQ-Me questions could be appropriate down to the age of 15” [ 1 ]; the Cronbach alphas for the 15–17-year-old subsample in this study ranged from 0.72 to 0.95, well within the acceptable range. The HRQOL measures used in the current study all fell into three primary domains: psychosocial, physical, and pain expression. For most scales, higher scores imply better functioning; the exceptions are the PROMIS fatigue , sleep related impairment and pain behavior scales. The psychosocial domain measures assessed emotional and social health aspects. From PROMIS, we used two scales: the global mental health scale and the satisfaction with social role scale. From ASCQ-ME, we used two scales: the emotional impact scale and the social functioning impact scale. The physical domain measures assessed movement, physicality, and impacts on waking and sleeping hours. From PROMIS, we used 3 scales: the physical functioning scale, the fatigue scale, and the sleep related impairment scale. From the ASCQ-ME, we used 2 scales: the stiffness scale and the sleep impact scale. The pain expression domain assessed actions and behaviors associated with being in pain. From the PROMIS, we used the pain behavior scale, and from the ASCQ-ME, we used the pain impact scale. SCD pain was self-report, characterized by pain intensity, frequency and duration as evaluated by the ASCQ-Me Pain Episode Frequency and Severity questionnaire. The frequency of “pain attacks” (also referred to as “crisis” on the form) was defined as the number of pain attacks/crises over a 12-month period (0, 1, 2, 3, or 4 or more attacks). Pain attack intensity was described on a scale from 0=no pain to 10=worst imaginable pain of the patient’s most recent pain attack. Duration of the most recent attack/crisis covered times from minutes to weeks (less than an hour, 1–12 hours, 13–23 hours, 1–3 days, 4–6 days, 1–2 weeks, >2 weeks). While “pain attack” was chosen to be used on the ASCQ-Me form, VOC is the prevalent term and will be used here. SCD severity was assessed using the ASCQ-Me Medical History Checklist (MHC)[ 26 , 28 ]. The MHC scored patients based on SCD induced organ damage, ulcers, daily use of pain medication, and regular (unplanned) blood transfusions. We excluded the item pertaining to use of pain medications because of its potential direct correlation with the VOC pain measures. The analysis sample was described with means (standard deviation), median (interquartile range) and frequency (percent) as appropriate. Spearman correlations were calculated between HRQOL measures and each of the three VOC pain measures. To check for potential bias due to subject exclusion, the analysis sample was compared to those excluded using Chi-square tests and Wilcoxon rank sum test as appropriate. Missing values for MHC were imputed using a fully conditional specification regression method with the final imputed baseline value being the mean of M=10 replications. Regressions of the HRQOL measures in a model simultaneously including all three of the VOC pain measures, controlling for demographics (age, sex) and clinical severity (genotype, MHC score) were carried out. Partial R 2 s, which reflect the added influence of VOC intensity, frequency and/or duration on each of the HRQOL measures, over and above demographics and clinical severity were reported. Statistical significance of the regression coefficient (B) and partial R 2 was based on a t-test. Canonical correlation was used to estimate the partial R 2 when considering all 3 pain measures simultaneously. Since pain profiles may change with age, we performed an analysis testing whether results varied with age. Interaction terms between age group and VOC frequency, intensity and duration were added to regression models for each HRQOL outcome, and we tested whether these interaction terms were statistically significant. We used previously defined age categories: <26 (n=101), 26–36 (n=54) and ≥37 (n=43)[ 33 ]. Study data were collected and managed using REDCap (Research Electronic Data Capture) electronic data capture tools hosted at Virginia Commonwealth University [ 23 , 24 ]. Analysis was performed using SAS 9.4 [ 44 ].

224 subjects were enrolled and randomized into the clinical trial, but 6 of these were later found not to qualify for the clinical trial and have been excluded from all study reports (5 had the wrong genotype and one had been on chronic transfusions at the time of enrollment). Of the 218 subjects qualified to be in the main study, 20 were missing the VOC pain measures (16 missing all 3, 4 only missing duration) and thus were excluded from analysis. A further 14 subjects were missing the MHC score (10 had no MHC data items at all, while an additional 4 participants were each missing a single item). After imputing the MHC scores the analysis data set consisted of n=198 observations. Subjects excluded from the analysis data set due to missing VOC data were of similar age, sex, genotype, education, household income, clinical site, marital status and HU status. Since participants did not always complete every HRQOL survey item, analysis of any particular HRQOL proceeded using all subjects with a computed scale value for that HRQOL. A description of the sample is in Table 1 . Subjects averaged 29 years of age, with 51% being 25 years and younger, 27% between ages 26 and 36 and 22% older than 36 years old. Fifty two percent were female and most were not married. Around 12% had not finished high school, while almost 48% had at least some college. Almost 42% had household income less than $10,000 while 16% had income $50,000 or more. Over 80% were on HU and 88% were of SS genotype. Subjects reported from 0 to 5 health problems on the MHC (median of 1), with hip or shoulder damage (25.5%), regular (unplanned) transfusions (24.6%) and spleen issues (21.4%) endorsed most frequently. Mean scores for the HRQOL measures ranged from 44.9 for physical function to 59.9 for pain behavior. Figure 1 shows the distribution of the VOC measures. VOC pain intensity averaged 7.4, with 8.6% reporting an intensity score of 3 or less, while 59% reported intensity scores of 8–10. Almost 15% reported no VOC over the past 12 months, while 39.9% reported four or more. Median duration of episodes was 1–3 days, with 6.1% reporting their VOC lasting more than 2 weeks. Table 2 has Spearman correlations between each of the three VOC pain measures of frequency, intensity and duration and the HRQOL measures. VOC frequency is statistically significantly correlated with all the HRQOL measures except sleep related impairment , with the smallest correlations for satisfaction with social roles (−0.20) and sleep impact (0.22), and largest correlations with emotional impact, social functioning impact and pain impact (correlations: −0.36, −0.36, −0.37 respectively). VOC intensity is statistically significantly correlated with all HRQOL measures except global mental health and satisfaction with social roles , with lowest correlations with sleep related impairment (0.14) and highest with pain behavior (0.30). VOC duration is similarly not statistically significantly correlated with global mental health and satisfaction with social roles as well as sleep related impairment . Lowest correlations are with pain impact (0.16) and highest with stiffness (−0.28). Table 3 shows the added influence (partial R 2 ) of VOC frequency, intensity, and duration on explaining the variation of each of the psychosocial function measures. VOC frequency was statistically significantly associated with all of the psychosocial measures except satisfaction with social roles . Partial R 2 for the statistically significant HRQOL measures ranged from 3.83% ( global mental health ) to 12.49% ( social functioning impact ). Neither VOC intensity nor duration were associated statistically with any of the psychosocial HRQOL measures, when the other VOC variables were in the model. Partial R 2 for VOC frequency, intensity and duration taken together ranged from 4.06% for satisfaction with social roles (not significant) to 17.36% for social functioning impact. Table 4 shows the added influence (partial R 2 ) of VOC frequency, intensity, and duration on each of the physical functioning HRQOL measures. VOC frequency was statistically significantly associated with all of the physical functioning measures except for sleep related impairment . Partial R 2 for the statistically significant HRQOL measures ranged from 2.82% ( sleep impact ) to 6.15% ( fatigue ). VOC intensity was not statistically significantly associated with any of the physical functioning HRQOL and VOC duration was only associated statistically with stiffness (partial R 2 =3.38%). Partial R 2 for VOC frequency, intensity and duration taken together ranged from 3.2% for sleep related impairment (not statistically significant) to 13.15% for fatigue . Table 5 shows the added influence (partial R 2 ) of VOC frequency, intensity, and duration on pain expression. VOC frequency was statistically significantly associated with both pain expression variables (partial R 2 =9.56%, 6.40% for pain impact and behavior respectively). Neither VOC intensity nor duration were statistically significantly associated with the pain expression HRQOL measures. Partial R 2 for VOC frequency, intensity and duration taken together was 13.55% for pain impact and 11.68% for pain behavior . We found no statistically significant interactions between age group and VOC frequency, intensity or duration for any of the HRQOL outcome measures except for VOC intensity and satisfaction with social roles (p=0.0158 for interaction). Partial R 2 values for intensity and satisfaction with social roles were 0.0%, 0.6% and 7.4% for subjects aged <26, 26–36 and ≥37 age groups respectively, although none were statistically significantly different, likely due to the smaller sample sizes in the age subgroups.

With these limitations, we conclude that variation in measures of HRQOL, including psychosocial, physical function and pain expression can be partially explained by frequency, intensity and duration of VOC pain, over and above differences in age, sex, genotype, and organ system damage. In particular, VOC frequency explains the most variation, when simultaneously considering VOC intensity and VOC duration. We recommend our results be weighed by those designing SCD VOC interventions, pharmacologic or non-pharmacologic, to improve SCD HRQOL. For some purposes, it may be more desirable or feasible to tailor a specific intervention to improve a specific VOC pain domain – pain frequency, intensity or duration. However, we reason that, for most purposes, since all three VOC domains contribute to variability in HRQOL, all three domains should be assessed and interventions should be targeted to improve all three domains in order to maximize HRQOL outcomes.

We found that SCD VOC intensity, frequency, and duration when included in the same model, each was statistically significantly associated with HRQOL in adolescents and adults with SCD. Specifically, each was associated with psychosocial function, physical function, and pain expression. This was true regardless of whether HRQOL was measured by the generic PROMIS measure, or by the SCD-specific measure, ASCQ-Me measure. The amount of HRQOL variation explained by the combined model ranged from around 3–4% for effects on sleep related impairment and satisfaction with social roles , neither of which was statistically significant, to as much as 17.4% for effects on social functioning impact, which were significant. However, no single domain measure, intensity, frequency nor duration, accounted for a partial R 2 higher than 12.5% (effects on social functioning impact ). We found that the most explanatory single domain was VOC frequency, both from individual correlations as well as when considered simultaneously with intensity and duration. The exception to this was effects on the HRQOL domain of stiffness . The joint explanatory capacity of intensity and duration was considerably diminished, and usually not significantly different from zero. This decrement was particularly notable for effects on physical functioning . Our analyses also tested whether VOC pain domains had different explanatory power for HRQOL among different age groups. Hypotheses were motivated by the growing evidence that SCD pain in some patients becomes far more frequent and complex with age, transforming from primarily acute pain in the form of short-lived VOCs in childhood, to multifactorial chronic pain or acute-on-chronic pain beginning in adolescence or adulthood[ 18 , 20 , 22 , 46 , 47 ]. We did not find in our sample that the relationship between VOC pain and HRQOL differed statistically by age. While we did find that VOC pain intensity appeared to explain more variation in satisfaction with social roles for the older group than the younger two groups, this age difference was not statistically significant. We found this difference even though in the youngest age group satisfaction with social role scores were significantly higher. Our primary results are somewhat consistent with those of Anie et al[ 4 , 16 , 42 ]. They studied VOC pain frequency, intensity and duration, and found that pain accounted for 7.1% - 10.5% of variation over and above demographics and clinical severity, for SF-36 physical function, role limitations due to physical problems, social functioning and general health perceptions. In contrast to our study, Anie and colleagues did not find pain effects for mental/emotional wellbeing. In both our study and that of Anie, results are consistent in the following ways. First, both studies show similar types of relationships between VOC pain and HRQOL. Second, both studies found only a modest partial R 2 , indicating that many factors other than VOC pain contribute to the variability in HRQOL. Our results also mirror the results of studies that have found that VOC frequency[ 4 , 16 , 42 ] and pain intensity[ 10 , 32 , 42 ] were significantly associated with HRQOL. However, our study results show more consistent effects compared with others[ 3 , 32 ]. Esham et al[ 19 ] did not find past VOC frequency to be significantly associated with ASCQ-Me HRQOL measures at time of hospitalization. Counter-intuitively, he found a significant decrement in PROMIS global mental health for those with lower VOC frequency. Except for Anie et al[ 4 ], none of these studies quantitatively assessed strength of association of VOC pain measures, nor did they compare the relative importance of intensity vs frequency vs duration. Unique to our study is examination of the relationship of HRQOL outcomes to the duration of VOC, when duration is measured simultaneously with the more often studied intensity and frequency. To do this, we specifically separated out duration of VOC from the ASCQ-Me severity index, in contrast to work of Rizio et al[ 42 ]. We were able to determine that duration was of less importance in general than frequency and intensity. Stiffness was the only HRQOL measure for which duration accounted for the most variation. While we noted no other SCD research that demonstrated this finding, we did note some research for migraines that indicate that duration (and frequency) of migraine attacks were statistically significantly related to HRQOL (role function and anxiety and depression scores)[ 29 , 40 ]. Duration of endometriosis pain episodes was found to be a better predictor of limitations in activities of daily living than either frequency or intensity of such episodes [ 31 ]. Several limitations to this work deserve emphasis. Chief among them is the definition of VOC used. It is self-reported, is self-defined by the responder and is based on recall over 12 months. For example, responders with degenerative joint disease, almost 25%, might have confused DJD pain with VOC pain, inappropriately changing their report of VOC frequency. VOC frequency was recorded categorically, with an upper limit of 4+, rather than as an integer, which would have made for a more realistic, continuous variable. This may be a legacy from the past when VOCs were defined as only those treated in the ED or hospital, rather than also self-defined and managed at home. Since 39% of patients reported 4+ VOC, we definitely underestimated the effect of VOC frequency on HRQOL. It should be noted that this emphasizes rather than questions our major finding of the dominance of frequency as the most important domain. Another measurement issue that could be considered a limitation is that intensity and duration were assessed with the ASCQ-Me questionnaire regarding the most recent VOC, rather than an average or typical VOC. One would expect though, that while some may have thus reported on a milder VOC than usual, others could have reported on one that was more serious, and that overall, on average, the typical VOC pain experience would be represented. Another limitation of our work is that most other measures are also self-reported. Also, not all SCD genotypes were represented in this study. Only genotypes SS and Sbeta0 qualified for this study, and these are considered the highest SCD severity genotypes. This limits the study’s generalizability. Further, we note that subjects were from 4 geographic locations mostly in a single state, and thus may not represent other experiences in other states or countries. Finally, our cross-sectional group analyses suggesting that SCD pain does not generally impact HRQOL differentially by age is not as strong as longitudinal data documenting whether the impact of SCD pain on HRQOL changes over time within individual subjects.

Sickle cell disease (SCD), a group of genetic hemoglobinopathies, affects 100,000 people in the United States, five million individuals worldwide, with approximately 300,000 live births annually[ 13 , 37 ]. SCD causes early mortality, chronic morbidity[ 30 ] and negatively impacts health-related quality of life (HRQOL). HRQOL in SCD is affected by pain[ 16 , 21 , 35 ], physical limitations and organ failure[ 41 , 53 , 35 ] and by psychosocial dysfunction[ 4 , 15 , 25 , 26 , 27 , 32 , 35 ]. By adulthood SCD patients have often endured years of assaults to each of these domains, reflected in lower HRQOL in each domain[ 21 ], and in lower HRQOL compared to national norms[ 32 ]. The hallmark of SCD pain is the vaso-occlusive crisis (VOC) [ 16 , 20 , 39 , 45 ] VOCs are often severe and often provoke action, including hospitalizations and emergency department visits[ 10 , 12 ], which are a harbinger of morbidity and mortality[ 38 , 50 ]. Although SCD clearly causes chronic pain as well [ 17 , 49 ], VOCs are what typically cause provider contact [ 39 ], prescription of disease modifiers [ 6 ] and worries about prognosis [ 38 ]. VOC pain measures have been associated with HRQOL. For example, previous VOC pain (yes/no) is associated with lower scores on the MOS-SF-36 bodily pain, general health, vitality, and overall physical health scales[ 16 ]. But researchers and clinicians have struggled with how to report VOC measures, and specifically which dimensions of VOC are best related to HRQOL. VOCs are unpredictable and variable. The domains of VOC pain most often measured are intensity, frequency and duration[ 2 , 18 ]. Of these, frequency of VOCs or VOC-related hospitalizations[ 2 , 5 , 8 , 10 , 36 , 52 ] have been the basis for Food and Drug Administration registration of three of the four available disease modifiers[ 6 , 14 , 34 ]. However, worse VOC intensity has also been related to worse HRQOL, along with more frequent VOCs[ 3 ]. Both palliative pharmacological and non-pharmacological interventions seek to improve pain intensity and thereby HRQOL[ 4 , 9 ]. Only recently has there been a focus on relating VOC duration to HRQOL [ 2 , 42 ] since consensus definitions of SCD pain, including VOC that involve duration, have been published[ 17 , 20 ]. It is possible that interpretations of results including only one of these VOC dimensions may not be straightforward, and that reporting such may be inadequate. Only a few studies have simultaneously investigated two or more SCD pain domains [ 4 , 19 , 32 , 42 ]. While some researchers have looked at pain frequency, intensity and duration in other pain conditions such as chronic pediatric pain [ 43 ], endometriosis [ 31 ] and migraine pain [ 29 , 40 ], only the studies of endometriosis [ 31 ] and migraine pain [ 29 ] reported on the relative relationship of each pain feature over and above the others as they relate to HRQOL. We thus studied adolescents and adults with SCD to measure the strength of relationship between HRQOL and the VOC pain domains intensity, frequency, and duration measured simultaneously. We hypothesized that each VOC pain domain would be associated with HRQOL independent of demographics or disease severity.