Association of obesity under different metabolic status with adverse outcomes in patients with chronic myeloid leukemia: A retrospective cohort study

2.1 Data sources and study population

Of the 35 460 557 (weighted) patients, we identified 23 008 patients diagnosed with CML in the 2018 NRD database using the International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) diagnostic (Table S1 in Data S1). The 2018 NRD database contains information from 28 states that represent 60% of the US population.¹⁵ The characteristics of the national readmission for patients can be studied based on a verified unique linkage number that enables tracking of patients across different hospitals within a state but does not across years.¹⁵ Meantime, it was publicly available and one of the largest inpatient national sample databases. The study was based on STROBE reporting guidelines and abided by the United States Agency for Healthcare Research and Quality's Healthcare Cost and Utilization Project Data Use Agreement and was exempt from research ethics board review. Consent to participate was not applicable because the data were extracted from the 2018 NRD database. The data were available from the Healthcare Cost and Utilization Project site, subject to a registration and application process.

We excluded patients who had missing baseline characteristic data (n = 245), who were aged <18 years (n = 103), who were pregnant (n = 71), who died during the initial hospitalization (n = 638), who had low body weight (BMI ≤ 19.9 kg/m², n = 371), and who were initially hospitalized between July 1, 2018 and December 31, 2018 for 180 days (n = 6043, Figure S1). Finally, we included 7931 adults with discharge diagnoses of CML from January 1, 2018 to June 30, 2018. The ICD-10-CM codes used are shown in Table S1 in Data S1.

2.2 Exposure measures

Patients were classified into two BMI categories: normal weight (BMI < 25 kg/m²) and obesity (BMI ≥ 25 kg/m²). Metabolic abnormalities were defined as having any two of the following according to the harmonized International Diabetes Federation criteria¹⁶: (1) hyperglycemia (including prediabetes, diabetes, and others), (2) dyslipidemia (including hypertriglyceridemia, hypercholesterolemia, and others), and (3) hypertension. The definition of metabolic abnormalities excluded waist circumference because of its collinearity with BMI. Patients were classified into four groups according to BMI categories and metabolic status: (1) metabolically healthy normal weight (MHNW); (2) metabolically unhealthy normal weight (MUNW); (3) metabolically healthy obesity (MHO); and (4) metabolically unhealthy obesity (MUO). The above diagnoses were determined by ICD-10-CM codes (Table S2 in Data S1).

To explore the impact of the number of metabolic risk factors and specific metabolic risk factors on adverse outcomes of CML, patients were classified into eight groups based on BMI and the number of metabolic risk factors: (1) normal weight with no metabolic risk factors; (2) normal weight with one metabolic risk factor; (3) normal weight with two metabolic risk factors; (4) normal weight with three metabolic risk factors; (5) obesity with no metabolic risk factors; (6) obesity with one metabolic risk factor; (7) obesity with two metabolic risk factors; and (8) obesity with three metabolic risk factors. In addition, patients were classified into eight groups based on BMI and specific metabolic risk factors: (1) normal weight with no metabolic risk factors; (2) normal weight only with hyperglycemia; (3) normal weight only with hypertension; (4) normal weight only with hyperlipidemia; (5) obesity with no metabolic risk factors; (6) obesity only with hyperglycemia; (7) obesity only with hypertension; and (8) obesity only with hyperlipidemia.

2.3 Outcome measures

The primary outcome was the adverse outcomes (nonremission [NR]/relapse and severe mortality risk) in patients with CML on different readmission days. Readmission was defined as diagnoses with CML and unplanned hospital readmissions from the discharge, and we only recorded the first readmission for patients with multiple readmissions.¹⁷ The NR/relapse in patients with CML was diagnosed by using ICD-10-CM. Severe mortality risk was defined as the major likelihood of dying based on the risk-of-mortality subclass in the 2018 NRD database.¹⁵ Disease burden was defined as hospital readmission costs due to illness. Severity of illness was defined as major or extreme loss of function based on the severity-of-illness subclass in the 2018 NRD database.¹⁵ Secondary outcome included disease burden, severity of illness, further analysis of the impact of the number of metabolic risk factors, and specific metabolic risk factors on the adverse outcomes in patients with CML.

2.4 Variables

We identified baseline characteristics of study participants using NRD variables such as age, sex, total charges, length of stay, admission types (whether nonelective), rehab transfer, same-day events (transfer flag indicating a combination of discharges involving same-day events), resident, insurance status (Medicare, Medicaid, private insurance, self-pay, no charge, or other), median household income by ZIP code, location of patient's residence, disposition of patient, hyperglycemia, hypertension, dyslipidemia, severity of illness, severe mortality risk, and comorbidities (e.g., ischemic heart disease, chronic obstructive pulmonary disease, neurologic condition, chronic kidney disease, and other oncologic diseases).¹⁸ The required ICD-10-CM codes are given in Table S2 in Data S1.

2.5 Statistical analysis

We summarized the baseline characteristics in patients with metabolically defined obesity. Frequency counts and percentages were appropriate for categorical variables and were compared using the chi-square test, whereas medians (interquartile range [IQR]) were used for continuous variables with non-normal distribution and uneven variances and were compared using the Kruskal–Wallis H test. Multivariable logistic regression was performed to analyze data. Results were denoted as odds ratios (ORs) with 95% CI. Discharge weights were used to generate national estimates. All hypothesis tests were two-sided, and a p < 0.05 was deemed statistically significant. All statistical analyses were performed using the SPSS software 26.0.

2.6 Sensitivity analysis

To evaluate the robustness of findings, the following sensitivity analyses were performed. We did the same analysis at 90 and 30 days (Figure S1).

3.1 Baseline characteristics of patients with CML with metabolically defined obesity

Based on our inclusion and exclusion criteria, we included 7931 (56.8%) hospital discharges for adult patients for the 180-day readmission analysis (MHNW [3585, 45.2%], MUNW [3060, 38.6%], MHO [525, 6.6%], MUO [761, 9.6%]), 10 980 (78.6%) discharges for the 90-day analysis (MHNW [4946, 45.0%], MUNW [4218, 38.4%], MHO [748, 6.8%], MUO [1068, 9.7%]), and 12 953 (92.7%) discharges for the 30-day analysis (MHNW [5801, 44.8%], MUNW [5040, 38.9%], MHO [878, 6.8%], MUO [1234, 9.5%]) (Figure S1). Baseline characteristics of study participants are summarized in Table 1. The average age was 68 years and males accounted for 56.7%. The patients with MUNW or MUO were older and had higher mortality risk compared to the patients with MHNW or MHO (all p < 0.01), meaning that patients with metabolic abnormalities were older and seemed to be at a higher mortality level. Moreover, patients with MHO or MUO, representing patients with obesity, had more hospitalization costs and longer lengths of stay (all p < 0.01). Compared with MHNW, the patients with MUNW, MHO, and MUO had higher severity of illness (all p < 0.01). Baseline characteristics of study participants were generally similar between the different readmission days (Tables S3 and S4 in Data S1).

3.2 Adverse outcomes and disease burden in patients with CML with metabolically defined obesity

Patients with MUNW or MUO had a significantly higher rate of adverse outcomes and disease burden than those with MHNW (all p < 0.01), suggesting that patients with metabolic abnormalities were more likely to experience disease progression and death (Figure 1D,E,F). Interestingly, even though the rate of NR/relapse and rate of disease burden did not reach statistical significance in the MHO group, it seemed to be higher in the total, in females, and in patients aged <55 years than the MHNW, and similar to MUNW. Hence, obesity appeared to affect NR/relapse and disease burden at least in some groups (Figure 1A,C).

After adjusting for age, sex, total charges, length of stay, admission types, rehab transfer, same-day events, resident, insurance status, income, location of patient's residence, disposition of patient, and total comorbidities, patients with MUNW and MUO were associated with higher risk of NR/relapse (adjusted odds ratio [aOR] = 1.18, 95% CI, 1.05–1.33; aOR = 1.32, 95% CI, 1.11–1.57), severe mortality (aOR = 1.36, 95% CI, 1.19–1.56; aOR = 1.63, 95% CI, 1.34–1.98) and disease burden (aOR = 1.25, 95% CI, 1.12–1.40; aOR = 1.44, 95% CI, 1.22–1.71), but patients with MHO showed no significant differences in the risk of adverse outcomes and disease burden (Table 2). Sensitivity analyses with different readmission days had similar results (Figure 2B, Table S5 and S6 in Data S1). Interestingly, we found that patients with MHO were related to a higher risk of NR/relapse (aOR = 1.33, 95% CI, 0.99–1.78; aOR = 1.38, 95% CI, 0.99–1.91), severe mortality (aOR = 1.55, 95% CI, 1.08–2.23; aOR = 1.53, 95% CI, 0.97–2.42) and disease burden (aOR = 1.33, 95% CI, 1.00–1.76; aOR = 1.47, 95% CI, 1.07–2.03) in females and in patients aged <55 years compared with MHNW, even if sometimes there was no statistically significant difference in the multivariable logistic regression at 180 days, meaning that obesity seemed to affect adverse outcomes and disease burden in some subgroups (Table 2).

In the subgroup analyses, we found that female patients with MUNW and MUO had 1.23-fold and 1.40-fold increased NR/relapse risk compared with MHNW at 180 days, while male patients with MUNW and MUO were not related to the NR/relapse risk (Figure 2A). Interestingly, male patients with MUNW (90-day readmission: aOR = 1.19, 95% CI, 1.04–1.36; 30-day readmission: aOR = 1.18, 95% CI, 1.04, 1.34) and MUO (90-day readmission: aOR = 1.27, 95% CI, 1.03–1.56; 30-day readmission: aOR = 1.32, 95% CI, 1.08–1.61) had a significantly higher risk of NR/relapse compared with MHNW. Similarly, female patients with MUNW (90-day readmission: aOR = 1.11, 95% CI, 0.96–1.30; 30-day readmission: aOR = 1.09, 95% CI, 0.94–1.25) and MUO (90-day readmission: aOR = 1.58, 95% CI, 1.27–1.97; 30-day readmission: aOR = 1.57, 95% CI, 1.28–1.92) had a higher risk of NR/relapse compared with MHNW. Even though the statistical significance was not reached in the female patients with MUNW, the risk of NR/relapse seemed to be higher than in those with MHNW. Therefore, we found that patients with MUNW and MUO had a higher risk of NR/relapse risk at 90 and 30 days compared with MHNW, regardless of gender (Table S5 and S6 in Data S1). We found that patients with MUNW and MUO had a significantly higher risk of severe mortality and disease burden compared with MHNW, regardless of gender (Figure 2C,D). While MHNW female patients were used as the reference group, female patients with MHO had a higher risk of severe mortality (aOR = 1.55, 95% CI, 1.08–2.23) and disease burden (aOR = 1.33, 95% CI, 1.00–1.76). In comparison, male patients with MHO were not related to the risk of severe mortality and disease burden. Therefore, the severe mortality risk and disease burden in the MHO group were different between males and females (Figure 2C,D). Age subgroups demonstrated similar relationships to the total population after adjusting for multiple variables (Table 2, Tables S5 and S6 in Data S1).

3.3 Adverse outcomes in patients with CML in further metabolic obesity analysis

The risk of adverse outcomes gradually increased with the number of metabolic risk factors compared to patients without metabolic abnormalities. In addition, obesity exacerbated the negative effects of metabolic abnormalities on adverse outcomes of CML compared to patients with normal weight (Table S7 in Data S1). Meanwhile, similar trends were observed in the sensitivity analyses (Table S8 and S9 in Data S1). We also observed that the risk of adverse outcomes in patients with normal weight only with hyperglycemia and obesity only with hypertension was higher than in normal weight without metabolic abnormalities. Furthermore, patients with obesity only with hyperlipidemia had a higher risk of severe mortality (Table S10 in Data S1). In the sensitivity analyses, we also found that simple hypertension was an independent risk factor for adverse outcomes, irrespective of obesity (Tables S11 and S12 in Data S1).

3.4 Severity of illness in patients with CML with metabolically defined obesity

After adjusting for age, sex, total charges, length of stay, admission types, rehab transfer, same-day events, resident, insurance status, income, location of patient's residence, disposition of patient, and total comorbidities, patients with MUNW (aOR = 1.32; 95% CI, 1.17–1.49), MHO (aOR = 1.36; 95% CI, 1.09–1.70), and MUO (aOR = 1.79; 95% CI, 1.50–2.14) had increased major or extreme loss-of-function risk compared with MHNW. Similar trends were also observed in the sensitivity analyses (Table S13 in Data S1).