Predictive Ability of Intermittent Daily Sickle Cell Pain Assessment: The PiSCES Project

Abstract

Background

Pain diary assessment in sickle cell disease (SCD) may be expensive and impose a high respondent burden.

Objective

To report whether intermittent assessment could substitute for continuous daily pain assessment in SCD.

Design

Prospective cohort study.

Setting

Academic and community practices in Virginia.

Patients. A total of 125 SCD patients age 16 years or older in the Pain in Sickle Cell Epidemiology Study.

Measurements. Using pain measures that summarized all diaries as the gold standard, we tested the statistical equivalence of four alternative strategies that summarized diaries only from the week prior or the month prior to study completion; one week per month; or one day per week (random day). Summary measures included percent pain days, percent crisis days (self-defined), mean pain (0–9 Likert scale) on all days, and mean pain on pain days. Equivalence tests included comparisons of means, regression intercepts, and slopes, as well as measurement of R².

Results

Compared with the gold standard, the one-day-per-week and one-week-per-month strategies yielded statistically equivalent means of six summary pain measures, and the week prior and month prior yielded equivalent means as some of the measures. Regression showed statistically equivalent slopes and intercepts to the gold standard using one-day-per-week and one-week-per-month strategies for percent pain days and percent crisis days, but almost no other equivalence. R² values ranged from 0.64 to 0.989.

Conclusions

It is possible to simulate five- to six-month daily assessment of pain in SCD. Either one-day-per-week or one-week-per-month assessment yields an equivalent mean and fair regression equivalence.

Introduction

Sickle cell disease (SCD) is an extraordinarily painful, vaso-occlusive, genetic erythrocyte disorder affecting those of African, Mediterranean, and Asian descent [1,2]. Caregivers and researchers have traditionally thought of pain as occurring mainly during vaso-occlusive pain “crises” that often lead to emergency department use or hospitalization [3]. Crisis frequency does predict mortality [4], and annual crisis frequency can be reduced with hydroxyurea [5], crizanlizumab [6], and pharmaceutical-grade L-glutamine [7,8].

But crises as a total pain measure vastly underestimate total pain burden. Most SCD pain is managed at home and is not called a crisis by patients. Yet pain may occur as often as daily [9].

Of the 480 SCD trials currently listed at ClinicalTrials.gov, some now are for medications designed in part to improve short-term crisis pain [10]. But others are for medications designed in part to reduce longer-term, noncrisis, daily SCD pain or symptoms [11], sometimes for pain that lasts more than a year, but often over periods shorter than the typical year or years required to see clinically significant changes in numbers of crises [12]. These or similar SCD trials would be improved by validated measures of daily pain outcomes in SCD. Daily pain assessment yields detail about individual daily variability and patterns of pain, but assessments can be summarized over time as well. One of the most robust set of measures for chronic disease patient-reported outcomes, the Patient-Reported Outcomes Measurement Information System (PROMIS) [13–15], includes among its multitude of item banks and scales the SCD-specific measure of health-related quality of life, comorbidities, utilization, and health care quality, the Adult Sickle Cell Quality of life Measurement System (ASCQ-Me) [16]. However, ASCQ-Me focuses only on summary scores of SCD pain and symptoms over months to a year, not the detail of pain on individual days. Cleeland’s Brief Pain Inventory [17] surveys a recall period of seven days, and thus is not recommended for use more frequently than weekly. Similarly, The McGill Pain Questionnaire’s [18] test-retest reliability is from one to seven days, suggesting it should not be used more frequently than weekly [19].

To get at this detail, the most prevalent SCD pain measures are standardized daily pain diaries [20–29,29–32], a form of Ecological Momentary Assessment (EMA) [33]. EMA methods such as diaries permit research participants to report on symptoms and behavior close in time to an experience or event and allow sampling of many events or time periods. EMA has a number of advantages over more traditional methods for the assessment of patients. It can both complement and, in part, replace existing methods. EMA methods permit more sensitive, and likely more accurate, assessments and enable more wide-ranging and detailed measurements. Diaries can be used to measure outcomes of therapy [34–37]. Diaries demonstrate sensitivity to both short-term and chronic illness changes [38].

For example, paper-and-pencil diaries were used in the Multicenter Study of Hydroxyurea [39], a trial of senicapoc [40], and cohort studies [41,42]. Paper and pencil SCD pain diaries are still in use [43]. But electronic diaries using the world wide web [44], handheld devices [45–47], or multimodal collection involving actigraphy [48] and other pain reporting methods are now the most popular. Regardless of the method of recording responses, continuous daily pain assessment is expensive and imposes a high respondent burden. Intermittent assessment strategies, if valid, might successfully substitute for continuous daily pain assessment in certain circumstances.

We were therefore interested in the accuracy of reports of daily pain that used intermittent daily assessment rather than continuous daily assessment, and thus resulted in a lower expense and respondent burden. We were particularly interested in whether one of several possible candidate strategies of intermittent pain assessment was better than another as alternate intermittent strategies might have similar costs and respondent burden. Our already-conducted daily diary pain assessment of a cohort of SCD patients afforded us the opportunity of retroactively testing various intermittent daily sampling strategies vs a continuous daily assessment strategy.

Methods

Study Design

The Pain in Sickle Cell Epidemiology Study (PiSCES) was a longitudinal study of pain in SCD with a particular emphasis on potentially mutable, causal, nonbiological variables. It was also a methodological study of the relationship among various measures of pain and utilization in SCD. The methods of PiSCES have been described in detail elsewhere [49,50]. Briefly, we enrolled 308 patients from July 2002 through August 2004. We collected baseline information (including demographic characteristics and medical history) and laboratory data (blood and urine samples) and encouraged patients to complete daily pain diaries for up to six months.

We recruited patients 16 years of age or older from across Virginia, mostly from the Richmond and Tidewater areas. Both the study and our recruitment methods were approved by the Institutional Review Board of Virginia Commonwealth University, Richmond, Virginia, and patients gave informed consent.

Patients received routine care for their SCD from either community-based physicians or sickle cell specialist physicians associated with academic medical centers (two physicians at Virginia Commonwealth University serving the Richmond area and one physician associated with Eastern Virginia Medical School, Norfolk, Virginia, serving the Tidewater region). Emergent care for the cohort was provided in emergency departments regardless of the patients’ usual source of ambulatory care. No day hospitals for SCD were located in the region.

Diary Data

Patients completed daily diaries for up to six months. They were encouraged (at the initial baseline visit and with reminder calls by study staff) to complete the diary each day and return it by mail using provided stamped envelopes. They received payment for each returned diary, with a higher payment in the latter two months of the study to encourage study completion. Among other things, we asked patients to report the following about the previous 24 hours in their diary: their worst sickle cell pain intensity on a scale from 0 (none) to 9 (unbearable). Also, in each daily diary entry, patients could endorse (using a check box) the item “Yesterday, I was in a crisis.” Crises were self-defined by each patient. Further, patients endorsed whether they had taken medication for their pain (and if so, what), had gone for an unscheduled physician visit or emergency department visit, or had been hospitalized because of sickle cell pain.

Measurements and Statistical Analysis

Assessment Strategies

We defined a gold standard strategy of summarizing continuous daily pain diary data for approximately six months (the entire PiSCES assessment period). We implemented this by using the subset of patients who submitted at least 153 diaries (5–6 months of diaries; the mean of our sample was 171 days or 5.6 months). Of the 308 patients who enrolled in PiSCES and completed baseline surveys, 23 patients sent in no diaries. Of the remaining 285 patients, we excluded 162 because they completed fewer than 153 (5 months) of the expected 183 diary entries. The remaining 125 patients constitute the analysis sample. We then selected and tested four candidate strategies of intermittent diary assessment. These included using only diaries from the last week of patient participation (week prior), using only diaries from the last month of participation (month prior), using diaries selected from the last week of each month (1 week per month), and using diaries from only one random day per week of participation (1 day per week). The rationale for selecting and testing the first two of these sampling strategies was to replicate a strategy that might be practical for use in a clinical trial. For example, investigators might ask a patient to keep a diary for only the week or month prior to their clinical trial visit. The rationale for selecting and testing the last two strategies was to try to sample a broader time frame during the six-month period.

Outcome Measures

Our outcome measures included percentage of days in pain (referred to as percentage of pain days), percentage of days with crisis (referred to as percentage of crisis days), and pain intensity on all days, on only days when pain was reported, on days with crisis, and on days with pain but no crisis. For each patient, all returned diaries consistent with their gold standard or intermittent (week prior, month prior, 1 week per month, or 1 day per week) strategy were counted. This total served as the denominator for measuring percent pain days, percent crisis days, and overall pain intensity. For the purposes of analysis, we defined a pain day as any day on which patients rated pain intensity as greater than 0. We defined a crisis day as any day when patients endorsed the above crisis item in a diary entry. We counted the number of pain days and crisis days during the study period, and this total served as the numerator for calculation of the percentage of pain days and crisis days. We calculated the mean pain intensity (0–9 Likert scale) over all days using the total of returned diaries as the denominator. We also calculated the mean pain intensity on all pain days, days when patients endorsed crisis in their diary, and noncrisis pain days. Note that when there were no pain days for a given patient, we could make no estimate of mean pain on pain days only—likewise for pain intensity on crisis days or noncrisis pain days.

Analysis

Our analyses sought to determine how well the candidate intermittent assessment strategies replicated the gold standard with regard to estimates of the percentage of pain days and crisis days, as well as the mean pain intensity over all days and over pain days—all pain days, crisis days only, and noncrisis pain days. To assess this, we used equivalence analysis. With equivalence analysis, we attempt to show that the means of the pain measures as assessed from the gold standard, continuous six-month daily sample vs one resulting from an intermittent sampling strategy, do not differ by more than a small amount, referred to as the tolerance and denoted by Δ. Statistically, the null hypothesis is that the difference in means is greater than Δ, with the alternative that the difference is smaller than by Δ. When rejecting the null hypothesis, we can conclude that the intermittent strategy is equivalent to the continuous strategy (gold standard). Based on the traditional Two One-sided Test (TOST) [51] equivalence analysis, we calculated the 90% confidence interval (CI) around the difference between measures from the gold standard and intermittent strategies samples. Equivalence is shown if the confidence interval is entirely contained within the interval (−Δ, Δ). Note that while the distribution of the original measures, such as percent crisis days, might be skewed themselves, the difference between means (GS vs intermittent strategy), which is the basis for the analysis, was reasonably symmetric, making it possible for Gaussian methods to be used.

A second analysis considered the regression of the gold standard values on the values based on the intermittent strategies. In this analysis, we tested for the equivalence of the intercept to 0 and the slope to 1. Thus, for example, we fit a model with mean pain intensity as assessed with the gold standard sample as the dependent variable, and main pain intensity as assessed on the one-day-per-week sample as the independent variable. The slope of the equation was compared with 1.0 and the intercept to 0, using TOST/confidence intervals as described above.

Tolerance values (Δ) selected for equivalence analysis for the mean difference between the gold standard sample and a given intermittent strategy were 3.0 for proportion of pain days (gold standard value = 52.2) and proportion of crisis days (gold standard value = 13.6) and 0.3 for mean overall pain intensity on pain days, crisis days, and noncrisis pain days (gold standard values of 2.39, 4.28, 5.42, and 4.06, respectively). Acceptable tolerances for the regression analyses were clinically estimated as 0.05 for both slope and intercept.

Analyses were conducted by using SAS, version 9, for Windows (SAS Institute, Cary, NC, USA).

Results

Table 1 shows that patients in the analysis sample were quite similar to those excluded when demographic variables were compared. The patients who did not complete sufficient diaries tended to have more pain days (61.3% vs 52.2%, P = 0.051) and somewhat higher pain intensity overall (2.94 vs 2.39, P = 0.047). We included a total of 21,393 diaries (patient-days) in this analysis. The mean number of diaries completed in the sample was 172 out of a possible 183.

Table 1.

Patient characteristics of included and excluded patients

Characteristic	Analysis Sample (N = 125)	Excluded Subjects (N = 183)	P
	Freq (%)	Freq (%)
Demographic characteristics
Sex			0.903
Female	76 (60.8)	110 (60.1)
Male	49 (39.2)	73 (39.9)
Education			0.810
Less than high school	17 (13.7)	24 (13.2)
High school graduate	43 (34.7)	73 (40.1)
Some college	47 (37.9)	63 (34.6)
College graduate	17 (13.7)	22 (12.1)
Age, y			0.334
16–24	30 (24.0)	49 (26.8)
25–34	31 (24.8)	61 (33.3)
35–44	37 (29.6)	46 (25.1)
45–54	21 (16.8)	21 (11.5)
55–64	6 (4.8)	6 (3.3)
Marital status			0.140
Married	33 (26.6)	34 (18.6)
Never married	72 (58.1)	126 (68.8)
Divorced/separated/widowed	19 (15.3)	23 (12.6)
Income			0.315
<$10,000	41 (33.6)	76 (43.2)
$10,001–$20,000	34 (27.8)	37 (21.0)
$20,001–$30,000	17 (13.9)	26 (14.8)
>$30,000	30 (24.6)	37 (21.0)
Sickle cell disease characteristics
Genotype			0.590
B-thal or SS	87 (73.1)	120 (75.9)
SB + thal or SC	32 (26.9)	38 (24.1)
	Mean ± SD	Mean ± SD
Pain-related variables*
Percent pain days	52.2 ± 40.3	61.3 ± 37.0	0.051
Percent crisis days	13.6 ± 22,9	16.2 ± 25.8	0.385
Mean pain: all days	2.39 ± 2.19	2.94 ± 2.34	0.047
Mean pain: pain days	4.28 ± 1.57	4.43 ± 1.70	0.452
Mean pain: crisis days	5.42± 1.95	5.63 ± 1.73	0.424
Mean pain: noncrisis pain days	4.06 ± 1.57	4.01 ± 1.65	0.810

^*N = 154 for the excluded group because not all excluded subjects provided diary data.

Table 2 shows the mean differences between the gold standard value and the values generated by the four intermittent strategies for the percentage of pain days and crisis days as well as mean pain, mean pain on pain days, mean pain on crisis days, and mean pain on noncrisis pain days. Compared with the gold standard, one-day-per-week and one-week-per-month strategies yielded statistically equivalent means of all summary pain measures except for pain intensity on crisis days with the one-week-per-month strategy (P < 0.05). The week prior and month prior strategies were equivalent to the gold standard for pain intensity on all pain days and on noncrisis pain days. For the week prior strategy, pain intensity on all days was also equivalent while for the month prior strategy the percent crisis days was equivalent to the gold standard measure.

Table 2.

90% confidence intervals for difference between outcomes, assessed using gold standard sample vs intermittent strategy samples

	Week Prior	Month Prior	1 Week/Month	1 Day/Week
Pain frequency
% pain days	(1.003 to 6.695)	(3.204 to 7.278)	(–0.037 to 1.606)	(0.067 to 1.436)
% crisis days	(–2.978 to 2.123)	(–1.540 to 1.574)	(–0.703 to 0.743)	(–0.697 to 0.372)
Pain intensity
All days	(–2.978 to 2.123)	(–1.540 to 1.574)	(–0.703 to 0.743)	(–0.697 to 0.372)
Pain days	(–0.272 to 0.129)	(–0.191 to 0.109)	(–0.095 to 0.063)	(–0.003 to 0.215)
Crisis days	(–0.350 to 0.231)	(–0.379 to 0.002)	(0.004 to 0.309)	(–0.134 to 0.185)
Noncrisis pain days	(–0.273 to 0.156)	(–0.063 to 0.269)	(–0.099 to 0.146)	(–0.032 to 0.192)

Week prior = using only diary entries from the last week of patient participation; month prior = using only diary entries from the last month of participation; 1 week/month = using diary entries from the last week of each month of participation; one day/week = using diary entries from only one random day per week of participation. Tolerance is Δ = 3% for pain days and crisis days; Δ = 0.30 for pain intensity. Reject the null hypothesis (and conclude equivalence) if the confidence interval is within (−Δ, Δ). Values in italics inidicate equivalence. if.

Table 3 shows regression fit values (slope, intercept, R² values) of the four intermittent strategy measures vs gold standard measures. The regressions showed that the one week per month and one day per week were perfectly calibrated (slope and intercept) for percent pain days, and the one-day-per-week strategy was also perfectly calibrated for percent crisis days. It also showed slope equivalency of mean pain on all days for these two measures.

Table 3.

Regression of measures based on gold standard (gold standard, 5–6 months daily) assessment of measures based on four intermittent assessment strategies of pain in sickle cell disease; coefficients and equivalence confidence intervals^*

	Week Prior	Month Prior	1 Week/Month	1 Day/Week
% pain days
Intercept	0.134	0.111	0.020	0.011
	(0.097 to 0.171)	(0.084 to 0.139)	(0.077 to 0.033)	(0.000 to 0.023)
Slope	0.802	0.874	0.976	0.992
	(–0.253 to –0.142)	(–0.169 to –0.083)	(–0.044 to 0.004)	(–0.025 to 009)
R2	0.820	0.901	0.981	0.987
% crisis days
Intercept	0.048	0.030	0.007	0.002
	(0.026 to 0.070)	(0.015 to 0.044)	(–0.002 to 0.015)	(–0.004 to 0.008)
Slope	0.629	0.781	0.953	0.973
	(–0.439 to –0.304)	(–0.267 to –0.171)	(–0.077 to –0.016)	(–0.049 to –0.004)
R2	0.657	0.854	0.956	0.976
Mean pain on all days
Intercept	0.564	0.498	0.054	0.052
	(0.372 to 0.575)	(0.347 to 0.649)	(–0.011 to 0.118)	(0.002 to 0.103)
Slope	0.811	0.865	0.992	0.996
	(–0.247 to –0.131)	(–0.182 to –0.088)	(–0.028 to 0.013)	(–0.019 to 0.012)
R2	0.812	0.882	0.982	0.989
Mean pain on pain days
Intercept	1.548	1.191	0.400	1.032
	(1.117 to 1.978)	(0.851 to 1.531)	(0.173 to 0.627)	(0.771 to 1.292)
Slope	0.645	0.718	0.904	0.783
	(–0.442 to –0.267)	(–0.355 to –0.210)	(–0.145 to –0.046)	(–0.274 to –0.156)
R2	0.668	0.751	0.899	0.834
Mean pain on crisis days
Intercept	0.752	0.754	0.887	1.081
	(0.539 to 2.095)	(0.253 to 1.256)	(0.445 to 1.330)	(0.670 to 1.492)
Slope	1.317	0.828	0.865	0.807
	(–0.380 to –0.116)	(–0.258 to –0.086)	(–0.212 to –0.058)	(–0.264 to –0.122)
R2	0.745	0.821	0.832	0.833
Mean pain on noncrisis pain days
Intercept	1.473	1.273	0.859	0.996
	(0.997 to 1.950)	(0.905 to 1.640)	(0.565 to 0.1.152)	(0.749 to 1.243)
Slope	0.641	0.715	0.797	0.777
	(–0.463 to –0.255)	(–0.367 to –0.202)	(–0.269 to –0.136)	(–0.279 to –0.167)
R2	0.643	0.732	0.807	0.847

Week prior = using only diary entries from the last week of patient participation; month prior = using only diary entries from the last month of participation; 1 week/month = using only diary entries selected from last week of each month of participation; 1 day/week = using diary entries from a random day per week of participation.

^*Equivalence confidence intervals are for slope = 1 and intercept = 0; Tolerance is Δ = 0.05 for both. Reject the null hypothesis (and conclude equivalence) if the confidence interval is within (−Δ, Δ). Italicized confidence intervals indicate equivalence.

However, the week prior and month prior strategies were not shown to be equivalent to the gold standard in these regressions for any of the measures—with the exception of percent crisis days, where the intercept was equivalent to 0 for the month prior strategy. R² values for the four tested measures were as low as 0.643 for the week prior strategy.

Discussion

For an intermittent daily sampling strategy to be a valid substitute for a continuous daily (gold standard) strategy for research purposes, we required statistical equivalence between the means of measures obtained using intermittent assessment, as compared with means obtained using continuous assessment. We found that either a random one-day-per-week or a one-week-per-month strategy performed best over a five- to six-month period, when considering percent pain days, crisis days, and mean pain intensity on all days, pain days, crisis days, and noncrisis pain days.

On the other hand, for an intermittent strategy to be recommended for clinical use, it must be perfectly calibrated with a slope of 1.0 and an intercept of 0 when regressing the gold standard measure on the measure from the intermittent strategy. In general, our intermittent strategies fell short on this account. Only the one-day-per-week strategy could reproduce both percent pain days and crisis days, and no strategy was adequate for any of our pain intensity measures. It might be possible to recalibrate using the regression coefficients.

There are several limitations to this work. First, we did not actually test an intermittent data collection strategy, but rather simulated one. Patients may respond differently when asked to report pain only on an intermittent basis. Also, the sample we used for the gold standard consisted of the most compliant of our patients. It is possible that “predictions” for less compliant patients would not be as accurate. We found that the patients included in our gold standard sample had slightly fewer pain days and less pain intensity on all days than the excluded patients.

Second, we only had data for six months. We do not have evidence as to how these strategies would fare if subjects were to be tracked over longer periods such as years.

Third, there are other measures of pain that we could have considered. Stone et al. [52] suggest that measures such as the standard deviation of pain intensity over time are also of interest. In fact, it may be that one should expect the variability of pain intensity to be less (hence a smaller SD) when looking at shorter periods of time. Thus, one might expect, for example, the week prior strategy to provide underestimates of standard deviation. We found (data not shown) that the week prior and month prior strategies tend to have a smaller SD than the full six-month continuous daily strategy (gold standard), but that the one-week-per-month and one-day-per-week strategies have somewhat similar values, particularly when considering all pain days. This may be because these strategies, while using fewer days than the gold standard, span the time frame covered.

Fourth, we found that some of the intermittent assessments were equivalent to continuous assessment when considering means but not regressions. This implies that while the intermittent assessments might be reasonably accurate for studies that wish to assess means of groups, their use for individual patients cannot be recommended unless equations are derived to recalibrate the results for each pain measure.

Fifth, tolerance values (Δ) for equivalence analysis are inherently subjective. We presented results as confidence intervals. Thus, the results can be readily re-interpreted using different values of Δ to determine whether the confidence interval is within (−Δ, Δ) and thus equivalence is achieved.

Sixth, our findings were based on paper-and-pencil diaries. We cannot generalize them to electronic diaries, though we believe intermittent strategies could easily be implemented today with electronic diary collection.

Last, while we can identify and characterize pain on days not called a crisis by patients, our diaries were not meant to define or study chronic SCD pain. Chronic SCD pain was recently defined by an expert panel constructed within the Analgesic, Anesthetic, and Addiction Clinical Trial Translations Innovations Opportunities and Networks–American Pain Society Pain Taxonomy initiative [53]. Chronic SCD pain must include a diagnosis of SCD confirmed by laboratory testing, plus reports of ongoing pain present on most days over the past six months either in a single location or in multiple locations, and one sign of pain sensitivity on palpation or with movement of the region of reported pain, decreased range of motion or weakness in the region of reported pain, or evidence of chronic disease complications (e.g., skin ulcer, splenic infarct, or bone infarction) associated with the region of reported pain.

Nonetheless, we believe our epidemiologic findings provide useful new methodologic information about the validity of pain diaries to assess pain in SCD and show that pain diary–based measures may not necessarily require high cost or be burdensome to patients. Our findings lend support for further use of pain diaries and other forms of EMA in clinical trials of SCD. We believe our findings show that these EMA measures can work for any trial measuring symptoms in SCD for up to six months.

Some of the most authoritative statements of how tools should be chosen to measure pain in clinical trials, and which pain domains might be important to measure, come from the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT) [54–57] and from the Analgesic, Anesthetic, and Addiction Clinical Trial Translations Innovations Opportunities and Networks–American Pain Society Pain Taxonomy initiative [58]. We believe these results may contribute to novel methods of pain assessment in clinical trials, including methods recommended by these expert bodies.

Authors’ contributions

Conception and design: W. R. Smith, V. E. Bovbjerg, D. K. McClish, J. D. Roberts, B. Dahman, J. L. Levenson. Analysis and interpretation of the data: W. R. Smith, V. E. Bovbjerg, D. K. McClish, J. D. Roberts, I. Aisiku, B. Dahman, J. L. Levenson, S. D. Roseff. Drafting of the article: W. R. Smith, McClish, B. Dahman, J. L. Levenson. Critical revision of the article for important intellectual content: W. R. Smith, V. E. Bovbjerg, D. K. McClish, I. Aisiku, B. Dahman, J. L. Levenson, S. D. Roseff. Final approval of the article: W. R. Smith, V. E. Bovbjerg, D. K. McClish, I. Aisiku, B. Dahman, J. L. Levenson, S. D. Roseff. Provision of study materials or patients: W. R. Smith, J. D. Roberts. Statistical expertise: D. K. McClish, B. Dahman. Obtaining of funding: W. R. Smith, D. K. McClish, J. D. Roberts. Administrative, technical, or logistic support: S. D. Roseff. Collection and assembly of data: W. R. Smith, B. Dahman, D. K. McClish.

Funding sources: SHIP-HU Study, 1 R18 HL 112737, National Heart, Lung, and Blood Institute, NIH. Pain in Sickle Cell Epidemiology Study, 1 R01 HL 64122, National Heart, Lung, and Blood Institute, NIH. Virginia Basic and Translational Research Program in Sickle Cell Disease. 1U54HL090516, National Heart, Lung, and Blood Institute. 1U10HL083732, Sickle Cell Disease Clinical Research Network, National Heart, Lung, and Blood Institute. Dr. Aisiku was a National Institutes of Health–funded Sickle Cell Scholar in the Basic and Translational Research Program.

Conflicts of interest: Grants received: W. R. Smith (National Heart, Lung, and Blood Institute, Pfizer, Emmaus, Global Blood Therapeutics, Novartis, Ironwood Pharmaceuticals), D. K. McClish (National Heart, Lung, and Blood Institute), J. D. Roberts (National Heart, Lung, and Blood Institute), I. P. Aisiku (National Heart, Lung, and Blood Institute), S. D. Roseff (National Heart, Lung, and Blood Institute).