The association of smoking with urinary and sexual function recovery following radical prostatectomy

Abstract

Objective

To investigate the association of smoking with post‐prostatectomy functional recovery in a large population‐based cohort using standardised outcome measures.

Patients and Methods

We conducted a cohort study and reported findings according to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines. We used a registry with prospectively gathered patient‐reported outcome measures (PROMs) regarding health and morbidity. We included all men who underwent radical prostatectomy (RP) for localised prostate cancer, without (neo)adjuvant hormone or radiotherapy treatment. In our analysis, we compared Expanded Prostate Cancer Index Composite‐26 (EPIC‐26) scores between people who had ever smoked and those who had not during 24 months’ follow‐up. We identified significant confounders by means of a causal directed acyclic graph, adjusted for these, and performed subgroup and sensitivity analyses.

Results

In total, 2676 patients were included in the analysis. PROM data were available for 61% of the total 4356 otherwise eligible participants. Multivariable regression analysis showed that patients who had ever smoked (ever smokers) scored 11 points lower for sexual function (95% confidence interval −15.0, −7.0) compared to never smokers during 24 months’ follow‐up. Subgroup analysis showed that individuals with baseline scores above 80 were more at risk of sexual function loss. Urinary incontinence scores were similar between ever smokers and never smokers.

Conclusion

Smoking was associated with lower sexual function scores in the first 2 years after RP, but not with changes in urinary incontinence outcomes.

Abbreviations

BMI

body mass index

CCI

Charlson comorbidity index

DAG

directed acyclic graph

EPIC‐26

Expanded Prostate Cancer Index Composite‐26

ICD‐10‐AM

International Statistical Classification of Diseases and Related Health Problems – Australian Modification

OR

odds ratio

PROM

patient‐reported outcome measure

SA‐PCCOC

South Australian – Prostate Cancer Clinical Outcomes Collaborative

RP

radical prostatectomy

Introduction

Prostate cancer is the second most common cancer among men and is diagnosed in over 1.4 million men globally each year [1]. Common complications after treatment with radical prostatectomy (RP) include the inability to sustain a functional erection and long‐term urinary incontinence [2, 3]. Sexual dysfunction and urinary incontinence are associated with reduced quality of life [4, 5] and 45% of men with prostate cancer reported moderate to large bother for sexual function after surgery, compared to 22% before surgery [6]. With increasing incidence and over 99% 5‐year survival rates for localised and regional prostate cancer [7], quality of life is an important patient priority.

Tobacco smoking is a known risk factor for developing sexual dysfunction [8] and is associated with many other functional outcomes after prostatectomy [9], while smoking cessation may improve postoperative functional outcomes [10]. Sexual dysfunction can have many causes; the pathophysiological pathways and both cardiovascular risk factors and hormonal factors play a role [11]. It is not fully understood how tobacco smoking contributes to sexual dysfunction and why smokers seem to have lower rehabilitation potential after prostatectomy. Supporting evidence from larger cohort studies and controlled intervention studies are currently lacking. Our recent systematic review [12] identified nine smaller observational studies and confirmed an association of smoking status with impaired sexual function at 2 years after RP in a meta‐analysis (odds ratio [OR] 0.7, 95% CI 0.6, 1.0). We also found some evidence suggesting that smoking cessation could improve sexual function over time, reaching clinically significant improvement at 18–24 months. We did not find an association between smoking and urinary incontinence (OR 1.2, 95% CI 0.8, 1.9). Major limitations included lack of ‘gold standard’ and validated measures to assess sexual function and urinary incontinence, as well as lack of sufficient follow‐up, as longer follow‐up studies have shown that at least 24 months of follow‐up after RP is needed to allow sufficient time for function recovery after surgery [13].

Sexual dysfunction and urinary incontinence have a big impact on quality of life and it was our objective to identify preventable lifestyle habits the avoidance of which may improve outcomes. The aim of this cohort study was to further investigate the association between smoking and urinary and sexual function after prostatectomy using a large prostate cancer patient database, validated gold standard instruments and a follow‐up of at least 24 months.

Methods

Study Design

We used a retrospective cohort study design, with participants recruited from a population‐level clinical quality registry in South Australia and reported our findings according to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) checklists (Appendix S1). We produced a study protocol prior to commencement of this study.

Study Population

Participants were identified using the South Australian – Prostate Cancer Clinical Outcomes Collaborative (SA‐PCCOC) database. The SA‐PCCOC database consists of over 22 000 male participants with diverse ethnic and socio‐economic characteristics, diagnosed with prostate cancer since 1998, who have been providing patient‐reported outcome measures (PROMs) regarding health and morbidity since 2007. PROMS were obtained using paper‐based questionnaires delivered by mail. We included men treated with RP for their localised prostate cancer who reported at least one PROM during follow‐up. Exclusion criteria were localised advanced and metastatic prostate cancer at diagnosis, previous bladder or prostate surgery, or any prior or additional treatment, such as hormone therapy or radiation therapy in the lower abdominal region.

Data Analysis

Exposure Data

Our parameter of interest was smoking status. We identified both hospital records and self‐reported smoking data in the SA‐PCCOC database. Hospital records were coded F17 (nicotine dependence), Z720 (tobacco use not otherwise specified) and Z878 (history of nicotine dependence), based on the International Statistical Classification of Diseases and Related Health Problems – Australian Modification (ICD‐10‐AM) [14]. Two additional questions were asked in the PROM questionnaires from 2017 onward at each follow‐up time point: Q1: ‘Over your lifetime would you have smoked at least 100 cigarettes or a similar amount of tobacco?’ and Q2: ‘Do you currently smoke cigarettes, cigars, pipes or any other tobacco products?’. We considered individuals with a registered ICD‐10‐AM code F17 or Z720 or who reported at least once ‘yes’ to Q1 as ‘ever smokers’. Participants reporting ‘daily’, ‘at least weekly’ and ‘less often than weekly’ to Q2 were classified as current smokers. Those without a registered ICD‐10‐AM code for smoking or answering ‘no’ for Q1 and ‘not at all’ for Q2 were classified as ‘never smokers’. Participants with registered ICD‐10‐AM code Z878 or reporting ‘yes’ for Q1 and ‘not at all’ for Q2 were classified as ‘past smokers’.

Outcome Data

Responses to the Expanded Prostate Cancer Index Composite‐26 (EPIC‐26) [15] health‐related quality‐of‐life questionnaire for sexual function and urinary incontinence were used. EPIC‐26 scores range from 0 to 100, with 0 indicating total function loss and 100 points indicating maximum function. Measurements were taken at baseline after diagnostic biopsy and before RP, and at 3, 6, 12 and 24 months post‐surgery. Where patients had returned multiple questionnaires at a given timepoint we selected either: (i) the closest to the specified timepoint; (ii) the survey with most data; or (iii) an average of the data available.

Covariates

We included the following variables in multivariable models: age at surgery; Charlson comorbidity index (CCI) score; body mass index (BMI); PSA value; prostate size; Gleason score; whether prostatectomy was the initial treatment or the patient was progressing from a period of active surveillance or watchful waiting; year of surgery; type of surgery (robot‐assisted laparoscopic vs open); whether or not the surgery was nerve‐sparing; surgical margin status; and the TNM classification of malignant tumours at diagnosis. Confounders were identified through a causal directed acyclic graph (DAG; Appendix S2) using the online tool for the R package ‘dagitty’ [16]. Age and year of surgery were identified as the minimal sufficient adjustment for the total effect of smoking on sexual dysfunction and urinary incontinence.

Analysis

Demographic data were compared between men who had ever smoked (ever smokers) and those who had not (never smokers) within our cohort using the chi‐squared test for categorical data (Fisher's exact test when one or more cells included fewer than five data points) and the t‐test for continuous data. Univariable analysis was performed comparing EPIC‐26 scores (0–100) between ever smokers and never smokers for both sexual function and urinary incontinence through linear regression.

Mixed effects linear regression analysis was performed to adjust for time since RP and identified confounders. Random effects were included at patient level through their unique study ID. Individual differences between baseline and follow‐up time points for both groups were analysed to compare the recovery rates during follow‐up. Interaction analyses (smoking vs age and time since surgery), subgroup analyses (classified by baseline EPIC‐26 score) and sensitivity analyses (comparing subcategories current, past and never smokers) were performed to test the consistency and robustness of the findings.

Ethics

The Southern Adelaide Clinical Human Research Ethics Committee (EC00188) has approved use of the SA‐PCCOC database (reference number HREC/14/SAC/315). Participants in this database have previously given consent for data usage and storage for current and new research.

Results

Of a total of 22 000 patients in the SA‐PCCOC database, 7452 were identified as having localised prostate cancer and were being treated with either observation (active surveillance or watchful waiting), RP, radiotherapy or a combination of those. In total, 5015 patients underwent RP in the period between September 1996 and December 2023. A total of 645 patients who underwent prior prostate or bladder surgery were excluded. Five patients were treated with radiation therapy or androgen deprivation therapy before or after undergoing RP and nine patients with lymph node or distant metastases were excluded, resulting in 4356 prostate cancer patients eligible for further analysis. Our database consisted of 8282 PROM questionnaires between November 2007 and December 2023. A total of 1680 eligible patients had no PROM responses on record, which resulted in a total of 2676 patients included in the final analyses (Fig. 1), of which 702 (26%) were identified as ‘ever smokers’.

Fig. 1.

Flow diagram of the study population. ADT, androgen deprivation therapy; PROM, patient‐reported outcome measure; SA‐PCCOC, South Australian – Prostate Cancer Clinical Outcomes Collaborative.

Patients with prostate cancer who had ever smoked (ever smokers) had a higher CCI score (0.54 vs 0.32; P < 0.01), had a higher BMI (28.4 vs 27.6 kg/m²; P < 0.01) and were older at time of surgery (median 66 vs 65 years; P ≤ 0.01) compared to never smokers. Ever smokers more often had more advanced disease at diagnosis, with more often a higher T stage (T3a 17% vs 11%; P < 0.01) and Gleason score above 8 (14% vs 8%; P < 0.01). There were no significant differences between PSA and prostate size between groups. Ever smokers underwent robotic surgery more often (86% vs 80%; P < 0.01), had positive surgical margins more often (32% vs 25%; P ≤ 0.01) and less often underwent nerve‐sparing surgery (64% vs 72%; P ≤0.01). The mean EPIC‐26 sexual function scores at baseline were significantly lower for ever smokers compared to never smokers (57 vs 62; P = 0.01), with similar results for urinary incontinence scores (91 vs 90; P = 0.78 [Table 1]).

Table 1.

Demographics stratified by smoking habit.

Variable	Category	Statistic	Never smoker	Ever smoker	P
n			1974	702
Age at diagnosis, years	Median (IQR)		65 (60, 69)	65 (60, 70)	0.01
Age at surgery, years	Median (IQR)		65 (60, 69)	66 (61, 71)	<0.01
CCI score	0–2	n (%)	1514 (97)	517 (94)	0.01
	3–4		44 (3)	21 (4)
	5+		9 (1)	11 (2)
BMI, kg/m2		Mean (sd)	27.6 (4.5)	28.4 (4.1)	<0.01
PSA, ng/mL		Mean (sd)	8.1 (5.8)	9.3 (23.3)	0.06
Prostate size, mL		Mean (sd)	41 (25)	37 (22)	0.38
Gleason score	<7	n (%)	168 (10)	31 (6)	<0.01
	7–8		1405 (82)	384 (80)
	>8		133 (8)	67 (14)
T‐stage	T1	n (%)	308 (34)	174 (40)	<0.01
	T2		494 (55)	183 (42)
	T3a		97 (11)	75 (17)
Initial treatment	Watchful waiting/active surveillance	n (%)	129 (7)	61 (9)	0.07
	RP		1845 (94)	641 (91)
Treatment year	2006–2011	n (%)	598 (30)	63 (9)	<0.01
	2012–2017		1090 (55)	354 (50)
	2018–2023		286 (15)	285 (41)
RP type	Open	n (%)	398 (20)	98 (14)	<0.01
	Robot‐assisted laparoscopic		1573 (80)	603 (86)
Nerve‐sparing	One‐sided	n (%)	92 (13)	27 (22)	<0.01
	Bilateral		410 (59)	51 (42)
	No		192 (28)	43 (36)
Surgical margins	Clear	n (%)	1277 (75)	329 (69)	<0.01
	Involved		419 (25)	151 (32)
Baseline EPIC‐26 score	Sexual	Mean (sd)	62 (31)	57 (32)	0.01
	Urinary		90 (18)	91 (17)	0.78

BMI, body mass index; CCI, Charlson comorbidity index; EPIC‐26, Expanded Prostate Cancer Index Composite‐26; IQR, interquartile range; RP, radical prostatectomy.

The mean EPIC‐26 sexual function scores were significantly different between groups at baseline and at 24 months after surgery: 57 vs 62 points at baseline (P = 0.01) and 31 vs 42 points at 24 months (P < 0.01). The mean urinary incontinence scores were similar at baseline and follow‐up, but were significantly different at 24 months’ follow‐up: 74 and 80 points, respectively (P < 0.01; Appendix S3). A steep decline in EPIC‐26 sexual and urinary function scores was observed from baseline to 3 months post‐surgery for both groups, with a subsequent gradual increase over time. Ever smokers (n = 702) scored consistently lower for sexual function during follow‐up compared to never smokers (n = 1974; Fig. 2A). When groups were further stratified based on PROM entries only and current (n = 67), never (n = 509) and past smokers (n = 494) were compared, never smokers scored better at baseline and follow‐up compared to current and past smokers, however, numbers were small (Fig. 2B). EPIC‐26 urinary incontinence scores were similar in the ever smoker and never smoker groups, except at 24 months.

Fig. 2.

Expanded Prostate Cancer Index Composite‐26 (EPIC‐26) urinary incontinence and sexual function score development over time following radical prostatectomy (RP). (A) EPIC‐26 scores in ever smokers vs never smokers over time after RP. (B) EPIC‐26 scores in never smokers vs past smokers vs current smokers over time after RP.

Univariable regression analysis showed a significant association between smoking and EPIC‐26 sexual function scores, with ever smokers scoring 4.6 points lower (P < 0.01) compared to never smokers. Multivariable regression analysis showed that smoking was associated with a decline of 11 points (95% CI −15.0, −7.0) in EPIC‐26 sexual function score over 24 months of follow‐up (Table 2). Age and year of surgery were identified as confounders by means of a causal DAG (Appendix S2) and were significantly associated with changes in EPIC‐26 scores over time. While univariable comparison of EPIC‐26 urinary incontinence scores at 24 months showed differences between the two groups (72 vs 80; P < 0.01), this difference was not statistically significant after adjustment for confounders (−2.7 [95% CI −6.6, 1.1]).

Table 2.

Mixed effects linear model for EPIC‐26 sexual function and urinary incontinence scores over 24 months of follow‐up.

Variable	Sexual function			Urinary incontinence
	Difference	95% CI	P	Difference	95% CI	P
Ever smoking	−11.0	−15.0, −7.0	<0.01	−2.7	−6.6, 1.1	0.08
Month of follow‐up	0.7	0.6, 0.8	<0.01	0.6	0.6, 0.7	<0.01
Age at surgery	−1.3	−1.4, −1.1	<0.01	−0.4	−0.6, −0.3	<0.01
Year of surgery	0.4	0.1, 0.7	<0.01	0.0	−0.3, 0.3	0.49

Month of follow‐up, age at surgery and year of surgery were significantly associated with post‐prostatectomy EPIC‐26 scores over time for both ever smokers and never smokers. Sexual function score improved by 0.7 points per month (95% CI 0.6, 0.8) and urinary incontinence score improved by 0.6 points per month (95% CI 0.6, 0.7) after RP (Table 2). Statistical interaction analysis between smoking and month of follow‐up showed no significant association with the outcome between the two variables (0.2 [95% CI −0.4, 0.0] for sexual function and −0.1 [95% CI −0.3, 0.1] for urinary incontinence). Age at surgery was associated with a mean decline of 1.3 points per year (95% CI −1.4, −1.1) for sexual function and a mean decline of 0.4 points per year (95% CI −0.6, −0.3) for urinary incontinence. Statistical interaction between smoking and age at surgery also showed no significant association with the outcome between the two variables (0.4 [95% CI −0.1, 0.9] for sexual function score and −0.3 [95% CI −0.7, 0.1] for urinary incontinence). Surgery performed in a more recent year was associated with an increase of 0.4 points per year (95% CI 0.1, 0.7) for EPIC‐26 sexual function scores and was not significantly associated with urinary incontinence scores.

We adjusted for other possible confounding variables by including all statistically significant variables on univariable analysis in a mixed effects multivariable model, which did not alter the overall results. We observed a 16‐point difference between nerve‐sparing and non‐nerve‐sparing surgical approaches (95% CI 11.7, 20.2); however, adding ‘nerve‐sparing’ to our mixed effects linear model did not alter the overall observed negative association of smoking and sexual function scores (Appendix S4).

Consistency was tested in a subgroup analysis of baseline EPIC‐26 scores, which showed that patients with both statistically (5% alpha) and clinically significant (reaching minimal clinically important difference) post‐surgical decline in EPIC‐26 sexual function scores were those with a high baseline EPIC‐26 sexual function score above 80 who had ever smoked (Appendix S5). Smoking was not associated with differences for individuals with lower baseline scores. Robustness was tested through sensitivity analysis of smoking status gathered through PROM responses, which showed a statistically significant difference among the groups at 3 months after surgery, with an average of 18 points for current smoking, 22 points for past smoking and 28 points for never smoking (P = 0.01; Fig. 2B). Mixed model regression analysis showed that current smoking was associated with a statistically significant decrease of 10 points (95% CI −16.6, −3.4, P < 0.01) at all time points for EPIC‐26 sexual function scores compared to never smoking (Table 3). Baseline EPIC‐26 sexual function scores were similar in the three groups (53, 59 and 62, respectively; P = 0.16). Smoking was not found to be associated with urinary incontinence scores in our cohort in subgroup and sensitivity analyses.

Table 3.

Subgroup analysis measuring difference in EPIC‐26 scores for never, past and current smokers.

Subgroup	Sexual function			Urinary incontinence
	Difference*	95% CI	P	Difference*	95% CI	P
Current vs never smoker	−10.0	−16.6, −3.4	<0.01	−0.2	−7.0, 6.6	0.48
Current vs past smoker	−7.9	−14.4, −1.3	0.01	−0.2	−7.3, 7.0	0.48
Past vs never smoker	−3.0	−6.1, 0.2	0.03	0.6	−3.9, 2.7	0.36

^*Adjusted for month of follow‐up, age at surgery and year of surgery.

Discussion

Smoking was associated with lower EPIC‐26 scores for sexual function compared to not smoking in our cohort. On average, ever smokers scored lower by 11 points (95% CI −15.0, −7.0). We identified age and year of surgery to be confounders by means of a causal DAG (Appendix S2) and adjusted accordingly. Subgroup analysis showed that the group with baseline scores above 80 was associated with smoking‐related post‐prostatectomy sexual function loss.

Sexual function outcomes are poor overall after treatment for prostate cancer, and even more so for RP, with only 23% of patients reporting fair, good or very good sexual function after surgery [6]. The observed initial drop in function score at 3 months after RP (Fig. 2) has been previously reported in the literature, as well as a consistent increase in function score between 3 and 24 months [5, 9, 13]. It seems that men who are current smokers or have smoked in the past recover at a lower rate throughout the follow‐up period than men who have never smoked.

Our analyses showed that patients with baseline EPIC‐26 sexual function scores above 80 who had ever smoked (ever smokers), experienced significantly less sexual function recovery after RP, compared to never smokers. No differences were observed in subgroups with lower baseline scores (Appendix S5). Recovery of sexual function will be dependent on an individual's baseline sexual function score, as recovery can only be expected in cases of sufficient prior function.

Improvements in EPIC‐26 sexual function score that are likely to be noticed by an individual (i.e., minimal clinically important difference) are between 10 and 12 points [17]. The average reduction identified in our study of 11 points for men who had ever smoked was within the 10–12 points considered to be noticeable and therefore clinically relevant to an individual. Considering previously conducted sexual function threshold studies, this increase could also mean that an individual would move from ‘poor function’ to ‘intermediate function’, or from ‘intermediate function’ to ‘good function’ [18, 19]. For example, an individual who smokes with a post‐prostatectomy EPIC‐26 sexual function score of 35 (i.e., poor function), could have had a score of 46–47 (i.e., intermediate function) if they were non‐smoking. Smoking was not associated with EPIC‐26 urinary incontinence scores in our study, which is consistent with findings from other studies [20, 21]. In a systematic literature review by John in 2020 [22], over 50 articles were analysed regarding the association between smoking and urinary function and the evidence for any association was considered to be weak. Our study corroborates the absence of such an association.

Age was significantly associated with EPIC‐26 scores for both sexual function and urinary incontinence, which is consistent with the literature [23, 24]. Age at surgery was associated with a mean decline of 1.3 points per year lived (95% CI −1.4, −1.1) for sexual function and a mean decline of 0.4 points per year lived (95% CI −0.6, −0.3) for urinary incontinence. This means that a 70‐year‐old patient with prostate cancer experienced a clinically significant 13‐point sexual function score reduction compared to a 60‐year‐old patient with similar disease and comorbidities. This effect was less strong for urinary incontinence; however, it becomes clinically significant with larger age differences [17].

Year of surgery was associated with an increase of 0.4 points per year (95% CI 0.1, 0.7) in EPIC‐26 sexual function score, which means that sexual function recovery is typically better after more recent prostatectomies. Probable reasons for this are the gradual shift from open to robot‐assisted RP and the progressive improvement in surgical proficiency with the robotic approach over time.

Nerve‐sparing surgery is known to positively influence sexual function scores after RP [4, 5] and we observed a 16‐point difference between nerve‐sparing and non‐nerve‐sparing approaches (95% CI 11.7, 20.2). The available data on nerve‐sparing were limited in our cohort and reporting varied among participating hospitals.

It is unclear, however, if smoking cessation before or after RP could also be of clinical importance to men with prostate cancer. In accordance with previous evidence [10], our study suggests that being a past smoker (i.e., smoking cessation) may improve recovery of sexual function scores compared to current smoking, although numbers were small. Evidence in the literature for post‐prostatectomy patients with localised disease is scarce, however, continuing smoking increases the likelihood of developing erectile dysfunction over time in both the general and the prostate cancer population [8, 11]. We hypothesise that smoking cessation potentially improves sexual function recovery after surgery, however, future studies would be needed to investigate this hypothesis.

Regardless of post‐prostatectomy function recovery, smoking cessation is known to improve many aspects of our health, however, it currently does not occupy a major role in the post‐surgical treatment of patients with prostate cancer [25], with only 7% of active smokers receiving smoking cessation counselling in the urological setting [26]. The AUA do not mention smoking cessation in their 2022 guidelines for management of localised prostate cancer [27]. The European Association of Urology guidelines mention smoking to be associated with worse prognosis and post‐prostatectomy stricture rates, and although smoking cessation is recommended for patients undergoing androgen deprivation therapy, smoking cessation counselling is currently not included in the context of prostatectomy [2]. Our research suggests that smoking cessation interventions should always be offered in urological practice to individuals who smoke, as those who received brief cessation counselling from their urologist were 2.3 times more likely to attempt to quit and had a 4.4 times higher 1‐year cessation rate compared to a usual care group [28]. Further research may include a prospective randomised controlled trial among men treated with RP who smoke to investigate the effects of smoking cessation on post‐prostatectomy urinary and sexual function, as well as other health outcomes.

This research contributes to the literature by addressing the main limitations identified in our systematic literature review [12]. Our research used the EPIC‐26 instrument, which is standardised and validated to measure erectile function and urinary incontinence. In addition, this study included a 24‐month follow‐up, the period during which most of the functional recovery is observed after RP [13], whereas five out of nine analysed studies in our meta‐analysis had shorter follow‐up periods of 3–12 months. Our results are therefore more robust and remain consistent with previous research [11, 12, 29]. Strengths of this study include the large sample size obtained from a large state‐wide prostate cancer registry, with sufficient data present to perform some subgroup and sensitivity analyses.

Our analytical method included identification of potential bias prior to statistical analysis by means of a causal DAG. This is a valuable tool in observational studies, providing systematic visualisation and clarification of complex hypothesised causal relationships among multiple variables, helping to explicitly define the assumptions underlying the study and guiding interpretation of results. DAGs facilitate the identification of mediators, confounders and colliders (i.e., potential biases), improve the clarity of causal relationships, and help guide appropriate statistical analysis, ultimately leading to more accurate and reliable conclusions. DAGs guide researchers in pinpointing which variables should be controlled for in analysis to avoid bias to ensure that the relationships between exposure and outcome are properly modelled. Our DAG identified a minimally sufficient adjustment set containing age and year of surgery for estimating the total effect of smoking on sexual dysfunction and incontinence. Adjusting for ‘nerve‐sparing’ in our model did not influence the overall results (Appendix S4), which is in accordance with the DAG outcome. We also performed several subgroup and sensitivity analyses to further investigate identified associations.

Study limitations included limited questionnaire response rates. More than a third (39%) of the 4356 eligible individuals were excluded because they did not return a PROM questionnaire and the majority of the remaining participants had at least one questionnaire missing during follow‐up. We compared our eligible cohort stratified by ‘having provided PROMs’ and observed several differences between groups (Appendix S6). People who did not provide PROMs were typically younger at diagnosis and surgery, and more often had lower‐grade disease. The current PROMs schedule was only introduced in 2007, so individuals who had a longer time between their prostatectomy and first PROM invitation may have been less inclined to participate or were already lost to follow‐up. Non‐responders also more often had bilateral nerve‐sparing surgery and less often had positive surgical margins, which may suggest that non‐responders experienced fewer symptoms. Furthermore, patients with a higher degree of bother may be more inclined to participate in quality‐of‐life‐related questionnaires, which is a potential source of bias.

Our follow‐up of 24 months was sufficient to expect recovery of most sexual and urinary function after prostatectomy and is a follow‐up period commonly used in other studies [3, 5]. Even longer follow‐up may be advised in future studies, as sexual function recovery continues beyond 24 months follow‐up [13]. Normalisation of sexual dysfunction with advancing age and acceptance of treatment‐induced function loss could also play a role [30].

Other health behaviour, such as alcohol consumption, was unknown in our study population, but may have impacted our outcomes. In addition, no information was available regarding sexual rehabilitation incidence, which includes the broader goal of preserving intimacy between partners after prostate surgery, and finding alternative ways to cope with the loss of function, as well as phosphodiesterase‐5 inhibitor use or other penile rehabilitation therapies. We expect, however, that rehabilitation incidence would be distributed equally among the patients, independent of smoking status.

The PROM questionnaire only included smoking questions from 2017. We had to rely on incomplete hospital records for classification of smoking prior to 2017. A cross‐tabulation analysis showed that, where hospital records report ‘ever smoker’, this is matched in 95% of PROMs. Where hospital records report ‘never smoker’, 53% were never smokers according to PROMs (Appendix S7). Furthermore, we did not have access to data relating to pack‐years and time since cessation. Future studies may include more detailed patient‐specific lifestyle and health behaviour data.

In conclusion, this study provides evidence that smoking is associated with sexual function declines in patients treated with RP for localised prostate cancer. An association between smoking and urinary incontinence was not detected.

Disclosure of Interests

A research scholarship was obtained by J.V. from the Freemasons Centre for Male Health & Wellbeing and the Flinders Foundation, who were not involved in the initiation or execution of this study. The authors have no other interests to declare.

Supporting information

Appendix S1. STROBE statement – checklist of items that should be included in reports of observational studies.

Appendix S2. Causal directed acyclic graph (DAG)*, showing (assumed) interactions between variables.

Appendix S3. Comparison of EPIC‐26 means during follow‐up never smoking vs ever smoking.

Appendix S4. Mixed effects linear model predicting sexual and urinary function EPIC‐26 score change, including adjustment for nerve sparing, with no significant change of the results for smoking compared to Table 2.

Appendix S5. Table of subgroup analysis of baseline EPIC‐26 scores.

Appendix S6. Comparison of cohort based on having provided PROMs or not.

Appendix S7. Ascertainment analysis cross tabulation of smoking status via hospital records or PROM records showing that 95% of PROM smoking records match the hospital smoking records and 47% of PROM smoking records did not match hospital smoking records, probably due to incomplete hospital records.