Multimorbidity in patients with monoclonal gammopathy of undetermined significance

Abstract

Monoclonal gammopathy of undetermined significance (MGUS), a precursor to multiple myeloma, is present in over 5% of adults aged 70 and older, a population with a high prevalence of multimorbidity. MGUS is often diagnosed incidentally when patients seek care for unrelated conditions. This study sought to examine patterns of multimorbidity among MGUS patients, as overall health may impact patient care and the prioritization of MGUS surveillance.

We examined patterns of comorbidities in 429 patients diagnosed with MGUS (2007–15) and 1,287 matched controls. Twenty-seven conditions were defined at diagnosis/index date using algorithms developed by the Centers for Medicare and Medicaid Chronic Conditions Warehouse. Patterns of common comorbidities were identified individually, in dyads and triads, and compared between MGUS cases and controls. We conducted a latent class analysis to identify comorbidity patterns among cases only. We also examined comorbidity patterns among a subset of 32 MGUS cases who progressed to cancer during the study period.

The most common comorbidities among both MGUS cases and controls included hypertension and hyperlipidemia. Anemia (Cases: 43%; Controls: 16%) and chronic kidney disease (CKD; Cases: 36%; Controls: 18%), and dyads and triads containing those conditions, were more common among cases. Latent class analysis identified three classes of comorbidity among MGUS cases: hypertension-hyperlipidemia plus anemia and CKD (31%); low comorbidity burden (17%); and hypertension-hyperlipidemia alone (52%). The higher prevalence among cases of anemia and CKD, which may be involved in the pathogenesis of, or surveillance for, MGUS, warrants additional investigation.

INTRODUCTION

Multiple myeloma, a malignancy of plasma cells, accounts for about 10% of new hematological cancer diagnoses in the US each year.¹ Although the 5-year survival rate for multiple myeloma has increased to 58% following improved treatment options since 2000, the disease remains incurable, with 95% of cases diagnosed at an advanced stage.^1–3 Multiple myeloma is preceded by the largely asymptomatic monoclonal gammopathy of undetermined significance (MGUS).^4–6 MGUS incidence increases with age, with retrospective studies estimating a prevalence of 5.3% in adults aged ≥70 years,⁷ and 7.5% in adults aged ≥85 years.^{7, 8,} Individuals with MGUS progress to multiple myeloma at an estimated rate of 1% per year, and have 25 times greater risk of developing multiple myeloma compared to the general population, yet few reliable markers predicting progression have been identified.^{9, 10} As a result of this uncertainty, patients with this premalignant condition endure the need for clinical follow-up every 6–12 months, along with related medical costs and heightened anxiety.^{6, 11, 12} With an aging US population, increasing numbers of MGUS and subsequent multiple myeloma diagnoses are likely over the coming decades.

Between 55–98% of all adults aged ≥65 experience multimorbidity, defined as two or more co-existing chronic diseases.¹³ Multimorbidity is likely common in MGUS patients at time of diagnosis, as the average age at MGUS diagnosis is 70 years,¹⁴ and patients tend to be diagnosed incidentally with MGUS when seeking care for other concerns. A recent study observed that, when compared to a control population without MGUS, patients with MGUS had shorter overall survival, with the majority of deaths due to cardiovascular and cerebrovascular disease.¹⁵ Furthermore, a population-based study confirmed previously-reported associations between 14 diseases and MGUS, including fracture, osteoporosis, neuropathy, and hyperlipidemia.¹⁶ Patterns of multimorbidity have recently been described among older US veterans (age ≥65 years) with multiple myeloma.¹⁷ In this analysis, patterns characterized as cardiovascular and metabolic disease, psychiatric and substance abuse disorders, chronic lung disease, and multisystem impairment were associated with higher overall mortality when compared to multiple myeloma patients with minimal comorbidity. However, it is unknown whether these same patterns of multimorbidity are prevalent among MGUS patients or non-veteran populations. In the clinic, multimorbidity may influence how MGUS patients are followed, and whether surveillance for MGUS progression is a priority. In this analysis, we sought to take the first step towards addressing these issues by characterizing multimorbidity at time of diagnosis in patients with MGUS, including the most prevalent comorbid conditions, common combinations, and clusters of conditions.

MATERIALS AND METHODS

Study Population

The study population consisted of 429 patients diagnosed with MGUS between January 2007 and December 2015 at a large multispecialty provider group in Worcester, Massachusetts. Cases were identified through a four step, case-finding algorithm developed by the study team based in electronic health record (EHR) data.¹⁸ Briefly, the algorithm identified all patients aged ≥50 years old who had at least two MGUS diagnosis codes (International Classification of Diseases (ICD)-9 273.1) in their EHR within 12 months, with the second diagnosis considered the index date to minimize rule-out diagnoses. All identified cases completed a serum or urine protein electrophoresis test and an immunofixation test within 90 days of the second MGUS diagnosis code and had ≥1 ambulatory oncologist visit within 90 days of MGUS diagnosis. Diagnosis and procedure codes were extracted from an EHR-based database. All patients had received care in the provider group for at least 12 months between January 2007 and December 2015 and did not have a prior diagnosis of multiple myeloma (ICD-O-3 morphology code 9732/3). A targeted, limited chart review of 206 algorithm-identified MGUS cases confirmed an MGUS diagnosis in 157 participants (Positive Predictive Value=76%), with cases adjudicated by a study oncologist specializing in hematological malignancies.

We also randomly selected a sample of 1,287 individuals aged ≥50 years who received care at the same provider group but did not have any MGUS diagnosis codes in their EHR within 1 year of index date to serve as a comparison group. Each case with MGUS was matched to three controls without MGUS by age (±2 years), sex, and length of enrollment in the health system (at least 12 months before and 6 months after index date). Patients without MGUS were assigned an index date equal to their matched MGUS cases’ earliest MGUS diagnosis date.

Data Source

This study used data extracted from EHRs and health claims, including diagnosis and procedure codes, and encounter types. Data were organized and standardized according to a common data model, the Health Care Systems Network (HCSRN) Virtual Data Warehouse (VDW).¹⁹ The VDW follows the specifications, definitions, and standards set by the HCSRN to facilitate pooling of data across healthcare systems participating in the HCSRN. Cancer diagnoses were primarily derived from the Massachusetts State Cancer Registry (1999–2017) and supplemented by an in-house cancer registry maintained by the medical group (2018–2020).

Definition of Variables

Twenty-seven chronic conditions were defined using ICD-9 and ICD-10 diagnosis codes according to algorithms developed by the Centers for Medicare and Medicaid Services (CMS) Chronic Conditions Warehouse (file years 1999–2020; available from: https://www2.ccwdata.org/web/guest/condition-categories).²⁰ Starting with diagnosis or index date, we looked back in time for relevant diagnosis codes as specified in the recommended reference period (between 1 and 3 years) for each algorithm. The algorithm for arthritis included a combination of ICD codes for both rheumatoid and osteoarthritis. We were unable to separately analyze the specific arthritis subtypes, as many patients diagnosed with rheumatoid arthritis also had codes indicating osteoarthritis. Cancer diagnoses were identified by ICD-O-3 morphology codes using tumor registry data. Multiple myeloma was defined by morphology code 9732.

Statistical Analysis

In descriptive analyses, we identified the most common individual comorbid diagnoses among MGUS cases and controls. We also examined the distribution of individual comorbid diagnoses separately by sex. We next determined the most common dyads and triads of comorbid conditions among the study population. Additionally, we calculated a Charlson comorbidity index score for all participants.²¹ We examined patterns of comorbidity separately among the full case population identified by the algorithm (n=429) and the subgroup of 157 validated MGUS cases. Chi-square tests were used to compare proportions of cases with each set of comorbid diagnoses to the proportion of controls with those same diagnoses.

We conducted a latent class analysis including the 14 most common comorbid conditions (≥10% prevalence in at least one case group) to identify different classes of comorbidities among cases and controls. We did not include 13 individual comorbid conditions with a prevalence among cases of <10% (Supplementary Table 1). We characterized each identified class by the predominant comorbid conditions and compared demographic and clinical characteristics of cases with the highest probability of belonging to each class. We compared the AIC and BIC for different numbers of classes (2–5) and decided on three classes.

In a secondary analysis, we investigated patterns of multimorbidity among 32 cases who progressed to cancer during the study period, and specifically among cases who progressed to multiple myeloma (66%). We also conducted a sensitivity analysis, excluding 11 cases who progressed to any type of cancer within 180 days of MGUS diagnosis (N=418).

RESULTS

The final study population included 429 MGUS cases identified by the initial case-finding algorithm, with a subset of 157 cases who were validated through chart review, and 1,287 controls. The demographic and clinical characteristics of the overall group of 429 MGUS cases were very similar to the subset of 157 validated cases (Table 1). Among both case groups, the mean age at diagnosis was about 75 years old, half the population was male, and the majority were white (85%). Controls were similar to the cases, with a slightly higher proportion of patients with other/unknown race.

Table 1.

Characteristics of the study population, patients diagnosed with monoclonal gammopathy of undetermined significance (MGUS) and matched controls in central Massachusetts, 2007–2015

Characteristic	Total MGUS cases (N=429)	Validated subset of MGUS cases (N=157)	Controlsa (N=1,287)
Age (mean, SD)	74.5 (10.4)	75.0 (10.3)	74.3 (10.4)
Sex
Female (N, %)	212 (49%)	78 (50%)	636 (49%)
Male	217 (51%)	79 (50%)	651 (51%)
Race
White	365 (85%)	132 (84%)	1041 (81%)
Black	6 (1%)	1 (1%)	19 (1%)
Other/Unknown	58 (14%)	24 (15%)	227 (18%)
Years of enrollment in health system prior to diagnosis or index date	6.1 (4.8)	5.9 (4.6)	5.9 (4.8)
Years of enrollment in health system after diagnosis or index date	3.4 (2.2)	3.4 (2.2)	3.5 (2.2)
Charlson comorbidity score
Mean (SD)	2.54 (2.54)	2.27 (2.30)	1.88 (2.14)
Median (Range)	2.00 (0–15)	2.00 (0–10)	1.00 (0–12)

^aControls matched to cases by age (±2 years), sex, and length of enrollment in the health system (at least 12 months before and 6 months after diagnosis/index date)

Abbreviations: SD – standard deviation

Hypertension and hyperlipidemia were the most common individual comorbid diagnoses among both MGUS cases and controls, followed by arthritis (Table 2). Hypertension and arthritis were significantly more common in cases (p<0.05). Interestingly, the prevalence of both anemia and chronic kidney disease, comorbidities that may signal MGUS-related disease activity, were significantly more common amongst cases (p<0.0001). There were no notable differences in the frequency of hyperlipidemia and hypertension between male and female cases or controls. In both cases (47% vs 41%) and controls (39% vs 30%), arthritis was slightly more common among women than men, and chronic kidney disease was more prevalent in men than women (cases: 41% vs 31%; controls: 22% vs 14%). Male cases were more likely to be diagnosed with anemia than female cases (47% vs 38%), although this was not observed among controls (males 17%; females 15%; Supplementary Table 4).

Table 2.

The most common individual comorbid conditions with a prevalence ≥10 percent among MGUS patients and matched controls, central Massachusetts 2007–2015

	All MGUS patients (N = 429)		Validated subset (N=157)		Controls (N=1,287)
Conditiona	N	Percentage	N	Percentage	N	Percentage
Hypertension	325	76%*	127	81%^	874	68%
Hyperlipidemia	314	73%	118	75%	899	70%
Arthritisb	189	44%^	68	43%*	448	35%
Anemia	183	43%^	65	41%^	207	16%
Chronic Kidney Disease	156	36%^	51	32.5%^	232	18%
Ischemic Heart Disease	128	30%	44	28%	351	27%
Diabetes	131	30.5%	43	27%	385	30%
Osteoporosis	90	21%^	29	18.5%	167	13%
COPD	77	18%*	26	17%	171	13%
Atrial Fibrillation	67	16%*	20	13%	148	11.5%
Acquired Hypothyroidism	59	14%	19	12%	191	15%
Depression	46	11%	18	11.5%	132	10%
Heart Failure	74	17%^	18	11.5%	137	11%
Benign Prostatic Hyperplasia	58	14%	14	9%	160	12%

^aComorbid conditions defined by algorithms from the Center for Medicare & Medicaid Services Chronic Conditions Data Warehouse (1999–2000; available from: https://www2.ccwdata.org/web/guest/condition-categories-chronic)

^bThe Arthritis category includes both rheumatoid and osteoarthritis

^*P-value from Chi-square test <0.05 comparing specified case category to controls

^{^}P-value from Chi-square test <0.001 comparing specified case category to controls

The most common dyads among cases and controls were hypertension-hyperlipidemia (62% cases; 57% controls), hypertension-arthritis (37% cases; 27% controls), and hyperlipidemia-arthritis (34% cases; 27% controls; Figure 1). Dyads with anemia and chronic kidney disease were more prevalent among cases (p<0.0001). Among triads of conditions (Figure 2), hyperlipidemia-hypertension-arthritis, a combination of the three most common individual comorbidities, and anemia-hyperlipidemia-hypertension were each present in 31% of cases, but only 23 and 11% of controls, respectively. The most common triad among controls was diabetes-hyperlipidemia-hypertension, which had a similar prevalence among cases (24–25%). Results among the larger case group were similar to those among the subgroup of validated cases. Exact frequencies and p-values from comparisons between cases and controls can be found in Supplementary Table 5.

Figure 1. Dyads of comorbid conditions among MGUS cases and controls in central Massachusetts, 2007–2015.

Black bar: Algorithm-identified MGUS cases (N=429); Medium gray bar: Validated MGUS cases (N=157); Light gray bar: Controls (N=1,287)

^a P ≤ 0.01 when comparing prevalence of the dyad between each case group and controls

^b P < 20.05 when comparing prevalence of the dyad between all cases and controls only

Figure 2. Triads of comorbid conditions among MGUS cases and controls in central Massachusetts, 2007–2015.

Black bar: Algorithm-identified MGUS cases (N=429); Medium gray bar: Validated MGUS cases (N=157); Light gray bar: Controls (N=1,287)

ᵅ P < 0.05 when comparing prevalence of the triad between each case group and controls

The latent class analysis identified three classes of comorbidities among cases (Table 3). The first class (31% of MGUS cases) was characterized by higher coefficients (≥0.70) for hypertension and hyperlipidemia, as well as anemia and chronic kidney disease, which may be indicative of MGUS-related disease processes. The second class (17% of cases) identified patients with low comorbidity count, with low correlation coefficients (≤0.32) for all conditions. The third class (52% of cases) was defined by hyperlipidemia and hypertension alone. Cases in class 1 were older (mean age 78.3 years), more likely to be male (60%), and had a higher Charlson comorbidity score (4.8) than cases who identified with the other classes. Cases in class 2 were the youngest (mean age 70), had the lowest Charlson score (0.9), and were slightly more likely to be diagnosed with multiple myeloma during follow-up (8%, compared to 5% in class 1 and 4% in class 3), although this difference is unlikely to be statistically significant due to the small number of myeloma diagnoses. Cases in class 3 were more likely to be female (56%) and white (90%).

Table 3.

Characteristics of 429 patients with MGUS according to pattern of multimorbidity as defined by latent class analysis (LCA)

Characteristic	LCA 1: “Myeloma-related comorbidities” (N=134)a	LCA 2: “Low comorbidity burden” (N=74)	LCA 3: “HL/HT only” (N=221)
Age (mean, SD)	78.3 (8.4)	70.0 (11.6)	73.6 (10.4)
Sex
Female (N, %)	53 (40%)	36 (49%)	123 (56%)
Male	81 (60%)	38 (51%)	98 (44%)
Race
White	111 (83%)	56 (76%)	198 (90%)
Black	3 (2%)	2 (3%)	1 (1%)
Other/Unknown	20 (15%)	16 (21%)	22 (10%)
Years of enrollment in health system prior to diagnosis or index date	6.8 (5.0)	6.6 (5.9)	6.0 (4.1)
Charlson comorbidity score
Mean (SD)	4.8 (2.7)	0.9 (1.3)	1.7 (1.7)
Median (Range)	4.00 (0–15)	1.00 (0–8)	1.00 (0–9)
Diagnosed with cancer during follow-up (N, %)	8 (6%)	7 (9%)	17 (8%)
Diagnosed with multiple myeloma during follow-upb	7 (5%)	6 (8%)	8 (4%)

SD – standard deviation

^aThe “myeloma-related comorbidities” pattern was characterized by hypertension, hyperlipidemia, anemia, and chronic kidney disease; the “low comorbidity burden” was characterized by low correlation coefficients with all 14 studied comorbid conditions; the “HL/HT only” pattern was characterized by hyperlipidemia and hypertension alone.

^bMultiple myeloma was identified by morphology code 9732 in tumor registry data following MGUS diagnosis date

Among the 32 MGUS cases who progressed to cancer during the study period, seven were diagnosed within 90 days of MGUS diagnosis, including three with multiple myeloma (Supplementary Table 2). Overall, these cases were diagnosed with cancer at a median of 343 days after MGUS diagnosis, with a median of 268 days to multiple myeloma (N=21) versus 562 days to other hematological cancers (N=11). Of the 11 patients with non-myeloma cancer, diagnoses included Waldenstrom macroglobulinemia and lymphoplasmacytic lymphoma (N=6), chronic lymphocytic leukemia/small lymphocytic lymphoma (N=2), and non-Hodgkin lymphoma (N=3). Patterns of multimorbidity in patients who progressed to cancer were similar to the entire MGUS patient population, with a slightly higher prevalence of anemia, but small numbers prohibited a more robust comparison (Supplementary Table 3). Of the seven patients who progressed to cancer within 90 days of MGUS diagnosis, five (71%) had existing anemia, but none had chronic kidney disease. Results from the latent class analysis for these 32 participants was similar to the larger population, with 25% defined by the class including anemia and chronic kidney disease, 22% by the low-comorbidity class, and 53% by the class featuring hyperlipidemia and hypertension.

Results from the sensitivity analysis excluding 11 cases who progressed to any type of cancer within 180 days of MGUS diagnosis (N=418) were nearly identical to the primary analysis, including the distribution of individual, dyads, and triads of comorbid conditions (Supplementary Table 6).

DISCUSSION

To our knowledge this is the first study to characterize patterns of multimorbidity among patients in the US diagnosed with MGUS. Although MGUS is often diagnosed incidentally when patients visit their physician for symptoms of other conditions, the distribution of comorbid conditions present at MGUS diagnosis has not been well-studied. MGUS is most commonly diagnosed among adults aged 50 years or older, an age group where multimorbidity is increasingly prevalent. Thus, it is not surprising that the most common comorbid conditions in both cases and controls were similar, including hypertension, hyperlipidemia and arthritis; however, the prevalence of hypertension and arthritis was statistically significantly higher among cases. The higher prevalence of hypertension and arthritis among MGUS cases could signal overall closer clinical surveillance among this patient population. However, we also found that anemia and chronic kidney disease were significantly more prevalent among cases than controls. Anemia is often the first symptom of multiple myeloma, with one study estimating that nearly 70% of patients with multiple myeloma have anemia at diagnosis.²² Future research should build upon this first step to determine if these patterns of comorbidity may be associated with MGUS progression, or if the comorbid diagnoses were related to the process of diagnosing MGUS. It is possible that symptoms from these comorbid diagnoses led care providers to order tests that are also used to diagnose MGUS, and thus testing bias may contribute to the overall higher burden of comorbidity in MGUS cases.

An analysis of the population-based Iceland Screens, Treats, or Prevents Multiple Myeloma (iStopMM) study found that patients with a clinical diagnosis of MGUS had a higher number of comorbidities, and a higher prevalence of certain comorbidities including chronic kidney disease, than those people whose MGUS was detected through screening conducted during the study.²⁵ This observation suggests that patterns of comorbidity in clinically diagnosed MGUS cases may not represent all people with MGUS, the majority of whom go undetected. As a result, our study results may be impacted by selection bias, and may only apply to patients with clinically diagnosed MGUS. In addition, our results are in line with the recent analysis of patterns of comorbidity in multiple myeloma patients at the VA, where patterns defined by cardiovascular and metabolic disease, and by hypertension, hyperlipidemia, and arthritis were the most common.¹⁷ However, other prevalent comorbidities in our non-veteran MGUS population differed from the multiple myeloma study, which observed a higher prevalence of psychiatric and substance use disorders and chronic lung disease. In addition, the prevalence of hypertension among patients with multiple myeloma is between 38–47% before multiple myeloma treatment.^{23, 24} Furthermore, monoclonal gammopathy of renal significance was recently classified as its own diagnosis due to an increase in kidney diseases associated with MGUS that do not meet the criteria for multiple myeloma.²⁶ Although there was a higher prevalence of chronic kidney disease in cases than controls in our population, we were unable to confirm if this was due to diagnoses of monoclonal gammopathy of renal significance, as currently there is no unique ICD diagnosis code for the condition. In addition, it is possible that providers treating patients with prevalent kidney disease may order a protein electrophoresis test to investigate whether other pathology (e.g., MGUS) may be contributing to symptoms.

This study was among the first to examine multimorbidity in detail within the context of patients with MGUS diagnosed clinically in the US, using EHR data to identify diagnoses. A strength of this analysis is the use of comprehensive EHR data, and the ability to examine participants’ medical history for diagnoses of comorbid conditions, with a median enrollment of 6 years prior to diagnosis/index date. Cases and controls in this population were members of a care provider group where patients tend to stay with the provider group for many years, allowing us to follow patients for the development of malignancies. Comorbidities were identified using algorithms from the CMS Chronic Conditions Data Warehouse that have been validated in the literature or defined by criteria used in federal sources.²⁰ However, exact dates of diagnosis for the comorbid conditions and MGUS were at times difficult to discern using EHR data without extensive review of individual records, which was beyond the scope of this study. As a result, a small amount of misclassification around timing of diagnosis date is possible but should not impact the findings of the study. We were unable to further investigate relationships with M-protein levels or isotype in MGUS cases, as text-based results from serum and urine protein electrophoresis tests were not available in the database.

Our study was restricted to patients with access to care from one geographic region in the US, and who are largely white, and as a result may not be generalizable to all patients with MGUS. Our findings should be replicated in larger and more geographically and demographically diverse populations of patients diagnosed with MGUS.

In conclusion, this analysis identified several patterns of comorbid conditions that were prevalent among patients diagnosed with MGUS in a central Massachusetts provider group. While some common comorbidities, such as hypertension and hyperlipidemia, were also highly prevalent among controls without MGUS, we observed a higher frequency of anemia and chronic kidney disease among MGUS cases. As these latter conditions may signal possible progression of MGUS towards multiple myeloma, these findings should be further investigated to determine if certain subgroups of patients could be identified who would benefit from closer clinical surveillance. Analyses of larger populations with longer follow-up periods and increased statistical power should further investigate possible associations between comorbidity patterns and later progression to multiple myeloma and other related malignancies.

Novelty and Impact:

This is the first study to characterize patterns of multimorbidity among patients with monoclonal gammopathy of undetermined significance (MGUS), a precursor to multiple myeloma. The most common comorbid conditions among all study participants included hypertension, hyperlipidemia, and arthritis, although hypertension and arthritis were significantly more prevalent among MGUS cases than controls. Notably, anemia and chronic kidney disease, and combinations of conditions including these diagnoses, were significantly more prevalent among patients with MGUS than among controls.

Abbreviations:

CKD

chronic kidney disease

CMS

Centers for Medicare and Medicaid Services

COPD

chronic obstructive pulmonary disease

EHR

electronic health record

HCSRN

Health Care Systems Research Network

ICD

International Classification of Diseases

LCA

latent class analysis

MGUS

monoclonal gammopathy of undetermined significance

SD

standard deviation

VDW

Virtual Data Warehouse

Data Availability Statement

The data generated in this study are available within the article and its supplementary files. Further information is available from the corresponding author upon request.