Fasting blood glucose and risk of prostate cancer: A systematic review and meta-analysis of dose-response

G Model

DIABET-929; No. of Pages 8 Diabetes & Metabolism xxx (2017) xxx–xxx

Available online at

ScienceDirect www.sciencedirect.com

Review

Fasting blood glucose and risk of prostate cancer: A systematic review and meta-analysis of dose-response A. Jayedi a, K. Djafarian b, F. Rezagholizadeh a, A. Mirzababaei a, M. Hajimohammadi a, S. Shab-Bidar a,* a

Department of Community Nutrition, School of Nutritional Sciences and Dietetics, Tehran University of Medical Sciences, P. O. Box 14155/6117, 14166/43931 Tehran, Iran Department of Clinical Nutrition, School of Nutritional Sciences and Dietetics, Tehran University of Medical Sciences, 14166/43931 Tehran, Iran

b

A R T I C L E I N F O

A B S T R A C T

Article history: Received 5 May 2017 Received in revised form 21 September 2017 Accepted 22 September 2017 Available online xxx

Aim. – This study aimed to test the dose-response relationship between fasting blood glucose (FBG) levels and risk of prostate cancer. Methods. – A systematic search was done of PubMed and Scopus from their inception up to January 2017. Prospective and retrospective studies reporting risk estimates of prostate cancer for two or more categories of blood glucose levels were identified, and two independent authors extracted the information. Relative risk (RR) was calculated using random-effects models and pooled. Results. – Ten prospective cohort studies, one nested case-control study, one case-cohort study and three case-control studies (total n = 1,214,947) involving 12,494 cases of prostate cancer were reviewed. The pooled RR of prostate cancer for the highest vs. lowest category of FBG was 0.88 (95% CI: 0.78–0.98, I2 = 25.5%, n = 15 studies). A 10 mg/dL increment in FBG level was not associated with risk of prostate cancer (0.98, 95% CI: 0.96–1.00, I2 = 45.4%, n = 11 studies). Subgroup analyses yielded a significant inverse association only in the subgroup of cohort studies. Non-linear dose-response meta-analysis showed a very slight decrement in risk with increasing FBG levels. Sensitivity analyses using cohort studies showed a steep decrease in risk along with an increase in FBG from baseline levels of  70 mg/dL across prediabetes and diabetes ranges. Conclusion. – Higher FBG levels are associated with lower risk of prostate cancer in cohort studies, but not in case-control studies, findings that limit interpretation of our present results.

C 2017 Elsevier Masson SAS. All rights reserved.

Keywords: Blood glucose Longitudinal studies Meta-analysis Prostate cancer Type 2 diabetes

Introduction Prostate cancer (PCa) is the most common cancer, and third leading cause of cancer deaths, in American men [1]. So far, only a few well-established risk factors, including ageing and being racially black, are thought to be associated with the risk of PCa [1,2]. However, given the substantially greater PCa prevalence in Western countries compared with other geographical regions [3], a potentially prognostic role for lifestyle-related factors has been proposed [4]. Evidence from epidemiological studies has indicated that people with type 2 diabetes (T2D) present with higher risks for Abbreviations: HbA1c, haemoglobin A1c; IGF-1, insulin-like growth factor 1; PCa, prostate cancer; PSA, prostate-specific antigen. * Corresponding author. E-mail address: [email protected] (S. Shab-Bidar).

several types of cancers in comparison to the general population [5,6]. Mainly through its accompanying metabolic features such as hyper-insulinaemia, hyperglycaemia and inflammation, T2D is strongly associated with the risk of cancer [7]. In the case of PCa, such an association appears to be time-dependent [8]. Large-scale epidemiological studies have revealed that T2D has variable timerelated effects on risk of PCa, with the risk decreasing as time passes and diabetes progresses [9,10]. The latest meta-analysis of 56 cohort and case-control studies suggested a 12% reduction in the risk of PCa in patients with T2D [11]. In fact, despite some discrepancies [12], subgroup analyses by diabetes duration have suggested that these protective effects may appear > 10 years after a diagnosis of T2D [11]. The protective effects of T2D against PCa have mainly been attributed to the lower serum concentration, bioavailability and bioactivity of insulin-like growth factor (IGF)-1 in the later stages

http://dx.doi.org/10.1016/j.diabet.2017.09.004 C 2017 Elsevier Masson SAS. All rights reserved. 1262-3636/

Please cite this article in press as: Jayedi A, et al. Fasting blood glucose and risk of prostate cancer: A systematic review and metaanalysis of dose-response. Diabetes Metab (2017), http://dx.doi.org/10.1016/j.diabet.2017.09.004

G Model

DIABET-929; No. of Pages 8 2

A. Jayedi et al. / Diabetes & Metabolism xxx (2017) xxx–xxx

of diabetes [7,8], whereas its elevated serum concentrations in early stages of diabetes [13,14] are associated with higher risk [15,16]. Lower levels of testosterone in the later stages of diabetes may also be contributing to these protective effects [17]. Up to now, several meta-analyses have examined the association between T2D and risk of PCa, and have proposed a 9–16% range of protective effects in association with T2D [11,18– 20]. In all the included studies, the incidence of PCa was compared with a non-diabetic population. Yet, concerning the prognostic or protective effects of different levels of serum glucose on risk of PCa, there has been no conclusive evidence, nor has it been clearly determined as to how the risk of PCa changes across different strata of serum glucose levels. Prediabetes, defined as the early stage of dysregulated glucose metabolism [21], is highly prevalent all over the world [22], and clarifying its longitudinal relationship to risk of PCa may shed more light on our relatively inconclusive understanding of the proposed timedependent association between T2D and risk of PCa. Therefore, our goal was to conduct a meta-analysis of dose-response to examine the linear and potentially non-linear association between fasting blood glucose (FBG) levels and risk of PCa. To our knowledge, this is the first meta-analysis to summarize data from the literature on the association between different levels of serum glucose and future risks of PCa.

Materials and methods The Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines were used to write this systematic review and report the results [23]. Data sources and searches A systematic search was done of the published literature in PubMed and Scopus from their inception up to 18 January 2017, using the following key search terms: ‘prostate cancer’ (MeSH) or ‘prostate carcinoma’ (MeSH) or ‘prostate neoplasm’ (MeSH) or ‘prostate neoplasms’ (MeSH) or ‘prostatic neoplasms’ (MeSH) or (‘prostate’ [All Fields] and ‘cancer’ [All Fields]) and (‘glucose’ [All Fields] or ‘glucose’ [MeSH]) and (‘risk’[All Fields] or ‘risk’ [MeSH]) in PubMed; and ‘prostate’ (Article title, Abstract, Keywords) and ‘glucose’ (Article title, Abstract, Keywords) and ‘cancer’ (Article title, Abstract, Keywords) in Scopus. The reference lists of all relevant articles and reviews were also manually searched. The search was restricted to only articles published in English. Study selection Two independent authors (A.J., A.M.) reviewed the titles and abstracts of all the studies identified. Relevant articles of prospective cohort, nested case-control, case-cohort and casecontrol studies were obtained and included in the present review if they: (1) measured fasting or post-load blood glucose (after oral glucose tolerance test [OGTT]) or haemoglobin A1c (HbA1c); (2) reported blood glucose concentrations as per at least two categories; (3) reported the outcome of interest as PCa incidence; (4) reported risk estimates (relative risk [RR], hazard ratio [HR] or odds ratio [OR]) and the corresponding 95% confidence interval (CI) of PCa incidence for each category of blood glucose level in participants aged > 18 years; and (5) reported the number of cases and participants/person-years in each blood glucose category, or reported sufficient information to allow estimation of those numbers. Excluded were cross-sectional studies, and studies measuring and reporting random blood glucose levels as exposure

as, in such cases, fasting times were not specified and there were no appropriate cut-off values to convert random blood glucose categories into those for FBG. Data extraction and quality assessment Two independent authors (M.H., F.R.) reviewed the full texts of the selected eligible studies and extracted the following information: first author’s name; publication year; study name; location; study design; follow-up duration; mean age and/or age range; number of participants/cases; adjusted covariates; and reported risk estimates and 95% CIs for PCa in each category of FBG. Models with the most covariate adjustments were then selected and included in the meta-analysis. The Newcastle-Ottawa Scale was used to assess the quality of the included studies, and studies with a score  7 were considered high quality [24] (Table S1; see supplementary material associated with this article online). Any discrepancies were resolved through discussion under the supervision of a third author (S.S.-B.). FBG concentration was used to measure exposure, as it is the most reliable test for diagnosis of T2D and categorization of diabetes stage [25]. For studies reporting blood glucose through HbA1c levels or post-load blood glucose concentrations after OGTT as exposure, the following was used to convert such measures to FBG: cut-off values for FBG in prediabetes (100 mg/dL) and diabetes (126 mg/dL) were assumed to be equivalent to cut-offs for HbA1c (5.7% and 6.5%, respectively) and post-load blood glucose (140 mg/dL and 200 mg/dL, respectively) [26,27]. RRs and 95% CIs were considered the effect size in all studies. The reported ORs from case-cohort, nested case-control or casecontrol studies and HRs from cohort studies were considered the equivalent of RRs. For highest vs. lowest category meta-analyses, the reported risk estimates for the highest vs. lowest category of blood glucose level were combined using the DerSimonian and Laird random-effects model [28]. Publication bias was tested for by funnel plots, and Egger’s and Begg’s tests (P < 0.10) [29]. Betweenstudy heterogeneity was explored using Cochrane’s Q test of heterogeneity and I2 statistics (P < 0.05), which provide the relative amount of variance in the summary effect due to between-study heterogeneity [30]. To determine sensitivity, the analysis was repeated after the exclusion of studies in which FBG levels were estimated from HbA1c or post-load blood glucose levels, as well as those with only two categories of FBG. In addition, to assess whether the results could have been affected by a single study, an influence analysis was carried out with one study removed at a time. Subgroup analyses were done based on study type, region, follow-up duration, number of cases and covariate adjustments. The linear dose-response relationship was estimated using a two-stage generalized least-squares trend estimation, according to methods developed by Greenland and Longnecker [31,32]. The method first estimates specific study slope lines, then combines these with studies for which slopes were specifically reported to obtain an overall average slope [32]. Study-specific results were combined using a random-effects model. The median point in each category of FBG was also identified. If medians were not reported, then approximate medians were estimated, using the midpoint of the lower and upper limits. If the highest study category was openended, its blood glucose concentration was calculated as the lower limit plus 1.5 times the size of the closest category [27]. If the lower boundary of the lowest category was not reported, it was considered to be 70 mg/dL, the lowest limit of normal fasting glucose [27]. Any potential non-linear association was examined by modelling blood levels, using restricted cubic splines with three knots at fixed percentiles (10%, 50% and 90%) of distribution [33]. A P value for non-linearity of the meta-analysis was calculated by

Please cite this article in press as: Jayedi A, et al. Fasting blood glucose and risk of prostate cancer: A systematic review and metaanalysis of dose-response. Diabetes Metab (2017), http://dx.doi.org/10.1016/j.diabet.2017.09.004

G Model

DIABET-929; No. of Pages 8 A. Jayedi et al. / Diabetes & Metabolism xxx (2017) xxx–xxx

testing the null hypothesis, where the coefficient of the second spline is equal to zero. All analyses were performed with Stata software, version 12 (StataCorp, College Station, TX, USA). A P value < 0.05 was considered statistically significant. Results Literature search and study characteristics As presented in Fig. 1, 2202 references were identified through electronic searches and 17 by manual searches, of which 130 were duplicates and 1993 non-relevant, and thus excluded on the initial screening of titles and abstracts. Of the remaining, another 81 articles were excluded (detailed reasons for the exclusions are given in Fig. 1). Ultimately, 15 studies involving a total of 1,214,947 participants and 12,494 PCa cases were included in our meta-analysis (Table 1). Reported risk estimates by the different categories of FBG in each study are presented in Table S2 (see supplementary materials associated with this article online). Four studies were conducted in the US [9,34–36], six in Europe [37–42] and five in Asia [43–47]. Ten studies were prospective cohort studies [9,34–36,38–41,44,45]; one was a nested case-control study [42]; another had a case-cohort design [37]; and a further three were case-control studies [43,46,47]. Twelve studies measured and reported FBG levels as exposure [34,36–40,42–47], in one

3

of which fasting time was the variable (> 8 years in about 48% of participants) [39]. FBG levels were estimated from post-load blood glucose in one study [9], and from HbA1c in two others [35,41]. One study included only patients with T2D [41], while another was conducted in patients with coronary heart disease [40]. Two studies reported risk estimates for PCa according to two categories (blood glucose levels < and  100 mg/dL) [35,38], the results of which were considered the cut-off values for the highest category, as they are considered to indicate the onset of prediabetes. However, the analysis was then repeated using the results of other categorizations to ensure accuracy of the results. Follow-up duration in the cohort studies ranged from 5 to 37 years (median: 12 years). Three studies included only two categories [38,44,47], and one reported neither the number of cases nor participants in three categories of blood sugar levels [35]. Thus, 11 studies were eligible for inclusion in the dose-response meta-analysis [9,34,36,37,39–43,45,46]. Highest vs. lowest category meta-analysis All studies were eligible for inclusion in the highest vs. lowest category meta-analysis, and the summary results indicated that higher FBG levels were associated with a 12% reduction in risk of PCa (pooled RR: 0.88, 95% CI: 0.78–0.98) with no evidence of heterogeneity (I2 = 25.5%, Pheterogeneity = 0.17; Fig. 2). When

Fig. 1. Flow chart of the literature search and study selection process for inclusion in the meta-analysis of fasting blood glucose and risk of prostate cancer (PCa).

Please cite this article in press as: Jayedi A, et al. Fasting blood glucose and risk of prostate cancer: A systematic review and metaanalysis of dose-response. Diabetes Metab (2017), http://dx.doi.org/10.1016/j.diabet.2017.09.004

G Model

DIABET-929; No. of Pages 8 A. Jayedi et al. / Diabetes & Metabolism xxx (2017) xxx–xxx

4

Table 1 Baseline characteristics of included studies in the meta-analysis of fasting blood glucose and risk of prostate cancer (PCa). Design

Follow-up duration (years)

Age range and/or mean age (years)

Participants

PCA incidence (cases, n)

Adjusted variables

Cohort

18.4 (median)

35–80

2744

Age, race, birth year, BMI

Cohort

13 (average)

54  18

45,050 male members of Kaiser Permanente Northern California 823 men

Cohort

 5a (from exam 7)

37.5  9.4

1215 men

Cohort

12.1 (mean)

45–64

6429 men in four US communities

385

Age, race

Case-cohort

9.2

50–69

400 non-case subjects without PCa from larger cohort

100

Age, BMI

Cohort

34

50

2313 Caucasian men

237

NA

Cohort

12 (mean)

44  11

289,866 men

Cohort

12.7 (median)

59 (median)

11,541 men with CHD

459

Age

Cohort

12

25–90

14,259 men with type 2 diabetes

740

Nested case-control

6.2 (mean)

59 (median)

392 controls

392

Age, diabetes duration, smoking, insulin treatment Leptin

Case-control

–

NA

306 controls

128

Age, total calories, BMI

Cohort

10.2 (mean)

40–69

9458 Japanese men

119

Jee [45], 2005 National Health Insurance Corp, Korea Pandeya [46], 2014 Hospital-based study, Nepal

Cohort

10

30–59

829,770 Korean men

140

Case-control

–

100 controls

25

Zhang [47], 2015 Xi’an Jiaotong University, Korea

Case-control

–

Cases: 69.4  10.2 Controls: 67.9  9.4 55–99

Age, study area, smoking status, ethanol intake, serum cholesterol Age, age-squared, smoking level, alcohol use NA

120 controls

101

Author [reference], year

Study name, country Darbinian [9], 2008 Multiphasic Health Checkup (MHC), US Hubbard [34], 2004 Baltimore Longitudinal Study of Aging, US Parekh [35], 2013 Framingham Heart Study– Offspring Cohort, US Tande [36], 2006 Atherosclerosis Risk in Communities (ARIC) Study, US Albanes [37], 2009 The Alpha-Tocopherol, BetaCarotene Cancer Prevention (ATBC) Study, Finland Grundmark [38], 2010 Uppsala Longitudinal Study of Adult Men, Sweden Ha¨ggstro¨m [39], 2012 The Metabolic Syndrome and Cancer Project (Me-Can), Norway, Sweden, Austria Lawrence [40], 2013 The BIP study, Israel Miao Jonasson [41], 2012 Swedish National Diabetes Register, Sweden Stocks [42], 2007 The Vasterbotten Intervention Project (VIP), Sweden Hsing [43], 2003 Case-control study in Shanghai, China Inoue [44], 2009 JPHCP, Japan

87

Age, smoking, WHR

75

Age, alcohol, smoking, BMI

6673

Age, smoking status, BMI, stratified for subcohort

Prostate volume, BMI, TG, LDL, HDL

BMI: body mass index; BIP: Bezafibrate Infarction Prevention; CHD: coronary heart disease; HDL: high-density lipoprotein; JPHCP: Japan Public Health Center-Based Prospective Study; LDL: low-density lipoprotein; NA: not available; TG: triglycerides; WHR: waist-to-hip ratio. a Estimated from text.

analysis was restricted to studies in which results were adjusted for more than one confounding variable, the pooled RR changed to 0.91 (95% CI: 0.83–0.98; I2 = 0%, n = 11 studies; Table 2). Influence and subgroup analyses, and publication bias None of the excluded studies changed the pooled RR in any meaningful way (Fig. S1; see supplementary materials associated with this article online). The exclusion of studies with only two categories of FBG [38,44,47] as well as those in which levels of FBG were estimated from post-load blood glucose [9] or haemoglobin A1c [35,41] also did not substantially alter the results (not shown here). In two studies [35,38], results were reported according to two categories, and substituting the reported risk estimates from these two groups, which used a cut-off value of 110 mg/dL for higher blood glucose levels instead of 100 mg/dL, had no impact

on the summary results. In the subgroup analyses, a closely similar result to the main analysis was found in the subgroup of cohort studies (pooled RR: 0.87, 95% CI: 0.80–0.94, I2 = 0%, n = 12 studies), whereas analysis of the case-control studies suggested a positive relationship (pooled RR: 1.70, 95% CI: 1.13– 2.28, I2 = 0%, n = 3 studies; Fig. S2; see supplementary materials associated with this article online). When the analysis was restricted to cohort studies only, a significant inverse association was found, but only in those with a follow-up duration > 12 years compared with < 12 years (Table 2). Begg’s (P = 0.08), but not Egger’s (P = 0.19), regression test revealed some indications of publication bias, and evidence of asymmetry was found in the funnel plot (Fig. S3; see supplementary materials associated with this article online). Repeating the publication bias tests using studies with multivariate analysis found evidence of bias with both Egger’s (P = 0.07) and Begg’s (P = 0.06) tests.

Please cite this article in press as: Jayedi A, et al. Fasting blood glucose and risk of prostate cancer: A systematic review and metaanalysis of dose-response. Diabetes Metab (2017), http://dx.doi.org/10.1016/j.diabet.2017.09.004

G Model

DIABET-929; No. of Pages 8 A. Jayedi et al. / Diabetes & Metabolism xxx (2017) xxx–xxx

5

Fig. 2. Forest plot of fasting blood glucose (FBS; highest vs. lowest category) and risk of prostate cancer. Pooled relative risk was obtained by random-effects model. *: estimated from post-load blood glucose or haemoglobin A1c; **: studies with only two categories of FBS. ES: estimated blood glucose.

Table 2 Subgroup analyses of fasting blood glucose (highest vs. lowest category) and relative risk (RR) of prostate cancer. Subgroup

Studies (n)

RR (95% CI)

I2 (%)

Pheterogeneity

Total Study type Cohort Case-control Region US Europe Asia Cases (n) < 300  300 Quality score < 7 scores  7 scores (high quality) Follow-up duration (years) in cohort studies < 12 years  12 years Adjustments None adjusted, or only age- or leptin-adjusted Multivariate-adjusted BMI-adjusted Yes No Smoking-adjusted Yes No Alcohol-adjusted Yes No

15

0.88 (0.78–0.98)

25.5

0.17

12 3

0.87 (0.80–0.94) 1.70 (1.13–2.28)

0 0

0.53 0.64

4 6 5

0.90 (0.81–0.99) 0.82 (0.66–0.98) 1.19 (0.80–1.57)

0 38.2 49.7

0.92 0.15 0.09

8 7

1.11 (0.84–1.39) 0.85 (0.76–0.94)

27.0 19.5

0.21 0.28

3 12

1.39 (0.91–1.88) 0.86 (0.77–0.95)

0 19.9

0.69 0.24

5 7

0.87 (0.70–1.05) 0.86 (0.77–0.95)

0 14.5

0.56 0.32

4

0.77 (0.47–1.06)

39.6

0.17

11

0.91 (0.83–0.98)

0

0.46

7 8

0.96 (0.77–1.15) 0.84 (0.72–0.97)

35.0 19.9

0.16 0.27

6 9

0.89 (0.77–1.01) 0.92 (0.72–1.12)

0 54.1

0.93 0.02

3 12

0.91 (0.71–1.11) 0.88 (0.76–1.00)

0 39.3

0.73 0.08

BMI: body mass index; CI: confidence interval.

associated with this article online), but there was evidence of heterogeneity (I2 = 45.4%, Pheterogeneity = 0.05). The pooled RR did not change when the analysis was restricted to studies with multivariate analyses (pooled RR: 0.98, 95% CI: 0.97–1.00; I2 = 12.2%, Pheterogeneity = 0.3, n = 8 studies). Sensitivity and subgroup analysis For the sensitivity analyses, none of the excluded studies changed the summary results materially (Fig. S5; see supplementary materials associated with this article online). However, with the exclusion of two case-control studies at a time [43,46], heterogeneity disappeared and, in both analyses, the pooled RR changed to 0.98 (95% CI: 0.96–0.99). When the analysis was repeated after the exclusion of two studies in which FBG levels were estimated from HbA1c or post-load blood glucose [9,41], the summary result changed to 0.98 (95% CI: 0.95–1.01, I2 = 52.7%, Pheterogeneity = 0.03). Subgroup analyses yielded a significant inverse association in the subgroup of cohort studies (pooled RR: 0.97, 95% CI: 0.96–0.99; I2 = 31.3%, Pheterogeneity = 0.16, n = 9 studies), but not of case-controls (pooled RR: 1.03, 95% CI: 0.99–1.07; I2 = 0%, Pheterogeneity = 0.61, n = 2 studies; Fig. S6; see supplementary materials associated with this article online). On examining the non-linear trend between FBG levels and risk of PCa, meta-analysis found a very slight decrement in risk with increasing FBG levels (Pnon-linearity = 0.86; Fig. 3). When the analysis was restricted to cohort studies only, the summary result found that the risk of PCa dropped sharply across all FBG strata (Pnon-linearity = 0.97, n = 9 studies; Fig. 4). Sensitivity analyses of seven cohort studies in the general population [9,34,36,37,39,42,45] showed a steep decrease in risk with FBG levels  100 mg/dL, considered the beginning of the prediabetes range (Pnon-linearity = 0.27; Fig. 5).

Dose-response meta-analysis

Discussion

Using data from 11 studies [9,34,36,37,39–43,45,46], the linear and non-linear trends were examined between FBG levels and risk of PCa. Linear dose-response meta-analysis showed that a 10 mg/ dL increment in FBG was not associated with risk of PCa (pooled RR: 0.98, 95% CI: 0.96–1.00; Fig. S4; see supplementary materials

Principal findings The present meta-analysis has provided supportive evidence regarding the proposed inverse association between long-term diabetes and risk of PCa, and has also indicated that higher FBG

Please cite this article in press as: Jayedi A, et al. Fasting blood glucose and risk of prostate cancer: A systematic review and metaanalysis of dose-response. Diabetes Metab (2017), http://dx.doi.org/10.1016/j.diabet.2017.09.004

G Model

DIABET-929; No. of Pages 8 6

A. Jayedi et al. / Diabetes & Metabolism xxx (2017) xxx–xxx

Fig. 3. Dose-response relationship between fasting blood glucose and risk of prostate cancer in 11 studies (P = 0.86 for non-linearity).

Fig. 4. Dose-response relationship between fasting blood glucose and risk of prostate cancer in nine cohort studies (P = 0.97 for non-linearity).

studies suggested a positive relationship. When the analysis was restricted to cohort studies, a significant inverse association was found, but only in the subgroup of cohort studies with follow-up durations > 12 years compared with < 12 years. So far, several meta-analyses have addressed the association between T2D and risk of PCa, and suggested protective effects ranging from 9% to 16% [11,18–20]. In the latest such report [11], subgroup analyses by diabetes duration indicated that the inverse association between T2D and PCa risk can appear > 10 years after T2D diagnosis, with a shorter duration of exposure to diabetes associated with a greater, but non-significant, risk. The results of three large US cohort studies of men have also confirmed that the risk of PCa decreases consistently with increasing time from T2D diagnosis [9,48,49]. Our present results are in line with these findings and suggest that higher FBG levels are associated with a lower risk of PCa in the long term. Interestingly, it was possible to test the non-linear doseresponse relationship between FBG levels and risk of PCa, although it has not been conclusively clear how the risk of PCa changes across the various strata of blood glucose levels. In our main analysis, a very slight decrement in risk was observed with increasing blood glucose levels. However, when the analysis was restricted to nine cohort studies, it showed that, as levels of FBG rose from the baseline of 70 mg/dL to 220 mg/dL, the risk of PCa fell sharply through both prediabetes and diabetes ranges. Of the nine cohort studies included in the dose-response meta-analysis, eight had follow-up durations > 9 years. Thus, our findings have appropriately revealed the long-term dose-response relationships within FBG ranges from 70 mg/dL to 220 mg/dL. Nevertheless, the available evidence thus far regarding the association between different levels of blood glucose — for example, in prediabetes — and risk of PCa are not conclusive, and individual studies have reported inconsistent results [34–37,39,45]. However, it is possible to present a unified picture of the relationships between different levels of blood glucose and risk of PCa. Sensitivity analyses of studies in the general population have shown that, following the relatively unchanged risk within the normal FBG range (70– 100 mg/dL), the risk then decreases consistently throughout the prediabetes and diabetes ranges. Mechanisms

Fig. 5. Dose-response relationship between fasting blood glucose and risk of prostate cancer in seven cohort studies of the general population (P = 0.27 for nonlinearity).

levels are associated with a 12% lower risk of PCa. Furthermore, the dose-response meta-analysis has indicated that a 10 mg/dL increment in FBG level is marginally and inversely associated with the risk of PCa (by 2%, with no significant greater risk). Subgroup analyses yielded a significant inverse relationship in the subgroup of cohort studies, whereas analysis of case-control

Although the biological and pathophysiological mechanisms involved in this time-dependent association are still not completely determined, the phenomenon has mostly been attributed to changes in serum insulin concentrations during the progression and worsening of T2D [4,8]. In the early stages of diabetes, insulin sensitivity leads to greater concentrations [50] and higher circulatory levels of insulin, both directly through stimulatory effects on the incidence of neoplasia and promotion of tumour growth [51,52], and indirectly through increasing the levels, bioactivity and bioavailability of IGF-1 [53,54], thereby leading, in turn, to a higher risk of PCa. It has been proposed that insulin is a secondary regulator of IGF-1 [14], which has previously been shown to be associated with a greater risk of PCa [55]. In parallel with worsening of hyperglycaemia and progression of T2D, insulin levels decrease due to pancreatic b-cell fatigue [50], thereby reversing these processes. Reduced levels of testosterone as time passes and diabetes progresses are another potential explanation for reversal of the trend in late diabetes [4,49]. In addition, a possible association between genetic susceptibility for T2D and lower risk of PCa has been proposed [56]. Despite this evidence, subgroup analyses have suggested a positive relationship between blood glucose levels and risk of PCa in the subgroup of case-control studies. This phenomenon may, at least in part, be due to a reverse causation bias in case-control

Please cite this article in press as: Jayedi A, et al. Fasting blood glucose and risk of prostate cancer: A systematic review and metaanalysis of dose-response. Diabetes Metab (2017), http://dx.doi.org/10.1016/j.diabet.2017.09.004

G Model

DIABET-929; No. of Pages 8 A. Jayedi et al. / Diabetes & Metabolism xxx (2017) xxx–xxx

studies. Tumour growth is accompanied by metabolic disturbances, including systemic inflammation and increased serum levels of inflammatory cytokines, as well as impaired insulin signalling, which may eventually lead to insulin resistance in patients with cancer which, in turn, may result in dysregulation of glucose metabolism [57–59]. Thus, blood glucose levels may have been influenced in cancer patients, which may have resulted in an overestimation of risk. Study strengths and weaknesses The present meta-analysis has both strengths and limitations. First, previous meta-analyses have examined the risk of PCa in patients with T2D in comparison to non-diabetic populations, whereas findings for the association between different levels of blood glucose and risk of PCa, especially in the general population, have been inconsistent. Using non-linear dose-response metaanalysis, our study has shown that the risk of PCa decreased linearly with increasing FBG from baseline levels up to > 200 mg/ dL. Second, this association could be tested in both cohort and casecontrol studies. As the latter are more likely to be affected by reverse causation bias and have substantially lower quality scores, testing the associations separately using cohort studies led to relatively high quality scores. Third, of the 12 included cohort studies, 10 had follow-up durations > 9 years, which allowed testing for the long-term association between different levels of blood glucose and risk of PCa. As a result, our present study was able to confirm previous findings regarding the inverse association between long-standing diabetes and risk of PCa. Nevertheless, our study also had some limitations. First, the time of T2D diagnosis in cohort studies was not specified and, as a result, it was not possible to test associations according to time of diagnosis. Second, the results of only two studies were adjusted for race [9,36], and only one study took family history of PCa into account [47], whereas being racially black and having a family history of PCa are two well-established risk factors for PCa [2,60]. Thus, this may have biased our conclusions, as these confounders were not adjusted for. The third consideration was the PCa detection method applied. Since the introduction and use of the prostate-specific antigen (PSA) test early in the 1990s for screening and detecting PCa, such cases could be found at earlier stages [25], thereby leading to shorter times since diagnosis of T2D, which may have eventually led to higher incidences and risks of short-term T2D [11]. In seven of the included cohort studies, follow-up durations covered both the pre- and post-PSA eras [9,34–36,38,39,42], whereas four cohort studies were conducted only in the post-PSA era [40,41,44,45], making it not possible to test the association before and after the advent of PSA testing. Fourth, the included studies failed to report sufficient data regarding medications for those with hyperglycaemia, which is relevant as the evidence indicates that metformin can reduce the risk of PCa in patients with T2D [61,62]. Therefore, although in five included studies the baseline prevalence of T2D was only 0–5.1% [9,37,39,42,45], and a further four studies reported baseline impaired FBG levels of 15–25% [34,35,40,44], the possible confounding effects of metformin use cannot be ignored. Fifth, when analyses were performed including case-control studies, the summary results indicated less PCa protection, thereby suggesting that the obtained results for the lower risk of PCa with higher blood glucose levels must be interpreted with caution. Finally, although Egger’s test showed no evidence of publication bias, evidence of asymmetry was found with Begg’s test and funnel plots. In addition, the included studies were all published relatively recently (between 2003 and 2015), thereby implying the possibility of unpublished articles. Thus, our results may have been affected by publication bias.

7

Conclusion The present meta-analysis provides supportive evidence for the inverse association between long-term diabetes and risk of PCa. Our results have shown that the risk of PCa decreases linearly in the long term as FBG levels increase from baseline levels. However, with the inclusion of case-control studies in the analysis, the summary results suggested less PCa protection. Thus, these results need to be interpreted with caution. Also, given the proposed twostage time-dependent relationship between T2D and risk of PCa, it may be helpful to examine the association between blood glucose levels and risk of PCa over shorter times since T2D diagnosis, especially when considering racial differences. Contribution statement Contributors: S.S.-B. and K.D. conceived and designed the study. A.J. and A.M. conducted the systematic search, screened articles and selected eligible articles. M.H. and F.R. extracted information from eligible studies. A.J., S.S.-B. and K.D. performed the analyses and interpreted the results. All authors contributed to writing, reviewing or revising the paper. S.S.-B. is the guarantor of the study, and all authors read and approved the final manuscript. Funding This research did not receive any specific grant from funding agencies in the public, commercial or not-for-profit sectors. Disclosure of interest The authors declare that they have no competing interest.

Appendix A. Supplementary data Supplementary data (Figs. S1–S5, Tables S1 and S2) associated with this article can be found, in the online version, at http://www.sciencedirect.com and http://dx.doi.10.1016/ j.diabet.2017.09.004. References [1] Edwards BK, Noone AM, Mariotto AB, Simard EP, Boscoe FP, Henley SJ, et al. Annual Report to the Nation on the status of cancer, 1975–2010, featuring prevalence of comorbidity and impact on survival among persons with lung, colorectal, breast, or prostate cancer. Cancer 2014;120:1290–314. [2] Hsing AW, Chokkalingam AP. Prostate cancer epidemiology. Front Biosci 2006;11:1388–413. [3] Haas GP, Delongchamps N, Brawley OW, Wang CY, de la Roza G. The worldwide epidemiology of prostate cancer: perspectives from autopsy studies. Can J Urol 2008;15:3866–71. [4] De Nunzio C, Aronson W, Freedland SJ, Giovannucci E, Parsons JK. The correlation between metabolic syndrome and prostatic diseases. Eur Urol 2012;61:560–70. [5] Coughlin SS, Calle EE, Teras LR, Petrelli J, Thun MJ. Diabetes mellitus as a predictor of cancer mortality in a large cohort of US adults. Am J Epidemiol 2004;159:1160–7. [6] Seshasai SR, Kaptoge S, Thompson A, Di Angelantonio E, Gao P, Sarwar N, et al. Diabetes mellitus, fasting glucose, and risk of cause-specific death. N Engl J Med 2011;364:829–41. [7] Giovannucci E, Harlan DM, Archer MC, Bergenstal RM, Gapstur SM, Habel LA, et al. Diabetes and cancer: a consensus report. CA Cancer J Clin 2010;60: 207–21. [8] Giovannucci E, Rimm EB, Stampfer MJ, Colditz GA, Willett WC. Diabetes mellitus and risk of prostate cancer (United States). Cancer Causes Control 1998;9:3–9. [9] Darbinian JA, Ferrara AM, Van Den Eeden SK, Quesenberry Jr CP, Fireman B, Habel LA. Glycemic status and risk of prostate cancer. Cancer Epidemiol Biomarkers Prev 2008;17:628–35.

Please cite this article in press as: Jayedi A, et al. Fasting blood glucose and risk of prostate cancer: A systematic review and metaanalysis of dose-response. Diabetes Metab (2017), http://dx.doi.org/10.1016/j.diabet.2017.09.004

G Model

DIABET-929; No. of Pages 8 8

A. Jayedi et al. / Diabetes & Metabolism xxx (2017) xxx–xxx

[10] Rodriguez C, Patel AV, Mondul AM, Jacobs EJ, Thun MJ, Calle EE. Diabetes and risk of prostate cancer in a prospective cohort of US men. Am J Epidemiol 2005;161:147–52. [11] Jian Gang P, Mo L, Lu Y, Runqi L, Xing Z. Diabetes mellitus and the risk of prostate cancer: an update and cumulative meta-analysis. Endocr Res 2015;40:54–61. [12] Will JC, Vinicor F, Calle EE. Is diabetes mellitus associated with prostate cancer incidence and survival? Epidemiology 1999;10:313–8. [13] Dahle SE, Chokkalingam AP, Gao YT, Deng J, Stanczyk FZ, Hsing AW. Body size and serum levels of insulin and leptin in relation to the risk of benign prostatic hyperplasia. J Urol 2002;168:599–604. [14] Stattin P, Bylund A, Rinaldi S, Biessy C, Dechaud H, Stenman UH, et al. Plasma insulin-like growth factor-I, insulin-like growth factor-binding proteins, and prostate cancer risk: a prospective study. J Natl Cancer Inst 2000;92:1910–7. [15] Dunn SE, Kari FW, French J, Leininger JR, Travlos G, Wilson R, et al. Dietary restriction reduces insulin-like growth factor I levels, which modulates apoptosis, cell proliferation, and tumor progression in p53-deficient mice. Cancer Res 1997;57:4667–72. [16] Jones JI, Clemmons DR. Insulin-like growth factors and their binding proteins: biological actions. Endocr Rev 1995;16:3–34. [17] Giovannucci E, Michaud D. The role of obesity and related metabolic disturbances in cancers of the colon, prostate, and pancreas. Gastroenterology 2007;132:2208–25. [18] Bansal D, Bhansali A, Kapil G, Undela K, Tiwari P. Type 2 diabetes and risk of prostate cancer: a meta-analysis of observational studies. Prostate Cancer Prostatic Dis 2013;16:151–8. [19] Bonovas S, Filioussi K, Tsantes A. Diabetes mellitus and risk of prostate cancer: a meta-analysis. Diabetologia 2004;47:1071–8. [20] Kasper JS, Giovannucci E. A meta-analysis of diabetes mellitus and the risk of prostate cancer. Cancer Epidemiol Biomarkers Prev 2006;15:2056–62. [21] Nathan DM, Davidson MB, DeFronzo RA, Heine RJ, Henry RR, Pratley R, et al. Impaired fasting glucose and impaired glucose tolerance: implications for care. Diabetes Care 2007;30:753–9. [22] van Dieren S, Beulens JW, van der Schouw YT, Grobbee DE, Neal B. The global burden of diabetes and its complications: an emerging pandemic. Eur J Cardiovasc Prev Rehabil 2010;17(Suppl 1):S3–8. [23] Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al., Metaanalysis Of Observational Studies in Epidemiology (MOOSE) group. Metaanalysis of observational studies in epidemiology: a proposal for reporting. Jama 2000;283:2008–12. [24] Peterson J, Welch V, Losos M, Tugwell P. The Newcastle-Ottawa scale (NOS) for assessing the quality of non-randomised studies in meta-analyses.; 2011. [25] ADA. Standards of medical care in diabetes – 2012. Diabetes Care 2012;35(Suppl 1):S11–63. [26] Kumaravel B, Bachmann MO, Murray N, Dhatariya K, Fenech M, John WG, et al. Use of haemoglobin A1c to detect impaired fasting glucose or Type 2 diabetes in a United Kingdom community based population. Diabetes Res Clin Pract 2012;96:211–6. [27] Liao WC, Tu YK, Wu MS, Lin JT, Wang HP, Chien KL. Blood glucose concentration and risk of pancreatic cancer: systematic review and dose-response metaanalysis. BMJ 2015;349:7371. [28] DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177–88. [29] Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315:629–34. [30] Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ 2003;327:557–60. [31] Berlin JA, Longnecker MP, Greenland S. Meta-analysis of epidemiologic doseresponse data. Epidemiol 1993;4:218–28. [32] Orsini N, Bellocco R, Greenland S. Generalized least squares for trend estimation of summarized dose-response data. Stata Journal 2006;6:40. [33] Orsini N, Li R, Wolk A, Khudyakov P, Spiegelman D. Meta-analysis for linear and nonlinear dose-response relations: examples, an evaluation of approximations, and software. Am J Epidemiol 2012;175:66–73. [34] Hubbard JS, Rohrmann S, Landis PK, Metter EJ, Muller DC, Andres R, et al. Association of prostate cancer risk with insulin, glucose, and anthropometry in the Baltimore longitudinal study of aging. Urology 2004;63:253–8. [35] Parekh N, Lin Y, Vadiveloo M, Hayes RB, Lu-Yao GL. Metabolic dysregulation of the insulin-glucose axis and risk of obesity-related cancers in the Framingham heart study-offspring cohort (1971–2008). Cancer Epidemiol Biomarkers Prev 2013;22:1825–36. [36] Tande AJ, Platz EA, Folsom AR. The metabolic syndrome is associated with reduced risk of prostate cancer. Am J Epidemiol 2006;164:1094–102.

[37] Albanes D, Weinstein SJ, Wright ME, Mannisto S, Limburg PJ, Snyder K, et al. Serum insulin, glucose, indices of insulin resistance, and risk of prostate cancer. J Natl Cancer Inst 2009;101:1272–9. [38] Grundmark B, Garmo H, Loda M, Busch C, Holmberg L, Zethelius B. The metabolic syndrome and the risk of prostate cancer under competing risks of death from other causes. Cancer Epidemiol Biomarkers Prev 2010;19:2088–96. [39] Ha¨ggstro¨m C, Stocks T, Ulmert D, Bjorge T, Ulmer H, Hallmans G, et al. Prospective study on metabolic factors and risk of prostate cancer. Cancer 2012;118:6199–206. [40] Lawrence YR, Morag O, Benderly M, Boyko V, Novikov I, Dicker AP, et al. Association between metabolic syndrome, diabetes mellitus and prostate cancer risk. Prostate Cancer Prostatic Dis 2013;16:181–6. [41] Miao Jonasson J, Cederholm J, Eliasson B, Zethelius B, Eeg-Olofsson K, Gudbjornsdottir S. HbA1C and cancer risk in patients with type 2 diabetes – a nationwide population-based prospective cohort study in Sweden. Plos One 2012;7:e38784. [42] Stocks T, Lukanova A, Rinaldi S, Biessy C, Dossus L, Lindahl B, et al. Insulin resistance is inversely related to prostate cancer: a prospective study in Northern Sweden. Int J Cancer 2007;120:2678–86. [43] Hsing AW, Gao YT, Chua Jr S, Deng J, Stanczyk FZ. Insulin resistance and prostate cancer risk. J Natl Cancer Inst 2003;95:67–71. [44] Inoue M, Noda M, Kurahashi N, Iwasaki M, Sasazuki S, Iso H, et al. Impact of metabolic factors on subsequent cancer risk: results from a large-scale population-based cohort study in Japan. Eur J Cancer Prev 2009;18:240–7. [45] Jee SH, Ohrr H, Sull JW, Yun JE, Ji M, Samet JM. Fasting serum glucose level and cancer risk in Korean men and women. Jama 2005;293:194–202. [46] Pandeya DR, Mittal A, Sathian B, Bhatta B. Role of hyperinsulinemia in increased risk of prostate cancer: a case control study from Kathmandu Valley. Asian Pac J Cancer Prev 2014;15:1031–3. [47] Zhang JQ, Geng H, Ma M, Nan XY, Sheng BW. Metabolic syndrome components are associated with increased prostate cancer risk. Med Sci Monit 2015;21: 2387–96. [48] Calton BA, Chang SC, Wright ME, Kipnis V, Lawson K, Thompson FE, et al. History of diabetes mellitus and subsequent prostate cancer risk in the NIHAARP Diet and Health Study. Cancer Causes Control 2007;18:493–503. [49] Kasper JS, Liu Y, Giovannucci E. Diabetes mellitus and risk of prostate cancer in the health professionals follow-up study. Int J Cancer 2009;124:1398–403. [50] Saad MF, Knowler WC, Pettitt DJ, Nelson RG, Charles MA, Bennett PH. A twostep model for development of non-insulin-dependent diabetes. Am J Med 1991;90:229–35. [51] Argiles JM, Lopez-Soriano FJ. Insulin and cancer (Review). Int J Oncol 2001;18:683–7. [52] Peehl DM, Stamey TA. Serum-free growth of adult human prostatic epithelial cells. In Vitro Cell Dev Biol 1986;22:82–90. [53] Musey VC, Goldstein S, Farmer PK, Moore PB, Phillips LS. Differential regulation of IGF-1 and IGF-binding protein-1 by dietary composition in humans. Am J Med Sci 1993;305:131–8. [54] Suikkari AM, Koivisto VA, Rutanen EM, Yki-Jarvinen H, Karonen SL, Seppala M. Insulin regulates the serum levels of low molecular weight insulin-like growth factor-binding protein. J Clin Endocrinol Metab 1988;66:266–72. [55] Roddam AW, Allen NE, Appleby P, Key TJ, Ferrucci L, Carter HB, et al. Insulinlike growth factors, their binding proteins, and prostate cancer risk: analysis of individual patient data from 12 prospective studies. Ann Intern Med 2008;149:w83–8 [461–471]. [56] Pierce BL, Ahsan H. Genetic susceptibility to Type 2 diabetes is associated with reduced prostate cancer risk. Hum Hered 2010;69:193–201. [57] Bertuzzi A, Conte F, Mingrone G, Papa F, Salinari S, Sinisgalli C. Insulin signalling in insulin resistance states and cancer: a modeling analysis. Plos One 2016;11:e0154415. [58] Griffith KA, Chung S-Y, Zhu S, Ryan AS. Insulin resistance and inflammation in Black Women with and without breast cancer: cause for concern. Ethnicity Disease 2016;26:513–20. [59] Ohsawa M, Murakami T, Kume K. Possible involvement of insulin resistance in the progression of cancer cachexia in mice, Yakugaku zasshi. J Pharma Soc Japan 2016;136:687–92. [60] Yanke BV, Carver BS, Bianco Jr FJ, Simoneaux WJ, Venable DD, Powell IJ, et al. African-American race is a predictor of prostate cancer detection: incorporation into a pre-biopsy nomogram. BJU Int 2006;98:783–7. [61] Tseng CH. Metformin significantly reduces incident prostate cancer risk in Taiwanese men with type 2 diabetes mellitus. Eur J Cancer 2014;50:2831–7. [62] 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;9:779–87.

Please cite this article in press as: Jayedi A, et al. Fasting blood glucose and risk of prostate cancer: A systematic review and metaanalysis of dose-response. Diabetes Metab (2017), http://dx.doi.org/10.1016/j.diabet.2017.09.004