br For all analyses P values were considered sig
For all analyses, P values < 0.05 were considered sig-nificant. Analyses were performed using R (version 3.5) using the packages dpylr for data management, compareGroups for descriptive tables, ggplot2 for graphics, polycor for tetrachoric correlations, cluster for cluster analyses,35 and MatchIt for matching.
For this study, 82 advanced CKD (not on dialysis) patients, 149 dialysis-dependent ESKD patients, and 606 patients with GI cancer were included. Table 1 shows the baseline patient characteristics. Patients with kidney disease were younger, had higher propor-tion of blacks, and had higher comorbidity burden of diabetes, hypertension, and cardiovascular disease as compared to the cancer group (P < 0.001 for all). An-tidepressant use was similar in the three groups (P ¼ 0.29). Pain medication use was similar in all three groups (P ¼ 0.15), but among those reporting signifi-cant levels of pain, CKD patients were significantly more likely to be on pain medications than ESKD or cancer groups (P ¼ 0.01). CKD and ESKD patients had lower hemoglobin but higher serum albumin than the cancer group (P < 0.001 and P ¼ 0.003, respectively).
Table 1 Baseline Characteristics
Variable or Mean (SD) or Mean (SD) or Mean (SD)
Dialysis vintage (months)
with depressive symptoms
reporting significant pain
CKD ¼ chronic kidney disease; ESKD ¼ end-stage kidney disease; BMI ¼ body mass index; eGFR ¼ estimated glomerular Galactose 1-phosphate rate. aRace missing for one patient; education status missing for 68 patients; marital status missing for seven patients; diabetes missing for 20 patients; HTN missing for 35 patients; cardiovascular disease missing for 18 patients; tobacco use missing for 68 patients; alcohol missing for 47 patients; antidepressant use missing for four patients; benzodiazepine missing for three patients; pain meds missing for 58 patients. bMedian (25th percentile, 75th percentile).
For the cancer patients, the most common etiol-ogies for cancer were hepatitis C (29%), cryptogenic (29%), alcohol alone or in addition to other causes
(25%), and nonalcoholic hepatic steatosis (10%). Cirrhosis was present in 42% of the patients, and 17% were noted to have vascular invasion. One-third of the patients had one lesion, 23% had more than five lesions, and average tumor size was 3.9 cm.
Table 2 shows the prevalence of patient-reported fatigue, pain, and depressive symptoms in the CKD, ESKD, and cancer patients. Fatigue was highly prevalent, and more than 75% of patients in each group reported having a significant level of fatigue. Surprisingly, the prevalence of fatigue was similar across the three groups (P ¼ 0.97). Moreover, the severity of fatigue in each group was very similar (P ¼ 0.17), and the average fatigue levels were much higher in these patients as compared to the U.S. gen-eral population (Fig. 1).
Similarly, prevalence of clinically significant pain was comparable across the three groups (P ¼ 0.21). However, moderate or severe depressive symptoms were much more common in cancer than in kidney disease patients (P < 0.001). Nearly 80% of patients in each group had at least one symptomdfatigue, depressive symptoms, or pain (Table 2). These symp-toms often coexisted, and 25%e30% of patients in each group had 2 or 3 of the symptoms independent of the patient group (Table 2 and Fig. 2).
Correlations Among Patient-Reported Symptoms
In CKD patients, fatigue was moderately correlated with pain but not with depressive symptoms, and there was no significant correlation between depressive symptoms and pain (Fig. 3). By contrast, in the ESKD group, all three symptoms were moderately correlated with each other (correlation coefficient
r ¼ 0.48e0.55, P < 0.001). Cancer patients also had significant correlations among all three symptoms; however, their strength of correlations was weaker