Obesity, metformin prescription and risk of advanced prostate cancer in non-diabetic men

Abstract

Background

The aim is to determine the effects of obesity and metformin-use in predicting prostate cancer (PC) risk.

Methods

We used male participants from the Sax Institute’s 45 and Up Study (Australia), recruited between 2005-2009. Participants completed a questionnaire at recruitment which included information on self-reported body mass index (BMI; kg/m²). Participants’ baseline data were linked by the Centre for Health Record Linkage to the NSW Cancer Registry and to Services Australia to identify index prescription claims for diabetic medications between January 2012 and December 2019. Multivariable Joint Cox regression analyses were used to examine associations between BMI, diabetic medications, and PC risk by cancer spread.

Results

Of the 94,674 eligible participants, there were 5265 incident PC cases (localised n = 2638, regional n = 925, metastatic n = 1514 and unknown; n = 1514) diagnosed between January 2012 and December 2019. BMI ≥ 30 kg/m² was associated with increased risk of metastatic PC (versus <30 kg/m²; HR_adjusted = 1.67;95%CI:1.10–2.54); metformin-use was associated with reduced risk of localised PC (versus non-users; HR_{-metformin-only} = 0.65;95%CI:0.50–0.84; HR_{metformin-combination} = 0.51;95%CI:0.34–0.77). Reduced risk of localised PC diagnosis in metformin-users (versus non-users) was evident across all BMI categories.

Conclusion

Metformin-use in obese men is associated with reduced PC risk, if detected early. Further research could inform the repurposing of metformin for PC control.

Introduction

Prostate cancer (PC) is the most commonly diagnosed cancer in men in 112 countries. It accounts for 15% of newly diagnosed cancers, with the number of new cases estimated to increase from 1·4 million in 2020 to 2·9 million by 2040 [1]. In Australia, there was a declining trend in the incidence rate of localised PC since 2008, although the incidence rate of regionally spread disease has increased over-time, with currently over 3500 deaths from PC reported annually [2]. Early diagnosis and treatment for asymptomatic men at highest risk of diagnosis of clinically significant PC, will reduce future PC deaths [3].

Obesity is an established risk factor for the diagnosis of advanced PC, which is defined as clinically significant cancers that are no longer organ-confined [4, 5]. The global increase in obesity is a significant public health issue, and its association with increased risk of advanced PC underscores the urgency of addressing this problem. There is emerging evidence that metformin, a safe low-cost generic drug that is commonly prescribed to improve insulin sensitivity in people with type 2 diabetes, as well as weight management of obese individuals, may also reduce cancer risk [6–10]. This is based on the premise that insulin can stimulate cancer proliferation, whereby metformin intake to lower insulin levels may inadvertently inhibit cancer growth [9, 11–15]. Levels of prostate specific antigen, a biomarker widely used in the opportunistic testing for PC, were lower in metformin-users compared with non-users, suggesting a regulatory role in carcinogenesis [16].

A meta-analysis of prospective studies reported a reduced risk of PC diagnosis with metformin-use although caution is advised in the interpretation of these findings, due to the potential for immortal time bias [17, 18]. Primary prevention trials showed no clear evidence that metformin offered protection against the development of any malignancies, including PC [19]. Trials within PC cases showed no survival benefit although metformin treatment was associated with metabolic benefits [20–25]. Obesity has a multifaceted relationship with metabolic effects, influencing nearly every aspect of metabolic health, and was not considered in the initial trial design but only at the post-hoc analysis [26]. The effect of obesity on cancer risk is currently under investigation in an ongoing trial (NCT04931017) [27].

The use of prescription records for diabetic medications offers an opportunity to examine the longitudinal relationship between type 2 diabetes treatment and PC risk. In this study, we aimed to evaluate the association between metformin-use, other diabetic medications and the risk of being diagnosed with PC, compared to non-use. We hypothesised that metformin-use would be associated with a lower risk of PC diagnosis. We also explored whether this relationship was influenced by obesity.

Method

Study population-cohort at baseline

The Sax Institute’s 45 and Up Study (45andUp) is a population-based cohort study of male and female residents of New South Wales (NSW), Australia, aged 45 years and above [28]. All participants were enrolled between 2005 and 2009. Participants were randomly sampled from Services Australia (formerly the Australian Government Department of Human Services) Medicare enrolment database: Services Australia is a publicly funded universal healthcare system which covers all citizens, permanent residents and some temporary residents and refugees. People 80+ years of age and residents of rural and remote areas were oversampled. Approximately 19% of those invited to the study participated by completing a postal questionnaire at recruitment, which included information on sociodemographic factors, health behaviours, and medical history. Participants provided consent for the linkage of their data to selected population health databases. Details of data items and response categories can be found elsewhere [29].

The 45andUp was approved by the University of NSW Human Research Ethics Committee, while approval for this analysis of PC risk factors was given by the NSW Population and Health Services Research Ethics Committee (HREC/14/CIPHS/54).

Record linkage

The Centre for Health Record Linkage (CHeReL) linked participants’ records with the following administrative health records: The NSW Cancer Registry (NSWCR: January 1994–December 2019), Cause of Death Unit Record File (CODURF: February 2006–December 2019), Registry of Births, Deaths and Marriages records (RBDM: February 2006–December 2019) and NSW Admitted Patient Data Collection (APDC: July 2001–June 2019).

Probabilistic record linkage with these administrative health databases provided information on participants’ cancer diagnoses, procedures during admissions to public and private hospitals, and death. In addition, the Sax Institute used a unique identifier to link study data deterministically with records of claims made to the Medicare Benefits Schedule (MBS: September 2005–December 2020) and Pharmaceutical Benefits Scheme (PBS: Sep 2005–Dec 2020) provided by Services Australia. This linkage provided information on PSA tests reimbursed by Medicare, General Practitioner (GP) consultations and relevant medications dispensed.

Case ascertainment

All pathology laboratories, hospitals, radiotherapy and medical oncology departments are required by the Public Health Act 2010 to report all newly diagnosed cancers to the NSW South Wales Cancer Registry (NSWCR), which includes information on degree of cancer spread at diagnosis which is assigned within 4 months of cancer diagnosis [30, 31]. Cases were defined as men with a record of a PC registration (ICD-10 C61) in the NSWCR, for diagnosis in the period between 45andUp enrolment and 31st December 2019 (the last date for which data used in these analyses were available). Information on degree of spread at diagnosis is used to categorise PC cases as localised, regional or metastatic (distant spread). However, if not all clinical information is received within this period of data acquisition, then these cases are classified as ‘Unknown’ spread which are closer in characteristic to localised cases than those with regional or metastatic spread [32, 33].

Data analysis

Prescription claims for diabetic medications were obtained from linked records with PBS, MBS and APDC for metformin (PBS ATC code A10BA02, A10BDXXX), and other diabetic medications (PBS ATC code beginning with A10 but excluding A10BA02, with subsidised measurement of glycosylated haemoglobin (MBS 66551). This analysis was restricted to participants with PBS records for claims made between January 2012 and end of follow-up (date of PC diagnosis or censor date) as this period covers claims for both general and concessional beneficiaries; prior to 2012 records for PBS claims were only available for concessional beneficiaries.

Participants were excluded if they (i) were diagnosed with PC prior to January 2012 which included men with a record for PC diagnosis (ICD-10 C61) registered by the NSWCR, with APDC records listing C61 diagnosis codes or a radical prostatectomy recorded in Medicare Benefits Schedule (MBS) claims, or (ii) had linkage errors in NSWCR, CODURF or RBDM records.

Men with a prescription record for only metformin and no other diabetic medications was classified as ’metformin-only’, record for metformin augmented with other diabetic medications as ‘metformin-combination’, record for diabetic medications that did not include metformin as ‘non-metformin’, and ‘non-users’ were those with no claims for dispensing of any diabetic prescriptions.

To account for immortal time bias, we employed a ‘new-user’ design where participants were identified by their index prescription claims for any AD medications from January 2012 and one year before end of follow-up. We excluded prescription claims before January 2012, and those with index prescription claims 12 months before PC diagnosis (or censor), to account for possible indication bias [34]. Follow-up started on the date of the index prescription claim for the initial drug which we define as an index drug, as described in Fig. 1 [17]. A self-reported body mass index (BMI; kg/m²) collected at time of recruitment; BMI ≥ 30 kg/m² was categorised as obese and <30 kg/m² were categorised as non-obese. All other covariates were based on self-reported information collected at time of recruitment, as previously described [35]. Those with missing information for a single variable were included in the analysis as a “missing” category.

Fig. 1. Scenarios in the new case design used to identify index prescription date for first prescription record for metformin (M) or diabetic treatment that is not metformin (NM), since recruitment to 45andUp.

Participants with no record for any diabetic prescriptions, the unexposed period (t₀) is defined as the follow-up period from time of recruitment to the end of follow-up, as shown for PC cases (a) or PC-free men (b). Participants for whom first diabetic prescription record is metformin (M), the follow-up time from recruitment to date of metformin prescription is defined as the unexposed period for PC-free men (t₀), and the follow-up time from date of metformin prescription to end of follow-up (t_m) is the exposure period to metformin for PC cases (c) or PC-free men (d). Participants for whom first diabetic prescription record is non-metformin (NM), the follow-up time from recruitment to date of non-metformin prescription is defined as the unexposed period for PC-free men (t₀), and the follow-up time from date of non-metformin prescription to end of follow-up (t_m) is the exposure period to non-metformin for PC cases (e) or PC-free men (f). Participants for whom first diabetic prescription record is not metformin, with a follow-up record for metformin, the follow-up time from recruitment to date of non-metformin prescription is defined as the unexposed period (t₀) for PC-free men, and the follow-up time from date of non-metformin diabetic prescription to date for metformin is defined as the exposure period to non-metformin prescription for PC-free men (t_nm) and the follow-up time from date of metformin prescription to PC diagnosis is the exposure period to metformin (t_m) for PC case (g), or PC-free men (h).

In the stratified analysis by BMI status (< ≥ 30 kg/m²), men with a dispensing record for metformin-only, metformin-combination or other-AD, were compared against non-users with a BMI < 25 kg/m².

Hazard ratios (HR) and 95% confidence intervals (CI) were estimated using Joint Cox Proportional Hazards regression, with age as the underlying time variable, to distinguish time to PC diagnosis by the spread of disease (localised, regional, metastatic and unknown), censoring on the earlier of December 31, 2019, date death or the diagnosis date of a cancer other than PC (SAS version 9 (SAS Institute Inc., Cary, NC, US)) [36]. The regression analyses were, at a minimum, adjusted for age at diagnosis (used as an underlying time variable), region of birth, health-insurance cover, income, qualifications, place of residence, marital status, Charlson’s comorbidity index, rate of PSA testing, rate of GP visits, and fully adjusted for family history, BMI, smoking, alcohol, physical activity, lower urinary tract symptoms, vasectomy, erectile dysfunction, prescriptions for enlarged prostate and diabetes dispensing (listed in footnotes of each table), as previously done [35].

Hazard ratios to investigate differences in PC risk between categories of diabetic drug-use (vs non-users), by BMI status (<, ≥ 30 kg/m²) were separately estimated using a Joint Cox model.

Sensitivity analysis was conducted to include diabetic prescription records in the 12 months preceding the end of follow-up, to determine if the inclusion of diabetic prescriptions in these periods altered the results.

Results

Of the 267,357 participants in the 45andUp, participants were excluded if they were female (n = 142,748), withdrew from the study after baseline or in the pilot study (n = 1019), or were aged below 45 years (n = 5)- Fig. 2. Of the remaining 123,585 participants, further exclusions were made for prevalent PC cases (n = 10, 508), those with data linkage errors (n = 42) died before 2012 (n = 6976) or had prevalent prescription record for diabetics (n = 11, 385). Of the remaining 94, 674 eligible participants, 5265 males were diagnosed with PC during the study period. Cases were classified as diagnosed with localised (n = 2638), regional (n = 925) or metastatic (n = 188) spread of disease at diagnosis; cases with missing status were classified as ‘Unknown’ (n = 1514). A previous analysis of their characteristics suggests these cases are closer in characteristic to localised cases than those with regional or metastatic spread.

Fig. 2.

Flow chart of participant selection for analysis from the 45 and Up Study.

The median age of PC diagnosis was 70.0 years for all participants (Table 1). There were 18.5% PC-free male participants and 16.9% of PC cases had BMI ≥ 30 kg/m². Of those with no record for any diabetes prescriptions, there were 16,6%, had BMI ≥ 30 kg/m². There were 40.0%, 45.3% and 24.2% men with a BMI ≥ 30 kg/m² with records for metformin-only, metformin-combination and other-AD, respectively. Details of participants’ other characteristics are summarised in Table 1.

Table 1.

Baseline characteristics of male participants and prostate cancer (PC) cases recruited between January 2012 and December 2019 (n = 94, 674).

Characteristics	Participantsa		PC cases		Prescription record for diabetic medications
					Non-usersb		Metformin-only		Metformin combination		Non-metforminc
	n	%	n	%	n	%	n	%	n	%	n	%
Total	89409	94.4%	5265	5.6%	87357	92.3%	4846	5.1%	2078	2.2%	393	0.4%
Median Age at recruitment, years (min, max)	60.0	(45.0, 102.8)	62.7	(45.3, 95.0)	60.3	(46.1, 102.8)	59.4	(45.1, 92.2)	57.8	(45.1, 94.0)	65.2	(45.5, 94.0)
Median age at diagnosis, years (min, max)			70	(50.0, 101.0)	70.0	(50.0, 101.0)	70	(56.0, 94.0)	68.5	(56.0, 93.0)	71	(59.0, 82.0)
Family history of prostate cancer (Q18), n (%)d	9061	10.1%	796	15.1%	9153	10.5%	462	9.5%	205	9.9%	37	9.4%
BMI (Q3) ≥ 30 kg/m2 n (%)	16568	18.5%	890	16.9%	14482	16.6%	1940	40.0%	941	45.3%	95	24.2%
Country of birth-Australia, n (%)d	65624	73.4%	4059	77.1%	64548	73.9%	3372	69.6%	1499	72.1%	264	67.2%
Health insurance-private (Q50), n (%)d	51199	57.3%	3311	62.9%	50380	57.7%	2767	57.1%	1168	56.2%	195	49.6%
Remoteness-major cities, n (%)d	46706	52.2%	2659	50.5%	45578	52.2%	2521	52.0%	1058	50.9%	208	52.9%
Annual Household income > =$70,000 (Q46), n (%)d	28770	32.2%	1660	31.5%	28184	32.3%	1504	31.0%	649	31.2%	93	23.7%
Highest qualification-University degree (Q5), n (%)d	24071	26.9%	1472	28.0%	23849	27.3%	1159	23.9%	444	21.4%	91	23.2%
Median #PSA screening testing / 5 years (min, max)e	0.7	(0, 4.9)	0.9	(0, 4.8)	0.7	(0.0, 4.9)	1.0	(0.0, 4.5)	0.7	(0.0, 3.8)	0.7	(0.0, 3.8)
Median #PSA monitoring testing / 5 years (min, max)e	0.7	(0, 26.5)	2.0	(0, 14.1)	0.7	(0.0, 26.5)	0.7	(0.0, 18.1)	0.6	(0.0, 16.7)	0.7	(0.0, 16.8)
Median rate GP visits (#visits / 5 years)f	25.6	(0, 428.5)	25.8	(0, 222.9)	25.1	(0.0, 428.5)	30.1	(0.0, 247.4)	30.4	(0.0, 174.1)	36.6	(0.0, 252.1)
Smoking status-Ever smoker (Q11), n (%)d	44584	49.9%	2415	45.9%	42994	49.2%	2643	54.5%	1153	55.5%	209	53.2%
Median weekly alcohol # drink (min, max)d	7	(0, 140)	7	(0, 100)	7.0	(0, 140)	5.0	(0, 110)	5.0	(0, 140)	5.0	(0, 100)
Median minutes of moderate activity (Q16) (min, max)d,g	120.0	(0, 6039)	120.0	(0, 6000)	120.0	(0, 6039)	120.0	(0, 5640)	90.0	(0, 5940)	120.0	(0, 3360)
Median minutes of vigorous activity (Q16) (min, max)d,h	0.0	(0, 6000)	0.0	(0, 5940)	0.0	(0, 6000)	0.0	(0, 3600)	0.0	(0, 5940)	0.0	(0, 3000)
No comorbidities, n (%)i	73789	82.5%	4791	91.0%	72535	83.0%	4149	85.6%	1659	79.8%	237	60.3%

^aExcluding PC cases.

^bNon-users are participants with no prescription claims for diabetic medications.

^cParticipants with prescription claims made for diabetic medication that is not metformin.

^dAs referenced in men’s baseline questionnaire.²⁵.

^eMBS records for PSA testing (MBS code 66655) and monitoring (MBS code 66656, 66659, 66660) before PC diagnosis or censoring date.

^fGP visits from MBS records before PC diagnosis or censor date.

^gphysical activity-moderate sessions/week.

^hphysical activity-vigorous sessions/week.

ⁱBased on Charlton comorbidity index for MBS records before PC diagnosis or censor date.

Risk of PC increased with increasing frequency of PSA testing, and this varied by risk groups (Supplementary Table 1). PSA testing and monitoring rates were lowest in men with metastatic PC, and highest in men with localised PC. Risk of localised PC was increased in men with a first-degree family history of PC, Australian-born and private health insurance. Participants with any comorbid health conditions had a reduced risk of PC diagnosis across all risk groups, while ever smokers appeared to have a reduced risk of localised disease. All other factors examined showed no association with PC risk.

A multivariable Joint Cox regression analysis showed men with a BMI ≥ 30 kg/m² (versus <25 kg/m²) was associated with increased risk of diagnosis of metastatic PC (Table 2; HR_adjusted = 1.73; 95%CI 1.13–2.64), and a non-significant increased risk of regional PC (HR_adjusted=1.19; 95%CI 0.98–1.45); there was no association with risk of diagnosis of localised disease (HR_adjusted = 0.95; 95%CI 0.85–1.07). There was a reduced risk of diagnosis of localised PC in users of metformin-only (HR = 0.65; 95%CI: 0.50, 0.84) or metformin-combination (HR = 0.51; 95%CI: 0.34, 0.77), compared to non-users. Non-metformin users showed no association with risk of PC diagnosis.

Table 2.

Multivariable-adjusted hazard ratios (95% CI) for risk of prostate cancer (PC) diagnosis is association with BMI collected at recruitment and anti-diabetic medications collected since January 2012, in male participants from the 45andUp Study (n = 94, 674).

Variable	Person years	n	HR (±95% CI)a
			Localisedb n=2638	Regionalc n=925	Metastatic d n=188	Unknowne n=1514
Diabetic prescriptionf
Non- users	652,624	87,357	1.00	1.00	1.00	1.00
Metformin only	21,907	4846	0.65 (0.50, 0.84)	0.79 (0.54, 1.18)	0.90 (0.40, 2.03)	0.76 (0.56, 1.05)
Metformin combination	12,104	2078	0.51 (0.34, 0.77)	0.51 (0.25, 1.02)	1.00 (0.32, 3.15)	0.32 (0.16, 0.65)
Non-metformin	1670	393	1.15 (0.55, 2.42)	1.01 (0.25, 4.03)	1.27 (0.18, 9.12)	na
p-value			0.0001	0.19	0.99	0.0051
BMI (kg/m2)
15.0–24.9	203,591	28,601	1.00	1.00	1.00	1.00
25.0–29.9	315,143	42,896	0.94 (0.86, 1.03)	1.13 (0.97, 1.32)	1.13 (0.81, 1.59)	1.12 (1.00, 1.26)
≥30.0	129,222	17,458	0.95 (0.85, 1.07)	1.19 (0.98, 1.45)	1.73 (1.13, 2.64)	1.07 (0.92, 1.26)
Missing	40,348	5719	1.00 (0.85, 1.19)	1.14 (0.85, 1.53)	1.51 (0.90, 2.55)	0.93 (0.73, 1.17)
p-valueg			0.54	0.30	0.052	0.15
p-valueh			0.39	0.16	0.035	0.16

^aJoint Cox regression adjusted for age (underlying time variable), region of birth, health cover, income, qualification, place of residence, marital status, comorbidity, frequency of PSA testing, frequency of GP visits, family history of prostate cancer, BMI, smoking alcohol, physical activity, lower urinary tract symptoms, vasectomy, erectile dysfunction, medications for BPH and medications for diabetes, unless variable is the exposure of interest.

^bProstate cancer that is still localised within the prostate capsule as identified by the cancer registry (NSWCR).

^cclinically significant PC that have spread of cancer to regional sites as identified by the NSWCR.

^dclinically significant PC that have spread of cancer to distant sites as identified by the NSWCR.

^ePC with no information on spread of disease and identified as ‘Unknown’ by the NSWCR.

^fPrescription claims for diabetic medication from PBS records between January 2012 and one year before PC diagnosis or censoring date.

^gp-value including missing category for BMI.

^hp-value excluding missing category for BMI.

In a stratified analysis by BMI categories, compared to non-user non-obese men (BMI < 30 kg/m²), there was a lower risk of diagnosis for localised PC in non-obese metformin-only users (Fig. 3; HR = 0.68; 95%CI 0.48–0.95) and metformin-combination users (HR = 0.57; 95%CI 0.33–0.98) users. A similar, but non-significant, reduced risk was observed with the diagnosis of regionally spread and metastatic PC, in participants with BMI < 30 kg/m². In a similar comparison, obese men with no record for diabetes prescriptions were at higher risk of metastatic PC (HR = 1.48; 95%CI 1.00–2.19). This risk was further increased in obese men with a prescription record for metformin-only (HR = 2.38; 95%CI 0.87–6.49) and metformin-combination (HR = 2.58; 95%CI 0.63–10.51). There were too few participants in the non-metformin category to obtain any result.

Fig. 3. Multivariable-adjusted hazard ratios.

Multivariable-adjusted hazard ratios (HR + -95% CI) for the interaction between body mass index (BMI) and prescription claims for diabetic medication (metformin-only, metformin combination, or non-metformin antidiabetic medication (AD), and risk of prostate cancer diagnosis of a localized, b regional and c metastatic disease, in male participants from the 45andUp Study, adjusted for age (underlying time variable), region of birth, health cover, income, qualification, place of residence, marital status, comorbidity, frequency of PSA testing, frequency of GP visits, family history of prostate cancer, smoking alcohol, physical activity, lower urinary tract symptoms, vasectomy, erectile dysfunction, medications for BPH, unless variable is the exposure of interest.

In the sensitivity analysis including prescription records in the year before end of follow-up, risk of being diagnosed with metastatic PC was 2- 3-fold higher in users of metformin-only (HR = 3.03; 95%CI1.23, 7.47; Supplementary Table 2), and metformin combination (HR = 2.24; 95%CI:0.55– 9.14).

Discussion

This study examined the association between obesity-status, use of diabetic medication, with a specific focus on metformin, and risk of PC diagnosis by spread of disease. Metformin-use was associated with a lower risk of diagnosis of localised PC, and a similar but non-significant lower risk was observed with the diagnosis of regional and metastatic PC. Obesity-status (BMI ≥ 30 kg/m²) was associated with increased risk of metastatic disease and remained high in metformin-users. However, metformin-use in obese as well as non-obese participants was associated with a lower risk of localised PC. Thus, metformin-use may lower the overall risk of PC diagnosis in non-obese participants however its use may not mitigate the adverse effect of obesity in the development of advanced PC. These findings underscore the importance of surveillance of obese individuals for prostate disease. Early detection strategies combined with metformin may contribute to lowering the risk of developing PC. However, the limited efficacy of metformin in mitigating the risk of advanced disease in obese men highlights the need for targeted interventions addressing obesity-related cancer risks.

A major strength this study is the large number of PC cases ( > 5000 PC cases), which allowed the advanced cases to be separately categorized as regional and metastatic cases, rather than combining both groups due to small numbers. Record linkage with Services Australia data for PBS claims provided participants’ information on the dispensing of metformin as well as other diabetic medications for index prescription since 2012. The use of ‘New-user’ design for the time dependent exposure in the regression analysis and the exclusion of index prescription claims within the 12 months before diagnosis (or censoring) aims to address immortal time bias, and indication bias, respectively, are additional strengths, although restricting to new users may reduce the sample size and may not represent long-term users, are limitations. Furthermore, this study adjusted for frequencies for PSA testing and consultations with primary care providers, for each participant through administrative record linkage, allowed us to minimise the risk of confounding related to differential levels of medical surveillance, is a strength [35]. Limitations include selection bias potentially effecting the participants representativeness of the entire Australian population, misclassification bias can weaken associations with outcome due to the potential under-reporting of weight or over-reporting of height leading to inaccurate BMI estimates, use of localised PC cases does not separate those with clinically significant disease from those with low risk or indolent cancers due to a lack of information on clinical pathology, the lack of information which did not allow the assessment for quantity of use, low number of prescriptions for non-metformin diabetic drugs, as well as requirement for English proficiency to complete the survey which may inadvertently under-represented culturally and linguistically diverse populations, considering that metabolic syndrome is a growing concern among migrants in high-income countries [28, 37]. One other limitation was use of missing indicator approach to address missing which may bias estimates [38].

Our findings for a higher risk of regional and metastatic PC in men ≥30 kg/m² are consistent with those previously reported by The World Cancer Research Fund network, various prospective studies, and The Pooling Project of Prospective studies of Diet and Cancer [39–45]. PSA testing is the first line of screening for the early detection of PC where PSA levels of 3 ng/mL are used as the threshold to warrant further investigation. It was apparent that the rates of PSA testing and monitoring, in our cohort, were lower in men diagnosed with advanced PC compared with localised disease (Supplementary Table 1). Androgen levels in obese men are generally lower than non-obese men, and it is driven by multiple factors including increased aromatase activity, and insulin resistance. [46] BMI and PSA levels are inversely associated in overweight to obese men, which is not evident in men with normal BMI ( < 25 kg/m²) [47–49]. PSA levels in overweight and obese men have been reported to be between 3.4% and and 12.9% lower, respectively, than men with BMI < 25 kg/m² [47, 50, 51]. Despite this, the current PSA testing guidelines recommendations in Australia (or elsewhere) rarely consider obesity status, which may partly explain the lower biopsy rates in obese men [52]. This missed opportunity for the early detection and monitoring of PC in obese men, may be one possible reason for the delayed detection of PC at an advanced stage Previous efforts to develop a BMI-adjusted PSA model for improving early detection of PC in obese men have shown little clinical benefit, likely due to the lack of adjustment for diabetes- a key confounder associated with obesity [47, 53, 54]. This needs serious consideration as obesity rates are growing with ~70% of Australians currently overweight or obese.

Metformin is widely used as a first line of treatment for type-2 diabetes to regulate insulin sensitivity as well as reduce body fatness [55]. A meta -analysis of 29 prospective studies showed a lower risk of PC incidence with metformin-use, more specifically, with the diagnosis of localised PC, which was consistent with our current findings [7, 10, 56]. We previously reported that users of diabetic medications were less like to have a prostate biopsy, and those that do have a biopsy were less likely to be diagnosed with PC [52]. Beckman et al. (2019) also reported lower biopsy rates by users of diabetic medications although they reported no association with PC diagnosis [57]. People with diabetes reportedly have lower PSA levels than those without diabetes which may explain the lower biopsy rates, however PSA levels appear to normalise following the initiation of diabetic medication which may explain some of the inconsistencies in PC diagnosis [16, 57]. Stratification by BMI status revealed a generally increased risk of metastatic disease among obese men (HR = 1.48) compared to their non-obese counterparts. This risk was further elevated among obese individuals who were metformin-users (HR = 2.38). Notably, when records for diabetic medications from the year preceding the end of follow-up were included, the observed risk increased further. This pattern may suggest a biological mechanism possibly involving hyperinsulinemia that may be driving the metastatic process in this population.

Hyperinsulinemia, a consequence of insulin resistance, is commonly observed in obese individuals. It can upregulate levels of IGFs along with their receptors (IGF-R) and binding proteins (IGFBPs) and subsequently increase prostate cell proliferation and invasiveness leading to the diagnosis of aggressive PC [58–64]. Circulating levels of these factors are positively associated with development of advanced PC, as well as increasing BMI [61, 62, 65, 66]. Metformin treatment can to improve insulin sensitivity, reduce body fatness, inhibit androgen-dependent IGF-1R production and subsequently reduce IGF-1R mediated cell proliferation, although if the PC and diabetes are not diagnosed early, the progression of the cancer may be ‘synergised’ by elevated insulin levels in obese men [63, 64, 67]. Biomarker analyses from metformin trials in non-diabetic PC cases showed changes in circulating levels of IGF markers related to improved insulin sensitivity, and more importantly, reduced levels of IGFBP3, which is associated with development of advanced disease [68]. This shift of the IGF-related markers in the direction that is expected if the treatment was converting the PC back to ‘normal’ phenotype. More importantly, this suggests obesity-status may differentially affect the development of PC and is conditional upon metformin-use. The normalisation of PSA levels in people with diabetes following the initiation of metformin may be explained by the lower risk of localised disease in obese metformin-users, highlighting a beneficial role of metformin in reducing risk of PC if both PC and diabetes are detected early and managed well.

Use of metformin as an anti-cancer agent has been tested in clinical trials, mainly as an adjuvant treatment to reduce the risk of cancer progression. Current evidence from metformin trials on PC risk remains inconclusive largely due to small sample sizes, uncertainties in dosing requirements for cancer treatment compared to conventional doses required to treat diabetes, and efficacy in non-diabetics is unknown [21, 23, 24]. Most recent findings from two larger trials that were focussed on locally advanced and metastatic hormone-naïve prostate cancer cases who were on hormone therapy (STAMPEDE) as well as men with low-risk disease on active surveillance (MAST), showed no survival benefit or reduced risk of progression, respectively, but showed improved metformin treatment with metabolic benefits [20, 22]. The results of these trials, while in different patient groups from our study, are likely to have wider implications to the disease course, and continued evaluation of the relationship between obesity, metformin-use and prostate cancer is necessary.

The global increase in obesity is a significant public health issue, and its association with higher risk of advanced PC underscores the urgency of addressing this problem. Metformin is a safe low-cost generic drug and its role in glucose and fatty acid metabolism is well-documented. Early identification and intervention of obese men with metformin might not only mitigate current risks but could also yield new insights into the role of obesity in the etiology and prevention of PC in high-risk populations.

Author contributions

VN-S wrote the main manuscript; All authors reviewed the manuscript.

Data availability

The authors confirm that, for approved reasons, some access restrictions apply to the data underlying the findings. We obtained the data for the project from a third party, namely the Sax Institute, which is the data custodian for the 45 and Up Study. Data are available through application to the Sax Institute. Details are available at https://www.saxinstitute.org.au/our-work/45-up-study/ or through contacting 45andUp.research@saxinstitute.org.au.

Conflict of interest

The authors declare no competing interests.

Ethics approval

Individuals gave written informed consent to take part in the study, including consent for follow-up through repeated data collection and linkage of their data to population health databases. Ethical approval for the conduct of the 45 and Up Study was provided by the University of New South Wales Human Research Ethics Committee. Ethical approval for the present study was provided by the NSW Population & Health Services Research Ethics Committee (14/CIPHS/54).

Supplementary information

The online version contains supplementary material available at 10.1038/s44276-025-00201-x.