SEARCH

SEARCH BY CITATION

Summary

  1. Top of page
  2. Summary
  3. What’s known
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Conclusion
  9. Acknowledgements
  10. Author contributions
  11. References

Aim:  To determine the rate of newly detected underlying disease in men receiving their first (index) phosphodiesterase type 5 inhibitor (PDE5i) prescription.

Methods:  This non-interventional, retrospective study used anonymised patient records from UK general practices identified from the THIN database. Records of men aged ≥ 18 years, who received an index PDE5i prescription between January 1999 and June 2008 and with a continuous medical history (≥ 60 months) before the index prescription were included. Primary end-points were the prevalence of underlying disease prior to the index prescription and to establish the detection rate, defined as cumulative incidence of such a diagnosis in the 3 months following the index prescription. Assessments included comparison with age-matched controls, comparison with identical time periods immediately before and 1 year after, index prescription, and changes over time during the study period. Descriptive statistics, analysis of proportions and multivariate logistic regression analysis were used.

Results:  Among the 24,708 patients receiving a PDE5i, the prevalence of any underlying diagnosis before the index prescription was 70.23%; prevalence of vasculogenic disease was highest (48.20%). The detection rate of any underlying disease was 11.53%, and again highest for vasculogenic disease (4.07%). Compared with an age-matched control population, the additional detection rate of an unknown underlying disease at PDE5i prescription was 45 for hypertension, 61 for hypercholesterolaemia, 38 for diabetes and 5 for hypogonadism per 10,000 men.

Conclusion:  Only a minority of men with erectile dysfunction have a previously undiagnosed important underlying disease that is uncovered at the time of an initial PDE5i prescription by a GP.


What’s known

  1. Top of page
  2. Summary
  3. What’s known
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Conclusion
  9. Acknowledgements
  10. Author contributions
  11. References

Erectile dysfunction (ED) can be caused by a variety of potentially underlying and reversible factors. Recently, ED has been identified as an independent risk factor for CVD, including angina, myocardial infarction, stroke and transient ischaemic attack. Management recommendations for ED suggest a clinical assessment to detect underlying conditions when prescribing a PDE5i.

What’s new

These findings suggest that most men with ED have known comorbidities, identified during routine healthcare assessments and not in the context of a PDE5i prescription. However, it may be beneficial that men are regularly asked about erection problems in clinical practice, as the opportunistic identification of ED can be a trigger to address CVD risk factors more effectively, enabling the targets of the UK Quality and Outcomes Framework to be achieved.

Introduction

  1. Top of page
  2. Summary
  3. What’s known
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Conclusion
  9. Acknowledgements
  10. Author contributions
  11. References

Erectile dysfunction (ED), defined as the inability to attain and/or maintain an erection sufficient for satisfactory sexual intercourse, is a prevalent medical condition affecting approximately 10–50% of the male population aged over 40 years (1–4). The origin of ED can be physical (organic ED), psychological or a combination of the two, and the condition is influenced by vasculogenic [including cardiovascular disease (CVD)], neurogenic, anatomical/structural, hormonal (e.g. hypogonadism), drug-induced, and psychogenic factors (5,6).

Since the launch of the first oral phosphodiesterase type 5 inhibitor (PDE5i) sildenafil more than a decade ago (7), PDE5is have become the first-line therapeutic option for the treatment of ED due to their established efficacy and favourable safety profile (8–10). ED has recently been identified as an independent risk factor for CVD, including angina, myocardial infarction, stroke and transient ischaemic attack. In an analysis of data from more than 9000 men aged 55 years or older enrolled in the Prostate Cancer Prevention Trial, men with ED were reported to have an approximate twofold increased risk of CVD compared with men without ED (11). Furthermore, ED has been reported to precede a major cardiovascular event by an average of about 5 years (12). These findings may be explained by the artery size hypothesis. Penile arteries are smaller than coronary arteries (1–2 mm vs. 3–4 mm), potentially explaining why ED presents in advance of coronary artery disease (CAD) symptoms; as a similar degree of endothelial dysfunction may be subclinical in the larger vessels. This could further explain the high frequency of ED symptoms observed in patients with CAD, and the lack of concomitant symptoms of CAD observed in most patients with ED (13). In addition, the current evidence suggests that organic ED and CAD share a common aetiology in endothelial dysfunction (14–16).

On the basis of the variety of potentially underlying and reversible factors leading to ED, and the strong association between ED and the subsequent development of clinical cardiovascular events, both international (9,10,17) and UK-based (8) treatment guidelines recommend a targeted medical and psychosexual assessment of ED (including cardiovascular risk factors) prior to treatment with a PDE5i. However, despite the known burden and extent of comorbid conditions in patients with ED (18), data on the detection rate of underlying medical conditions in men assessed for the treatment of ED are limited to specialist single-centre studies (19,20). So far, no study has assessed the detection rate for previously unknown underlying medical conditions in men with ED undergoing medical evaluation for a PDE5i prescription in general practice in the UK.

The aim of this study was to determine the outcome of diagnostic activities performed in the temporal context of a PDE5i prescription by assessing the detection rate of newly diagnosed medical conditions following the initial prescription. Several analyses were conducted to account for potential influencing factors including the known difference in prevalence rates for cardiovascular risk factors in the ED population compared with a male population who had not previously received a PDE5i prescription, and the diagnostic assessments prior to and post prescription. Differences in diagnostic activities performed at the time when PDE5is were introduced (in the late 1990s) and in more recent years (when ED became recognised as a potential harbinger for CVD), were also investigated.

Methods

  1. Top of page
  2. Summary
  3. What’s known
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Conclusion
  9. Acknowledgements
  10. Author contributions
  11. References

The study was a non-interventional, retrospective database of medical records registered in The Health Improvement Network (THIN) database in the UK (http://www.thin-uk.com), a primary-care administrative database containing the electronic medical records of 5% of the UK population from the clinical systems of over 385 GP practices. As a managed database, records undergo systematic quality assessments. This database has previously been validated for epidemiological research (21) and has recently been used for epidemiological research in the field of urology (22–24). The THIN Data Collection Scheme was approved by the South-East Multicentre Research Ethics Committee (SE-MREC) in the UK. This study protocol was approved by the independent scientific review committee of THIN.

The primary objectives of the study were to determine the prevalence of underlying disease in male patients prior to receiving a first (index) PDE5i prescription, to establish the detection rate of new diagnoses (defined as the cumulative incidence of such a diagnosis in the 3 months (day 0–day 91) following the index prescription), and to determine the role of ED assessment in the identification of underlying disease in men receiving their first PDE5i prescription. The diagnoses of interest investigated in this study were adapted from the diagnostic evaluations recommended in the European Urology Association Guidelines on Male Sexual Dysfunction (9,25) and were divided into a vasculogenic, neurogenic, anatomic/structural, hormonal, pharmacologic (drug prescribed < 1 year before index prescription) and psychogenic category (Table 1). In addition, diagnoses of hypertension, hypercholesterolaemia, diabetes and hypogonadism were specifically assessed. The PDE5i prescription index date was used as a surrogate parameter for the time when a complete medical check would have been indicated. The 3-month time period following the index prescription was considered sufficient to include any new diagnosis identified based on additional tests requested at the time of PDE5i prescription. In order to fully capture the prevalence of underlying diseases, only patients with a continuous medical history of ≥ 60 months prior to the index prescription were included in the study.

Table 1.   Categories of conditions potentially leading to ED (adapted from the European Urology Association Guidelines)(9,25)
CategorySubcategoryMedical diagnoses
  1. *Drug must have been given in the year prior to initial PDE5i prescription/reference index date. ACE, angiotensin-converting enzyme; BPH, benign prostatic hypertrophy; LUTS, lower urinary tract symptoms; MAO, monoamine oxidase; NSRP, nerve-sparing radical prostatectomy; SSRI, selective serotonin reuptake inhibitor; TURP, transurethral resection of the prostate.

VasculogenicCardiovascular diseaseAtherosclerosis
Ischaemic vascular disease
Ischaemic heart disease
Renal failure
Cardiovascular risk factorsHypertension
Diabetes
Hypercholesterolaemia
Smoking
Obesity
UrologicAnatomicalPeyronie’s disease
Hypospadiasis
BPH/LUTS
IatrogenicProstatectomy (NSRP, TURP)
Pelvic and retroperitoneal surgery/radiotherapy
NeurogenicCentralMultiple sclerosis
Multiple system atrophy
Parkinson’s disease
Spinal cord injury
PeripheralPolyneuropathy
Hormonal Hypogonadism (primary/secondary), including hypopituitarism
Hyperprolactinaemia
Thyroid disease
Cushing’s and Addison’s disease
Pharmacologic* Thiazides, spironolactone, calcium channel blockers, ACE-inhibitors, beta-blockers, SSRI, tricyclic antidepressants, lithium, MAO inhibitors, benzodiazepines, ranitidine, cimetidine, methotrexate, interferon-alpha, alcohol, cocaine
Psychogenic Anxiety
Dementia
Depression
Neurotic disorders
Psychosis
Schizophrenia

Secondary study objectives included comparison of the prevalence and detection rate of underlying disease identified in the study subjects at index prescription with an age-matched control population not receiving a PDE5i. In addition, the detection rate of new diagnoses at index prescription was compared with that in a 3-month period immediately before the index PDE5i prescription (day −92 to day −1) and a 3-month period 1 year after the index prescription (day 366–day 457) in patients with a continuous medical history of ≥ 457 days following their index prescription (target sub-group population) to assess the detection of underlying conditions pre- and post PDE5i prescription. Finally, temporal trends were analysed among the target population by comparing the prevalence and 3-month post index detection rate of underlying medical conditions in subjects with an index prescription between 1st January 1999 and 31st December 2001 with that in subjects with a more recent index prescription (between 1st July 2006 and 30th June 2008). This was done to see if the increased awareness of ED as a marker of underlying disease had made any impact on the detection rate.

Patient population

Study participants were identified from anonymised THIN patient records. Records of acceptable quality (THIN patflag ‘A’ and ‘C’) of men aged ≥ 18 years, who received an index PDE5i prescription (sildenafil, tadalafil or vardenafil) between 1st January 1999 and 30th June 2008 were identified from the database (target population). Men disqualifying these criteria and women were excluded.

The reference population of age-matched controls was identified from the database (control population). The same inclusion criteria applied with an exception of a PDE5i prescription during the study period. For each subject of the target population, three control subjects were randomly selected based on the age at the first data entry date in any given year, and strata were filled starting at the end of the study period and working backwards (towards 1999).

Data collection and study end-points

The relevant READ codes for all diagnoses and Multilex drug codes for the pharmacological compounds were used to search the records of study participants and age-matched controls for instances of diagnoses of the medical conditions detailed in Table 1. To avoid multiplicity, only one code per diagnosis was captured, and the presence of at least one diagnosis defined whether the disease of a given category was present. All ever-coded diagnoses during the 60 months prior to the first PDE5i prescription (or date used for matching of control patients) were used to define the prevalence of underlying medical conditions at the time of index prescription. The cumulative incidence, i.e., the proportion of new diagnoses of interest within the 3-month period (0–91 days) following the index PDE5i prescription (or date used for matching controls) in each population was used to define the detection rate.

Statistical analysis

Categorical data were reported as counts and proportions, whereas continuous data were reported using descriptive statistics (mean, standard deviation). For the three age-matched controls per study participant, an average of the proportion or value was computed.

Comparisons of the detection rates between groups were made using the difference in proportions for categorical data and the difference in the means for continuous data, along with 95% confidence intervals (CI). For inferential analysis based on large samples, the χ2 test was used to compare categorical data for the groups of interest. The analysis of proportions and the two-sample t-test were performed to compare the overall mean prevalence, detection rates and the mean age at which men received a new diagnosis for the study participants and the control population. A two-sided p-value of < 0.05 was considered statistically significant.

For study participants who died after their index prescription date or reference index date (day 0), the last observation carried forward imputation method was used i.e. all observations (diagnostic codes) between day 0 and the date of death were carried forward and defined as a new diagnosis during the time period from day 0 until day 91.

To adjust for potential confounding effects on the difference in incidence of new diagnoses between the target and control populations, a multivariate regression analysis was performed using cardiovascular risk factors [hypertension, diabetes, hypercholesterolaemia, smoking and obesity (body mass index > 30)] and year of inclusion in the study cohort as covariates.

All statistical analyses were conducted using SAS 9.1 and 9.2 of the SAS Institute Inc., Cary, NC (http://www.sas.com).

Results

  1. Top of page
  2. Summary
  3. What’s known
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Conclusion
  9. Acknowledgements
  10. Author contributions
  11. References

A total of 89,781 men who received an index PDE5i prescription between 1st January 1999 and 30th June 2008 were identified from the THIN database. Among these patients, 24,708 met the inclusion criteria and were included in the target population. A total of 24,479 men in the target population had a continuous medical history of ≥ 457 days and were included in the subgroup target population. The control group comprised of 74,124 male patients. The baseline characteristics of the target population and control group are summarised in Table 2.

Table 2.   Baseline characteristics of the study and control populations
 Target population (= 24,708)Control population (= 74,124)
  1. BMI, body mass index; SD, standard deviation. Smoking was excluded as data on this was not routinely recorded until QOF commenced in 2005. If weight and BMI values were outside of pre-defined ranges (< 40 to > 200 kg and < 14 to > 48 kg/m2, respectively), they were considered missing.

Age (years)
 Mean (SD)57.63 (11.59)57.63 (11.59)
 Median (range)59.0 (18.0–91.0)59.0 (18.0–91.0)
Weight (kg)
 Mean (SD)87.95 (16.08)86.72 (16.82)
 Median (range)86.0 (44.0–200.0)85.0 (40.0–199.4)
BMI (kg/m2)
 Mean (SD)28.28 (4.56)28.05 (4.80)
 Median (range)27.7 (14.3–48.0)27.5 (14.0–48.0)

Prevalence of existing diagnoses

The prevalence of any coded diagnosis before the index PDE5i prescription was 70.23% (95% CI: 69.66–70.80). When each specific diagnosis was considered, vasculogenic diagnoses and co-medication (pharmacologic category) were found to have the highest prevalence rates of 48.20% (95% CI: 47.58–48.83) and 42.97% (95% CI: 42.36–43.59), respectively. The prevalence of all other categories was low (< 10%). The most prevalent vasculogenic diagnoses were hypertension 19.93% (95% CI: 19.43–20.43), diabetes 9.69% (95% CI: 9.32–10.06) and hypercholesterolaemia 8.26% (95% CI: 7.92–8.61). Hypogonadism was present in 0.19% (95% CI: 0.17–0.25) of the men prior to receiving a PDE5i (Figure 1).

image

Figure 1.  Prevalence (% and 95% CI) of underlying disease in men receiving a PDE5i (target population) and in age-matched controls without a PDE5i prescription (control population). All differences between target and control populations were significant (p < 0.001)

Download figure to PowerPoint

Incidence of new diagnoses

The detection rate of any previously unknown underlying condition in the 3 months following the index PDE5i prescription was 11.53% (95% CI: 11.13–11.93). The detection rate of a vasculogenic diagnoses was 4.07% (95% CI: 3.82–4.32), including new detection of hypertension, hypercholesterolaemia or diabetes in 1.21% (95% CI: 1.07–1.34), 1.04% (95% CI: 0.92–1.17) and 0.69% (95% CI: 0.59–0.80) of men, respectively (Table 3).

Table 3.   Detection rate of underlying conditions (over 3 months) in men receiving a PDE5i prescription (target population) and in age-matched controls (control population).
Category or diagnosisDetection rate, (%) (95% CI)Absolute difference in detection rate (%)*Multivariate analysis
Target population (= 24,708)Control population (= 74,124)p-valueOdds ratio† (95% CI)p-value
  1. *Difference between detection rate of the target and the control population. †Adjusted for pre-existing cardiovascular risk factors (hypertension, hypercholesterolaemia, diabetes, smoking and obesity) as well as time of inclusion into the study cohort. ‡Not included as a variable in the multivariate analysis.

Vasculogenic4.07 (3.82–4.32)2.89 (2.77–3.01)< 0.0011.18 (1.46–0.90)0.96 (0.77–1.18)0.67897
Hypertension‡1.21 (1.07–1.34)0.76 (0.70–0.82)< 0.0010.45 (0.29–0.60)1.81 (1.49–2.19)< .00001
Hypercholesterolaemia‡1.04 (0.92–1.17)0.43 (0.39–0.48)< 0.0010.61 (0.47–0.75)3.45 (2.74–4.34)< .00001
Diabetes‡0.69 (0.59–0.80)0.31 (0.27–0.35)< 0.0010.38 (0.12–0.65)2.83 (2.15–3.72)< .00001
Urologic0.45 (0.36–0.53)0.22 (0.19–0.25)< 0.0010.23 (0.32–0.13)2.72 (2.01–3.68)< .00001
Neurogenic0.09 (0.05–0.13)0.04 (0.02–0.05)0.0020.06 (0.10–0.01)2.21 (1.07–4.59)0.03317
Hormonal0.17 (0.11–0.22)0.08 (0.06–0.10)< 0.0010.09 (0.15–0.03)2.29 (1.36–3.87)0.00194
Hypogonadism0.05 (0.02–0.08)0.00 (0.00–0.01)< 0.0010.05 (0.02–0.08)9.92 (2.31–42.65)0.00206
Pharmacogenic4.16 (3.91–4.41)3.08 (2.95–3.20)< 0.0011.09 (1.37–0.80)1.43 (1.31–1.55)< .00001
Psychogenic0.75 (0.64–0.86)0.28 (0.24–0.32)< 0.0010.47 (0.59–0.35)2.36 (1.83–3.03)< .00001

Comparison vs. an age-matched control population

Compared with the control population, the target population had a significantly higher prevalence of existing diagnoses as overall (70.23% vs. 55.73%; p < 0.001) and in each category (Figure 1). The overall detection rate of new diagnoses was also significantly higher in the target population compared with the control population (11.53% vs. 7.49%; p < 0.001), and similar findings were seen when each specific category or diagnosis was considered. The absolute difference between the detection rates was calculated and shown in Table 3. The mean age of men who received a new diagnosis of diabetes or urologic, neurogenic, hormonal or psychogenic ED in the 3 months post prescription was significantly lower in the target population compared with the age-matched controls (Table 4). No significant difference was found for vasculogenic diagnoses or hypogonadism.

Table 4.   Mean age of men with newly detected diagnosis 3 months post index prescription (day 0–day 91)
CategoryTarget population (= 24,708)Control population (= 74,124)p-value
nMean ageSDnMean ageSD
  1. SD, standard deviation; p-values were based on two-sided two-sample t-tests.

Vasculogenic100556.711.7214057.311.40.1447
Hypertension29860.19.556360.88.50.2618
Hypercholesterolaemia25859.18.832260.39.10.1226
Diabetes17158.410.422962.49.0< 0.0001
Urologic11062.17.816364.48.30.0199
Neurogenic2355.013.12863.49.40.0112
Hormonal4158.212.25863.711.30.0221
Hypogonadism1360.512.1368.07.80.3315
Pharmacogenic102857.711.8228058.411.20.1132
Psychogenic18551.212.020855.413.20.0013

Adjustment for potential confounders

Table 3 summarises the odds ratio based on multivariate logistic regression analysis for the detection of a new diagnosis after adjustment of pre-existing cardiovascular risk factors (hypertension, hypercholesterolaemia, diabetes, smoking and obesity) and the year of inclusion in the study cohort. Odds ratios remained statistically significant for all diagnostic categories, except for vasculogenic disease.

Comparison of detection rate with a 3-month period prior to the index date and a 3-month period 1 year after the index date

The detection rates in the 3-month period prior to index prescription (day −92 to −1) and the 3-month period 1 year after prescription (day 366–457), were lower than that in the 3 months post index prescription for all diagnoses except for the urologic diagnoses and diabetes (which were diagnosed with a significantly higher frequency in the 3 months prior to index prescription) (Figure 2). While the detection rate for any new diagnosis immediately prior to the prescription was not statistically different to that in the 3 months post index prescription (11.08% vs. 11.49%; p = 0.153), it was significantly lower 1 year later when compared with the time of prescription (5.65% vs. 11.49%; p < 0.001).

image

Figure 2.  Comparison of detection rates (%) during the 3 months prior to index prescription (day −92 to day −1), the 3 months post index prescription (day 0 to day 91) and a 3-month period 1 year after the index prescription (day 366 to day 457). Data for the subgroup target population (= 24,479) with continuous recording over the full time period are presented. *p < 0.05 compared with the detection rate at index prescription

Download figure to PowerPoint

Change of prevalence and detection rate of underlying conditions over time

Prevalence of any underlying condition was significantly lower in the population receiving a PDE5i in the early study period included between 1999 and 2001 (= 892) than with the population included between mid 2006 and mid 2008 (= 9862), the end of the study period (67.49% vs. 72.14%; p = 0.004). The proportion of men with vasculogenic disease (44.39% vs. 54.66%; p < 0.001) or co-medication (43.50% vs. 48.74%; p = 0.003) at index prescription was significantly lower in the early study population, whereas the prevalence of psychogenic conditions (12.33% vs. 8.27%; p < 0.001) was significantly higher compared with the later period. However, detection rates overall (11.32% vs. 11.06%; p = 0.813), and for each category, were not statistically different over time (Figure 3).

image

Figure 3.  Comparison of prevalence (%) and detection rate (%) of underlying conditions in patients prescribed a PDE5i during the early study period (1999–2001; = 892) and the end of the study period (mid 2006 to mid 2008; = 9862). *p < 0.05; no significant differences were seen in the detection rates

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. What’s known
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Conclusion
  9. Acknowledgements
  10. Author contributions
  11. References

This retrospective database study using records from THIN found that seven out of ten men have an established diagnosis of underlying disease before they receive their first PDE5i prescription. The most common underlying medical conditions were vasculogenic diagnoses and co-medication, with prevalence rates of 48% and 43%, respectively. As expected, the prevalence of underlying disease (all medical conditions combined and considered separately) which could potentially lead to ED was significantly higher among patients receiving a PDE5i prescription compared with the reference population of age-matched controls without PDE5i prescription. Similarly, the detection rate of conditions potentially underlying ED was statistically significantly higher compared with the detection rate of a control population (11.5% vs. 7.5% for any diagnosis).

The outcome of diagnostic activities recommended for the assessment of ED at a PDE5i prescription was further assessed using the absolute differences in detection rates between the target and the control population. The study results suggest that following the clinical assessment for ED, the absolute number of newly detected underlying diagnoses (per 10,000 men) above the normal detection rate is 45 for hypertension, 61 for hypercholesterolaemia and 38 for diabetes. Furthermore, previously undiagnosed hypogonadism was detected in an additional five out of 10,000 men assessed for an initial PDE5i prescription. The odds ratios remained statistically significant for participants in the target group vs. the control group, even when adjusted for potential confounding factors in the ED population (resulting from a higher prevalence of pre-existing CV risk factors) and the skewed inclusion date of the control population (towards the end of the study period), with the exception of new vasculogenic diagnoses where the odds ratio suggested no difference when adjusted for pre-existing CV risk factors. These findings imply that only a relatively small number of men with ED will have a previously undiagnosed, important underlying disease uncovered at the time of an initial PDE5i prescription. However, it may remain beneficial that men are regularly asked about erection problems in clinical practice, given that a substantial proportion of men with ED do not seek medical advice or treatment (26,27) or access treatment from sources outside of the healthcare system (28).The opportunistic identification can be a trigger to assess and manage ED according to the accepted treatment recommendations (8–10), and to address CVD risk factors more effectively, enabling the targets of the UK Quality and Outcomes Framework (QOF) to be achieved.

The detection rates reported in this study are similar or somewhat lower than those described in two single-centre studies. Hatzichristou et al. reported that blood tests identified a previously undiagnosed medical condition in 6.2% of 1276 evaluable patients with ED (mean duration 4.9 years) (19). Diabetes mellitus and abnormal lipid profiles were newly detected in 11 (0.9%) and 37 (2.9%) patients, respectively, and hypogonadism and hyperprolactinaemia were uncovered in 22 (1.7%) and 8 (0.6%) patients, respectively. In 28 patients (2.2%), a previously unknown urinary tract disease was diagnosed and in eight patients (0.6%), tests revealed a previously unknown neurologic disorder. Solomon et al. described even higher rates following an assessment of 174 participants in a joint urologic and cardiologic referral practice using the Princeton Consensus Guideline. The detection rate of vasculogenic conditions was high, with a new diagnosis of hyperlipidaemia, hypertension or diabetes, and suspected significant angina reported in 26%, 10% and 6% of men, respectively (20). In addition, a potential urologic disorder was identified in five men and hyperprolactinaemia was reported in one man.

The overall mean age of the target population of men at the time of their first index prescription for PDE5i in this analysis of general population data was 59 years. For some diagnoses, the mean age of men in the target group in the period 3 months post prescription was significantly lower than that in the control population. As this study was not specifically intended to investigate the age at which these diagnoses were made, it remains questionable whether ED assessment in the context of a PDE5i prescription currently would lead to an earlier detection of some potentially underlying conditions.

Two factors may have contributed to the lower diagnosis rates reported in this retrospective database study compared with the two specialist single-centre studies. First, patients who complained about ED might have received a PDE5i prescription only after all the results of the clinical assessments were available, although guidelines do not explicitly recommend a deferred treatment. Interestingly, a considerable number of new diagnoses were made in the 3 months prior to the PDE5i prescription, and were substantially lower in the period 1 year after the index prescription. The detection rate seen in this later period is in a similar range to the rate seen in an age-matched control population without a PDE5i prescription.

The second factor which may account for the lower detection rates reported in this observational study, compared with previous studies from specialist centres which specifically investigated the presence of underlying conditions, is the extent to which these investigations are actually performed by physicians in general practice. Although no information about this is available from general practitioners, a survey among urologists suggests these investigations are often abbreviated or deferred: of 1590 participating delegates of the annual congress of the European Association of Urologists 2008, 33% reported that they would perform further interviews with 19%, 15% and 7%, respectively, reporting that they would assess cardiovascular risk factors, order laboratory tests and perform blood pressure measurements in a man presenting with ED (29). It is difficult to assess the level of clinical investigation performed during routine GP practice; however, the comparable detection rates of underlying conditions during the early phase and the later phase of the study suggest that there was no change over time, despite the statistically higher prevalence of known vasculogenic diagnoses and co-medication in the cohort receiving a PDE5i between July 2006 and June 2008. The 2004 UK QOF targets would likely have contributed to this finding (30).

This study has limitations inherent to any retrospective database study (31), including the potential effect that under- and misreporting would have no prevalence and detection rates. Further limitations of this study include the fact that the data represent a temporal rather than a causal relationship between the detection rate of new diagnoses and the PDE5i prescription. Furthermore, it can not be ruled out that some men in the control population may have had ED (diagnosed or undiagnosed) but did not receive treatment. Current knowledge of the pharmacologic effect and the safety of PDE5i does not support an inversion of the interpretation that hypertension or diabetes is induced by a PDE5i prescription (17,32,33). More importantly, it is impossible to determine in this study, whether the symptoms of ED were the cause of the GP visit, or whether ED was identified as a part of other health-related assessments. This could lead to an overestimation of the detection rates for underlying conditions. Alternatively, and as discussed above, the presented detection rates could be lower than that in reality, because the PDE5i prescription index date was used as a surrogate parameter for the time when a complete medical check would be performed. This excludes the detection of underlying conditions diagnosed before the PDE5i was actually prescribed. Finally, patients with private PDE5i prescriptions were not included in the analysis and the impact of this on the study outcome remains unknown.

Conclusion

  1. Top of page
  2. Summary
  3. What’s known
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Conclusion
  9. Acknowledgements
  10. Author contributions
  11. References

The high prevalence and low detection rates observed in this study suggest that in current clinical practice, many of the diagnoses associated with ED, including vasculogenic conditions, are identified during routine healthcare assessments and not as a result of a diagnostic work-up for ED. In consequence, a vast majority of men with ED have a previously known medical condition when asking for a PDE5i treatment. Routine clinical assessments in the context of a PDE5i prescription appear to uncover an additional case of hypertension, hypercholesterolaemia or diabetes in, approximately, one out of 200 men and a hypogonadism in one out of 2000 men compared with the normal detection rate in an age-matched control population. No change was seen in the detection rate over the past 10 years.

Acknowledgements

  1. Top of page
  2. Summary
  3. What’s known
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Conclusion
  9. Acknowledgements
  10. Author contributions
  11. References

The authors would like to thank John McPeek and William Moritz of Eliassen, Katie Elsworth and Cornea Venter of Quanticate, and Alexa Parliyan of Pfizer Inc for programming support for database analysis, and Dr Jamie Ashman of Prism Ideas for providing editorial support during the preparation of this article. Funding for editorial support and for programming support was provided by Pfizer.

Author contributions

  1. Top of page
  2. Summary
  3. What’s known
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Conclusion
  9. Acknowledgements
  10. Author contributions
  11. References

Design of the study: MK, GS, KHZ, TS; analysis of the data: GS, KHZ; interpretation of the data: MK, GS, KHZ, TS; drafting of the manuscript and review for content: MK, GS, KHZ, TS.

References

  1. Top of page
  2. Summary
  3. What’s known
  4. Introduction
  5. Methods
  6. Results
  7. Discussion
  8. Conclusion
  9. Acknowledgements
  10. Author contributions
  11. References