’Review article

Association between insulin therapy and risk of liver

cancer among diabetics: a meta-analysis of
epidemiological studies
Xiao-Li Liu, Hua Wu, Long-Gang Zhao, Hong-Li Xu, Wei Zhang and Yong-Bing Xiang

As the results of the association between insulin therapy and risk of liver cancer among diabetics have been inconsistent in
epidemiological studies, we conducted a meta-analysis to quantify this issue. Data of relevant epidemiological studies were
collected by searching articles in PubMed, Web of Science, and Embase till 29 June 2017. Random-effects models were
employed to combine study-specific risks. Five cohort studies and nine case–control studies were included in our meta-analysis
with 285 008 patients with diabetes mellitus and 4329 liver cancer cases. When we compared insulin-use group with noninsulin
use group in patients with diabetes mellitus, we observed a statistically significant association between insulin therapy and liver
cancer, with an overall relative risk of 1.90 (95% confidence interval: 1.44–2.50, I2 = 76.1%). We did not find heterogeneity
between subgroups stratified by study characteristics and adjusted confounders, except for subgroups related to ‘follow-up
years’ of cohort studies. The combined estimate was robust across sensitivity analysis, and no publication bias was detected.
Our results indicated that insulin therapy was associated with elevated incidence of liver cancer among diabetics. Given the high
prevalence of diabetes, avoiding excess or unnecessary insulin use to control the blood glucose may offer a potential public
health benefit in reducing liver cancer risk. Further studies are warranted to investigate the types, doses, and treatment duration
of insulin use in large sample size or cohort of diabetic patients. Eur J Gastroenterol Hepatol 00:000–000
Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved.

Introduction case–control and cohort studies have demonstrated that

The prevalence of diabetes has been steadily increasing over
the past three decades, and there were 422 million patients
with diabetes all over the world in 2016 [1]. Diabetes
mellitus (DM), especially the blood glucose control, is
becoming a great public health issue. When the controlling
targets of blood glucose fail to be achieved by healthy
lifestyles, antidiabetic medicines are prescribed to diabetic
patients. However, some of these glucose-lowering medi-
cines may be associated with cancer risk [2–4], which has
attracted public concerns regarding their use.
Insulin, including exogenous insulin and insulin analo-
gues, is one of the common types of glucose-lowering drugs
administrated to almost all of the patients with type 1
diabetes and some with type 2 diabetes whose blood glu-
cose fails to be controlled by the first-line or second-line
oral antidiabetic drugs. During the past two decades, many
European Journal of Gastroenterology & Hepatology 2017, 00:000–000
Keywords: diabetes mellitus, epidemiological studies, insulin, insulin
analogues, liver cancer, meta-analysis
State Key Laboratory of Oncogenes and Related Genes, Department of
Epidemiology, Shanghai Cancer Institute, Renji Hospital, Shanghai Jiaotong
University School of Medicine, Shanghai, People’s Republic of China
Correspondence to Yong-Bing Xiang, MD, MSc, State Key Laboratory of
Oncogenes and Related Genes, Department of Epidemiology, Shanghai Cancer
Institute, Renji Hospital, Shanghai Jiaotong University School of Medicine, No. 25,
Lane 2200, Xie Tu Road, Shanghai 200032, People’s Republic of China
Tel: + 86 216 443 7002; fax: + 86 216 404 6550; e-mail: ybxiang@shsci.org
Received 20 July 2017 Accepted 12 September 2017
Supplemental Digital Content is available for this article. Direct URL citations
appear in the printed text and are provided in the HTML and PDF versions of this
article on the journal's website, www.eurojgh.com.
insulin injection among patients with DM conferred a
substantially elevated risk for both overall cancer and many
specific-site cancers, including colorectal, pancreatic,
endometrial, ovarian, and breast cancers [5–13].
Although the relationship between insulin use and liver
cancer risk has been investigated in many epidemiologic
studies, inconsistent and inconclusive reports make the
association obscure. Some studies indicated more than
two-fold risks of liver cancer when they compared insulin
users with noninsulin use diabetics [14,15], whereas other
studies suggested that insulin injection did not significantly
associate with liver cancer [16–18]. To quantify the asso-
ciation between insulin and liver cancer, we carried out
this meta-analysis by pooling all available publications.
Materials and methods
Literature search
Articles published before 29 June 2017 were systematically
searched on the PubMed, Web of Science, and Embase
using the following key words: ‘insulin OR insulin analo-
gues’ AND ‘liver OR hepatic OR hepatocellular’ AND
‘cancer OR neoplasm OR tumor OR carcinoma’ AND
‘diabetes’. Additionally, we manually reviewed the cited
references of previous relevant meta-analyses or reviews to
identify more target studies that we failed to obtain from
preliminary literature searches.
Study selection criteria
Original studies published in English were included in our
meta-analysis if they met with the following criteria:

0954-691X Copyright © 2017 Wolters Kluwer Health, Inc. All rights reserved. DOI: 10.1097/MEG.0000000000001001 1

Copyright r 2017 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
2 European Journal of Gastroenterology & Hepatology
(a) had a case–control or cohort study design; (b) eval-
uated the association between insulin therapy (included
any kind of exogenous insulin and insulin analogues) and
liver cancer in patients with DM; and (c) presented relative
risks (RRs), odds ratios (ORs) or hazard ratios (HRs), and
their corresponding 95% confidence intervals (CIs). When
multiple publications were derived from the same study
population, the publication with longer follow-up and
more applicable information was used.
Data extraction and quality assessment
Important information of the selective articles was
extracted independently by two researchers (X-L. L. and
W.H.), and discrepancies were resolved through discus-
sion. Data were abstracted from each eligible study
regarding name of the first author, year of publication,
region, study design, sample size (number of liver cancer
cases and total number of participants), follow-up years (if
cohort design), type of control participants (if case–control
design), DM type, age range or mean age of participants,
sex, exposure of insulin, most fully adjusted estimates
(RRs, ORs, or HRs) with 95% CIs, and adjusted variables
in each study as well.
Newcastle–Ottawa scale, a nine-star scoring system,
was adopted to assess the study quality judged from the
selection of study subjects, comparability of cohorts or
compared groups, as well as ascertainment of outcomes or
exposures of interest for cohort or case–control studies
[19]. A high-quality study is required to gain at least seven
scores according to this evaluation standard.
Statistical analysis
It is reported that RRs are close to ORs, when the inci-
dence of an outcome of interest is less than 10% [20]. As
the incidence of liver cancer in patients with DM is rela-
tively low, both ORs from case–control studies and HRs
from cohort studies were all assumed as valid estimated
values of RRs. Thus, RRs were used to report all results for
simplicity [21]. In one study, RRs were reported separately
in sex groups [22]; therefore, we pooled the classified sex
RRs first, and then combined the results with risk estimates
from the rest of the studies.
Random-effects models were performed to calculate
summary RRs and 95% CIs for liver cancer risk in insulin
users versus noninsulin use diabetics [23]. Cochran’s
Q test and I2 statistics were used to assess the hetero-
geneity of the results among independent studies (P < 0.10
or I2 > 50% was used as a threshold indicating statistically
significant heterogeneity) [24]. Heterogeneity between
subgroups was evaluated by meta-regression. To test the
robustness of the combined estimates, we conducted
separate meta-analysis stratified by study design, follow-up
years, type of control participants, study population, study
quality, DM type, publication year, and adjustment for
potential confounders or risk factors [BMI, smoking and/
or alcohol intake, liver cirrhosis, infection with hepatitis B
virus (HBV) and/or hepatitis C virus (HCV), other medi-
cines, and DM duration]. Sensitivity analyses were exe-
cuted by excluding one study at a time to explore whether
the omitted specific study strongly influenced the summary
RRs [25]. Potential publication bias was detected using the
Egger’s test and Begg’s test [26,27]. All statistical analyses
Copyright r 2017 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Month 2017 • Volume 00 • Number 00
were carried out using Stata, version 12 (StataCorp.,
College Station, Texas, USA). P values with two-sided tests
less than 0.05 (if not specified) were considered statistically
significant.
Results
Literature search
After removing the duplicates, we obtained 5554 poten-
tially relevant articles, for which we performed a pre-
liminary screen through titles and abstracts for inclusion.
Full texts of 129 articles were further reviewed, and 12
original articles plus two retrieved articles from reference
lists eventually met the inclusion criteria (Fig. 1). The main
reasons for exclusion were as follows: the exposure was
not insulin (n = 35), the outcome was not liver cancer
occurrence (n = 18), review articles (n = 27), conference
abstracts (n = 15), articles did not report RRs/ORs or did
not provide 95% CIs about risk estimates (n = 11), articles
repeatedly reported the same study population (n = 9), and
articles were not published in English (n = 2). The latest
study in Taiwan by Lin was replaced by Chiu’s study
[28,29], because we were unable to abstract the risk estimate
of interest as Lin provided RRs according to four insulin
types instead of overall insulin use versus noninsulin use.
Study characteristics
Nine case–control studies and five cohort studies (Table 1)
[14–18,22,29–36], published between 1991 and 2017,
were recruited for a total of 285 008 patients with DM and
4329 liver cancer cases in our meta-analysis. The median
follow-up time of five cohort studies varied from 5 to
15 years. Both hospital-based (n = 5) and population-
based (n = 4) case–control studies were included. Five
studies were conducted in Europe, five in the USA, as well
as four in Asia. In addition to one study reporting out-
comes for men and women separately [22], the other stu-
dies all provided only one risk estimate of insulin use
versus noninsulin use among diabetics. Confounding fac-
tors such as sex and age were adjusted in all of 14 studies,
whereas other potential confounders or risk factors,
including smoking and/or alcohol intake (n = 9), BMI
(n = 3), liver cirrhosis (n = 5), HBV and/or HCV (n = 5),
other medicines (n = 7), and DM duration (n = 3), were
considered in some of the studies. Study-specific quality
scores were summarized in Supplementary Table 1 and 2
(Supplemental digital content 1, http://links.lww.com/
EJGH/A230).
Insulin-use versus noninsulin use
By combining estimates from 14 observational studies over
insulin-use versus noninsulin use, diabetic patients treated
with insulin might have a higher risk for liver cancer
occurrence (pooled RR = 1.90; 95% CI: 1.44–2.50), with
substantial heterogeneity among studies (I2 = 76.1%,
Pheterogeneity < 0.001) by using a random-effects model
(Table 2 and Fig. 2). According to the results of funnel plot
(Supplementary Fig. 1, Supplemental digital content 2,
http://links.lww.com/EJGH/A231), Begg’s test (P = 0.913),
and Egger’s test (P = 0.677), no significant publication bias
was observed in our meta-analysis.
 Insulin therapy and liver cancer risk Liu et al.
Fig. 1 . Flow diagram for selection of studies in meta-analysis of insulin use and liver cancer risk. CI, confidence interval; OR, odds ratio; RR, relative risk.
Subgroup and sensitivity analyses
We observed slightly fluctuant pooled RRs of liver cancer
when comparing insulin user with noninsulin users in
subgroups stratified by selected study characteristics
(Table 2). However, the heterogeneity was only significant
in the strata of follow-up years (Pheterogeneity = 0.018),
which implied that the association between insulin therapy
and liver cancer was not consistent depending on whether
cohort studies had longer follow-up (≥8 years). Cohort
studies with longer follow-up (≥8 years) showed an ele-
vated pooled risk estimate (RR = 2.48; 95% CI: 2.05–3.01;
I2 = 0%) with less heterogeneity (Pheterogeneity = 0.827).
Studies conducted in European countries (RR = 2.36;
95% CI: 1.55–3.59; I2 = 55.7%; Pheterogeneity = 0.061) and
the USA (RR = 1.93; 95% CI: 1.23–3.02; I2 = 46.8%;
Pheterogeneity = 0.111) suggested an elevated liver cancer risk
in diabetics treated with insulin, whereas the significant
Copyright r 2017 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
www.eurojgh.com 3
association was not found in Asian studies (RR = 1.45;
95% CI: 0.69–3.04; I2 = 73.7%; Pheterogeneity = 0.010).
However, heterogeneity was not detected between differ-
ent study population subgroups (Pheterogeneity = 0.285).
Moreover, studies that failed to adjust confounders or risk
factors, such as ‘BMI’ and ‘DM duration’, might provide
significant combined RRs within each unadjusted sub-
groups, whereas no association between insulin therapy
and liver cancer was observed in the adjusted subgroups
over each of the aforementioned confounders or risk fac-
tors. However, we did not find significant heterogeneity
between subgroups regarding any adjustment factors.
To assess the influence of each independent study on the
summary RR, sensitivity analysis was conducted by
sequentially excluding one study and then recalculating
combined results of other remaining studies. The pooled
results ranged from 1.73 (95% CI: 1.32–2.27; I2 = 69.8%;
Pheterogeneity < 0.001) to 2.06 (95% CI: 1.53–2.76;
 4
Copyright r 2017 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.

European Journal of Gastroenterology & Hepatology

Table 1 . Characteristics of studies of insulin use and liver cancer risk
References, study time, location, Cases/DM subjects (baseline age), Insulin Exposure
study design duration of follow-up (cohort only) categories RR/OR (95% CI) Cases assessment assessment Matched/adjusted factors
Chiu et al. [29], 2000–2007, 347/39 515 (58.52 ± 14.9 years), Insulin vs. 1.14 (0.87–1.49) Electronic medical Electronic medical Sex, age, selected comorbidities (coronary artery disease,
Taiwan, CO 7 years noninsulin record record arrhythmia, hyperlipidemia, heart failure, valvular heart disease,
and ischemic stroke)
Luo et al. [14], 1993–2009, USA, 5/8154 (64.3 years), 10.3 years Insulin vs. 3.21 (1.22–8.44) Electronic medical Self-reported Age, ethnicity, education, smoking status, BMI, WHR, physical
CO noninsulin record insulin shots activity, alcohol intake, total daily energy intake, percent of daily
dietary calories from fat, history of hormone therapy use,
NSAIDs use, and liver disease
Gu et al. [15], 2001–2011, 27/8774 (median: 62.2, 61.0 years), Insulin vs. 2.84 (1.12–7.17) Electronic record Electronic medical Age, sex, smoking status, diabetes duration, macrovascular,
Shanghai, CO 10 years noninsulin record HbA1c and concomitant oral glucose-lowering agents
Carstensen et al. [22], 1995 521/22 826, 15 years Insulin vs. 2.44 (1.99–2.98) a
Electronic record Electronic record Age, sex
January to 2009 December, noninsulin
Denmark, CO
Oliveria et al. [30], 2001 January to 12/191 223, 5 years Insulin vs. 1.19 (0.54–2.65) Electronic medical Electronic medical Age, sex, hepatitis B/hepatitis C, cirrhosis, alcoholism
2004 December, USA, CO noninsulin record record
Kasmari et al. [31], 2008–2012, 2541/7955 (57 years) Insulin vs. 1.640 (1.482–1.814) Electronic medical Electronic medical Age, sex, metformin, rosiglitazone, pioglitazone, sulfonylureas,
USA, PCC noninsulin record record cholesterol medication, hypertension, hyperlipidemia, and HCV
Bosetti et al. [32], 2005 January to 190/3962 (65 years) Insulin vs. 3.73 (2.52–5.51) Pathologically Electronic medical Age, sex, date at cohort entry, and duration of follow-up,
2012 December, Italy, PCC noninsulin confirmed diagnosis record Charlson’s comorbidity index, cardiovascular/cerebrovascular
diseases, use of antihypertensive drugs, use of antiplatelet
drugs, and use of statins in the 5 years before cohort entry
Miele et al. [18], 2005 January to 69/121 (median: 60–69 years) Insulin vs. 1.90 (0.74–4.88) Pathologically Self-reported Age, sex, tobacco smoking, and alcohol drinking
2012 June, Italy, HCC noninsulin confirmed diagnosis insulin shots
Hagberg et al. [17], 1988–2011, 305/1456 (69.6 years) Insulin vs. 0.72 (0.18–2.84) Electronic medical Electronic medical Age, sex, BMI, smoking, alcohol abuse, HBV or HCV, rare
UK, PCC noninsulin record record metabolic disorders, statin and paracetamol use, and duration
of diabetes, conditional on matching factors
Gao et al. [16], 2003 January to 16/59 Insulin vs. 0.349 (0.097–1.256) Pathologically NR Age, sex, cigarette smoking and alcohol drinking
2009 June, China, HCC noninsulin confirmed diagnosis
Kawaguchi et al. [33], 2004 138/241 (66.8–64.7 years) Insulin vs. 2.969 (1.293–6.819) Pathologically Self-reported Age, sex, alcohol intake, cirrhosis, albumin, AST, LDH, ALP,
January to 2008 December, noninsulin confirmed diagnosis insulin shots γ-GTP, platelet count, fasting plasma glucose, HbA1c, and
Japan, HCC sulphonylurea
Hassan et al. [34], 2000 January to 122/208 (40–70 years) Insulin vs. 1.9 (0.8–4.6) Pathologically Self-reported Age, sex, race, educational level, cigarette smoking, alcohol
2008 July, USA, HCC noninsulin confirmed diagnosis insulin shots drinking, HCV, HBV, and family history of cancer

Month 2017 • Volume 00 • Number 00

Mannucci et al. [35], 1998 January 29/482 (68.9–68.0 years) Insulin vs. 0.58 (0.06–5.87) Electronic medical Electronic medical Age, sex, BMI, follow-up time, Charlson comorbidity score,
to 2007 December, Italy, HCC noninsulin record record exposure to metformin, and total insulin mean daily dose
Yu et al. [36], 1984 January to 7/32 (50.6–53.4 years) Insulin vs. 18.5 (2.2–156.0) Pathologically Self-reported Age, sex, race, cirrhosis, HBV, HCV, alcohol use, smoking, and
1990 February, USA, PCC noninsulin confirmed diagnosis insulin shots DM duration

ALP, alkaline phosphatase; AST, aspartate aminotransferase; CI, confidence interval; CO, cohort study; DM, diabetes mellitus; γ-GTP, γ glutamyl transpeptidase; HBV, hepatitis B virus; HCC, hospital-based case–control study;
HCV, hepatitis C virus; LDH, lactate dehydrogenase; NR, not reported; NSAIDs, non-steroidal antiinflammatory drugs; PCC, population-based case–control study; WHR, waist–hip ratio.
a
RR was recalculated by fixed effect model.
 Insulin therapy and liver cancer risk Liu et al. www.eurojgh.com 5

Table 2 . Summary risk estimates of the association between insulin use and liver cancer risk
No. of Studies Summary RR (95% CI) Q statistics I2 (%) Pheterogeneitya Pheterogeneityb
All studies 14 1.90 (1.44–2.50) 54.44 76.1 < 0.001
Subgroup analyses
Study design 0.982
Cohort study 5 1.87 (1.15–3.05) 22.89 82.5 < 0.001
Case–control study 9 1.88 (1.17–3.00) 30.77 74.0 < 0.001
Follow-up (years) 0.018
<8 2 1.15 (0.89–1.48) 0.01 0.0 0.920
≥8 3 2.48 (2.05–3.01) 0.38 0.0 0.827
Type of control participants 0.421
Hospital-based 5 1.44 (0.70–2.96) 8.56 53.2 0.073
Population-based 4 2.39 (1.14–5.02) 22.20 86.5 < 0.001
Study population 0.285
Europeans 5 2.36 (1.55–3.59) 9.02 55.7 0.061
Americans 5 1.93 (1.23–3.02) 7.52 46.8 0.111
Asians 4 1.45 (0.69–3.04) 11.42 73.7 0.010
Quality 0.746
<7 6 1.70 (0.92–3.14) 9.99 50.0 0.075
≥7 8 1.95 (1.41–2.70) 44.35 84.2 < 0.001
DM type 0.528
T2DM 8 2.09 (1.38–3.17) 23.44 70.1 0.001
T1DM + T2DM 3 1.73 (0.94–3.17) 19.67 89.8 < 0.001
Unknown 3 1.93 (0.31–11.95) 10.55 81.0 0.005
Publication year 0.553
< 2013 6 2.19 (1.42–3.39) 8.51 41.2 0.130
2013 + 8 1.74 (1.19–2.54) 34.23 79.6 < 0.001
Potential confounders or risk factors
BMI 0.675
Yes 3 1.38 (0.43–4.47) 3.99 49.9 0.136
No 11 1.94 (1.45–2.60) 50.45 80.2 < 0.001
Smoking and/or alcohol intake 0.939
Yes 9 1.86 (1.12–3.08) 17.29 53.7 0.027
No 5 1.93 (1.34–2.77) 37.01 89.2 < 0.001
Liver cirrhosis 0.485
Yes 5 2.16 (1.31–3.58) 9.31 57.0 0.054
No 9 1.69 (1.10–2.62) 41.18 80.6 < 0.001
Infection with HBV and/or HCV 0.621
Yes 5 1.62 (1.02–2.58) 7.07 43.4 0.132
No 9 2.00 (1.31–3.05) 40.73 80.4 < 0.001
Other medicines 0.331
Yes 7 2.26 (1.42–3.60) 22.56 73.4 0.001
No 7 1.59 (0.96–2.61) 31.76 81.1 < 0.001
DM duration 0.596
Yes 3 2.79 (0.64–12.26) 6.58 69.6 0.037
No 11 1.85 (1.39–2.45) 47.13 78.8 < 0.001

CI, confidence interval; DM, diabetes mellitus; HBV, hepatitis B virus; HCV, hepatitis C virus; RR, relative risk; T1DM, type 1 diabetes mellitus; T2DM, type 2 diabetes
mellitus.
a
P-value for heterogeneity within each subgroup.
b
P-value for heterogeneity between subgroups in meta-regression analysis.

I2 = 72.3%; Pheterogeneity < 0.001) when excluding the stu-
dies by Bosetti et al. [32] and Chiu et al. [29], respectively.
None of the studies considerably affected the summary
results.
Discussion
The main finding from our meta-analysis implied that
insulin therapy was associated with a higher risk of liver
cancer among diabetic population. Compared with non-
insulin use group, the risk of liver cancer in patients with
DM of insulin-use group increased by 90%. The summary
estimate for liver cancer was robust in sensitivity analyses,
and no publication bias was detected in our study.
Our meta-analysis was consistent with a previous sys-
tematic review which reported that insulin exposure might
increase risk for pancreas, liver, stomach, and respiratory
cancers. The result of liver cancer from this systematic
review was combined from three cohort studies and two
case–control studies (RR = 1.84; 95% CI: 1.32–2.58;
Pheterogeneity < 0.001) [37]. Furthermore, an expanding
body of epidemiological evidence has emerged over the
past few years in support of the hypothesis that insulin is
associated with overall cancer risk or some specific-site
cancer risk [5–9,11,12]. Insulin therapy, predominantly
insulin analogues, is increasingly recognized as a potential
factor for cancer risk in patients with DM. Many obser-
vational studies from German, Sweden, the UK, and
Taiwan area suggested that patients using insulin glargine
were more likely to develop a solid tumor [7,38–40],
whereas those using pioglitazone or metformin might be
safer [41,42].
The results of the current meta-analysis seemed biolo-
gically plausible. The first developed insulin analogue,
B10Asp, was withdrawn for having an observed promo-
tion effect of mammary tumors in rats [43]. By modifying
the amino acid strains, insulin analogues have different
structures from human insulin. These changed structures
allow insulin analogues to have higher affinity and longer
binding time to insulin-like growth factor 1 receptors

Copyright r 2017 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
6 European Journal of Gastroenterology & Hepatology
Fig. 2 . Results from meta-analysis of the association between insulin and liver cancer (random-effects model), 1991–2017. Relative risks (RRs) were estimated
by comparing insulin use with noninsulin use. Squares represent study-specific estimates (size of the square reflects the study-specific statistical weight);
horizontal lines represent 95% confidence intervals (CIs); diamonds represent the summary estimate with corresponding 95% CI.
which is mainly involved in mitogenic potency [44,45].
Some studies also suggested that the increased mitogenic
potency is attributable to both insulin-like growth factor 1
and insulin receptor, although the latter tends to have a
growth effect on cells [46]. As insulin analogues, especially
the long-acting kinds, can activate the mitogenic signal
pathway more effectively than human insulin and induce
antiapoptotic activities, they might cause increased pro-
liferation of tumor cells and accelerate progression of
tumors [47–49]. Various insulin analogues might have
disproportionate mitogenic potency owing to the increased
half-life, in various degrees, of the ligand-receptor complex
which induces sustained activation of the insulin receptor
tyrosine kinase and sustained phosphorylation of Shc
protein [50]. To sum up, our findings of a positive asso-
ciation between insulin therapy and liver cancer is credible
for the support from experimental evidence.
Significant heterogeneity might discourage us to inter-
pret the association between insulin therapy and liver
cancer. Theoretically, a large number of cases and popu-
lation and longer follow-up time of cohort studies
enhanced the power to verify our hypothesis, whereas
limited cases, various control groups, and potential recall
bias may distort the risk estimate in case–control studies.
However, when we focused our attention on five cohort
studies and nine case–control studies in this meta-analysis,
we failed to detect the heterogeneity between different
study design subgroups (Pheterogeneity = 0.982). Of note, we
observed significant heterogeneity in the subgroups
regarding follow-up years of cohort, in which the group
‘ ≥ 8 years’ tended to present a positive pooled result
(Pheterogeneity = 0.018). We assumed that diabetic patients
in these cohort studies might have a larger amount of
insulin exposure which might lead to higher liver cancer
risk. The results might be explained by the fact that par-
ticipants in cohort study with longer follow-up might have
Copyright r 2017 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Month 2017 • Volume 00 • Number 00
a longer DM duration and poor blood glucose control
events, which occurred after using baseline oral glucose-
lowering agents, might be more frequent; therefore, in this
case, insulin shot might be a more effective choice for
blood glucose control. Although diabetes duration was a
potential confounder, Gu et al. [15] conducted a cohort
study of 10 follow-up years and found a significant asso-
ciation between insulin and liver cancer (RR = 2.84; 95%
CI: 1.12–7.17) after fully adjusting confounders including
diabetes duration.
Heterogeneity might also ascribe to adjustments of
various factors in different researches. Apart from one
study derived from linkage data of the Danish National
Diabetes Register and Cancer Registry by Carstensen and
another Taiwan study based on the National Health
Insurance Research Database [22,29], the remaining stu-
dies in our meta-analysis adjusted for important con-
founders or risk factors such as BMI, smoking and/or
alcohol intake, liver cirrhosis, infection with HBV and/or
HCV, other medicines, and DM duration. The risk esti-
mate from these two studies might be overstated so that we
conducted a sensitive analysis by excluding them [22,29];
however, we still found a robust summary estimate with
decreasing heterogeneity (RR = 1.96; 95% CI: 1.35–2.85;
I2 = 67.9%; Pheterogeneity < 0.001).
Additionally, different ‘study population’ is likely to have
an effect on heterogeneity. Although we did not detect sig-
nificant heterogeneity between subgroups stratified by study
population, studies conducted among Asians presented a
higher heterogeneity (I2 = 73.7%; Pheterogeneity = 0.010),
which could be explained by more participants having a
HBV and/or HCV infection in Asia. Among the four Asian
studies, one case–control study from Japan only included
patients with HCV [33], whereas another case–control study
from China only referred to patients with HBV [16], and
both of these studies were hospital-based designs.
 Insulin therapy and liver cancer risk Liu et al.
Nonetheless, when we excluded these two studies
Kawaguchi et al. [33] and Gao et al. [16], the heterogeneity
of the remaining 12 studies did not change materially
(I2 = 76.5%; Pheterogeneity < 0.001).
The relatively larger study population (285 008 DM
patients and 4329 liver cancer cases) gave this meta-
analysis the strength to confirm the association between
insulin therapy and liver cancer with sufficient statistical
power. To evaluate the heterogeneity, we carried out
subgroup and sensitivity analyses in which our findings
were moderately stable and robust. Furthermore, our
analysis, which included more literature studies (n = 14),
can update and support the previous results [37].
However, the number of high-quality cohort studies is
relatively small in our analysis. Cohort studies with longer
follow-up years, large number of DM participants, and
well-controlled bias are in urgent need.
Enrolling only observational studies in our meta-
analysis might present some limitations to test the asso-
ciation between insulin therapy and liver cancer. For
individual study, potential confounders could not to be
perfectly controlled, which might produce bias in unpre-
dictable directions and increase the heterogeneity.
Moreover, unknown confounders and imprecise adjust-
ment could have made an influence on our findings.
Finally, the categories of insulin exposure are not much
detailed as most studies only reported insulin use and
noninsulin use. Thus, we failed to analyze the effect of
different kinds of insulin on liver cancer risk.
Conclusion
Results from this meta-analysis support the hypothesis that
insulin therapy is associated with an increased risk of liver
cancer among diabetic patients. As the number of type 2
diabetes is growing and as many as 40–80% of diabetic
individuals will be treated with insulin within 10 years
after diagnosis [51], attention toward liver cancer pre-
vention should be paid with respect to excessive use of
exogenous insulin and insulin analogous. Furthermore, the
link between insulin types, doses, and treatment duration
and live cancer risk is need to be evaluated in a
future study.
Acknowledgements
The authors thank the original studies for their contribu-
tion to this review.
Yong-Bing Xiang obtained the funding, conducted the
research design and also had primary responsibility for the
final content; Xiao-Li Liu, Hua-Wu analyzed the data and
interpreted the results; Xiao-Li Liu drafted first manu-
script; all authors critically reviewed and approved the
manuscript.
This work was supported by funds from the National Key
Basic Research Program ‘973 project’ (2015CB554000), the
State Key Project Specialized for Infectious Diseases of China
(no. 2008ZX10002-015 and 2012ZX10002008-002).
Conflicts of interest
There are no conflicts of interest.
Copyright r 2017 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
www.eurojgh.com 7
References
1 World Health Organization. 10 facts on diabetes. 2016. Available at:
http://www.who.int /features/factfiles/diabetes/en/. [Accessed 1 June
2017].
2 Kao CH, Sun LM, Chen PC, Lin MC, Liang JA, Muo CH, et al. A
population-based cohort study in Taiwan – use of insulin sensitizers can
decrease cancer risk in diabetic patients? Ann Oncol 2013;
24:523–530.
3 Tuccori M, Filion KB, Yin H, Yu OH, Platt RW, Azoulay L. Pioglitazone
use and risk of bladder cancer: population based cohort study. BMJ
2016; 352:i1541.
4 Wang CP, Lehman DM, Lam YF, Kuhn JG, Mahalingam D, Weitman S,
et al. Metformin for reducing racial/ethnic difference in prostate cancer
incidence for men with type II diabetes. Cancer Prev Res (Phila) 2016;
9:779–787.
5 Ruiter R, Visser LE, van Herk-Sukel MP, Coebergh JW, Haak HR,
Geelhoed-Duijvestijn PH, et al. Risk of cancer in patients on insulin
glargine and other insulin analogues in comparison with those on human
insulin: results from a large population-based follow-up study.
Diabetologia 2012; 55:51–62.
6 Buchs AE, Silverman BG. Incidence of malignancies in patients with
diabetes mellitus and correlation with treatment modalities in a large
Israeli health maintenance organization: a historical cohort study.
Metabolism 2011; 60:1379–1385.
7 Currie CJ, Poole CD, Gale EA. The influence of glucose-lowering
therapies on cancer risk in type 2 diabetes. Diabetologia 2009;
52:1766–1777.
8 Tripkovic I, Tripkovic A, Ivanisevic Z, Capkun V, Zekan L. Insulin increase
in colon cancerogenesis: a case–control study. Arch Med Res 2004;
35:215–219.
9 Yang YX, Hennessy S, Lewis JD. Insulin therapy and colorectal cancer
risk among type 2 diabetes mellitus patients. Gastroenterology 2004;
127:1044–1050.
10 Bodmer M, Becker C, Meier C, Jick SS, Meier CR. Use of antidiabetic
agents and the risk of pancreatic cancer: a case–control analysis. Am J
Gastroenterol 2012; 107:620–626.
11 Li D, Yeung SC, Hassan MM, Konopleva M, Abbruzzese JL.
Antidiabetic therapies affect risk of pancreatic cancer. Gastroenterology
2009; 137:482–488.
12 Shao Y, Cheng S, Hou J, Zuo Y, Zheng W, Xia M, et al. Insulin is an
important risk factor of endometrial cancer among premenopausal
women: a case–control study in China. Tumour Biol 2016;
37:4721–4726.
13 Swerdlow AJ, Laing SP, Qiao Z, Slater SD, Burden AC, Botha JL, et al.
Cancer incidence and mortality in patients with insulin-treated diabetes:
a UK cohort study. Br J Cancer 2005; 92:2070–2075.
14 Luo J, Chlebowski R, Liu S, McGlynn KA, Parekh N, White DL, et al.
Diabetes mellitus as a risk factor for gastrointestinal cancers among
postmenopausal women. Cancer Causes Control 2013; 24:577–585.
15 Gu Y, Wang C, Zheng Y, Hou X, Mo Y, Yu W, et al. Cancer incidence
and mortality in patients with type 2 diabetes treated with human insulin:
a cohort study in Shanghai. PLoS One 2013; 8:e53411.
16 Gao C, Fang L, Zhao HC, Li JT, Yao SK. Potential role of diabetes
mellitus in the progression of cirrhosis to hepatocellular carcinoma: a
cross-sectional case–control study from Chinese patients with HBV
infection. Hepatobiliary Pancreat Dis Int 2013; 12:385–393.
17 Hagberg KW, McGlynn KA, Sahasrabuddhe VV, Jick S. Anti-diabetic
medications and risk of primary liver cancer in persons with type II
diabetes. Br J Cancer 2014; 111:1710–1717.
18 Miele L, Bosetti C, Turati F, Rapaccini G, Gasbarrini A, La Vecchia C,
et al. Diabetes and Insulin Therapy, but Not Metformin, Are Related to
Hepatocellular Cancer Risk. Gastroenterol Res Pract 2015;
2015:570356.
19 Ottawa Hospital Research Institute. The Newcastle–Ottawa Scale
(NOS) for assessing the quality of nonrandomised studies in meta-
analyses. 2000. Available at: http://www.ohri.ca /programs/clinical_
epidemiology/oxford.asp. [Accessed 1 June 2017].
20 Zhang J, Yu KF. What’s the relative risk? A method of correcting the
odds ratio in cohort studies of common outcomes. JAMA 1998;
280:1690–1691.
21 Greenland S. Quantitative methods in the review of epidemiologic lit-
erature. Epidemiol Rev 1987; 9:1–30.
22 Carstensen B, Witte DR, Friis S. Cancer occurrence in Danish diabetic
patients: duration and insulin effects. Diabetologia 2012; 55:948–958.
23 DerSimonian R, Kacker R. Random-effects model for meta-analysis of
clinical trials: an update. Contemp Clin Trials 2007; 28:105–114.
8 European Journal of Gastroenterology & Hepatology
24 Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-
analysis. Stat Med 2002; 21:1539–1558.
25 Wallace BC, Schmid CH, Lau J, Trikalinos TA. Meta-analyst: software
for meta-analysis of binary, continuous and diagnostic data. BMC Med
Res Methodol 2009; 9:80.
26 Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis
detected by a simple, graphical test. BMJ 1997; 315:629–634.
27 Begg CB, Mazumdar M. Operating characteristics of a rank correlation
test for publication bias. Biometrics 1994; 50:1088–1101.
28 Lin CM, Huang HL, Chu FY, Fan HC, Chen HA, Chu DM, et al.
Association between gastroenterological malignancy and diabetes
mellitus and anti-diabetic therapy: a nationwide, population-based
cohort study. PLoS One 2015; 10:e0125421.
29 Chiu CC, Huang CC, Chen YC, Chen TJ, Liang Y, Lin SJ, et al.
Increased risk of gastrointestinal malignancy in patients with diabetes
mellitus and correlations with anti-diabetes drugs: a nationwide
population-based study in Taiwan. Intern Med 2013; 52:939–946.
30 Oliveria SA, Koro CE, Ulcickas Yood M, Sowell M. Cancer incidence
among patients treated with antidiabetic pharmacotherapy. Diabetes
Metab Syndr Clin Res Rev 2008; 2:47–57.
31 Kasmari AJ, Welch A, Liu G, Leslie D, McGarrity T, Riley T. Independent
of cirrhosis, hepatocellular carcinoma risk is increased with diabetes and
metabolic syndrome. Am J Med 2017; 130:746.e1–746.e7.
32 Bosetti C, Franchi M, Nicotra F, Asciutto R, Merlino L, La Vecchia C,
et al. Insulin and other antidiabetic drugs and hepatocellular carcinoma
risk: a nested case–control study based on Italian healthcare utilization
databases. Pharmacoepidemiol Drug Saf 2015; 24:771–778.
33 Kawaguchi T, Taniguchi E, Morita Y, Shirachi M, Tateishi I, Nagata E,
et al. Association of exogenous insulin or sulphonylurea treatment with
an increased incidence of hepatoma in patients with hepatitis C virus
infection. Liver Int 2010; 30:479–486.
34 Hassan MM, Curley SA, Li D, Kaseb A, Davila M, Abdalla EK, et al.
Association of diabetes duration and diabetes treatment with the risk of
hepatocellular carcinoma. Cancer 2010; 116:1938–1946.
35 Mannucci E, Monami M, Balzi D, Cresci B, Pala L, Melani C, et al. Doses
of insulin and its analogues and cancer occurrence in insulin-treated
type 2 diabetic patients. Diabetes Care 2010; 33:1997–2003.
36 Yu MC, Tong MJ, Govindarajan S, Henderson BE. Nonviral risk factors
for hepatocellular carcinoma in a low-risk population, the non-Asians of
Los Angeles County, California. J Natl Cancer Inst 1991; 83:1820–1826.
37 Karlstad O, Starup-Linde J, Vestergaard P, Hjellvik V, Bazelier MT,
Schmidt MK, et al. Use of insulin and insulin analogs and risk of cancer -
systematic review and meta-analysis of observational studies. Curr Drug
Saf 2013; 8:333–348.
38 Hemkens LG, Grouven U, Bender R, Gunster C, Gutschmidt S,
Selke GW, et al. Risk of malignancies in patients with diabetes treated
Copyright r 2017 Wolters Kluwer Health, Inc. Unauthorized reproduction of this article is prohibited.
Month 2017 • Volume 00 • Number 00
with human insulin or insulin analogues: a cohort study. Diabetologia
2009; 52:1732–1744.
39 Jonasson JM, Ljung R, Talback M, Haglund B, Gudbjornsdottir S,
Steineck G. Insulin glargine use and short-term incidence of
malignancies-a population-based follow-up study in Sweden.
Diabetologia 2009; 52:1745–1754.
40 Chang CH, Toh S, Lin JW, Chen ST, Kuo CW, Chuang LM, et al.
Cancer risk associated with insulin glargine among adult type 2 diabetes
patients – a nationwide cohort study. PLoS One 2011; 6:e21368.
41 Donadon V, Balbi M, Mas MD, Casarin P, Zanette G. Metformin and
reduced risk of hepatocellular carcinoma in diabetic patients with
chronic liver disease. Liver Int 2010; 30:750–758.
42 Chang CH, Lin JW, Wu LC, Lai MS, Chuang LM, Chan KA. Association
of thiazolidinediones with liver cancer and colorectal cancer in type 2
diabetes mellitus. Hepatology 2012; 55:1462–1472.
43 Jorgensen LN, Dideriksen LH, Drejer K. Carcinogenic effect of the
human insulin analogue B10Asp in female rats. Diabetologia 1992; 35
(Suppl 1):A3.
44 Kurtzhals P, Schaffer L, Sorensen A, Kristensen C, Jonassen I,
Schmid C, et al. Correlations of receptor binding and metabolic and
mitogenic potencies of insulin analogs designed for clinical use.
Diabetes 2000; 49:999–1005.
45 Sommerfeld MR, Muller G, Tschank G, Seipke G, Habermann P,
Kurrle R, et al. In vitro metabolic and mitogenic signaling of insulin
glargine and its metabolites. PLoS One 2010; 5:e9540.
46 Glendorf T, Knudsen L, Stidsen CE, Hansen BF, Hegelund AC,
Sorensen AR, et al. Systematic evaluation of the metabolic to mitogenic
potency ratio for B10-substituted insulin analogues. PLoS One 2012; 7:
e29198.
47 Yehezkel E, Weinstein D, Simon M, Sarfstein R, Laron Z, Werner H.
Long-acting insulin analogues elicit atypical signalling events mediated
by the insulin receptor and insulin-like growth factor-I receptor.
Diabetologia 2010; 53:2667–2675.
48 Wada T, Azegami M, Sugiyama M, Tsuneki H, Sasaoka T.
Characteristics of signalling properties mediated by long-acting insulin
analogue glargine and detemir in target cells of insulin. Diabetes Res Clin
Pract 2008; 81:269–277.
49 Sciacca L, Cassarino MF, Genua M, Pandini G, Le Moli R, Squatrito S,
et al. Insulin analogues differently activate insulin receptor isoforms and
post-receptor signalling. Diabetologia 2010; 53:1743–1753.
50 Hansen BF, Danielsen GM, Drejer K, Sorensen AR, Wiberg FC,
Klein HH, et al. Sustained signalling from the insulin receptor after sti-
mulation with insulin analogues exhibiting increased mitogenic potency.
Biochem J 1996; 315 (Pt 1):271–279.
51 DeWitt DE, Hirsch IB. Outpatient insulin therapy in type 1 and type 2
diabetes mellitus: scientific review. JAMA 2003; 289:2254–2264.