Association Between DPP4 Inhibitor Use and the Incidence of Cirrhosis, ESRD, and Some Cancers in Patients With Diabetes

Abstract

Context

There are relatively few data on noncardiovascular (non-CV) long-term clinical outcomes of dipeptidyl peptidase 4 inhibitor (DPP4i) treatment.

Objective

We aimed to evaluate some non-CV effects of DPP4is in patients with diabetes.

Methods

Based on data from the National Health Insurance Service database in Korea (2007-2018), we conducted 3 pairwise comparisons of metformin-combined antidiabetic therapies in adult patients with diabetes: DPP4is vs (1) all other oral antidiabetic agents, (2) sulfonylureas/glinides, and (3) thiazolidinediones (TZDs). Major outcomes were liver cirrhosis, end-stage renal disease (ESRD), and cancers in the liver, kidney, and pancreas. Adjusted hazard ratios (HRs) and 95% CIs for the outcomes were estimated using an adjusted Cox model.

Results

Of the 747 124 patients included, 628 217 had received DPP4i therapy for a mean duration of 33.8 ± 25.0 months. Compared with TZD therapy, DPP4i therapy was associated with higher adjusted HRs [95% CIs] for liver cirrhosis (1.267 [1.108-1.449]), ESRD (1.596 [1.139-2.236]), liver cancer (1.117 [1.011-1.235]), and pancreatic cancer (1.158 [1.040-1.290]). Furthermore, apart from liver cirrhosis, a higher risk of each of these outcomes was associated with DPP4i use than with non-DPP4i use. The higher adjusted HRs associated with DPP4i use further increased when patients with long-term exposure to DPP4is were analyzed.

Conclusion

DPP4i therapy in patients with diabetes was associated with a higher risk of liver cirrhosis and cancer, ESRD, and pancreatic cancer than TZD therapy and, except for liver cirrhosis, the risk of these outcomes was greater with DPP4i treatment than with non-DPP4i treatment.

big data, cancers, diabetes mellitus, dipeptidyl peptidase 4 inhibitors, end stage renal disease, liver cirrhosis

Dipeptidyl peptidase 4 (DPP4) inhibitors (DPP4is) are the most frequently prescribed antidiabetic agent either in initial combination with or as an add-on agent to metformin treatment in patients with type 2 diabetes (1). DPP4 itself has proinflammatory and profibrotic effects via its enzymatic actions on many substrates and through other direct and indirect pathways (2–4). In subjects with insulin resistance, obesity, or diabetes, the expression and activity of DPP4 in various tissues and cells as well as the circulating levels of soluble DPP4 are increased (5–7). Therefore, in patients with diabetes, chronic low-grade inflammation, metabolic stress- and inflammation-related cell injury and death, tissue repair, and profibrotic progression in conditions such as liver cirrhosis and end-stage renal disease (ESRD) may be related to increased DPP4 levels and activity (8, 9). Increased serum soluble DPP4 levels were shown to be associated with nonalcoholic steatohepatitis and liver fibrosis in patients with type 2 diabetes (9). Markedly increased expression/activity of DPP4 has also been observed in human cirrhotic nodules and liver cancer tissue (10, 11), which may be related to extracellular matrix collagen binding and cytokine activation. In patients with diabetic kidney disease, DPP4 expression in glomerular podocytes and DPP4 activity in urine were reported to be increased compared with control subjects (12).

DPP4is have been shown to have antiinflammatory, antifibrotic, and even antitumor effects in many preclinical studies (2, 13–15). Although an increased incidence of cancers specific to several organs has been reported in relation to type 2 diabetes, obesity, or specific diet patterns (16), the effects of DPP4is on chronic fibrotic tissue remodeling, advanced organ-specific diseases, and cancer development have not been studied systematically in patients with type 2 diabetes. Major clinical trials involving DPP4is have failed to show a decrease in major adverse cardiovascular (CV) events in patients with type 2 diabetes despite moderate glucose control and a low risk of hypoglycemia or weight gain (17–19). This issue has become a very difficult riddle to solve. Therefore, we believe it is appropriate to question whether DPP4is have beneficial effects on other major non-CV clinical outcomes. A large-scale clinical study capable of addressing all of these issues would be unrealistic. Currently, real-world big data may be a useful alternative. The National Health Insurance Service (NHIS) in Korea delivers a government-controlled single-payer obligatory insurance plan that covers almost the entire Korean population of approximately 50 million (20). Encrypted customized data tailored for a specific study can be extracted from the NHIS data, which consist of 6 major files, namely, general information, health care services (including inpatient prescriptions), diagnoses, outpatient prescriptions, drug master, and provider information (20).

To determine whether various non-CV beneficial effects of DPP4is that have been observed in many experimental or preclinical studies can be translated into actual clinical practice, we evaluated the effects of DPP4i therapy on the incidence of liver cirrhosis and ESRD as well as cancer incidence in these specific organs in patients with diabetes by analyzing data from the NHIS database. As a parallel study, we also evaluated the incidence of pancreatic cancer in relation to the use of DPP4is.

Materials and Methods

Design, Setting, and Participants

Information on patients with new-onset diabetes after the age of 30 years between January 1, 2004, and December 31, 2018, was extracted from the NHIS database. Then, we randomly selected 50% of patients from the retrospective cohort. The exclusion criteria were as follows: missing major demographic variables, outcome events before or within 30 days of study medication, and type 1 diabetes. To ensure a washout period, we also excluded patients with every type of diabetes diagnosed before 2007. The diagnoses were coded in accordance with the Korean Standard Classification of Diseases Version 6, which is based on the International Classification of Diseases 10th Revision (20). Three pairwise comparisons between DPP4is and other oral antidiabetic medication(s) were performed for treatments involving use of these agents for 90 days or more in combination therapy with metformin (for 30 days or more) as follows: (1) vs other oral antidiabetic agent(s) except for DPP4is; (2) vs sulfonylureas/glinides (SUR/Gs); and (3) vs thiazolidinediones (TZDs). The classes of other oral antidiabetic agents included in the present study were metformin, SUR/Gs, TZDs, alpha-glucosidase inhibitors, and sodium-glucose cotransporter 2 inhibitors (SGLT2is). The main analysis of the 3 active comparator groups and respective DPP4i treatment groups included patients who received allocated therapy for 90 days or more, whereas, in the sensitivity analyses, the treatment period was extended to at least 365 days.

Data on the prescription of DPP4is and other drugs were extracted by using drug codes based on the Anatomical Therapeutic Chemical Classification in the claim data of the study period and were analyzed in conjunction with other information, including demographic parameters, comorbidities, and procedures, based on the patients’ medical records (21).

The primary outcomes of the present study were the incidence of liver cirrhosis, ESRD, and liver and kidney cancers. The secondary outcome was the incidence of pancreatic cancer. To support the robustness of the NHIS data, our study design, and the results, we predefined a negative control outcome, lung cancer, for which evidence of a relationship with DPP4is or other antidiabetic agents is lacking and has few specific concerns associated with diabetes (22). The study outcomes were defined if predefined requirements were met as detailed in Table 1: (1) an International Classification of Diseases 10th Revision code corresponding to the diseases was available; (2) a specific disease-related procedure code was available (for example, for hemodialysis); and (3) a condition was supported by a special Korean NHIS program under the copayment assistance policy that covers rare, incurable, malignant, or severe and burdensome diseases (21, 23). Smoking status and alcohol intake data were obtained from questionnaires completed during biennial health check-ups; however, data were available for only approximately one-third of all patients.

Table 1.

Study outcome definitions used for data extraction

Outcomes	Definition	Exclusion
Lung cancer	C34	Preexisting lung cancer
Liver cirrhosis	K74	Preexisting liver cirrhosis or liver cancer
Liver cancer	C22	Preexisting liver cirrhosis or liver cancer
ESRD	N185, HD, PD, or KT	Preexisting ESRD or kidney cancer
Kidney cancer	C64	Preexisting kidney cancer
Pancreatic cancer	C25	Preexisting pancreatic cancer

Abbreviations: ESRD, end-stage renal disease; HD, hemodialysis; KT, kidney transplantation; PD, peritoneal dialysis.

Ethics Committee Approval

The study was reviewed by the institutional review board of the Gachon University Gil Medical Center, which decided that the study protocol qualified for waiving ethics approval in compliance with governmental laws and regulations (protocol GFIRB2019-425). The requirement for informed consent was also waived because we only accessed deidentified, previously collected data.

Statistical Analysis

Baseline characteristics were compared according to the DPP4i use status. Categorical variables were expressed as numbers and percentages, and continuous variables were expressed as the mean ± SD. The tests for single variables between groups were performed through t tests and χ² tests. For the outcome analyses, we used time-dependent Cox proportional hazards models to estimate crude and adjusted hazard ratios (HRs) and 95% CIs of each outcome associated with the use of DPP4is vs non-DPP4is, SUR/Gs, and TZDs. We adjusted all models for the potential confounders listed below: age, sex, body mass index (BMI), specific cotreatments (insulins, statins, and renin-angiotensin-aldosterone-system [RAAS] blockers), year of study medication (median year), smoking status, alcohol drinking, and the use of SGLT2is. In the sensitivity analyses, we repeated the primary analyses by including patients who were prescribed DPP4is or comparators for at least 365 days. SAS EG (SAS Institute Inc., Cary, NC, USA) and R version 3.5.2 (R Foundation for Statistical Computing, Vienna, Austria) software were used for the analyses. J.J. and D.H.L. had full access to all the data in the study and assume responsibility for the integrity of the data and the accuracy of the data analysis.

Results

Characteristics of the Study Population

From the NHIS database, we randomly extracted customized data from 50% of the total patients with diabetes by age (≥30 years) and sex between January 1, 2004, and December 31, 2018 (n = 7 535 003). As presented in the flow chart in Fig. 1, we excluded 6 787 879 patients according to the following exclusion criteria: (1) patients who were prescribed antidiabetic drugs in combination with metformin for less than 30 days (N = 5 702 037); (2) patients who were diagnosed with diabetes before 2007 and patients with type 1 diabetes (N = 1 026 458); and (3) patients who were prescribed the study drugs for less than 90 days (N = 59 384).

Figure 1.

Flow diagram of inclusion of participants in the study.

Therefore, 747 124 patients were ultimately included in the present study. Of these, 628 217 patients were included in the DPP4i group, whereas 118 907 patients were included in the comparator group. When we analyzed the distribution of specific types of diabetes after the exclusion of type 1 diabetes in the whole cohort, the proportion of definite or presumed type 2 diabetes was approximately 97.3%. Thus, nearly all of the study subjects in the present study were patients with type 2 diabetes.

The clinical characteristics of the study subjects by oral antidiabetic agent (DPP4i and non-DPP4i agents) are summarized in Table 2. The mean ages of the 2 groups were 56.6 ± 11.6 and 58.4 ± 12.4 years at enrollment. The mean duration of diabetes at entry and the mean duration of exposure to DPP4is in the DPP4i group were 6.94 ± 3.25 years and 2.82 ± 2.08 years, respectively. During the follow-up period, significant proportions of patients in the DPP4i group and non-DPP4i group had been exposed to other oral antidiabetic agents: SUR/Gs 63.5% vs 79.8%, respectively, P < 0.0001; TZDs 21.8% vs 19.2%, respectively, P < 0.001; and SGLT2is 11.8% vs 22.0%, respectively, P < 0.001. Approximately one-third of patients in both groups had been exposed to insulin therapy during the follow-up period.

Table 2.

Clinical characteristics of patients with diabetes by DPP4i treatment status

Parameters	DPP4is	Non-DPP4is	P values
Sample size, n (%)	628 217 (100)	118 907 (100)
Male sex, n (%)	382 880 (61.0)	71 998 (60.6)	0.0101
Age at entry, years	56.6 (11.6)	58.4 (12.4)	<0.0001
BMI, kg/m2	26.0 (3.6)	26.1 (3.8)	<0.0001
Duration of diabetes, y	6.94 (3.25)	6.42 (3.34)	<0.0001
Smoking, n (%)			<0.0001
Never	73 573 (11.7)	13 721 (11.5)
Past	28 577 (4.6)	5229 (4.4)
Current	105 455 (16.8)	18 311 (15.4)
No data	420 612 (67.0)	81 646 (68.7)
Alcohol drinking, n (%)			<0.0001
≥2 days or ≥1 bottle/wk	61 052 (9.7)	11 437 (9.6)
Less	146 553 (23.3)	25 824 (21.7)
No data	420 612 (67.0)	81 646 (68.7)
Comedications
Insulins, n (%)	210 730 (33.5)	36 507 (30.7)	<0.0001
Statins, n (%)	497 485 (79.2)	81 746 (68.8)	<0.0001
RAAS blockers, n (%)	387 172 (61.6)	75 111 (63.2)	<0.0001
Healthcare supported by
National health insurance	587 723 (93.6)	109 291 (91.9)	NS
Medical aid program	38 721 (6.2)	9263 (7.8)	NS
Period of medication usage, months
DPP4is
n (%)	628 217 (100)	—
Months	33.8 (25.0)	—
Metformin
n (%)	628 217 (100)	118 907 (100)	1.0000
Months	50.6 (34.2)	42.0 (33.1)	<0.0001
SUR/Gs
n (%)	398 780 (63.5)	94 859 (79.8)	<0.0001
Months	35.1 (34.6)	38.3 (34.9)	<0.0001
TZDs
n (%)	136 660 (21.8)	22 860 (19.2)	<0.0001
Months	17.1 (18.3)	27.0 (28.2)	<0.0001
SGLT2is
n (%)	74 371 (11.8)	26 175 (22.0)	<0.0001
Months	12.2 (11.2)	16.9 (12.0)	<0.0001

Data are expressed as the mean (SD) or n (%), unless otherwise specified.

Abbreviations: BMI, body mass index; DPP4is, dipeptidyl peptidase 4 inhibitors; NS, not significant; RAAS, renin-angiotensin-aldosterone-system; SGLT2is, sodium-glucose cotransporter 2 inhibitors; SUR/Gs, sulfonylureas/glinides; TZDs, thiazolidinediones.

The clinical characteristics of subgroups of comparators (ie, TZDs and SGR/Gs) among non-DPP4i users are also summarized in Table 3. Although comparator drugs were also prescribed in a significant proportion of the DPP4i group, each target drug exposure period in the TZD and SUR/G subgroups was longer in the comparator groups than in the DPP4i group. In addition, a key difference between the DPP4i treatment group and the 2 comparative drug groups was whether they used DPP4is.

Table 3.

Clinical characteristics of patients with diabetes by SUR/G and TZD treatment status in comparison to the DPP4i group

Parameters	DPP4is	SUR/Gs	P valuesa	TZDs	P valuesa
Sample size, n (%)	628 217 (100)	88 172 (100)		19 117 (100)
Male, n (%)	382 880 (61.0)	53 773 (61.0)	0.8220	11 789 (61.7)	0.0442
Age at entry, y	56.6 (11.6)	59.7 (12.3)	<0.0001	58.3 (11.6)	<0.0001
BMI, kg/m2	26.0 (3.6)	25.8 (3.7)	<0.0001	26.1 (3.6)	0.0017
Duration of diabetes, y	6.94 (3.25)	6.71 (3.24)	<0.0001	7.25 (3.23)	<0.0001
Smoking, n (%)			<0.0001		0.8970
Never	73 573 (11.7)	9782 (11.1)		2380 (12.5)
Past	28 577 (4.6)	3804 (4.3)		1003 (5.3)
Current	105 455 (16.8)	13 103 (14.9)		3080 (16.1)
No data	420 612 (67.0)	61 483 (69.7)		12 654 (66.2)
Alcohol drinking, n (%)			<0.0001		0.8970
≥2 days or ≥1 bottle/wk	61 052 (9.7)	8375 (9.5)		2009 (10.5)
Less	146 553 (23.3)	18 314 (20.8)		4454 (23.3)
No data	420 612 (67.0)	61 483 (69.7)		12 654 (66.2)
Comedications
Insulins, n (%)	210 730 (33.5)	29 747 (33.7)	0.2522	5318 (27.8)	<0.0001
Statins, n (%)	497 485 (79.2)	58 008 (65.8)	<0.0001	14 793 (77.4)	<0.0001
RAAS  Blockers, n (%)	387 172 (61.6)	56 618 (64.2)	<0.0001	12 411 (64.9)	0.0366
Healthcare supported by
National health insurance	587 723 (93.6)	80 402 (91.2)	<0.0001	17 641 (92.3)	<0.0001
Medical aid program	38 721 (6.2)	7495 (8.5)		1421 (7.4)
Period of medication usage, months
DPP4is
n (%)	628 217 (100)	—	—	—	—
Months	33.8 (25.0)	—	—	—	—
Metformin
n (%)	628 217 (100)	88 172 (100)	<0.0001	19 117 (100)	1.0000
Months	50.6 (34.2)	45.4 (33.7)	<0.0001	53.0 (35.8)	<0.0001
SUR/Gs
n (%)	398 780 (63.5)	88 172 (100)	<0.0001	11 534 (60.3)	<0.0001
Months	35.1 (34.6)	40.7 (34.8)	<0.0001	42.3 (39.1)	<0.0001
TZDs
n (%)	136 660 (21.8)	12 335 (14.0)	<0.0001	19 117 (100)	<0.0001
Months	17.1 (18.3)	23.9 (26.2)	<0.0001	32.7 (29.3)	<0.0001
SGLT2is
n (%)	74 371 (11.8)	7880 (8.9)	1.0000	1764 (9.2)	<0.0001
Months	12.2 (11.2)	12.2 (11.2)	<0.0001	16.8 (13.2)	<0.0001

Data are expressed as the mean (SD) or n (%), unless otherwise specified.

Abbreviations: BMI, body mass index; DPP4is, dipeptidyl peptidase 4 inhibitors; RAAS, renin-angiotensin-aldosterone-system; SGLT2is, sodium-glucose cotransporter 2 inhibitors; SUR/Gs, sulfonylureas/glinides; TZDs, thiazolidinediones.

a

Versus DPP4i-treated group.

Primary Outcomes Associated With DPP4I Therapy in the Liver and Kidney

Considering that clinical characteristics were not statistically the same between the treatment groups, presumably because of the large sample size, we addressed multiple adjusted results only in the present study.

Regarding liver outcomes, although the DPP4i group was associated with a comparable risk of liver cirrhosis relative to the entire non-DPP4i group and the SUR/G group, after adjustment for multiple factors, there was an increased risk of liver cirrhosis compared with the TZD group (Table 4, Fig. 2A and 3A). In addition, DPP4i therapy was associated with a higher risk of liver cancer compared with the entire non-DPP4i group and the TZD group after adjustment for multiple factors (Table 4; Fig. 2B and 3B). Thus, as a combination therapy in patients with diabetes, DPP4is showed a consistent inferiority to TZDs in major liver outcomes (cirrhosis and cancer).

Figure 2.

Multiple adjusted survival probability plots for study outcomes related to DPP4i therapy versus non-DPP4i therapy in patients with diabetes. Both treatment arms included patients treated with metformin in combination. The number of patients at risk in each arm is indicated. *Adjusted for potential confounders listed below: age, sex, BMI, specific cotreatments (insulins, statins, and RAAS blockers), year of study medication (median year), smoking status, alcohol drinking, and the use of SGLT2is.

Figure 3.

Multiple adjusted survival probability plots for study outcomes related to DPP4i therapy versus TZD therapy in patients with diabetes. Both treatment arms included patients treated with metformin in combination. The number of patients at risk in each arm is indicated. *Adjusted for potential confounders as in Fig. 2.

Table 4.

Association between DPP4i use and major outcomes in patients with diabetes

Treatment groups and outcomes	Unadjusted HRs			Adjusted HRsa
	HRs	95% CIs	P value	HRs	95% CIs	P value
Liver cirrhosis
DPP4is vs non-DPP4is	0.840	0.803-0.879	<0.0001	1.045	0.991-1.103	0.1039
DPP4is vs SUR/Gs	0.784	0.748-0.822	<0.0001	1.003	0.947-1.062	0.9259
DPP4is vs TZDs	1.218	1.118-1.438	0.0002	1.267	1.108-1.449	0.0005
Liver cancer
DPP4is vs non-DPP4is	0.926	0.892-0.960	<0.0001	1.054	1.010-1.101	0.0157
DPP4is vs SUR/Gs	0.889	0.855-0.925	<0.0001	1.035	0.988-1.084	0.1432
DPP4is vs TZDs	1.156	1.051-1.270	0.0027	1.117	1.011-1.235	0.0396
ESRD
DPP4is vs non-DPP4is	1.039	0.937-1.152	0.4667	1.258	1.101-1.437	0.0007
DPP4is vs SUR/Gs	0.982	0.881-1.095	0.7465	1.250	1.086-1.438	0.0018
DPP4is vs TZDs	1.817	1.323-2.495	0.0002	1.596	1.139-2.236	0.0066
Kidney cancer
DPP4is vs non-DPP4is	1.064	0.943-1.200	0.3122	1.157	1.004-1.333	0.0439
DPP4is vs SUR/Gs	1.035	0.909-1.178	0.6035	1.143	0.981-1.333	0.0873
DPP4is vs TZDs	1.176	0.874-1.582	0.2854	1.191	0.866-1.640	0.2827
Pancreatic cancer
DPP4is vs non-DPP4is	1.065	1.022-1.109	0.0027	1.129	1.077-1.183	<0.0001
DPP4is vs SUR/Gs	1.051	1.006-1.098	0.0264	1.129	1.073-1.189	<0.0001
DPP4is vs TZDs	1.224	1.106-1.356	<0.0001	1.158	1.040-1.290	0.0074

Abbreviations: DPP4is, dipeptidyl peptidase 4 inhibitors; HRs, hazard ratios; RAAS, renin-angiotensin-aldosterone-system; SGLT2is, sodium-glucose cotransporter 2 inhibitors; SUR/G, sulfonylureas/glinides; TZDs, thiazolidinediones.

a

All models were adjusted for the potential confounders listed below: age, sex, body mass index, specific cotreatments (insulins, statins, and RAAS blockers), year of study medication (median year), smoking status, alcohol drinking, and the use of SGLT2is.

With regard to kidney outcomes, compared with non-DPP4i, SUR/G, and TZD users, all of them in combination with metformin after adjustments for multiple parameters, patients treated with a DPP4i and metformin combination had an increased risk of ESRD (Table 4; Fig. 2C and 3C). Compared with the entire non-DPP4i group, the DPP4i group was associated with an increased risk of kidney cancer (Table 4; Fig. 2D).

Because our primary outcomes in the liver and kidney might not be independent of comorbidities or unknown factors, we also predefined a negative outcome control (lung cancer). Currently, there is no evidence or clinical implication for an association between antidiabetic agents and lung diseases, including lung cancer (24). Compared with non-DPP4i users, SUR/G users, and TZD users, all of whom used the drugs in combination with metformin, patients treated with a DPP4i and metformin combination did not have an altered incidence of lung cancer after adjustments for multiple parameters (Table 5; Figs. 2E and 3E).

Table 5.

Association between DPP4i use and lung cancer in patients with diabetes in the main analysis and sensitivity analysis

Treatment groups and outcomes	Unadjusted HRs			Adjusted HRs
	HRs	95% CIs	P value	HRs	95% CIs	P value
Main analysis
DPP4is vs non-DPP4is	0.752	0.718-0.787	<0.0001	0.989	0.936-1.045	0.7033
DPP4is vs SUR/Gs	0.707	0.673-0.742	<0.0001	0.981	0.925-1.040	0.5157
DPP4is vs TZDs	0.954	0.849-1.073	0.4368	0.961	0.849-1.088	0.5802
Sensitivity analysis
DPP4is vs non-DPP4is	0.754	0.716-0.794	<0.0001	0.999	0.939-1.064	0.9867
DPP4is vs SUR/Gs	0.715	0.677-0.756	<0.0001	0.993	0.929-1.062	0.8438
DPP4is vs TZDs	0.945	0.824-1.082	0.4107	0.945	0.820-1.088	0.4305

Abbreviations: DPP4is, dipeptidyl peptidase 4 inhibitors; HRs, hazard ratios; SUR/Gs, Sulfonylureas/glinides; TZDs, thiazolidinediones.

a

All models were adjusted for the potential confounders as in Table 2.

DPP4i Therapy in Patients With Diabetes was Associated With a Higher Risk of Pancreatic Cancer

DPP4i and metformin combination therapy was consistently associated with a higher risk of pancreatic cancer compared with the metformin-combined non-DPP4i, SUR/G, and TZD therapies after adjustments for multiple parameters (Table 4; Figs. 2F and 3F).

Sensitivity Analyses

In the main analysis, we included patients with diabetes who received DPP4i therapy for 3 months or more and in combination with metformin for at least a month. We performed a sensitivity test in which we included patients who received DPP4i or comparator therapy for 1 year or more and in combination with metformin for at least a month. The clinical characteristics of the study subjects by oral antidiabetic agent (DPP4i and non-DPP4i agents) are summarized in Table 6.

Table 6.

Clinical characteristics of patients with diabetes by DPP4i treatment status included in the long-term observation cohort (≥365 days)

Parameters	DPP4is	Non-DPP4is	P value
Sample size, n (%)	481 105 (100)	118 687 (100)
Male, n (%)	289 846 (60.3)	71 848 (60.5)	0.0676
Age at entry, y	56.6 (11.6)	58.4 (12.4)	<0.0001
BMI, kg/m2	25.9 (3.6)	26.1 (3.8)	<0.0001
Duration of diabetes, y	7.42 (2.99)	6.42 (3.34)	<0.0001
Smoking, n (%)			<0.0001
Never	57.150 (11.9)	13 674 (11.5)
Past	22 407 (4.7)	5.217 (4.4)
Current	80 535 (16.7)	18 281 (15.4)
No data	321 013 (66.7)	81 495 (68.7)
Alcohol drinking, n (%)			<0.0001
≥2 days or ≥1 bottle/wk	47 001 (9.8)	11 417 (9.6)
Less	113 091 (23.5)	25 775 (21.7)
No data	321 013 (66.7)	81 495 (68.7)
Comedications
Insulins, n (%)	166 170 (34.5)	36 398 (30.7)	<0.0001
Statins, n (%)	389 927 (81.1)	81 602 (68.8)	<0.0001
RAAS Blockers, n (%)	305 433 (63.5)	74 962 (63.2)	0.0366
Healthcare supported by
National health insurance	449 674 (93.5)	109 103 (91.9)	NS
Medical aid program	30 126 (6.3)	9231 (7.8)	NS
Period of medication usage, months
DPP4is
n (%)	481 105 (100)	—
Months	41.88 (22.99)	—
Metformin
n (%)	481.105 (100)	118 687 (100)	1.0000
Months	58.18 (32.3)	41.95 (33.1)	<0.0001
SUR/Gs
n (%)	323 589 (67.3)	94 673 (79.8)	<0.0001
Months	37.89 (34.92)	38.35 (34.9)	<0.0001
TZDs
n (%)	115 593 (24.0)	22 823 (19.2)	<0.0001
Months	17.88 (18.19)	27.07 (28.19)	<0.0001
SGLT2is
n (%)	56 281 (11.7)	26 148 (22.0)	<0.0001
Months	12.04 (11.03)	16.88 (12.04)	<0.0001

Data are expressed as the mean (SD) or n (%), unless otherwise specified.

Abbreviations: BMI, body mass index; DPP4is, dipeptidyl peptidase 4 inhibitors; NS, not significant; RAAS, renin-angiotensin-aldosterone-system; SGLT2is, sodium-glucose cotransporter 2 inhibitors; SUR/Gs, sulfonylureas/glinides; TZDs, thiazolidinediones.

Compared with TZD therapy, DPP4i therapy for a longer period was associated with an increased risk of liver cirrhosis (1.368; 95% CI, 1.184-1.581; P < 0.0001), liver cancer (1.150; 95% CI, 1.035-1.278; P = 0.0093), ESRD (2.206; 95% CI, 1.501-3.243; P < 0.0001), and pancreatic cancer (1.204; 95% CI, 1.061-1.366; P = 0.0040) after adjusting for multiple factors (Table 7). Additionally, the DPP4i group had increased adjusted HRs for liver cancer (vs non-DPP4is 1.073; 95% CI, 1.022-1.127; P = 0.0044 vs SUR/Gs 1.055; 95% CI, 1.001-1.111; P = 0.0470), ESRD (vs non-DPP4is 1.345; 95% CI, 1.154-1.567; P = 0.0002 vs SUR/Gs 1.292; 95% CI, 1.102-1.516; P = 0.0016), and pancreatic cancer (vs non-DPP4is 1.212; 95% CI, 1.147-1.279; P < 0.0001 vs SUR/Gs 1.239;, 95% CI, 1.167-1.316; P < 0.0001) compared with the non-DPP4i and SUR/G groups after multiple adjustments (Table 7). Notably, all of the increased risks observed in the sensitivity analysis for the 1-year period were numerically higher than those risks in the main analysis (for 90 days) (Tables 4 and 7). Adjusted HRs for lung cancer and kidney cancer were comparable between the treatment groups (Tables 5 and 7).

Table 7.

Adjusted outcome results associated with at least 1 year of DPP4i therapy compared to other therapies

Treatment groups and outcomes	Unadjusted HRs			Adjusted HRsa
	HRs	95% CIs	P value	HRs	95% CIs	P value
Liver cirrhosis
DPP4is vs non-DPP4is	0.858	0.816-0.903	<0.0001	1.060	0.997-1.126	0.0616
DPP4is vs SUR/Gs	0.800	0.758-0.844	<0.0001	1.017	0.953-1.085	0.6090
DPP4is vs TZDs	1.366	1.191-1.566	<0.0001	1.368	1.184-1.581	<0.0001
Liver cancer
DPP4is vs non-DPP4is	0.945	0.906-0.985	0.0073	1.073	1.022-1.127	0.0044
DPP4is vs SUR/Gs	0.907	0.867-0.947	<0.0001	1.055	1.001-1.111	0.0470
DPP4is vs TZDs	1.156	1.047-1.277	0.0043	1.150	1.035-1.278	0.0093
ESRD
DPP4is vs non-DPP4is	1.075	0.957-1.208	0.2234	1.345	1.154-1.567	0.0002
DPP4is vs SUR/Gs	0.997	0.883-1.125	0.9591	1.292	1.102-1.516	0.0016
DPP4is vs TZDs	2.012	1.409-2.874	<0.0001	2.206	1.501-3.243	<0.0001
Kidney cancer
DPP4is vs non-DPP4is	1.037	0.918-1.172	0.5559	1.127	0.975-1.302	0.1057
DPP4is vs SUR/Gs	1.007	0.874-1.160	0.9265	1.112	0.939-1.317	0.2199
DPP4is vs TZDs	1.026	0.746-1.411	0.8742	1.047	0.743-1.475	0.7933
Pancreatic cancer
DPP4is vs non-DPP4is	1.148	1.094-1.203	<0.0001	1.212	1.147-1.279	<0.0001
DPP4is vs SUR/Gs	1.158	1.100-1.219	<0.0001	1.239	1.167-1.316	<0.0001
DPP4is vs TZDs	1.285	1.138-1.450	<0.0001	1.204	1.061-1.366	0.0040

Abbreviations: DPP4is, dipeptidyl peptidase 4 inhibitors; HRs, hazard ratios; SUR/Gs, sulfonylureas/glinides; TZDs, thiazolidinediones.

a

All models were adjusted for the potential confounders as in Table 2.

Furthermore, even when we analyzed the data after extending the washout period to 3 months for the occurrence of liver cirrhosis and ESRD and to 1 year for the occurrence of cancer, the trends in the study outcome results were completely consistent with the results of the above sensitivity analyses (data not shown).

Discussion

By analyzing real-world data from the Korean NHIS in the present study, we found that DPP4i therapy, combined with metformin, in adult patients with diabetes had differential effects compared with other metformin-combined oral antidiabetic therapies on the risk of liver cirrhosis, ESRD, and organ-specific cancers in the liver, kidney, and pancreas. Specifically, when compared with TZDs, DPP4is showed increased adjusted HRs for several outcomes (liver cirrhosis, liver cancer, ESRD, and pancreatic cancer), with a further increased risk after long-term exposure, as shown in the sensitivity analyses. In addition, adjusted HRs for liver cancer, ESRD, and pancreatic cancer were higher in the DPP4i group than in the non-DPP4i and SUR/G groups; the risk of liver cancer was higher than in the SUR/G group in the sensitivity analyses, but not in the main analyses. In contrast, adjusted HRs for the predefined negative control outcome (ie, lung cancer) were comparable between the DPP4i group and other therapeutic groups, supporting the robustness of our study results.

There have been many expectations for the long-term clinical benefits of DPP4is because of their low risk of hypoglycemia and pleiotropic effects in preclinical and small-scale clinical studies (4, 7). However, several large-scale randomized clinical trials (RCTs) (17–19, 25, 26) and real-world data analyses, including Korean NHIS data (27–29), of DPP4is have failed to show a decrease in major CV and renal outcomes in patients with type 2 diabetes and elevated CV risk and/or renal risk. The question of why there are discrepancies between preclinical and clinical study results on DPP4is remains to be answered. Additionally, the so-called pleiotropic effects of DPP4is have yet to be clinically confirmed. Among them, the antifibrotic and antitumor actions observed in preclinical studies are of clinical interest (3, 15). However, the antitumor effect has not been clinically proven (30, 31). Rather, there have been reports of positive or negative associations between the use of DPP4is and cancer incidence in various types of cancers, with pancreatic cancer attracting much attention (32, 33).

Chronic fibrotic diseases, ESRD, and tumors are long-standing diseases that may not be evident in short-term clinical studies. Real-world big data analysis may be a useful alternative. The NHIS data in Korea contain clinical data, including those from the first use of DPP4is in 2007 in Korea, covering almost the entire Korean population (21). Thus, using NHIS data, we evaluated non-CV outcomes of DPP4i therapy that may be dependent or independent of the metabolic actions of DPP4is and that might not be evident in relatively short-term or focused RCTs evaluating major CV outcomes.

Many in vitro and in vivo studies have shown the antifibrotic effects of DPP4is in various organs, including the liver and kidney (3, 13, 14, 34). Type 2 diabetes can increase the risk of nonalcoholic fatty liver disease (NAFLD) and its progression to liver cirrhosis and even, finally, to hepatocellular carcinoma (HCC) (35). However, the antifibrotic effects of DPP4is have not been systematically verified in clinical studies. Our results showed that DPP4i therapy was associated with a higher risk of liver cirrhosis and liver cancer compared with TZDs and an increased risk of liver cancer compared with other non-DPP4 groups. In line with our results, a recent study using the US Medicare Fee-for-Service database showed that DPP4i use was associated with a higher risk of liver cirrhosis than TZD use (36). A retrospective cohort study that enrolled patients with type 2 diabetes and chronic hepatitis C virus infection from the National Health Insurance Research Database in Taiwan showed a significantly lower risk of HCC in patients who took DPP4is than in those who did not (37). In contrast, another study on the National Health Insurance Research Database cohort in Taiwan recently reported that the incidence of decompensated cirrhosis in patients with type 2 diabetes and liver cirrhosis was higher in DPP4i users than in nonusers (38), while the risk of HCC between DPP4i users and nonusers was not significantly different (38). Although study design and comparative medications are different between studies, we believe that the inferiority of DPP4is relative to TZDs in liver outcomes may have clinical implications, considering current evidence of the effect of pioglitazone on NAFLD (39).

There have been many clinical issues regarding the risk of pancreatic cancer associated with DPP4i therapy (33), whereas some large-scale meta-analyses of clinical trials of DPP4is have reported no association between DPP4is and pancreatic cancer (40, 41). In the present study, DPP4i therapy in combination with metformin was associated with a higher risk of pancreatic cancer than other metformin-combined non-DPP4i therapies. Similarly, a recent real-world data analysis in Korea showed an increased risk of pancreatic cancer with the use of DPP4is in patients with newly diagnosed type 2 diabetes (42). Furthermore, our sensitivity analysis results showed that adjusted HRs for pancreatic cancer increased as the duration of exposure to DPP4is increased. Considering that longstanding type 2 diabetes is also associated with pancreatic cancer (43), the choice of antidiabetic agent has important clinical implications in reducing the risk of both CV and other diseases. Because new onset of diabetes can be a harbinger of underlying pancreatic cancer (43), in the present study, we excluded patients with outcomes within 30 days of DPP4i treatment initiation. The sensitivity analyses were performed with focus on long-term exposure to DPP4is and extension of the washout period before outcome review showed results that were consistent with the main analyses.

A recent large RCT showed that linagliptin had a neutral effect on the risk of progressive renal disease, including ESRD, in patients with type 2 diabetes at high CV and renal risk (26). Another long-term RCT on linagliptin, CAROLINA, showed no difference in major CV outcomes and incidence of cancers, including pancreatic cancer, between the DPP4i group and the SUR group. In contrast to our results, in an observational study of real-world data from the US Department of Veterans Affairs on patients with type 2 diabetes, compared with the new use of SURs, new use of DPP4is was associated with a lower risk of composite renal outcomes (24). There have been few studies on kidney cancer in relation to DPP4is or other antidiabetic agents (40). In the present study, no difference in the risk of kidney cancer between the DPP4i group and other groups was observed in long-term observation, whereas there was an increased adjusted HR in the main analysis for the DPP4i group compared with the non-DPP4i group.

A difference between our findings and previous RCT results is that although DPP4is were observed to have a neutral effect on major CV and renal outcomes in previous large-scale RCTs (25, 26), our study showed that DPP4i therapy was associated with an increased risk of liver cirrhosis and ESRD depending on the comparative agents and an increased incidence of liver and pancreatic cancers compared with other antidiabetic agents. This may be partly explained by the fact that the duration of diabetes was longer and the observation period was relatively short in the previous RCT studies, whereas in the present study, the duration of diabetes was relatively short and the observation period was longer than in those RCTs (7, 18, 26). In most RCTs, except for the CAROLINA trial, the mean study duration was 3 years or less (7, 25). Meta-analyses also reviewed data from relatively short study periods (40). If the study period is relatively short, it would be difficult to observe rare and hard endpoints other than the main study outcomes. In the present study, 628 217 patients with diabetes were included in the DPP4i group, with a mean exposure period of 33.8 ± 25.0 months for DPP4is and up to 50.6 ± 34.2 months for metformin treatment as a combination agent. In the non-DPP4i group, the mean metformin combination period was 42.0 ± 33.1 months. However, the duration of diabetes was 10 or more years in many RCTs, and the majority of patients had CV disease, chronic kidney disease, or multiple risks or were older than 60 years of age (7, 18, 26). Thus, old age and multiple combined risks may have diluted adverse or beneficial effects in those studies.

We believe that our retrospective cohort with diabetes is one of the largest real-world datasets to evaluate important issues related to DPP4is. Although not a randomized study, it involved random selection of 50% of patients with diabetes from the nationwide NHIS database and performance of main and sensitivity analyses after multiple adjustments, considering the classes of antidiabetic agents.

Collectively, our results showed that DPP4is may have differential effects on some outcomes according to drug combination and active comparators, with a relatively clear inferiority to TZDs in relation to the risk of liver cirrhosis, liver cancer, ESRD, and pancreatic cancer as well as to other non-DPP4is in relation to liver cancer, ESRD, and pancreatic cancer.

The present study has several limitations. First, the customized NHIS data did not provide detailed information pertaining to laboratory data, lifestyle, or family history that could affect the outcomes or the selection of antidiabetic agents in the present study (20). However, we adjusted for important clinical parameters, including diabetes duration and medications such as insulin, statins, and RAAS blockers, which were also related to the status of metabolic control and disease progression. Second, additional outcomes related to other organs/systems or factors were not systematically evaluated together. Third, metformin combination therapy was analyzed because health insurance in Korea primarily covers metformin combination in most prescriptions for oral antidiabetic agents, with a few exceptions. Currently, information is limited regarding whether DPP4i monotherapy or other combinations would lead to different results. Fourth, in additional comparisons between DPP4i users and non-DPP4i drug users, there were proportions of patients who used comparator drugs (SUR/Gs or TZDs) even if they were DPP4i users. Despite these limitations, there were significant differences in many outcomes between the DPP4i group and the TZD group. The exposure period in the comparator groups for these specific agents was longer than that for DPP4i users in both the main and sensitivity analyses. The reason for our analysis of TZD use in comparison with DPP4i use was that NAFLD can be found in more than 70% of patients with type 2 diabetes and that TZDs may have some benefit in NAFLD (44). Our study design and findings seemed to have clinical relevance in relation to TZD results.

In conclusion, our observational study using a nationwide cohort showed that DPP4i therapy in patients with diabetes was associated with a higher risk of liver cirrhosis, liver cancer, ESRD, and pancreatic cancer than TZD therapy and, similarly, a higher risk than non-DPP4i therapy except for liver cirrhosis.

Funding

This work was supported by grants from the Korea Health Technology Research & Development Project through the Korea Health Industry Development Institute, funded by the Ministry of Health & Welfare, Korea (grant numbers HI14C1135, HI16C1997, and HI18C0331); the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (No. NRF-2021R1A5A2030333); and grants from Gachon University Gil Medical Center (FRD2021-03 and 2019-11).

Author Contributions

J.J. and D.H.L. conceptualized and designed the study, interpreted the data, critically revised the manuscript for important intellectual content, supervised the study, and drafted the manuscript. Y.N. analyzed and interpreted the data and wrote the manuscript. S.W.K., Y.I.P., S.J.C., and S.N. reviewed the study design, interpreted the data, supervised the study, and critically edited the manuscript for important content. J.J. and D.H.L. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. All authors approved the final manuscript.

Disclosures

No potential conflicts of interest in relation to this article were reported.

Data Availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Role of Study Sponsor or Funder

The sponsor was not involved in the study design, analysis, and interpretation of data, writing of the report, or the decision to submit the manuscript for publication.

Abbreviations

 
• BMI

  body mass index

 
• CV

  cardiovascular

 
• DPP4

  dipeptidyl peptidase 4

 
• DPP4i

  dipeptidyl peptidase 4 inhibitor

 
• ESRD

  end-stage renal disease

 
• HCC

  hepatocellular carcinoma

 
• HR

  hazard ratio

 
• NAFLD

  nonalcoholic fatty liver disease

 
• NHIS

  National Health Insurance Service

 
• RAAS

  renin-angiotensin-aldosterone-system

 
• RCT

  randomized controlled trial

 
• SGLT2i

  sodium-glucose cotransporter 2 inhibitor

 
• SUR/G

  sulfonylurea/glinide

 
• TZD

  thiazolidinedione