GINGER & ROSA

American Journal of Epidemiology
Copyright ª 2006 by the Johns Hopkins Bloomberg School of Public Health
All rights reserved; printed in U.S.A.
Vol. 164, No. 10
DOI: 10.1093/aje/kwj311
Advance Access publication October 5, 2006
Original Contribution
Prostate-specific Antigen Values in Diabetic and Nondiabetic US Men, 2001–2002
David M. Werny1, Mona Saraiya1, and Edward W. Gregg2
1
Division of Cancer Prevention and Control, National Center for Chronic Disease Prevention and Health Promotion, Centers
for Disease Control and Prevention, Atlanta, GA.
2
Division of Diabetes Translation, National Center for Chronic Disease Prevention and Health Promotion, Centers for
Disease Control and Prevention, Atlanta, GA.
Recent studies have shown that diabetic men have a lower risk of prostate cancer and that this association may
be related to time since diagnosis. The authors examined the association between diabetes and prostate-specific
antigen (PSA) levels, controlling for potential confounders, in a nationally representative cross-sectional survey of
the US population (National Health and Nutrition Examination Survey 2001–2002). Diabetes classification was
self-reported, and undiagnosed diabetes was determined with fasting plasma glucose measurements. Controlling
for age, men with self-reported diabetes had a 21.6% lower geometric mean PSA level than men without diabetes.
The difference increased with years since diagnosis (>10 years: 27.5% lower geometric mean PSA level). Overweight men who had had diabetes for more than 10 years had a predicted geometric mean PSA level 40.8% lower
than that of nondiabetic, normal-weight men. These results are consistent with the hypothesis that long-term
diabetes is associated with a lower risk of prostate cancer. The mechanism of this association may involve the
regulation of PSA by androgens, although the authors are unable to confirm this assertion. Better understanding of
the determinants of PSA level is needed to make the distinction between factors affecting the PSA test’s accuracy
and those altering the risk of prostate cancer.
diabetes mellitus, type 2; prostate-specific antigen; prostatic neoplasms; testosterone
Abbreviations: BMI, body mass index; IGF-1, insulin-like growth factor 1; NHANES, National Health and Nutrition Examination
Survey; PSA, prostate-specific antigen.
Recent studies have suggested an association between
type 2 diabetes mellitus and lower risk of prostate cancer
(1, 2). It has been hypothesized that men with long-term
diabetes have a lower risk of prostate cancer than nondiabetic men, and recently diagnosed men have a higher risk (3,
4). In biologic models proposed to explain this association,
researchers note the higher concentrations of insulin and
insulin-like growth factor 1 (IGF-1) in early diabetes and
the lower testosterone and IGF-1 levels and higher estrogen
concentrations in long-term diabetes (5, 6). Whether diabetes influences levels of biomarkers such as prostate-specific
antigen (PSA), which is involved in the detection of prostate
cancer, is unknown.
Factors influencing serum PSA levels in men include age,
benign prostatic hyperplasia, prostatitis, and body mass index (BMI) (7, 8). Still, we understand little about PSA, and
its relations with comorbid conditions remain unexplored
(9). Diabetes and PSA screening are prevalent among men
aged 50 years or older, and no doubt many men this age with
diabetes undergo PSA testing (10, 11). In this analysis, we
examined whether PSA concentrations varied by diabetes
status and duration of diabetes.
Correspondence to Dr. Mona Saraiya, National Center for Chronic Disease Prevention and Health Promotion, Mail Stop K-55, 4770 Buford
Highway NE, Atlanta, GA 30341 (e-mail: [email protected]).
978
Am J Epidemiol 2006;164:978–983
Downloaded from http://aje.oxfordjournals.org/ by guest on June 15, 2014
Received for publication February 8, 2006; accepted for publication April 28, 2006.
PSA Values in Diabetic and Nondiabetic US Men
MATERIALS AND METHODS
Survey
Measures
We categorized participants by measured BMI (weight
(kg)/height (m)2) as normal-weight (BMI <25), overweight
(BMI 25–<30), or obese (BMI 30) (15). Persons with
a measured systolic blood pressure of 140 mmHg or a diastolic blood pressure of 90 mmHg were considered hypertensive (16). We defined measured high density lipoprotein
cholesterol with a gender-specific 0.9-mmol/liter threshold
(17). Serum insulin concentrations were measured for the
men who had fasted. The QUICKI index of insulin sensitivity was calculated for men in the fasting subsample (1/[log
insulin (lU/ml) þ log glucose (mg/dl)]) (18). Ever having
been diagnosed with benign prostatic hyperplasia was selfreported. We categorized age by decade (40–49, 50–59, 60–
69, 70–79, and 80–85 years) in the univariate analysis but
controlled for age continuously in our multivariate models.
We categorized race/ethnicity as Mexican-American, nonHispanic White, non-Hispanic Black, or other. For all variAm J Epidemiol 2006;164:978–983
ables, responses such as ‘‘refused’’ or ‘‘don’t know’’ were
considered missing data.
Statistical analyses
PSA values were natural-log-transformed to improve normality. Log PSA was used as the dependent variable for the
bivariate and linear regression analyses. In all analyses, we
used SUDAAN, version 9.0 (Research Triangle Institute,
Research Triangle Park, North Carolina), to account for the
complex sampling design. All p values (from t tests) were
two-sided. We analyzed log-transformed PSA values, and
accordingly we present geometric means in table 1. We controlled for age when presenting results because of the known
association between PSA and age. Variables for the categories of years since diagnosis of diabetes were coded as separate terms in the multivariate analyses. In constructing the
multiple linear regression model (table 2), we first assessed
variables using a series of models containing only the variable in question, the diabetes terms, including time since
diagnosis, and an interaction term for the interaction between the two. Variables that were themselves significant
predictors in the presence of the diabetes terms were included in the multiple linear regression model, as were any
confounders that altered the predicted effect of the diabetes
terms by more than 10 percent (19). Interactions between
the diabetes terms and BMI, high density lipoprotein cholesterol, race/ethnicity, blood pressure, age, and benign
prostatic hyperplasia were nonsignificant. The race/ethnicity terms were included in the model because of previous
reports that showed variation in PSA levels by race/ethnicity (20).
RESULTS
Geometric mean PSA levels increased with age (40–49
years: 0.75 ng/ml; 50–59 years: 0.93 ng/ml; 60–69 years:
1.11 ng/ml; 70–79 years: 1.76 ng/ml; 80–85 years: 2.15 ng/
ml). Results are shown in table 1 after controlling for age
(median age, 51 years). Both the overweight and obese
BMI groups showed lower predicted geometric mean PSA
levels than the comparison group (BMI <25) (table 1). No
significant differences in predicted geometric mean PSA levels were seen by race/ethnicity, blood pressure, high density
lipoprotein cholesterol, or past diagnosis of benign prostatic
hyperplasia. The predicted geometric mean PSA value was
lower in the diabetic group than in the nondiabetic group (p <
0.001) and was lowest in men who had been diagnosed with
diabetes more than 10 years previously. The trend for decreased PSA by diabetes status (beginning with nondiabetic
men and ending with men diagnosed more than 10 years
previously) was statistically significant (p ¼ 0.001). Diabetic
men taking medication to control blood glucose levels had
lower PSA levels than nondiabetic men, but in a comparison
with diabetic men not taking medication, the p value was not
statistically significant (p ¼ 0.073).
We examined the association between PSA and diabetes
covariates in a fasting subsample of our study population,
adjusting for age as in table 1. There was no association
Downloaded from http://aje.oxfordjournals.org/ by guest on June 15, 2014
The 2001–2002 National Health and Nutrition Examination Survey (NHANES) was weighted to provide a nationally
representative sample of the noninstitutionalized, civilian
US population (12). In 2001–2002, participation rates for
the NHANES interview and medical examination were 83.9
percent and 79.7 percent, respectively. An institutional review board approved the survey. Participants provided informed consent for PSA testing.
PSA assays were conducted on serum samples taken from
all men aged 40 years who did not meet any of the PSA
testing exclusion criteria, including current prostatitis or diagnosis of prostate cancer (n ¼ 1,308) (13). PSA was measured using the Hybritech assay (Hybritech, Inc., San Diego,
California) in the Beckman immunoassay system (Beckman
Instruments, Inc., Fullerton, California).
Sample adults were asked, ‘‘Have you ever been told by
a doctor or health professional that you have diabetes or sugar
diabetes?’’ Men answering affirmatively (n ¼ 173) were classified as having diabetes. We calculated time since diagnosis
(in years) by subtracting age at diagnosis from current age,
producing categories of 0–5, 6–10, and >10 years (3). Men
who reported being diabetic were also asked whether they
were currently taking ‘‘diabetic pills to lower blood sugar.’’
A subset of our 1,308 participants were instructed to fast
before the medical examination (n ¼ 605). In NHANES,
separate weights are assigned to ensure that this fasting sample is nationally representative. Among participants not selfclassified as diabetic, we designated a fasting plasma glucose
level of 126 mg/dl as undiagnosed diabetes, and we created
separate groups for persons with fasting plasma glucose levels of <100 mg/dl, 100–110 mg/dl, and 111–125 mg/dl. Selfclassified diabetic men were categorized by years since
diagnosis, as noted above. Fasting has not been shown to
affect PSA levels (14), and there was no statistically significant difference between PSA values in the fasting and nonfasting groups (p ¼ 0.415).
979
980 Werny et al.
TABLE 1. Predicted geometric mean prostate-specific antigen values for men aged 51 years, by
demographic and clinical characteristics, National Health and Nutrition Examination Survey 2001–2002*
Predicted
geometric mean
(ng/ml)
% change in
predicted
geometric mean
1.02
0 (reference)
95% confidence interval
for change in predicted
geometric mean
p valuey
Body mass indexz (total n ¼ 1,239)§
<25 (n ¼ 296)
25–<30 (n ¼ 559)
0.84
17.9
28.0, 6.2
0.006
30 (n ¼ 375)
0.85
16.4
28.4, 2.3
0.027
0.87
0 (reference)
Race/ethnicity (total n ¼ 1,308)
Non-Hispanic White (n ¼ 767)
Non-Hispanic Black (n ¼ 225)
0.95
þ9.5
6.8, 28.6
0.250
Mexican-American (n ¼ 235)
0.95
þ9.1
7.6, 28.8
0.283
Other (n ¼ 81)
0.83
4.2
25.9, 23.9
0.727
21.9, 4.0
0.141
16.1, 10.3
0.650
High blood pressure (total n ¼ 1,300){
0.81
10.0
0.89
0 (reference)
High density lipoprotein cholesterol
level (total n ¼ 1,308)
0.9 mmol/liter (n ¼ 1,103)
0.87
2.7
<0.9 mmol/liter (n ¼ 205)
0.90
0 (reference)
0.70
21.6
30.5, 11.5
0.65
27.5
41.7, 9.9
Diabetes status (total n ¼ 1,308)
Diabetic (n ¼ 173)#
<0.001
Duration of diabetes
>10 years (n ¼ 63)
0.007
6–10 years (n ¼ 31)
0.67
25.5
40.8, 6.2
0.016
0–5 years (n ¼ 77)
0.74
17.1
32.7, 2.0
0.073
0.63
29.4
41.7, 14.4
0.002
22.6, 36.1
0.848
Use of diabetes medication
Currently taking medication
(n ¼ 123)
Not taking medication (n ¼ 49)
Nondiabetic (n ¼ 1,135)
0.92
0.90
2.7
0 (reference)
* Results are from a multiple linear regression model weighted analysis with the complex sample design taken
into account.
y Two-sided t test for the difference in mean log-transformed prostate-specific antigen value compared with the
reference group.
z Weight (kg)/height (m)2.
§ Data on height and/or weight were missing for 69 participants.
{ Persons with a measured systolic pressure 140 mmHg or a diastolic pressure 90 mmHg were considered
hypertensive. Eight participants had missing values for either measure.
# Two diabetic participants did not report their age at diagnosis; one diabetic participant did not report on
medication use.
between PSA and fasting insulin levels (p ¼ 0.182; data not
shown), as well as no change in predicted geometric mean
PSA by category of fasting plasma glucose among the nondiabetic men (p ¼ 0.800; data not shown). Additionally,
there was no association between the QUICKI index of insulin sensitivity and PSA (p ¼ 0.122; data not shown). The
predicted geometric mean PSA for self-reporting diabetic
men was lower than that of self-reporting nondiabetic men
(n ¼ 540) (0.62 ng/ml vs. 0.88 ng/ml; p ¼ 0.012). There was
no statistically significant difference between self-reporting
diabetic men (n ¼ 65) and men with undiagnosed diabetes
(n ¼ 45) (0.57 ng/ml vs. 0.77 ng/ml; p ¼ 0.146). When we
compared the group of diagnosed and undiagnosed diabetic
men (n ¼ 109) with those self-reporting nondiabetic men
who had a fasting plasma glucose level less than 100 mg/dl
(n ¼ 250), there was no statistically significant difference
(0.70 ng/ml vs. 0.86 ng/ml; p ¼ 0.120). In contrast to the
results in the larger sample, the lowest predicted geometric
mean PSA level in the fasting subsample was for persons
diagnosed with diabetes within the previous 5 years (n ¼ 28)
Am J Epidemiol 2006;164:978–983
Downloaded from http://aje.oxfordjournals.org/ by guest on June 15, 2014
Yes (n ¼ 359)
No (n ¼ 941)
PSA Values in Diabetic and Nondiabetic US Men
981
TABLE 2. Predictors of prostate-specific antigen (PSA) concentration obtained using
multivariate linear regression, National Health and Nutrition Examination Survey
2001–2002*
Variable
Age (per year)
% change in
predicted
geometric mean
þ2.7
95% confidence interval
for change in predicted
geometric mean
þ2.2, 3.2
p valuey
<0.001
Body mass index (BMI)z
<25
0 (reference)
25–<30
17.8
27.6, 6.7
0.005
30
15.2
26.9, 1.6
0.032
8.1, 28.4
0.310
Race/ethnicity
Non-Hispanic White
0 (reference)
Non-Hispanic Black
þ8.6
Mexican-American
þ10.9
7.0, 32.2
0.231
6.0
29.2, 25.5
0.666
Other
Diabetes status
Any duration
20.2
31.2, 7.5
0.005
Duration >10 years
29.3
43.6, 11.4
0.005
Duration 6–10 years
24.4
43.5, 1.1
0.058
Duration 0–5 years
14.0
32.0, 8.9
0.193
Not diabetic
0 (reference)
B
B
* Final model: log PSA ¼ B0 þ B1 3 age þ B2–3 3 BMI1–2 þ B4–6 3 race1–3 þ B7–9 3
diabetes1–3; R2 ¼ 0.13; n ¼ 1,237. Results were obtained from a weighted analysis with the
complex sample design taken into account.
y Two-sided t test for the probability that the predicted percent change in the geometric mean
differs from 0.
z Weight (kg)/height (m)2.
(0.56 ng/ml; in comparison with nondiabetic men with a
fasting plasma glucose level less than 100 mg/dl, p ¼
0.019).
In the multiple linear regression model, neither BMI nor
race/ethnicity confounded the association between the diabetes terms and PSA (table 2). We used the model to examine the joint associations of BMI and diabetes with PSA.
The predicted geometric mean PSA for all 51-year-old nondiabetic men with a BMI less than 25 was 1.03 ng/ml—
significantly higher than the value for their same-age peers
with diabetes of more than 10 years’ duration who had a
BMI greater than or equal to 25 (0.61 ng/ml, a 40.8 percent
reduction; p ¼ 0.001).
DISCUSSION
Diabetes and an overweight BMI are independently associated with lower geometric mean PSA levels. Together they
are associated with as much as a 40.8 percent lower predicted geometric mean PSA level relative to nondiabetic,
normal-weight men. Diabetic men have lower androgen levels than nondiabetic men, and this may partially explain
their lower PSA levels (6). Our finding of a significant downward trend of PSA (from no diabetes to diabetes of more
Am J Epidemiol 2006;164:978–983
than 10 years’ duration) is consistent with a possible association between progressing diabetes and decreasing testosterone concentrations, which may lower the risk of prostate
cancer (21). Despite the existence of functional androgenresponsive elements upstream of the PSA gene promoter,
previous investigations have not shown an association between serum testosterone and PSA levels (22). However,
PSA may be associated with testosterone in men with subnormal levels of the sex hormone (23), as is the case with
diabetic men. Diabetes might also alter PSA values through
impaired kidney function (24) or as a consequence of diabetes medication use. Metformin, a hypoglycemic agent,
has been shown to decrease testosterone levels in nondiabetic men but not in diabetic men (25, 26). In our study,
diabetic men taking medication appeared to have lower PSA
levels than diabetic men not taking medication, but our
small sample size limited our statistical power to detect a
difference. This association can be examined more appropriately with additional data.
In diabetes, IGF-1 levels increase during hyperinsulinemia
and decrease as insulin production drops (27). Previous studies have found a peak in prostate cancer risk during early
diabetes, and a recent study (28) found a modest positive
association between serum PSA and serum IGF-1 that was
proposed to be due to IGF-1-induced benign prostatic
Downloaded from http://aje.oxfordjournals.org/ by guest on June 15, 2014
Diabetic
982 Werny et al.
The analysis was repeated with these men categorized as
diabetic; no substantial changes were noted.
Some investigators have recommended age- and racespecific thresholds for prostate cancer detection based on
population-level differences (20). While this study was unable to demonstrate that modified PSA thresholds for diabetic men would result in greater accuracy, future studies
should investigate whether diabetes status, duration of diabetes, and obesity should be considered when interpreting
a PSA test result. Better understanding of the determinants
of PSA levels is needed to make the distinction between
factors affecting the test’s accuracy and those altering the
risk of prostate cancer.
ACKNOWLEDGMENTS
This research was performed while David Werny was
a fellow in the Centers for Disease Control and Prevention
Research Participation Program, administered by the Oak
Ridge Institute for Science and Education under contract
DE-AC05-00OR22750 between the US Department of
Energy and Oak Ridge Associated Universities.
Mr. Werny thanks Dr. Kevin Sullivan for his guidance.
Conflict of interest: none declared.
REFERENCES
1. Bonovas S, Filioussi K, Tsantes A. Diabetes mellitus and risk
of prostate cancer: a meta-analysis. Diabetologia 2004;47:
1071–8.
2. Zhu K, Lee IM, Sesso HD, et al. History of diabetes mellitus
and risk of prostate cancer in physicians. Am J Epidemiol
2004;159:978–82.
3. Giovannucci E, Rimm EB, Stampfer MJ, et al. Diabetes mellitus and risk of prostate cancer (United States). Cancer Causes
Control 1998;9:3–9.
4. Rodriguez C, Patel AV, Mondul AM, et al. Diabetes and risk
of prostate cancer in a prospective cohort of US men. Am J
Epidemiol 2005;161:147–52.
5. Djavan B, Waldert M, Seitz C, et al. Insulin-like growth factors
and prostate cancer. World J Urol 2001;19:225–33.
6. Betancourt-Albrecht M, Cunningham GR. Hypogonadism and
diabetes. Int J Impot Res 2003;15(suppl 4):S14–20.
7. Baillargeon J, Pollock BH, Kristal AR, et al. The association of
body mass index and prostate-specific antigen in a populationbased study. Cancer 2005;103:1092–5.
8. Game X, Vincendeau S, Palascak R, et al. Total and free serum
prostate specific antigen levels during the first month of acute
prostatitis. Eur Urol 2003;43:702–5.
9. Thompson I, Leach RJ, Pollock BH, et al. Prostate cancer and
prostate-specific antigen: the more we know, the less we understand. J Natl Cancer Inst 2003;95:1027–8.
10. Diabetes Public Health Resource, Division of Diabetes
Translation, Centers for Disease Control and Prevention. Data
and trends. National Diabetes Surveillance System. Prevalence
of diabetes: age-specific prevalence of diagnosed diabetes,
by race/ethnicity and sex, United States, 2003. Atlanta,
Am J Epidemiol 2006;164:978–983
Downloaded from http://aje.oxfordjournals.org/ by guest on June 15, 2014
hyperplasia in older men (3, 4). We did not observe a positive
association between insulin and PSA, even after restricting
the analysis to men aged 60 years or older. Insulin levels may
be a poor proxy for long-term IGF-1, however, and this misclassification may have weakened our chances of finding
an association. We also found no association between PSA
and the QUICKI measure of insulin resistance (p ¼ 0.121).
However, interpreting these results is made difficult by our
inability to control for serum testosterone, which is associated inversely with insulin, fasting glucose, and insulin resistance (29).
It is unclear whether the lowered PSA levels in diabetic
men accurately reflect a decreased risk of prostate cancer in
the diabetic population or whether their lower PSA levels
result in a reduced likelihood of receiving a diagnostic
workup for detection of asymptomatic prostate cancers, as
has been suggested for obese men (7). If the latter were true,
diabetic men might well be diagnosed with later-stage tumors
and have poorer treatment outcomes, and overweight diabetic men would have later-stage tumors than normal-weight
diabetic men. However, data from CaPSURE (Cancer of the
Prostate Strategic Urologic Research Endeavor), a registry of
men with prostate cancer, have not shown later stages at presentation or worse clinical outcomes for diabetic men (30).
The principal limitation of this analysis is the small number of men self-reporting with diabetes. These men had a
lower PSA level than men without diabetes in both our fasting
(n ¼ 605) and nonfasting (n ¼ 703) subsamples, although
only in the fasting subsample did the difference attain statistical significance (fasting: p ¼ 0.012; nonfasting: p ¼ 0.078).
Examining the trend in PSA by duration of diabetes in the two
subsamples revealed two different trends. We do not know
whether these discrepancies arise from less stable estimates
in the separated groups, whether fasting itself influences the
relation between PSA and diabetes, or whether there are other
differences between the groups that may explain these results. Future data from NHANES surveys will enable investigators to make more stable estimates of geometric mean
PSA levels among diabetic men.
An additional limitation of this study is that the number of
participants with undiagnosed cancer was unknown, and
thus we do not know whether differences in PSA were due
to undiagnosed prostate cancer in the diabetic and nondiabetic groups. Furthermore, without clinical data, we cannot
be sure that these differences in PSA result in missed biopsy
opportunities, nor can we recommend alternate biopsy
thresholds for diabetic men. A third limitation is that selfreported measures of diabetes status have been shown to be
highly specific but only 66 percent sensitive (31). We addressed this concern by analyzing the fasting subsample,
although there may have been additional diabetic men in
the NHANES who reported being nondiabetic. This misclassification would have biased our results towards the null.
Although we were unable to separate men with type 1 diabetes from men with type 2 diabetes in our study, excluding
the six persons who were diagnosed before age 30 years and
were currently taking insulin did not significantly alter our
results. In another potential source of misclassification, participants were given the option to report having ‘‘borderline’’
diabetes. We categorized these men as nondiabetic (n ¼ 29).
PSA Values in Diabetic and Nondiabetic US Men
11.
12.
13.
14.
15.
16.
18.
19.
20.
Am J Epidemiol 2006;164:978–983
21. Barrett-Connor E, Khaw KT, Yen SS. Endogenous sex hormone levels in older adult men with diabetes mellitus. Am J
Epidemiol 1990;132:895–901.
22. Monath JR, McCullough DL, Hart LJ, et al. Physiologic variations of serum testosterone within the normal range do not
affect serum prostate-specific antigen. Urology 1995;46:58–61.
23. Tenover JS. Effects of testosterone supplementation in the
aging male. J Clin Endocrinol Metab 1992;75:1092–8.
24. Bruun L, Ekberg H, Bjork T, et al. Rapid elimination by glomerular filtration of free prostate specific antigen and human
kallikrein 2 after renal transplantation. J Urol 2004;171:1432–5.
25. Shegem NS, Alsheek Nasir AM, Batieha AM, et al. Effects
of short term metformin administration on androgens in diabetic men. Saudi Med J 2004;25:75–8.
26. Shegem NS, Nasir AM, Jbour AK, et al. Effects of short
term metformin administration on androgens in normal men.
Saudi Med J 2002;23:934–7.
27. Giovannucci E. Nutrition, insulin, insulin-like growth factors
and cancer. Horm Metab Res 2003;35:694–704.
28. Oliver SE, Barrass B, Gunnell DJ, et al. Serum insulin-like
growth factor-I is positively associated with serum prostatespecific antigen in middle-aged men without evidence of prostate cancer. Cancer Epidemiol Biomarkers Prev 2004;13:163–5.
29. Kapoor D, Malkin CJ, Channer KS, et al. Androgens, insulin
resistance and vascular disease in men. Clin Endocrinol (Oxf)
2005;63:239–50.
30. Chan JM, Latini DM, Cowan J, et al. History of diabetes,
clinical features of prostate cancer, and prostate cancer recurrence-data from CaPSURE (United States). Cancer Causes
Control 2005;16:789–97.
31. Okura Y, Urban LH, Mahoney DW, et al. Agreement between
self-report questionnaires and medical record data was substantial for diabetes, hypertension, myocardial infarction and
stroke but not for heart failure. J Clin Epidemiol 2004;57:
1096–103.
Downloaded from http://aje.oxfordjournals.org/ by guest on June 15, 2014
17.
GA: Centers for Disease Control and Prevention, 2003. (http://
www.cdc.gov/diabetes/statistics/prev/national/fig2003.htm).
Ross LE, Coates RJ, Breen N, et al. Prostate-specific antigen
test use reported in the 2000 National Health Interview Survey.
Prev Med 2004;38:732–44.
National Center for Health Statistics. NHANES 1999–2000
addendum to the NHANES III analytic guidelines. Hyattsville,
MD: National Center for Health Statistics, 2002. (http://www.
cdc.gov/nchs/data/nhanes/guidelines1.pdf).
Saraiya M, Kottiri BJ, Leadbetter S, et al. Total and percent
free prostate-specific antigen levels among U.S. men, 2001–
2002. Cancer Epidemiol Biomarkers Prev 2005;14:2178–82.
Tuncel A, Aksut H, Agras K, et al. Is serum prostate-specific
antigen level affected by fasting and nonfasting? Urology
2005;66:105–7.
Physical status: the use and interpretation of anthropometry.
Report of a WHO Expert Committee. World Health Organ
Tech Rep Ser 1995;854:1–452.
Chobanian AV, Bakris GL, Black HR, et al. Seventh report
of the Joint National Committee on Prevention, Detection,
Evaluation, and Treatment of High Blood Pressure. Hypertension 2003;42:1206–52.
Isomaa B, Almgren P, Tuomi T, et al. Cardiovascular morbidity and mortality associated with the metabolic syndrome.
Diabetes Care 2001;24:683–9.
Katz A, Nambi SS, Mather K, et al. Quantitative insulin sensitivity check index: a simple, accurate method for assessing
insulin sensitivity in humans. J Clin Endocrinol Metab 2000;
85:2402–10.
Kleinbaum DG. Applied regression analysis and other multivariable methods. 3rd ed. Pacific Grove, CA: Duxbury Press,
1997.
Morgan TO, Jacobsen SJ, McCarthy WF, et al. Age-specific
reference ranges for prostate-specific antigen in black men.
N Engl J Med 1996;335:304–10.
983
`