WO2001071355A2

WO2001071355A2 - Igf-i, igfbp-3 and psa in prostate cancer prediction

Info

Publication number: WO2001071355A2
Application number: PCT/US2001/040355
Authority: WO
Inventors: Christos S. Mantzoros; Alicja Wolk; Swen-Olof Andersson; Hans-Olov Adami; Dimitrious Trichopoulos
Original assignee: Beth Israel Deaconess Medical Center, Inc.; Karolinskas Innovations Ab; Presidents And Fellows Of Harvard College
Priority date: 2000-03-23
Filing date: 2001-03-22
Publication date: 2001-09-27
Also published as: AU2001250051A1; WO2001071355A3

Abstract

It was determined that the analytes IGF-I and IGFBP-3 can be used to predict the risk of incurring prostate cancer when PSA levels are within certain ranges. Thus, the levels of PSA must also be determined in order to determine how these analytes impact risk. It was found that, when PSA is ≤ 3 ng/ml, the risks for IGF-I and IGFBP-3 are additive. When PSA is > 10 ng/ml, the level of IGFBP-3 can be used to predict risk.

Description

TITLE OF THE INVENTION

USE OF IGF-I, IGFBP-3 AND PSA TO PREDICT THE RISK OF PROSTATE CANCER

CROSS REFERENCE TO RELATED APPLICATIONS This claims priority from United States provisional patent application Serial No. 60/191,678, filed March 23, 2000, which is incorporated in its entirety herein.

STATEMENT REGARDING FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

N/A

■BACKGROUND OF THE INVENTION

Prostate cancer is one of the higher incidence cancers, affecting 320,000 men per year and being the cause of death of 40,000 men per year. Diagnosis of prostate cancer has relied to a large extent on the measurement of levels of prostate-specific antigen

(PSA) . However, it is recognized that PSA testing is not specific, since as many as 25% of men with cancer have "normal" PSA levels, while as many as half of men with elevated PSA levels are, in reality, cancer-free.

(Garnick et al, Scientific American, 12/98, 75) Other risk factors (such as diet, height, etc.) have been less consistently associated with prostate cancer risk and are not included in prediction models. Recent studies (e.g., Mantzoros et al, 76 Br . J. Cancer (1997) 115) have shown that, when there is an increase in the analyte insulin-like growth factor I (IGF-I, also referred to as IGF-1) in a patient's serum, there is a higher likelihood that the patient will develop prostate cancer. Studies have also investigated insulin-like growth factor binding protein 3 (IGFBP-3), but they have not been consistent in demonstrating a relationship between IGFBP-3 and the risk of prostate cancer, with one (Chan et al, 279 Science (1998) 563) showing that there was an inverse relationship, while another ( olk et al, 90 J. Natl. Cancer Inst. (1998) 911) indicated no relationship. One (Chan et al) considered the level of prostate specific antigen (PSA) but did not consider IGFBP-3 as a predictor of risk according to PSA levels. Because of the uncertainty as to whether there was a relationship between these analytes, and whether the lack of agreement with respect to the role of IGF-1 and IGFBP-3 is due to a modifying effect of PSA levels, a study was undertaken to investigate whether there was such a relationship and whether measuring IGF-I and IGFBP-3 levels would provide information regarding risk for developing prostate cancer above and beyond that provided by PSA.

BRIEF SUMMARY OF THE INVENTION It was determined that the analytes IGF-I and IGFBP-3 can be used to predict the risk of incurring prostate cancer when PSA levels are within certain ranges. Thus, the levels of PSA must also be determined in order to determine how these analytes impact risk. It was found that, when PSA is ≤ 3 ng/ml, the risks for IGF-I and IGFBP-3 are additive. When PSA is > 10 ng/ml, the level of IGFBP-3 can be used to predict risk.

BRIEF DESCRIPTION OF THE FIGURES

Fig. 1 is a curve showing the relationship between risk of prostate cancer (expressed as percent change of risk in relation to an average person) and level of IGF- I (in ng/ml) , when PSA is ≤ 3 ng/ml and IGFBP-3 is equal to the mean value for the population. The average value for IGF-I in the reference group was found to be about 154 ng/mL.

Fig. 2 is a curve showing the relationship between risk of prostate cancer (expressed as percent change of risk in relation to an average person) and level of IGFBP-3 (in ng/ml), when PSA is ≤ 3 ng/ml and IGF-I is equal to the mean value for the population. The average value for IGFBP-3 in the reference group was found to be about 3000 ng/mL. Fig. 3 is a curve showing the relationship between risk of prostate cancer (expressed as percent change of risk in relation to an average person) and level of IGFBP-3 (in ng/ml), when PSA is > 10 ng/ml. The average value for IGFBP-3 in the reference group was found to be about 3000 ng/mL.

Fig. 4 is a decision tree for risk of prostate cancer - considering PSA, IGF-I, and IGFBP-3.

DETAILED DESCRIPTION OF THE INVENTION It has been found that all three analytes, PSA,

IGF-I and IGFBP-3, are needed to predict the incidence of prostate cancer. In addition, the risk values for both the IGF-I and IGFBP-3 increase and decrease in a linear fashion. That means, for example, for each fraction or percent increase of IGF-I, risk of prostate cancer increases or decreases by a similar fraction or percent risk.

The discussion herein refers to the concept of "average risk of a man getting prostate cancer" or the "risk of an average man getting prostate cancer". This concept refers to an adjusted risk for the specific population to which the subject belongs, which includes the risk for the population as a whole (as determined by several factors, including age, race and diet) plus the following: 1. an increase in risk if any first degree relatives of the subject are positive for prostate cancer,

2. an increase in risk if the subject has a high testosterone level (in the highest quartile of the population) , and

3. an increase in risk if the subject has low level of sex hormone binding globulin (SHBG) in the lowest quartile of the population. Each of these factors increases/decreases risk twofold. (Shaneyfelt, 18 J. Clinical Oncology (2000) 847.)

Furthermore, it was learned that the risks for IGF-I and IGFBP-3 are additive, when examining sera from patients where the PSA is ≤ 3 ng/ml. This means, for example, that if the risk based on IGF-I was estimated to be 200% (i.e., twice the average risk), while the risk for IGFBP-3 was found to be -50% (i.e., half the risk of the control sample) , the combined risk would be 150% (i.e., 1.5 times the average).

The fact that the risks are linear makes it possible to predict risk via the use of a calibration curve. Examples of such curves for IGF-I and IGFBP-3 when the PSA value is ≤ 3 ng/ml is shown in Figures 1-2. (Note: The average value for IGF-I in the reference group was found to be about 154 ng/mL. (which is the mean value for IGF-I in the DSL kit insert (see below) for the 50-70 year age group, that group having the highest incidence of prostate cancer) . The DSL literature was also used for the IGFBP-3, which showed a mean value for 2970 for the 50-70 year age group. (See also Figure 3 herein.) The risks were determined by considering the expected values for these analytes, which have been determined previously and can be found by examining literature, including that available from some manufacturers of kits for these analytes. (See, for example, literature from kits produced by Diagnostic Systems Laboratories, Inc. (DSL) , Webster, TX. In addition to relying on these published expected values, analysts should develop their own expected values, in order to account for lab to lab differences due to technique, reagents used, etc.) Furthermore, variation from the expected mean value of the analyte needs to be determined.

As indicated above, since the risk values are additive, and both analytes provide guidance when PSA is ≤ 3 ng/ml, an equation adding the predicted risk for each analyte provides the most comprehensive prediction using the 2 analytes. This actually provides the best prognostic tool, since it is most helpful to be able to predict the likelihood of prostate cancer before the disease occurs (i.e., when the PSA value is low), so that the patient and his physician can be alert for signs of the disease.

When PSA is ≤ 3 ng/ml, an equation to predict the risk of prostate cancer is:

Risk of prostate cancer = risk for an average person

+ risk based on IGF-I determination

+ risk based on IGFBP-3 determination

An example showing the calculation for representative patients is shown in Example 4.

Similarly, when PSA is > 10 ng/ml, the relationship between risk of incidence of prostate cancer vs. concentration of IGFBP-3 is shown in Figure 3. A value above 10 makes it highly likely, but not certain, that a person already has prostate cancer. Current recommendation is to proceed with biopsy and ultrasound to prove or disprove. Our data suggest that IGFBP-3 can provide additional levels of confidence before biopsy. It should be noted that in using this assay for measuring IGFBP-3, one should include both intact and fragments of IGFBP-3.

Thus, one can use the Decision Tree found in Figure 4 as a preferred embodiment to predict the odds of a patient developing prostate cancer. The PSA value is used to determine which of the other analytes should be used to determine the odds. When the PSA is ≤ 3 ng/ml, both IGF-I and IGFBP-3 should be considered. Note that, since the calculations for the 2 analytes are linear

(i.e., additive), the sign (i.e., positive or negative) must be considered. For PSA values > 10 ng/ml, only the IGFBP-3 should be considered. (Note, any method of determining concentration of these analytes can be used, including the use of commercial products, such as those available from Diagnostic Systems Laboratories, Inc.

(DSL) , Webster, TX. Fragments can be measured using similar methodology, including ELISA and IRMA

(immunoradiometric assay).)

The following examples are intended to further illustrate, but not limit, aspects of the invention.

Example 1. Subject pool and serum analysis

Subjects. All men under the age of 80 years, born in Sweden and living in Orebro County at any time from January 1989 through September 1991 formed the study base. Newly diagnosed, cytologically and histologically confirmed prostate cancer cases in this population were eligible and immediately reported to the investigators. The completeness of clinical records was confirmed through cross checking with the regional cancer registry. Control subjects were randomly selected every third month from the county population register, frequency-matched to case subjects in ten-year age groups. All potential control subjects underwent a digital rectal examination; and those with a palpable nodule and/or increased serum levels of prostate specific antigen were further investigated through ultrasound-guided biopsy. Only individuals whose biopsies showed no evidence of cancer were deemed eligible as control subjects. All participants filled out an extensive questionnaire while height and weight were obtained at a physical examination. All eligible patients and control subjects gave their informed consent to take part in the study.

Serum analyses. Blood samples were drawn, between 08:00 to 10:00 am, from 240 case subjects (86% of those eligible) and 235 control subjects (82% of those eligible) , before digital rectal examination was performed, or any treatment to the case subjects was given. The majority of blood samples among case subjects was collected within 4-6 weeks after diagnosis, none later than 3 months. Potential control subjects who did not provide blood were generally similar (age, body mass index) to those who did. Blood samples were centrifuged; a part was used for PSA analysis while the rest was stored at -70^°C.

For IGF-I and IGFBP-3 analysis immunoradiometric assays were conducted. (Kits from DSL were used. It should be noted, however, that other means of determination of IGF-I and IGFBP-3 can be utilized, including ELISA, radioimmunoassay or radioimmunometric assays sold by several companies for all three tests.) Serum samples were analyzed without knowledge of case- control status. Among 210 cancer subjects and 224 control subjects, PSA tests performed at time of inclusion were available from 209 and 221 subjects, respectively. Example 2. Statistical analysis

Simple pairwise comparisons of means in two groups were performed by t-tests (standard t-tests and the Welch approach, which allows unequal variances) and by nonparametric Mann-Whitney tests. They yielded similar results. The unconditional logistic regression model was used in both the univariate and multivariate modelling to estimate odds ratios (ORs - see below) with 95% confidence intervals (CIs) . Models were obtained with continuous variables in their original untransformed form. First, we performed analyses using all 221 control subjects. Then we restricted the analyses to 141 controls with a PSA classified as normal

(<3 ng/mL) . The ratio of odds for developing prostate cancer of the two groups is

(odds of cases)

(odds of controls)

Odds ratio is described in numerous references, including Breslow et al, Statistical Methods in Cancer Research, Vol. I, The Analysis of Case-Control Studies, in IARC Scientific Publication 32, International Agency for Research on Cancer, Lyon, 1980. (Further discussion of odds ratio appears in Example 3, below.)

The mean serum level of IGF-I was approximately 8% higher among case patients than among all control subjects (p=0.02) . Among case patients IGF-I levels did not differ appreciably between subgroups defined by serum levels of prostate specific antigen (PSA) at time of diagnosis (p for trend = 0.77) (See Table 1) . Nor was there any overall difference in serum levels of IGFBP-3 between case patients and all control subjects (p=0.09). However, mean levels of this binding protein were lower among patients with low levels of PSA and increased substantially with increasing PSA (p for trend = 0.004) .

Table 1. Mean value and standard deviation (SD) of insulin-like growth factor-I (IGF-I) serum levels and insulin-like growth factor binding-protein-3 (IGFBP-3) levels among control subjects and prostate cancer subjects by prostate specific antigen (PSA) levels.

Group PSA Number IGF-1, ng/mL IGFBP- 3, ng/mL mean (SD)^a mean (SD)^b

Controls all 221 146.9 (47.5) 2501 (760)

≤3 ng/mL 141 147.7 (48.5) 2474 (745)

Cases all 209 158.4 (53.8) 2668 (1038)

≤3 ng/mL 50 158.5 (54.7) 2345 (801)

3.1-10 50 158.6 (44.1) 2611 (943)

10.1-30 43 153.2 (53.3) 2739 (987)

30.1-100 32 161.8 (71.0) 2926 (1179)

>100 34 161.8 (51.2) 2906 (1301)

P trend^c 0. 77 0 004

^a Overall difference between cases and controls, p=0.02 ° Overall difference between cases and controls, p=0.09 ^c Tests for linear trend for IGF-I and IGFBP-3 values through strata of PSA Please note that the predictive value of IGF-I and IGF BP-3 in prostate cancer patients vs. normal subjects is examined in prespecified strata according to PSA values, as indicated in the table.

The associations between IGF-I, IGFBP-3, PSA and risk for prostate cancer were first evaluated in logistic regression models including all case subjects and control subjects with normal PSA level (<3 ng/mL)

(Table 2) . For IGF-I, it was noted that there was a significant positive association (OR 1.6; 95% CI 1.0-2.5 for 100 ng/mL IGF-I increment corresponding to two standard deviations in the study group - see Figure

1), which was attenuated after adjustment for IGFBP-3, but strengthened when adjusted also for PSA (OR 1.8; 95%

CI 0.9-3.4). A different pattern was seen for IGFBP-3

(See Table 2 and Figures 1 and 2); a positive association in a model adjusted only for age, height and body mass index (OR 1.5; 95% CI 1.0-2.2) became inverse after further adjustment for IGF-I and PSA (OR 0.7; 95%

CI 0.4-1.3) .

Table 2. Odds ratios for prostate cancer patients in relation to controls (with 95% CI) for specified increments of IGF-I and IGFBP-3 serum concentration and by PSA levels³; 141 controls with PSA <3 ng/mL

Cases IGF-I per p for IGFBP-3 p for

100 ng/mL trend⁰ per 1500 ng/mL^b trend⁰

OR (95%) CI) OR (95% CI)

All cases

Multivariate^d 209 1.6 (1.0-2.5) 0.05 1.5(1.0-2.2) 0.04

Multivariate^d mutually 209 1.4 (0.8-2.3) 0.25 1.3(0.9-2.0) 0.22 adjusted IGF-I and IGFBP-3

Multivariate^ mutually and 209 (0.9-3.4) .10 0.7(0.4-1.3) 0.26 PSA adjusted⁰

By PSA subgroups⁸

Multivariate^d ≤3 ng/mL 50 1.5 (0.8-3.1) 0.23 0.7(0.3-1.3) 0.23

3.1-10 50 1.6 (0.8-3.3) 0.19 1.5(0.8-2.8) 0.21

>10 109 1.6 (0.9-2.7) 0.09 2.1(1.4-3.3) 0.002

Multivariate^d and mutually adjusted IGF-I and IGFBP-3 ≤3 ng/mL 50 2.5 (1.1-5.9) 0.03 0.4(0.2-1.0) 0.04

3.1-10 50 1.4 (0.6-3.1) 0.40 1.3(0.6-2.7) 0.46

>10 109 1.0 (0.6-2.0) 0.91 2.0(1.2-3.5) 0.01

^a OR = odds ratio; CI = confidence interval; IGF-I = insulinlike growth factor I; IGFBP-3 = insulin-like growth factor- binding protein 3; PSA = prostate-specific antigen ^b Unit of increment corresponds to approximately two standard deviations of IGF-I (2 x 48.5 ng/mL) and IGFBP-3 (2 x 745 ng/mL) serum levels in control subjects. ^c Trend analysis based on continuous variables ^d Adjusted for age, height, and body mass index ^e Notice that all the 141 controls were included in each PSA subgroup in these analyses. Please note that the predictive value of IGF-I and IGF BP-3 in prostate cancer patients vs. normal subjects is examined in prespecified strata according to PSA values, as indicated in the table. Bold numbers indicate statistically significantly increased or decreased odds ratios of cancer patients in relation to controls .

To further explore the hypothesis that PSA, as a protease, is an effect modifier, different models for patients with PSA levels considered normal (<3 ng/mL) , moderately elevated (3.1-10 ng/mL) or markedly elevated

(>10 ng/mL) were considered (Table 2) . In these analyses, no evidence of effect modification was seen for IGF-I in models adjusting for age, height and body mass index; in each category of PSA, an increase in IGF-I levels of two standard deviations (100 ng/mL) was associated with a non-significant 50-60% increase in prostate cancer risk. However, when further adjusted for IGFBP-3, a different pattern emerged. Among patients with normal serum levels of PSA, a strong positive association between IGF-I and risk for prostate cancer (OR 2.5; 95% 1.1-5.9 per 100 ng/mL IGF-I) was found. This association was substantially weaker and statistically non-significant among men with a moderately elevated PSA level, and it was altogether eliminated among those with markedly elevated PSA levels

(OR 1.0; 0.6-2.0) . Analyses of IGFBP-3 also showed a clear evidence of effect modification by PSA (Table 2) . Indeed, while the association between IGFBP-3 and prostate cancer was apparently inverse among men with normal PSA values (OR 0.7; 95% CI 0.3-1.3), it was statistically highly significantly positive among men with markedly elevated PSA levels (OR 2.1; 95% CI 1.4-3.3). The inverse association between IGFBP-3 and prostate cancer risk among cases with normal PSA levels became further strengthened when IGFBP-3 was adjusted not only for age, height and body mass index, but also for IGF-I. Among men with PSA levels <3 ng/mL an increment of two standard deviation increase (1500 ng/mL) of IGFBP-3 was associated with statistically significant 60% reduction in risk for prostate cancer. Among men with PSA levels >3 ng/mL additional adjustment for IGF-I changed odds ratio estimates negligibly. When all the 221 controls in above described analyses were used, i.e. even those 80 with PSA >3 ng/mL, all associations were slightly attenuated. It should be noted that there are differences between specific population groups for the parameters used in the calculations. For example, 2 standard deviations for IGFBP-3 for the Swedish population study was about 1500 ng/mL, while the population in the U.S. examined for the DSL data had a value of about 880 ng/mL (for age group 50-70) . Thus, it is important that the characteristics of the specific population be determined.

To examine the phenomenon that underlies the powerful interaction among IGF-I, IGFBP-3 and PSA in relation to prostate cancer, the quartiles of IGF-I and IGFBP-3 were defined among controls. Then the 141 controls, the 50 prostate cancer cases with low PSA levels, the 50 prostate cancer cases with intermediate PSA levels, and the 109 prostate cancer cases with high PSA levels were cross-classified by quartile. In all four cross-tabulations, a strong positive association between IGF-I and IGFBP-3 was evident, the Pearson correlation coefficient being 0.5 (controls). Among controls, 61 belonged to the same quartile with respect to IGF-I and IGFBP-3, whereas for 41 the IGF-I quartile was higher than the IGFBP-3 quartile and for 39 the opposite was true. Among the 50 cases with low PSA levels, 23 occupied the same quartile, whereas for 22 the IGF-I quartile was higher than the IGFBP-3 quartile and for 5 the opposite was true. Among the 50 cases with intermediate PSA levels, 18 occupied the same quartile, whereas for 20 the IGF-I quartile was higher than the IGFBP-3 quartile and for 12 the opposite was true. Last, among the 109 cases with high PSA levels, 49 occupied the same quartile, whereas for 22 the IGF-I quartile was higher than the IGFBP-3 quartile and for 38 the opposite was true. It is apparent that, among cases with low PSA levels, relatively higher IGF-I levels correspond, as a rule, to relatively lower IGFBP-3 levels, whereas, among cases with high PSA levels, relatively lower IGF-I levels correspond to relatively higher IGFBP-3 levels. The case-control study has several strengths. By studying directly the cases with prostate cancer and controls without prostate cancer, one can study directly risk factors for developing the disease. The case and control subjects were generated from a well-defined study base. The lack of screening practices for prostate cancer in the study area, at the time of subject recruitment, minimizes selective overrepresentation of health conscious men and contamination of the case series with prevalent cancers with an unknown potential to progress. Blood sampling and laboratory methods were highly standardized. Evidence of interaction of IGF-I, IGFBP-3 and PSA in relation to prostate cancer is striking. The biology that underlies the phenomenon, however, is not immediately obvious. The results indicate that at the time of diagnosis of a prostate cancer characterized by low levels of PSA (≤3 ng/mL) , the disease is associated positively with IGF-I and inversely with IGFBP-3. Indeed, when the positively associated IGF-I and IGFBP-3 are mutually adjusted for, their respective relations to incident prostate cancer with low PSA levels become both statistically significant. For prostate cancer with high PSA levels (>10 ng/mL) , both IGF-I and in particular IGFBP-3 are positively associated with the disease (ORs 1.6 and 2.1, respectively). The increase of IGFBP-3, however, is disproportionally large and, after adjustment for IGFBP-3 level, the association of IGF-I with high PSA levels prostate cancer appears to become essentially null. The latter results, however, refers to IGF-I increments over and beyond those predicted by the substantial increase noted for IGFBP-3, a compound which is strongly positively associated with IGF-I. In reality, incident cases of prostate cancer with high PSA levels are characterized by both elevated IGF-I levels and substantially elevated levels of IGFBP-3.

IGF-I and IGFBP-3 are positively associated and it is assumed that this reflects a positive feedback system. As circulating PSA increases, however, larger amounts of IGFBP-3 need to be secreted for a given increment of IGF-I, because a substantial fraction of IGFBP-3 is being fragmented by the increasing PSA. IGF-I itself remains elevated and might continue to influence the growth of prostate cancer, even though its levels are substantially lower than those predicted by the increasing serum levels of IGFBP-3.

Other established or suspected processes may play a role, including differential effects of IGFBP-3 on the growth of normal and malignant prostate epithelial cells and the IGFBP-3 induced prostate cell apoptosis recently reported. Nevertheless, it appears that neither PSA nor IGF-I and/or IGFBP-3 levels can be properly interpreted without taking into account the values of the other two compounds. Among subjects with PSA levels less or equal to 3 ng/mL, the likelihood of being truly prostate cancer free increases when the value of IGFBP-3 is increased relative to the value of IGF-I. On the contrary, among subjects with PSA >3 ng/mL the likelihood of being prostate cancer free decreases when the value of IGFBP-3 is increased relative to the value of IGF-I. The observed interrelation between PSA, IGF-I and IGFBP-3 in cancer development of prostate cancer risk might potentially have use in clinical evaluation of men for prostate cancer.

Example 3: Determination of odds ratio

Odds of a specific person to develop prostate cancer vs. not

The formulas below indicate the odds of a specific person to develop prostate cancer in relation to that of an average person, i.e., percent increase over an average person's odds.

A) if PSA < 3

Odds of a specific person = odds of average risk of an average person

+ [(this person's IGF-1 level - population mean

IGF-1 in ng/ml) * 2.5%] - [(this person's IGF-BP3 levels - population mean

IGF-BP3 in ng/mL) x (2.5/15)%]

B) if PSA > 10

Odds of a specific person = odds of an average person + [(this person's IGF-BP3 levels - population mean IGF-BP3 in ng/mL) x (2/15)%].

(See sample data plotted in Figure 3.)

Example 4: Sample calculations

Expected values from DSL kit inserts have been used in the following calculations.

1. A 60 year old man with PSA < 3, IGF - 1 = 253.7 (i.e. 100 above the mean) and IGFBP3 = 3466 (i.e. 500 above the mean) The population means for this age according to package inserts are IGF-1= 153.7 and IGF BP-3= 2966 the calculated risk would be: odds of an average person +

(100 x 2.5%) - (500 x 2.5/15 %) = 167% or 1.67 times the odds of an average person.

2. A 60 year old person with PSA < 3, IGF-1=253.7 (i.e. 100 above the mean) and IGF BP3= 1466 (i.e. 1500 below the mean) the calculated risk would be: odds of an average person + (100 x 2.5%) - (-1500 x 2.5/15 %) = the odds of an average person + 250% + 250% = the odds of an average person increased by 500% or in other words increased five times.

3. A 50 year old person with PSA > 10 and IGF BP3= 4466 The calculated risk would be: the odds of an average person + (1500 x 2/15%) = odds of an average person increased by 200%.

In other words, obtaining an IGF BP-3 level tells us that this person's expected risk based on his PSA is increased by two fold (based on the information provided by his IGF BP-3 level) .

Example 5 : Fragment data The Descriptive Statistics of total and IGFBP-3 fragments in a population of prostate cancer patients are as follows:

Minimum Maximum Mean Std. Deviation F FRRAAGGMMEENNTTSS 1 1..4400 4 4..8800 2.8111 .7885

TOTAL 1.00 5.60 3.0404 1.0961 Total and fragments of IGFBP-3 were determined as indicated above (see page 7) .

Those with skill in this technology area will recognize variations in the above disclosure which are consistent with this invention.

Claims

1. A method for predicting the risk for prostate cancer in a patient comprising: a. collecting a serum sample from said patient, b. analyzing said sample for the analytes (1) insulin-like growth factor I (IGF-I), (2) insulinlike growth factor binding protein 3 (IGFBP-3) , and (3) prostate specific antigen (PSA) , c. comparing the value of said analytes to those of a control group, and d. predicting the risk for prostate cancer.

2. The method of claim 1 where, when PSA is less than or equal to 3 ng/ml, said values for the analytes IGF-I and IGFBP-3 are quantitatively determined, are additive, and are used to predict the risk of prostate cancer.

3. The method of claim 1 in which the risk of prostate cancer where PSA is less than or equal to 3 ng/ml, is predicted by the equation:

Risk of prostate cancer = risk for an average person + risk based on IGF-I determination + risk based on IGFBP-3 determination

4. The method of claim 3 in which said risk for an average person is the sum of: the risk for the population as a whole plus the risk derived by incidence of prostate cancer in first degree relatives plus the risk derived by testosterone levels plus the risk derived from level of sex hormone binding globulin.

5. The method of claim 1 where, when PSA is greater than 10 ng/ml, said value for the analyte IGFBP-3 is quantitatively determined and used to predict the risk for prostate cancer.

6. The method of claim 1, in which the risk of prostate cancer where PSA is greater than 10 ng/ml, is predicted by the equation:

Risk of prostate cancer = risk of an average person + risk based on IGFBP-3 determination.

7. The method of claim 1, in which the risk of prostate cancer where PSA is greater than 10 ng/ml, is predicted by the equation:

Risk of prostate cancer = risk of an average person + risk based on IGFBP-3 fragment determination.

8. The method of claim 6 in which said risk for an average person is the sum of: the risk for the population as a whole plus the risk derived by incidence of prostate cancer in first degree relatives plus the risk derived by testosterone levels plus the risk derived from level of sex hormone binding globulin.