Dietary Antioxidants and Longitudinal Changes in Lower Urinary Tract Symptoms in Elderly Men: The Osteoporotic Fractures in Men Study

We previously reported on descriptive AUA-SI data for this analytic cohort [26] . Among men with mild baseline symptoms, men with LUTS progression had slightly lower physical activity and were more likely to report mobility limitations and a history of dizziness and non–prostate cancer when compared to men with stable LUTS. (As mentioned earlier, those with a history of prostate cancer were excluded from analyses.) In those with moderate baseline symptoms, men with LUTS progression were marginally younger, had a significantly lower SF-12 mental score, were less likely to live alone, had lower overall energy intake, and were more likely to suffer from depression and back pain when compared to men with stable LUTS. By contrast, men with LUTS remission had a slightly higher SF-12 mental score (closer to those with stable symptoms); had less education; were less likely to have a history of problem drinking, hypertension, or angina; were less likely to use diuretics; and consumed less caffeine when compared to men with LUTS progression ( Table 1 ).

Among men with mild baseline symptoms, LUTS progression was not associated with the antioxidants vitamin C, vitamin E, β-carotene, α-carotene, β-cryptoxanthin, lycopene, and lutein/zeaxanthin when comparing the highest quartile (Q4) to the lowest quartile (Q1) for intake ( Table 2 ).

Among men with moderate baseline symptoms, higher antioxidant intake was similarly not associated with a lower likelihood of LUTS progression ( Table 3 ). In analyses of LUTS remission relative to LUTS progression ( Table 4 ), higher lycopene intake was associated with a lower likelihood of LUTS remission after adjustment for energy intake, history of problem drinking, angina, and educational level (OR 0.37, 95% CI 0.16–0.82 for Q4 vs Q1; p = 0.04 for trend). However, this association was no longer statistically significant after adjustment for multiple comparisons.

This study used data collected in the Osteoporotic Fractures in Men Study (MrOS), a prospective study of community-dwelling men enrolled from six US sites [16] and [17]. The study was designed to evaluate risk factors for fracture, falls, and other conditions relevant to aging men, including prostate disease and LUTS [17] . From March 2000 to April 2002, 5994 men age 65–100 yr who could walk unassisted and had at least one natural hip for bone density measurement were enrolled. The six sites were Birmingham, AL; Minneapolis, MN; Palo Alto, CA; Pittsburgh, PA; Portland, OR; and San Diego, CA. The study protocol was approved by the institutional review boards at all participating institutions, and all men gave written informed consent.

Baseline measures collected at clinic visits included basic demographic data, lifestyle information (alcohol use, cigarette smoking), medical conditions, self-rated health, quality of life (short-form 12, SF-12 [18] ), and physical activity information based on the Physical Activity Scale for the Elderly (PASE) [19] . Height and weight were measured at the clinic visit and body mass index (BMI) was computed (kg/m²) [20] . Medications and supplements were brought to the baseline clinic visit and were inventoried by study staff and matched to ingredients according to on the Iowa Drug Information Service drug vocabulary (College of Pharmacy, University of Iowa, Iowa City, IA, USA) [21] .

Dietary data were collected at baseline using a brief Block food frequency questionnaire (FFQ) [22] and [23], which contains 69 items specifically drawn from foods most frequently consumed by elderly US men according to an analysis of data from the Third National Health and Nutrition Examination Survey (NHANES III) [22] . The Block FFQ is a validated, robust instrument used in observational research to measure nutrient intake, including antioxidant micronutrients. The brief Block FFQ asks about consumption frequency and portion size for each item. Gram estimates of foods were calculated as the gram weight for the chosen portion size multiplied by the consumption frequency. Nutrients were then estimated using the average amount of a nutrient in a food multiplied by the gram weight consumed. Nutrients from supplements were calculated separately from food estimates. Total intake was calculated as the sum of the nutrient estimates from food plus supplement nutrient estimates, when applicable. Supplement information was available only for the antioxidants vitamin C, vitamin E, and β-carotene.

At baseline and approximately every 2 yr (2002–2004, 2005–2006, and 2007–2009), follow-up data were collected on lower urinary tract health, including the American Urological Association Symptom Index (AUA-SI) and history of LUTS treatment. Additional follow-up occurred every 4 mo via mailed questionnaires to collect reports of deaths and incident prostate cancer cases, which were adjudicated by study physicians using death certificates and pathology reports.

The following baseline variables were classified into categories for the analysis. BMI was classified as normal (<25.0 kg/m²), overweight (25.0–29.9 kg/m²), or obese (≥30.0 kg/m²) [20] . Depressed mood was defined as an SF-12 mental component score ≤45 points [24] . Alcohol consumption was classified as never, ≤14 drinks/wk, or >14 drinks/wk; problem drinking was defined as a CAGE (Cutting down, Annoyance by criticism, Guilty feeling, and Eye-openers) [25] score of >1. Central nervous system (CNS) medication was defined as use of antiepileptics, benzodiazepines, antidepressants, opioids, or sedatives at baseline. Participants missing medication information were coded as nonusers after analyses showed no difference between nonusers and those with missing data.

We first restricted the MrOS cohort to 3594 men who had no history of prostate cancer and no prior or current treatment for LUTS, including surgery or medication (α-blockers, antispasmodics, anticholinergics, 5α-reductase inhibitors). These men were then followed through the fourth AUA-SI assessment (2007–2009). During follow-up, the cohort for trajectory analysis was further restricted to 1740 men who remained free from diagnosed prostate cancer, reported no surgery or medication use for LUTS, and completed all four AUA-SI assessments. A figure demonstrating how the analytic cohort was ascertained for the trajectory analysis has been previously published [26] . The same study demonstrated no significant differences in LUTS trajectory among men with a history of stroke and those who used CNS medications [26] , so we did not exclude them from the present analysis.

Outcomes for this analysis were LUTS trajectories. Details of the trajectory method were reported previously [26] . In brief, group-based trajectory modeling was performed using the AUA-SI data from all four time points. Trajectory modeling uses the maximum likelihood method and applies a semiparametric mixed model to longitudinal data (PROC TRAJ for SAS 9.1) [27], [28], and [29]. The four trajectory types identified were stable (n = 1277), progressing (n = 345), and remitting (n = 98) LUTS groups, and one very small group (n = 20) in which the LUTS trajectory increased and then decreased during follow-up. We identified three LUTS progression trajectories during follow-up: men who progressed from mild to moderate LUTS; men who progressed from moderate to severe LUTS; and men who progressed from low-moderate to high-moderate LUTS [26] .

2.1. Selection of the analytic cohort

For the present study, our goal was to determine associations between antioxidant intake and the likelihood of LUTS progression relative to LUTS stability in groups stratified by baseline clinical cut points of mild (AUA-SI 0–7) and moderate (AUA-SI 8–19) LUTS. We excluded men with severe symptoms at baseline (AUA-SI >19) and those who underwent LUTS treatment during follow-up to minimize the potential for confounding and focus on elucidating the natural history of LUTS in elderly men. In addition, we examined the likelihood of LUTS remission relative to LUTS progression in those with moderate baseline symptoms. Therefore, we restricted our analysis to men who had a similar baseline AUA-SI and who were not missing any dietary data (n = 1670; Fig. 1 ).

Fig. 1 Selection of the analytic cohort. MrOS = Osteoporotic Fractures in Men Study; LUTS = lower urinary tract symptoms; BPH = benign prostatic hyperplasia.

Overall, we observed no significant associations between antioxidant intake and LUTS progression or remission over approximately 7 yr of follow-up. To the best of our knowledge, this is the first study to examine associations between dietary antioxidants and LUTS progression. It is also the first study to describe associations between dietary factors and spontaneous remission of LUTS by analyzing whether dietary factors are associated with LUTS remission in otherwise untreated men.

Our results differ from those for prior cross-sectional studies and fail to confirm beneficial associations between higher antioxidant intake and lower risk of LUTS and BPH previously observed in some populations. Rohrman and colleagues [8] reported that dietary intakes of vitamin C, lutein/zeaxanthin, and β-cryptoxanthin in the Health Professionals Follow-up Study were inversely associated with total BPH (the combined endpoint of either surgery for BPH or high/moderate to severe LUTS). Similarly, Maserejian and colleagues [10] reported 40–50% lower odds of LUTS for high dietary intake of vitamin C and β-cryptoxanthin among men with high iron intake. A study examining serum concentrations of micronutrients from NHANES III data also reported lower serum concentrations of vitamin E in men with compared to men without LUTS [9] . It is possible that the null association between antioxidant intake and LUTS in the older men in this study could be due to a lack of effect of antioxidants in older individuals. Future research may demonstrate different effects in younger populations. It is also possible that prior studies were limited by a cross-sectional design or that factors such as recall bias or measurement error confounded the results. In fact, the longitudinal nature of our analysis was a prominent strength of our study that could have provided a more robust characterization of LUTS over time compared to other cohorts.

Unexpectedly, for those with moderate LUTS at baseline, there was a lower likelihood of LUTS remission among men with the highest lycopene intakes, although this observation did not reach statistical significance after correction for multiple comparisons. Further research is needed to determine whether similar results are observed in longitudinal analyses in other large cohorts. Lycopene was associated with lower LUTS odds in two cross-sectional studies [9] and [10] and one BPH study [11] ; therefore, if this association is found to be valid, it would be contrary to previous research on lycopene.

Our study has multiple strengths. We used a robust definition of LUTS progression obtained via a group-based trajectory method, which reduced the inherent variability in symptom scores and allowed clear identification of symptom progression over time. Another strength is the exclusion of men treated for LUTS with medication or surgery from the analytic cohort to provide an unbiased natural history of the disease, free of potential treatment effects that would otherwise change the symptom trajectory and confound the findings. Finally, the study included a geographically diverse group of community-dwelling elderly men in the USA who had an extended follow-up of approximately 7 yr.

One main limitation of our study is use of the brief Block FFQ, which has fewer food groups than other FFQs and may have biased results toward the null hypothesis. For example, pumpkin is not included in the instrument, yet it is the highest dietary source of β-cryptoxanthin and the second highest source of β-carotene. Thus, if a man consumed pumpkin regularly, he may have been incorrectly categorized as a lower rather than a higher consumer of these antioxidants. However, any misclassification should have been random with regard to LUTS progression, since men did not know their future LUTS status when filling out the baseline FFQ, and thus would have biased results toward the null hypothesis. A second limitation is the relatively small number of non-Caucasian men in the cohort, which probably diminished the external validity of the results with respect to other ethnic and racial groups. A third limitation is that dietary information was only collected at baseline; it is possible that some study participants may have subsequently altered their diets during follow-up, leading to a potential for bias. Finally, we were unable to assess objective urologic measures such as urinary flow, prostate volume, prostate-specific antigen (PSA), and voiding diaries.

Lower urinary tract symptoms (LUTS) are common among elderly men and have substantial global adverse effects on male health [1] and [2]. LUTS have been associated with higher mortality and morbidity [3] and billions of US dollars in annual health care expenditure [4] . Since obesity and exercise have been associated with higher and lower risks of LUTS, respectively [5] and [6], lifestyle changes might potentially prevent LUTS progression. Dietary constituents, especially antioxidants, are candidate lifestyle targets for LUTS prevention. Since antioxidants have potent anti-inflammatory properties, an increase in dietary antioxidant consumption might decrease LUTS via modulation of inflammatory pathways involved in the pathogenesis of LUTS and benign prostatic hyperplasia (BPH) [7] .

At least three studies have reported inverse associations between LUTS or BPH and consumption of antioxidants or foods rich in antioxidants, including β-carotene [8], [9], and [10], lutein [8] , lycopene [9] and [10], total carotenoids [10] , vitamin C [8] , vitamin E and selenium [9] , vegetables [11], [12], [13], and [14], and fruits [13] and [15]. Higher consumption of fruits and vegetables (including those rich in β-carotene and lycopene) has also been associated with a lower risk of BPH incidence [8] and [11].

Since most of the prior studies were cross-sectional, temporal associations between dietary antioxidants and LUTS remain unclear. Moreover, to the nest of our knowledge, no studies have examined potential associations between dietary antioxidants and the risk of LUTS progression in elderly men. Given the high prevalence of LUTS in this population [2] , the development of relatively straightforward dietary interventions to prevent LUTS progression in elderly men may substantially inform the clinical care of men with LUTS. Therefore, we examined the association between baseline dietary antioxidant intake and subsequent LUTS progression over a 7-yr period in elderly men. We hypothesized that higher baseline consumption of dietary antioxidants is associated with a lower probability of LUTS progression.

We previously reported on descriptive AUA-SI data for this analytic cohort [26] . Among men with mild baseline symptoms, men with LUTS progression had slightly lower physical activity and were more likely to report mobility limitations and a history of dizziness and non–prostate cancer when compared to men with stable LUTS. (As mentioned earlier, those with a history of prostate cancer were excluded from analyses.) In those with moderate baseline symptoms, men with LUTS progression were marginally younger, had a significantly lower SF-12 mental score, were less likely to live alone, had lower overall energy intake, and were more likely to suffer from depression and back pain when compared to men with stable LUTS. By contrast, men with LUTS remission had a slightly higher SF-12 mental score (closer to those with stable symptoms); had less education; were less likely to have a history of problem drinking, hypertension, or angina; were less likely to use diuretics; and consumed less caffeine when compared to men with LUTS progression ( Table 1 ).

Among men with mild baseline symptoms, LUTS progression was not associated with the antioxidants vitamin C, vitamin E, β-carotene, α-carotene, β-cryptoxanthin, lycopene, and lutein/zeaxanthin when comparing the highest quartile (Q4) to the lowest quartile (Q1) for intake ( Table 2 ).

Among men with moderate baseline symptoms, higher antioxidant intake was similarly not associated with a lower likelihood of LUTS progression ( Table 3 ). In analyses of LUTS remission relative to LUTS progression ( Table 4 ), higher lycopene intake was associated with a lower likelihood of LUTS remission after adjustment for energy intake, history of problem drinking, angina, and educational level (OR 0.37, 95% CI 0.16–0.82 for Q4 vs Q1; p = 0.04 for trend). However, this association was no longer statistically significant after adjustment for multiple comparisons.

This study used data collected in the Osteoporotic Fractures in Men Study (MrOS), a prospective study of community-dwelling men enrolled from six US sites [16] and [17]. The study was designed to evaluate risk factors for fracture, falls, and other conditions relevant to aging men, including prostate disease and LUTS [17] . From March 2000 to April 2002, 5994 men age 65–100 yr who could walk unassisted and had at least one natural hip for bone density measurement were enrolled. The six sites were Birmingham, AL; Minneapolis, MN; Palo Alto, CA; Pittsburgh, PA; Portland, OR; and San Diego, CA. The study protocol was approved by the institutional review boards at all participating institutions, and all men gave written informed consent.

Baseline measures collected at clinic visits included basic demographic data, lifestyle information (alcohol use, cigarette smoking), medical conditions, self-rated health, quality of life (short-form 12, SF-12 [18] ), and physical activity information based on the Physical Activity Scale for the Elderly (PASE) [19] . Height and weight were measured at the clinic visit and body mass index (BMI) was computed (kg/m²) [20] . Medications and supplements were brought to the baseline clinic visit and were inventoried by study staff and matched to ingredients according to on the Iowa Drug Information Service drug vocabulary (College of Pharmacy, University of Iowa, Iowa City, IA, USA) [21] .

Dietary data were collected at baseline using a brief Block food frequency questionnaire (FFQ) [22] and [23], which contains 69 items specifically drawn from foods most frequently consumed by elderly US men according to an analysis of data from the Third National Health and Nutrition Examination Survey (NHANES III) [22] . The Block FFQ is a validated, robust instrument used in observational research to measure nutrient intake, including antioxidant micronutrients. The brief Block FFQ asks about consumption frequency and portion size for each item. Gram estimates of foods were calculated as the gram weight for the chosen portion size multiplied by the consumption frequency. Nutrients were then estimated using the average amount of a nutrient in a food multiplied by the gram weight consumed. Nutrients from supplements were calculated separately from food estimates. Total intake was calculated as the sum of the nutrient estimates from food plus supplement nutrient estimates, when applicable. Supplement information was available only for the antioxidants vitamin C, vitamin E, and β-carotene.

At baseline and approximately every 2 yr (2002–2004, 2005–2006, and 2007–2009), follow-up data were collected on lower urinary tract health, including the American Urological Association Symptom Index (AUA-SI) and history of LUTS treatment. Additional follow-up occurred every 4 mo via mailed questionnaires to collect reports of deaths and incident prostate cancer cases, which were adjudicated by study physicians using death certificates and pathology reports.

The following baseline variables were classified into categories for the analysis. BMI was classified as normal (<25.0 kg/m²), overweight (25.0–29.9 kg/m²), or obese (≥30.0 kg/m²) [20] . Depressed mood was defined as an SF-12 mental component score ≤45 points [24] . Alcohol consumption was classified as never, ≤14 drinks/wk, or >14 drinks/wk; problem drinking was defined as a CAGE (Cutting down, Annoyance by criticism, Guilty feeling, and Eye-openers) [25] score of >1. Central nervous system (CNS) medication was defined as use of antiepileptics, benzodiazepines, antidepressants, opioids, or sedatives at baseline. Participants missing medication information were coded as nonusers after analyses showed no difference between nonusers and those with missing data.

We first restricted the MrOS cohort to 3594 men who had no history of prostate cancer and no prior or current treatment for LUTS, including surgery or medication (α-blockers, antispasmodics, anticholinergics, 5α-reductase inhibitors). These men were then followed through the fourth AUA-SI assessment (2007–2009). During follow-up, the cohort for trajectory analysis was further restricted to 1740 men who remained free from diagnosed prostate cancer, reported no surgery or medication use for LUTS, and completed all four AUA-SI assessments. A figure demonstrating how the analytic cohort was ascertained for the trajectory analysis has been previously published [26] . The same study demonstrated no significant differences in LUTS trajectory among men with a history of stroke and those who used CNS medications [26] , so we did not exclude them from the present analysis.

Outcomes for this analysis were LUTS trajectories. Details of the trajectory method were reported previously [26] . In brief, group-based trajectory modeling was performed using the AUA-SI data from all four time points. Trajectory modeling uses the maximum likelihood method and applies a semiparametric mixed model to longitudinal data (PROC TRAJ for SAS 9.1) [27], [28], and [29]. The four trajectory types identified were stable (n = 1277), progressing (n = 345), and remitting (n = 98) LUTS groups, and one very small group (n = 20) in which the LUTS trajectory increased and then decreased during follow-up. We identified three LUTS progression trajectories during follow-up: men who progressed from mild to moderate LUTS; men who progressed from moderate to severe LUTS; and men who progressed from low-moderate to high-moderate LUTS [26] .

2.1. Selection of the analytic cohort

For the present study, our goal was to determine associations between antioxidant intake and the likelihood of LUTS progression relative to LUTS stability in groups stratified by baseline clinical cut points of mild (AUA-SI 0–7) and moderate (AUA-SI 8–19) LUTS. We excluded men with severe symptoms at baseline (AUA-SI >19) and those who underwent LUTS treatment during follow-up to minimize the potential for confounding and focus on elucidating the natural history of LUTS in elderly men. In addition, we examined the likelihood of LUTS remission relative to LUTS progression in those with moderate baseline symptoms. Therefore, we restricted our analysis to men who had a similar baseline AUA-SI and who were not missing any dietary data (n = 1670; Fig. 1 ).

Fig. 1 Selection of the analytic cohort. MrOS = Osteoporotic Fractures in Men Study; LUTS = lower urinary tract symptoms; BPH = benign prostatic hyperplasia.

Overall, we observed no significant associations between antioxidant intake and LUTS progression or remission over approximately 7 yr of follow-up. To the best of our knowledge, this is the first study to examine associations between dietary antioxidants and LUTS progression. It is also the first study to describe associations between dietary factors and spontaneous remission of LUTS by analyzing whether dietary factors are associated with LUTS remission in otherwise untreated men.

Our results differ from those for prior cross-sectional studies and fail to confirm beneficial associations between higher antioxidant intake and lower risk of LUTS and BPH previously observed in some populations. Rohrman and colleagues [8] reported that dietary intakes of vitamin C, lutein/zeaxanthin, and β-cryptoxanthin in the Health Professionals Follow-up Study were inversely associated with total BPH (the combined endpoint of either surgery for BPH or high/moderate to severe LUTS). Similarly, Maserejian and colleagues [10] reported 40–50% lower odds of LUTS for high dietary intake of vitamin C and β-cryptoxanthin among men with high iron intake. A study examining serum concentrations of micronutrients from NHANES III data also reported lower serum concentrations of vitamin E in men with compared to men without LUTS [9] . It is possible that the null association between antioxidant intake and LUTS in the older men in this study could be due to a lack of effect of antioxidants in older individuals. Future research may demonstrate different effects in younger populations. It is also possible that prior studies were limited by a cross-sectional design or that factors such as recall bias or measurement error confounded the results. In fact, the longitudinal nature of our analysis was a prominent strength of our study that could have provided a more robust characterization of LUTS over time compared to other cohorts.

Unexpectedly, for those with moderate LUTS at baseline, there was a lower likelihood of LUTS remission among men with the highest lycopene intakes, although this observation did not reach statistical significance after correction for multiple comparisons. Further research is needed to determine whether similar results are observed in longitudinal analyses in other large cohorts. Lycopene was associated with lower LUTS odds in two cross-sectional studies [9] and [10] and one BPH study [11] ; therefore, if this association is found to be valid, it would be contrary to previous research on lycopene.

Our study has multiple strengths. We used a robust definition of LUTS progression obtained via a group-based trajectory method, which reduced the inherent variability in symptom scores and allowed clear identification of symptom progression over time. Another strength is the exclusion of men treated for LUTS with medication or surgery from the analytic cohort to provide an unbiased natural history of the disease, free of potential treatment effects that would otherwise change the symptom trajectory and confound the findings. Finally, the study included a geographically diverse group of community-dwelling elderly men in the USA who had an extended follow-up of approximately 7 yr.

One main limitation of our study is use of the brief Block FFQ, which has fewer food groups than other FFQs and may have biased results toward the null hypothesis. For example, pumpkin is not included in the instrument, yet it is the highest dietary source of β-cryptoxanthin and the second highest source of β-carotene. Thus, if a man consumed pumpkin regularly, he may have been incorrectly categorized as a lower rather than a higher consumer of these antioxidants. However, any misclassification should have been random with regard to LUTS progression, since men did not know their future LUTS status when filling out the baseline FFQ, and thus would have biased results toward the null hypothesis. A second limitation is the relatively small number of non-Caucasian men in the cohort, which probably diminished the external validity of the results with respect to other ethnic and racial groups. A third limitation is that dietary information was only collected at baseline; it is possible that some study participants may have subsequently altered their diets during follow-up, leading to a potential for bias. Finally, we were unable to assess objective urologic measures such as urinary flow, prostate volume, prostate-specific antigen (PSA), and voiding diaries.

Lower urinary tract symptoms (LUTS) are common among elderly men and have substantial global adverse effects on male health [1] and [2]. LUTS have been associated with higher mortality and morbidity [3] and billions of US dollars in annual health care expenditure [4] . Since obesity and exercise have been associated with higher and lower risks of LUTS, respectively [5] and [6], lifestyle changes might potentially prevent LUTS progression. Dietary constituents, especially antioxidants, are candidate lifestyle targets for LUTS prevention. Since antioxidants have potent anti-inflammatory properties, an increase in dietary antioxidant consumption might decrease LUTS via modulation of inflammatory pathways involved in the pathogenesis of LUTS and benign prostatic hyperplasia (BPH) [7] .

At least three studies have reported inverse associations between LUTS or BPH and consumption of antioxidants or foods rich in antioxidants, including β-carotene [8], [9], and [10], lutein [8] , lycopene [9] and [10], total carotenoids [10] , vitamin C [8] , vitamin E and selenium [9] , vegetables [11], [12], [13], and [14], and fruits [13] and [15]. Higher consumption of fruits and vegetables (including those rich in β-carotene and lycopene) has also been associated with a lower risk of BPH incidence [8] and [11].

Since most of the prior studies were cross-sectional, temporal associations between dietary antioxidants and LUTS remain unclear. Moreover, to the nest of our knowledge, no studies have examined potential associations between dietary antioxidants and the risk of LUTS progression in elderly men. Given the high prevalence of LUTS in this population [2] , the development of relatively straightforward dietary interventions to prevent LUTS progression in elderly men may substantially inform the clinical care of men with LUTS. Therefore, we examined the association between baseline dietary antioxidant intake and subsequent LUTS progression over a 7-yr period in elderly men. We hypothesized that higher baseline consumption of dietary antioxidants is associated with a lower probability of LUTS progression.

We previously reported on descriptive AUA-SI data for this analytic cohort [26] . Among men with mild baseline symptoms, men with LUTS progression had slightly lower physical activity and were more likely to report mobility limitations and a history of dizziness and non–prostate cancer when compared to men with stable LUTS. (As mentioned earlier, those with a history of prostate cancer were excluded from analyses.) In those with moderate baseline symptoms, men with LUTS progression were marginally younger, had a significantly lower SF-12 mental score, were less likely to live alone, had lower overall energy intake, and were more likely to suffer from depression and back pain when compared to men with stable LUTS. By contrast, men with LUTS remission had a slightly higher SF-12 mental score (closer to those with stable symptoms); had less education; were less likely to have a history of problem drinking, hypertension, or angina; were less likely to use diuretics; and consumed less caffeine when compared to men with LUTS progression ( Table 1 ).

Among men with mild baseline symptoms, LUTS progression was not associated with the antioxidants vitamin C, vitamin E, β-carotene, α-carotene, β-cryptoxanthin, lycopene, and lutein/zeaxanthin when comparing the highest quartile (Q4) to the lowest quartile (Q1) for intake ( Table 2 ).

Among men with moderate baseline symptoms, higher antioxidant intake was similarly not associated with a lower likelihood of LUTS progression ( Table 3 ). In analyses of LUTS remission relative to LUTS progression ( Table 4 ), higher lycopene intake was associated with a lower likelihood of LUTS remission after adjustment for energy intake, history of problem drinking, angina, and educational level (OR 0.37, 95% CI 0.16–0.82 for Q4 vs Q1; p = 0.04 for trend). However, this association was no longer statistically significant after adjustment for multiple comparisons.

This study used data collected in the Osteoporotic Fractures in Men Study (MrOS), a prospective study of community-dwelling men enrolled from six US sites [16] and [17]. The study was designed to evaluate risk factors for fracture, falls, and other conditions relevant to aging men, including prostate disease and LUTS [17] . From March 2000 to April 2002, 5994 men age 65–100 yr who could walk unassisted and had at least one natural hip for bone density measurement were enrolled. The six sites were Birmingham, AL; Minneapolis, MN; Palo Alto, CA; Pittsburgh, PA; Portland, OR; and San Diego, CA. The study protocol was approved by the institutional review boards at all participating institutions, and all men gave written informed consent.

Baseline measures collected at clinic visits included basic demographic data, lifestyle information (alcohol use, cigarette smoking), medical conditions, self-rated health, quality of life (short-form 12, SF-12 [18] ), and physical activity information based on the Physical Activity Scale for the Elderly (PASE) [19] . Height and weight were measured at the clinic visit and body mass index (BMI) was computed (kg/m²) [20] . Medications and supplements were brought to the baseline clinic visit and were inventoried by study staff and matched to ingredients according to on the Iowa Drug Information Service drug vocabulary (College of Pharmacy, University of Iowa, Iowa City, IA, USA) [21] .

Dietary data were collected at baseline using a brief Block food frequency questionnaire (FFQ) [22] and [23], which contains 69 items specifically drawn from foods most frequently consumed by elderly US men according to an analysis of data from the Third National Health and Nutrition Examination Survey (NHANES III) [22] . The Block FFQ is a validated, robust instrument used in observational research to measure nutrient intake, including antioxidant micronutrients. The brief Block FFQ asks about consumption frequency and portion size for each item. Gram estimates of foods were calculated as the gram weight for the chosen portion size multiplied by the consumption frequency. Nutrients were then estimated using the average amount of a nutrient in a food multiplied by the gram weight consumed. Nutrients from supplements were calculated separately from food estimates. Total intake was calculated as the sum of the nutrient estimates from food plus supplement nutrient estimates, when applicable. Supplement information was available only for the antioxidants vitamin C, vitamin E, and β-carotene.

At baseline and approximately every 2 yr (2002–2004, 2005–2006, and 2007–2009), follow-up data were collected on lower urinary tract health, including the American Urological Association Symptom Index (AUA-SI) and history of LUTS treatment. Additional follow-up occurred every 4 mo via mailed questionnaires to collect reports of deaths and incident prostate cancer cases, which were adjudicated by study physicians using death certificates and pathology reports.

The following baseline variables were classified into categories for the analysis. BMI was classified as normal (<25.0 kg/m²), overweight (25.0–29.9 kg/m²), or obese (≥30.0 kg/m²) [20] . Depressed mood was defined as an SF-12 mental component score ≤45 points [24] . Alcohol consumption was classified as never, ≤14 drinks/wk, or >14 drinks/wk; problem drinking was defined as a CAGE (Cutting down, Annoyance by criticism, Guilty feeling, and Eye-openers) [25] score of >1. Central nervous system (CNS) medication was defined as use of antiepileptics, benzodiazepines, antidepressants, opioids, or sedatives at baseline. Participants missing medication information were coded as nonusers after analyses showed no difference between nonusers and those with missing data.

We first restricted the MrOS cohort to 3594 men who had no history of prostate cancer and no prior or current treatment for LUTS, including surgery or medication (α-blockers, antispasmodics, anticholinergics, 5α-reductase inhibitors). These men were then followed through the fourth AUA-SI assessment (2007–2009). During follow-up, the cohort for trajectory analysis was further restricted to 1740 men who remained free from diagnosed prostate cancer, reported no surgery or medication use for LUTS, and completed all four AUA-SI assessments. A figure demonstrating how the analytic cohort was ascertained for the trajectory analysis has been previously published [26] . The same study demonstrated no significant differences in LUTS trajectory among men with a history of stroke and those who used CNS medications [26] , so we did not exclude them from the present analysis.

Outcomes for this analysis were LUTS trajectories. Details of the trajectory method were reported previously [26] . In brief, group-based trajectory modeling was performed using the AUA-SI data from all four time points. Trajectory modeling uses the maximum likelihood method and applies a semiparametric mixed model to longitudinal data (PROC TRAJ for SAS 9.1) [27], [28], and [29]. The four trajectory types identified were stable (n = 1277), progressing (n = 345), and remitting (n = 98) LUTS groups, and one very small group (n = 20) in which the LUTS trajectory increased and then decreased during follow-up. We identified three LUTS progression trajectories during follow-up: men who progressed from mild to moderate LUTS; men who progressed from moderate to severe LUTS; and men who progressed from low-moderate to high-moderate LUTS [26] .

2.1. Selection of the analytic cohort

For the present study, our goal was to determine associations between antioxidant intake and the likelihood of LUTS progression relative to LUTS stability in groups stratified by baseline clinical cut points of mild (AUA-SI 0–7) and moderate (AUA-SI 8–19) LUTS. We excluded men with severe symptoms at baseline (AUA-SI >19) and those who underwent LUTS treatment during follow-up to minimize the potential for confounding and focus on elucidating the natural history of LUTS in elderly men. In addition, we examined the likelihood of LUTS remission relative to LUTS progression in those with moderate baseline symptoms. Therefore, we restricted our analysis to men who had a similar baseline AUA-SI and who were not missing any dietary data (n = 1670; Fig. 1 ).

Fig. 1 Selection of the analytic cohort. MrOS = Osteoporotic Fractures in Men Study; LUTS = lower urinary tract symptoms; BPH = benign prostatic hyperplasia.

Overall, we observed no significant associations between antioxidant intake and LUTS progression or remission over approximately 7 yr of follow-up. To the best of our knowledge, this is the first study to examine associations between dietary antioxidants and LUTS progression. It is also the first study to describe associations between dietary factors and spontaneous remission of LUTS by analyzing whether dietary factors are associated with LUTS remission in otherwise untreated men.

Our results differ from those for prior cross-sectional studies and fail to confirm beneficial associations between higher antioxidant intake and lower risk of LUTS and BPH previously observed in some populations. Rohrman and colleagues [8] reported that dietary intakes of vitamin C, lutein/zeaxanthin, and β-cryptoxanthin in the Health Professionals Follow-up Study were inversely associated with total BPH (the combined endpoint of either surgery for BPH or high/moderate to severe LUTS). Similarly, Maserejian and colleagues [10] reported 40–50% lower odds of LUTS for high dietary intake of vitamin C and β-cryptoxanthin among men with high iron intake. A study examining serum concentrations of micronutrients from NHANES III data also reported lower serum concentrations of vitamin E in men with compared to men without LUTS [9] . It is possible that the null association between antioxidant intake and LUTS in the older men in this study could be due to a lack of effect of antioxidants in older individuals. Future research may demonstrate different effects in younger populations. It is also possible that prior studies were limited by a cross-sectional design or that factors such as recall bias or measurement error confounded the results. In fact, the longitudinal nature of our analysis was a prominent strength of our study that could have provided a more robust characterization of LUTS over time compared to other cohorts.

Unexpectedly, for those with moderate LUTS at baseline, there was a lower likelihood of LUTS remission among men with the highest lycopene intakes, although this observation did not reach statistical significance after correction for multiple comparisons. Further research is needed to determine whether similar results are observed in longitudinal analyses in other large cohorts. Lycopene was associated with lower LUTS odds in two cross-sectional studies [9] and [10] and one BPH study [11] ; therefore, if this association is found to be valid, it would be contrary to previous research on lycopene.

Our study has multiple strengths. We used a robust definition of LUTS progression obtained via a group-based trajectory method, which reduced the inherent variability in symptom scores and allowed clear identification of symptom progression over time. Another strength is the exclusion of men treated for LUTS with medication or surgery from the analytic cohort to provide an unbiased natural history of the disease, free of potential treatment effects that would otherwise change the symptom trajectory and confound the findings. Finally, the study included a geographically diverse group of community-dwelling elderly men in the USA who had an extended follow-up of approximately 7 yr.

One main limitation of our study is use of the brief Block FFQ, which has fewer food groups than other FFQs and may have biased results toward the null hypothesis. For example, pumpkin is not included in the instrument, yet it is the highest dietary source of β-cryptoxanthin and the second highest source of β-carotene. Thus, if a man consumed pumpkin regularly, he may have been incorrectly categorized as a lower rather than a higher consumer of these antioxidants. However, any misclassification should have been random with regard to LUTS progression, since men did not know their future LUTS status when filling out the baseline FFQ, and thus would have biased results toward the null hypothesis. A second limitation is the relatively small number of non-Caucasian men in the cohort, which probably diminished the external validity of the results with respect to other ethnic and racial groups. A third limitation is that dietary information was only collected at baseline; it is possible that some study participants may have subsequently altered their diets during follow-up, leading to a potential for bias. Finally, we were unable to assess objective urologic measures such as urinary flow, prostate volume, prostate-specific antigen (PSA), and voiding diaries.

Lower urinary tract symptoms (LUTS) are common among elderly men and have substantial global adverse effects on male health [1] and [2]. LUTS have been associated with higher mortality and morbidity [3] and billions of US dollars in annual health care expenditure [4] . Since obesity and exercise have been associated with higher and lower risks of LUTS, respectively [5] and [6], lifestyle changes might potentially prevent LUTS progression. Dietary constituents, especially antioxidants, are candidate lifestyle targets for LUTS prevention. Since antioxidants have potent anti-inflammatory properties, an increase in dietary antioxidant consumption might decrease LUTS via modulation of inflammatory pathways involved in the pathogenesis of LUTS and benign prostatic hyperplasia (BPH) [7] .

At least three studies have reported inverse associations between LUTS or BPH and consumption of antioxidants or foods rich in antioxidants, including β-carotene [8], [9], and [10], lutein [8] , lycopene [9] and [10], total carotenoids [10] , vitamin C [8] , vitamin E and selenium [9] , vegetables [11], [12], [13], and [14], and fruits [13] and [15]. Higher consumption of fruits and vegetables (including those rich in β-carotene and lycopene) has also been associated with a lower risk of BPH incidence [8] and [11].

Since most of the prior studies were cross-sectional, temporal associations between dietary antioxidants and LUTS remain unclear. Moreover, to the nest of our knowledge, no studies have examined potential associations between dietary antioxidants and the risk of LUTS progression in elderly men. Given the high prevalence of LUTS in this population [2] , the development of relatively straightforward dietary interventions to prevent LUTS progression in elderly men may substantially inform the clinical care of men with LUTS. Therefore, we examined the association between baseline dietary antioxidant intake and subsequent LUTS progression over a 7-yr period in elderly men. We hypothesized that higher baseline consumption of dietary antioxidants is associated with a lower probability of LUTS progression.

We previously reported on descriptive AUA-SI data for this analytic cohort [26] . Among men with mild baseline symptoms, men with LUTS progression had slightly lower physical activity and were more likely to report mobility limitations and a history of dizziness and non–prostate cancer when compared to men with stable LUTS. (As mentioned earlier, those with a history of prostate cancer were excluded from analyses.) In those with moderate baseline symptoms, men with LUTS progression were marginally younger, had a significantly lower SF-12 mental score, were less likely to live alone, had lower overall energy intake, and were more likely to suffer from depression and back pain when compared to men with stable LUTS. By contrast, men with LUTS remission had a slightly higher SF-12 mental score (closer to those with stable symptoms); had less education; were less likely to have a history of problem drinking, hypertension, or angina; were less likely to use diuretics; and consumed less caffeine when compared to men with LUTS progression ( Table 1 ).

Among men with mild baseline symptoms, LUTS progression was not associated with the antioxidants vitamin C, vitamin E, β-carotene, α-carotene, β-cryptoxanthin, lycopene, and lutein/zeaxanthin when comparing the highest quartile (Q4) to the lowest quartile (Q1) for intake ( Table 2 ).

Among men with moderate baseline symptoms, higher antioxidant intake was similarly not associated with a lower likelihood of LUTS progression ( Table 3 ). In analyses of LUTS remission relative to LUTS progression ( Table 4 ), higher lycopene intake was associated with a lower likelihood of LUTS remission after adjustment for energy intake, history of problem drinking, angina, and educational level (OR 0.37, 95% CI 0.16–0.82 for Q4 vs Q1; p = 0.04 for trend). However, this association was no longer statistically significant after adjustment for multiple comparisons.

This study used data collected in the Osteoporotic Fractures in Men Study (MrOS), a prospective study of community-dwelling men enrolled from six US sites [16] and [17]. The study was designed to evaluate risk factors for fracture, falls, and other conditions relevant to aging men, including prostate disease and LUTS [17] . From March 2000 to April 2002, 5994 men age 65–100 yr who could walk unassisted and had at least one natural hip for bone density measurement were enrolled. The six sites were Birmingham, AL; Minneapolis, MN; Palo Alto, CA; Pittsburgh, PA; Portland, OR; and San Diego, CA. The study protocol was approved by the institutional review boards at all participating institutions, and all men gave written informed consent.

Baseline measures collected at clinic visits included basic demographic data, lifestyle information (alcohol use, cigarette smoking), medical conditions, self-rated health, quality of life (short-form 12, SF-12 [18] ), and physical activity information based on the Physical Activity Scale for the Elderly (PASE) [19] . Height and weight were measured at the clinic visit and body mass index (BMI) was computed (kg/m²) [20] . Medications and supplements were brought to the baseline clinic visit and were inventoried by study staff and matched to ingredients according to on the Iowa Drug Information Service drug vocabulary (College of Pharmacy, University of Iowa, Iowa City, IA, USA) [21] .

Dietary data were collected at baseline using a brief Block food frequency questionnaire (FFQ) [22] and [23], which contains 69 items specifically drawn from foods most frequently consumed by elderly US men according to an analysis of data from the Third National Health and Nutrition Examination Survey (NHANES III) [22] . The Block FFQ is a validated, robust instrument used in observational research to measure nutrient intake, including antioxidant micronutrients. The brief Block FFQ asks about consumption frequency and portion size for each item. Gram estimates of foods were calculated as the gram weight for the chosen portion size multiplied by the consumption frequency. Nutrients were then estimated using the average amount of a nutrient in a food multiplied by the gram weight consumed. Nutrients from supplements were calculated separately from food estimates. Total intake was calculated as the sum of the nutrient estimates from food plus supplement nutrient estimates, when applicable. Supplement information was available only for the antioxidants vitamin C, vitamin E, and β-carotene.

At baseline and approximately every 2 yr (2002–2004, 2005–2006, and 2007–2009), follow-up data were collected on lower urinary tract health, including the American Urological Association Symptom Index (AUA-SI) and history of LUTS treatment. Additional follow-up occurred every 4 mo via mailed questionnaires to collect reports of deaths and incident prostate cancer cases, which were adjudicated by study physicians using death certificates and pathology reports.

The following baseline variables were classified into categories for the analysis. BMI was classified as normal (<25.0 kg/m²), overweight (25.0–29.9 kg/m²), or obese (≥30.0 kg/m²) [20] . Depressed mood was defined as an SF-12 mental component score ≤45 points [24] . Alcohol consumption was classified as never, ≤14 drinks/wk, or >14 drinks/wk; problem drinking was defined as a CAGE (Cutting down, Annoyance by criticism, Guilty feeling, and Eye-openers) [25] score of >1. Central nervous system (CNS) medication was defined as use of antiepileptics, benzodiazepines, antidepressants, opioids, or sedatives at baseline. Participants missing medication information were coded as nonusers after analyses showed no difference between nonusers and those with missing data.

We first restricted the MrOS cohort to 3594 men who had no history of prostate cancer and no prior or current treatment for LUTS, including surgery or medication (α-blockers, antispasmodics, anticholinergics, 5α-reductase inhibitors). These men were then followed through the fourth AUA-SI assessment (2007–2009). During follow-up, the cohort for trajectory analysis was further restricted to 1740 men who remained free from diagnosed prostate cancer, reported no surgery or medication use for LUTS, and completed all four AUA-SI assessments. A figure demonstrating how the analytic cohort was ascertained for the trajectory analysis has been previously published [26] . The same study demonstrated no significant differences in LUTS trajectory among men with a history of stroke and those who used CNS medications [26] , so we did not exclude them from the present analysis.

Outcomes for this analysis were LUTS trajectories. Details of the trajectory method were reported previously [26] . In brief, group-based trajectory modeling was performed using the AUA-SI data from all four time points. Trajectory modeling uses the maximum likelihood method and applies a semiparametric mixed model to longitudinal data (PROC TRAJ for SAS 9.1) [27], [28], and [29]. The four trajectory types identified were stable (n = 1277), progressing (n = 345), and remitting (n = 98) LUTS groups, and one very small group (n = 20) in which the LUTS trajectory increased and then decreased during follow-up. We identified three LUTS progression trajectories during follow-up: men who progressed from mild to moderate LUTS; men who progressed from moderate to severe LUTS; and men who progressed from low-moderate to high-moderate LUTS [26] .

2.1. Selection of the analytic cohort

For the present study, our goal was to determine associations between antioxidant intake and the likelihood of LUTS progression relative to LUTS stability in groups stratified by baseline clinical cut points of mild (AUA-SI 0–7) and moderate (AUA-SI 8–19) LUTS. We excluded men with severe symptoms at baseline (AUA-SI >19) and those who underwent LUTS treatment during follow-up to minimize the potential for confounding and focus on elucidating the natural history of LUTS in elderly men. In addition, we examined the likelihood of LUTS remission relative to LUTS progression in those with moderate baseline symptoms. Therefore, we restricted our analysis to men who had a similar baseline AUA-SI and who were not missing any dietary data (n = 1670; Fig. 1 ).

Fig. 1 Selection of the analytic cohort. MrOS = Osteoporotic Fractures in Men Study; LUTS = lower urinary tract symptoms; BPH = benign prostatic hyperplasia.

Overall, we observed no significant associations between antioxidant intake and LUTS progression or remission over approximately 7 yr of follow-up. To the best of our knowledge, this is the first study to examine associations between dietary antioxidants and LUTS progression. It is also the first study to describe associations between dietary factors and spontaneous remission of LUTS by analyzing whether dietary factors are associated with LUTS remission in otherwise untreated men.

Our results differ from those for prior cross-sectional studies and fail to confirm beneficial associations between higher antioxidant intake and lower risk of LUTS and BPH previously observed in some populations. Rohrman and colleagues [8] reported that dietary intakes of vitamin C, lutein/zeaxanthin, and β-cryptoxanthin in the Health Professionals Follow-up Study were inversely associated with total BPH (the combined endpoint of either surgery for BPH or high/moderate to severe LUTS). Similarly, Maserejian and colleagues [10] reported 40–50% lower odds of LUTS for high dietary intake of vitamin C and β-cryptoxanthin among men with high iron intake. A study examining serum concentrations of micronutrients from NHANES III data also reported lower serum concentrations of vitamin E in men with compared to men without LUTS [9] . It is possible that the null association between antioxidant intake and LUTS in the older men in this study could be due to a lack of effect of antioxidants in older individuals. Future research may demonstrate different effects in younger populations. It is also possible that prior studies were limited by a cross-sectional design or that factors such as recall bias or measurement error confounded the results. In fact, the longitudinal nature of our analysis was a prominent strength of our study that could have provided a more robust characterization of LUTS over time compared to other cohorts.

Unexpectedly, for those with moderate LUTS at baseline, there was a lower likelihood of LUTS remission among men with the highest lycopene intakes, although this observation did not reach statistical significance after correction for multiple comparisons. Further research is needed to determine whether similar results are observed in longitudinal analyses in other large cohorts. Lycopene was associated with lower LUTS odds in two cross-sectional studies [9] and [10] and one BPH study [11] ; therefore, if this association is found to be valid, it would be contrary to previous research on lycopene.

Our study has multiple strengths. We used a robust definition of LUTS progression obtained via a group-based trajectory method, which reduced the inherent variability in symptom scores and allowed clear identification of symptom progression over time. Another strength is the exclusion of men treated for LUTS with medication or surgery from the analytic cohort to provide an unbiased natural history of the disease, free of potential treatment effects that would otherwise change the symptom trajectory and confound the findings. Finally, the study included a geographically diverse group of community-dwelling elderly men in the USA who had an extended follow-up of approximately 7 yr.

One main limitation of our study is use of the brief Block FFQ, which has fewer food groups than other FFQs and may have biased results toward the null hypothesis. For example, pumpkin is not included in the instrument, yet it is the highest dietary source of β-cryptoxanthin and the second highest source of β-carotene. Thus, if a man consumed pumpkin regularly, he may have been incorrectly categorized as a lower rather than a higher consumer of these antioxidants. However, any misclassification should have been random with regard to LUTS progression, since men did not know their future LUTS status when filling out the baseline FFQ, and thus would have biased results toward the null hypothesis. A second limitation is the relatively small number of non-Caucasian men in the cohort, which probably diminished the external validity of the results with respect to other ethnic and racial groups. A third limitation is that dietary information was only collected at baseline; it is possible that some study participants may have subsequently altered their diets during follow-up, leading to a potential for bias. Finally, we were unable to assess objective urologic measures such as urinary flow, prostate volume, prostate-specific antigen (PSA), and voiding diaries.