Composite screener for androgen deficiency related to the aging males’ symptoms scale

Abstract

Objective. Methodological characteristics of the Aging Males’ Symptoms (AMS) scale point towards a high standard measurement and comparison of health-related quality of life (HRQoL) over time or intervention. However, the scale was neither developed nor standardized as a screening instrument for androgen deficiency.

Methods. Data of the Austrian ANDROX study suggested to develop a composite screener for androgen deficiency based on AMS, age, and BMI, to determine sensitivity and specificity to detect low total testosterone (TT) levels. The findings were compared with those of an independent sample of urological patients with suspicion of androgen deficiency (AD) from Germany (n = 803).

Results. A graphical solution for a composite-screening tool was proposed, with three levels of suspicion for AD: positive screening result (high suspicion), equivocal result, and negative screening result (no suspicion, AD unlikely). The percentage of TT values < 4 ng/ml were 18.7%, 40.7% and 58.8% in the three categories: negative, equivocal and positive, respectively. The findings were confirmed in the independent German sample. In both instances, neither sensitivity nor specificity was very high, i.e., between about 50% and 75%.

Conclusion. The results of the development and initial validation of the new screener are promising. Further information, and experience from the practice, is needed to confirm or refute the hypothesis that this screener is a useful tool for medical practice.

Keywords:

• AMS
• androgen deficiency
• screening
• testosterone
• validation

Background

The Aging Males' Symptoms (AMS) scale is a health-related quality of life scale (HRQoL) and was originally developed in Germany in 1999 [1]. The scale was designed to be a self-administered scale to: (a) assess symptoms of aging (independent from those which are disease-related) between groups of males under different conditions, (b) evaluate the severity of symptoms over time, and (c) measure changes pre- and post-androgen therapy [1]. It was developed in response to the lack of fully standardized scales available to measure the severity of aging symptoms, and their impact on HRQoL in males, specifically [2],[3]. The AMS scale was internationally well received: it is now available in 21 languages, in both hardcopy and electronic versions [2],[4],[5], i.e., it can be downloaded from the internet (either through http://www.aging-males-symptom-scale.info or from a relevant publication [4]). The scale has also shown its utility to measure effects of treatment for androgen deficiency [6].

The evaluation of the AMS scale is simple, and the scheme has recently been published [4]. Norm values for comparison were determined [1],[3]. The conventional psychometric requirements of test reliability and validity, as well as demands regarding clinical utility and outcome sensitivity, were also recently published [7].

The scale was not developed as a screening instrument for androgen deficiency. However, many complaints considered in the AMS scale are common symptoms in androgen-deficient males. Therefore, we compared the AMS scale with internationally well-known screening instruments for androgen deficiency in adult males (ADAM scale of Morley et al. [8] and the screener of Smith et al. [9]). We found that the AMS has obviously similar test characteristics as both screening instruments [10]. In addition, Kratzik et al. [11] observed, in a population-based cross-sectional study in Vienna, an impressive association between subscales of the AMS and free testosterone level when age and body mass index (BMI) was multivariately taken into account. Recently, a Japanese research groups under Itoh et al. [12] and Soh et al. [13] observed a correlation between AMS scores and testosterone level. A Polish research group found a similar, but less apparent, result [14]. Other studies, however, could not find associations between AMS scores and testosterone level [15],[16].

These observations and findings by Kratzik et al. [11] led to the working hypothesis that the AMS score, in conjunction with other variables (age and BMI), may have the potential to screen for androgen deficiency, despite our conceptual hesitation to use a symptom profile HRQoL scale for screening purposes.

The aim of this paper is to propose a simple, graphical screening tool for androgen deficiency–consisting of AMS total score, age, and BMI–based on data already available from previous research [6],[11], and to provide the very first test characteristics of the proposed screening tool, such as sensitivity and specificity.

Material and methods

The ANDROX study from Vienna was used to develop the screening tool. It is a community based cross-sectional study and has been described earlier in this journal [11]. In brief, a voluntary health program was offered to men aged 40–60 years, who were working in the Vienna municipality as manual workers between May 2001 and November 2001. The subjects were asked to complete a questionnaire (including the AMS); body weight and height were measured, and a blood sample taken between 8:00 a.m. and 10:30 a.m. Moreover, a urologist performed a complete examination. All 2400 employees were invited to participate in the study; 752 attended the interview and examination survey. Data of 669 subjects were included in this analysis. The testosterone determinations of two samples were done in one central laboratory under standardized conditions, i.e., the laboratory methods described in a previous issue of this journal [11].

For a crude comparison, a large German sample of urological patients (2000/2001) was used to check the plausibility of the results obtained in the ANDROX study sample. This German sample was a database of those patients with any suspicion of androgen deficiency, who were attending urologist outpatient departments, and were included in an open, uncontrolled post-marketing surveillance study described elsewhere [6]. Patients with cancer, or other terminal illnesses, were not included. One hundred and ninety-three urologists from all parts of Germany also participated in this study. A questionnaire was completed, as well as the AMS questionnaire, a physical examination performed and a blood sample taken. The testosterone determinations were decentrally performed in the routine labs of the participating physicians. No standardized TT determination was applied. The data of the initial visit in the surveillance were used for the purposes of this paper. One thousand, one hundred and seventy-four patients (no age limit) participated in the study; 803 patients aged 40 years and more were available for our analyses.

As a first, explorative analysis, we checked how closely the three envisaged screener variables (AMS + age + BMI) were associated with total testosterone in the ANDROX sample. This was done using linear regression analysis.

The development of the graphical screening tool was based on a simple frequency table: how many subjects with low total testosterone, i.e., potentially suspicious for AD, were observed in a three-dimensional table of AMS total score, age, and BMI? In order to get an easy, usable graphical tool, we used the variables transformed into the following categories: four BMI categories (< 24, 24–26, 27–29, 30 + ), five AMS total score groups (< 18, 18–30, 31–38, 39–50, 50+ scoring points), and two broad age groups (under/equal to 50 years and over 50 years of age). We distinguished only men with ‘normal’ and ‘suspicious' level of total testosterone, and intentionally defined these alternative categories as over 4.0 ng/ml and under/equal 4.0 ng/ml for the purpose of a screening scale.

The screening outcome was categorized in three levels:

(+) Positive (= suspicion of AD): 50% or more males with total testosterone under 4 ng/ml,(?) Equivocal screening result: 25% to 50% subjects with TT under 4 ng/ml observed, and(−) Negative (no/little suspicion of AD): fewer than 25% subjects with TT less than 4 ng/ml observed.

Thereafter, the model was applied to an independent sample of urological patients for a plausibility check, or preliminary validation (in the absence of data from another community sample): the test characteristics of the screening model were described as sensitivity and specificity. For this purpose, 2 × 2 tables were used. The screening outcome was dichotomized: positive results (+) and non-positive (i.e., combined equivocal and negative results). The TT level was also classified into two categories: ‘normal TT’ (over/equal 4 ng/ml (−)) or ‘suspicious TT’ (under 4 ng/ml (+)). This preliminary validation should help to determine whether the proposed screener could be considered for further research and testing in medical practice.

All analyses were done with the statistical packages SPSS 10.0 for Windows and STATA 6.0.

Results

We analyzed the association between total testosterone (TT) and the AMS total score with age, and BMI in the community-based ANDROX sample [11] (n = 669), by means of a linear regression model (Table I). The association of the total AMS score was not significantly associated with TT. However, the association between TT and BMI was significant, as was the association with BMI.

Table I.  Association between total testosterone and the AMS total score, age in years, and body mass index (BMI). Explorative linear regression analysis done for the ANDROX study sample (Austria)

	Regression coefficient	CI95%	p-value
AMS, total score	−0.007	(−0.018; 0.004)	0.20
Age, years	−0.048	(−0.074;−0.023)	0.00
BMI (kg/m^2)	−0.132	(−0.162; −0.101)	0.00

Table II shows the graphical scheme of the proposed screening tool following the rational that the likelihood of suspicious TT values (< 4 ng/ml) should be high (over 50%) in the category of ‘screening positives', low in ‘screening negatives' (less than 25%), and an equivocal group in between. The model works with five categories of increasing AMS total score (< 18, 18–30, 31–38, 39–50, 50 +) and with four categories of increasing BMI (< 24, 24–26, 27–29, 30+) in two broad age groups (< = 50, 50+). One sees that in the age group including 50 years, the screening probability of ‘screening negatives' (white boxes = −) declined with increasing BMI and higher AMS total scores. The higher the BMI and AMS total score category, the more ‘screening positives' (grey marked boxes = +) were found. In between is a region of equivocal results (= ?). The number of boxes of ‘screening negatives' (−) declined and the number of ‘equivocal’ (?) and ‘screening positives' (+) increased in the over 50 age group, which is an expected finding for a screening tool in the higher age segment.

Table II.  The screening tool: Graphic scheme to display the three categories of increasing suspicion of androgen deficiency in two age groups based on data of AMS total score, BMI and two age categories.

Age ⩽50					Age > 50
	BMI					BMI
AMS Total Score	< 24	24–26	27–29	30 +	AMS Total Score	< 24	24–26	27–29	30 +
< 18	−	−	−	−	< 18	−	−	?	?
18–30	−	−	?	+	18–30	?	?	?	+
31–38	−	−	?	+	31–38	?	?	?	+
39–50	−	?	+	+	39–50	?	+	+	+
51 +	?	?	+	+	51 +	+	+	+	+

+ Screening positive (suspicious for AD). ? Equivocal screening result. − Screening negative (no suspicion of AD).

The three possible outcomes of the screening tool (−, ?, + ) have a different proportion of subjects with TT level below or equal or higher than 4 ng/ml in the ANDROX sample. The percentages of low TT levels (<4 ng/ml), and were 18.7%, 40.7%, and 58.8% in the three screening categories ‘−’, ‘?’, and ‘ + ’, respectively.

We used an accessible sample of urological patients in Germany (see Methods section) to check the plausibility of the results of the proposed screener. The test characteristics of the screening model were described as sensitivity and specificity. For this purpose, two 2×2 tables were built. As can be seen in Table III, the test characteristics – sensitivity and specificity – to detect ‘screening positives' (under 4 ng/ml TT) as compared with the rest of the sample (equivocal and negative combined) are not so different in the ANDROX sample (used to develop the screening tool) and the independent sample of patients (used to preliminarily validate the proposed screener).

Table III.  Test characteristics (sensitivity and specificity) to predict a low androgen level in two samples, number of males over 40 years in each cells of the 2 × 2 table.

ANDROX sample	Total testosterone			Patient sample	Total testosterone
Austria	+ (<4 ng/ml)	− (⩾4 ng/ml)		Germany	+ (<4 ng/ml)	− (⩾4 ng/ml)
Screening result				Screening result
+	141	99	Sensitivity = 48.3%	+	388	58	Sensitivity = 69.9%
? / −	151	278	Specificity = 73.7%	? / −	167	105	Specificity = 64.4%
Total	292	377		Total	555	163

The two samples are:

• The ANDROX sample from Austria – used to develop this screener, and

• An independent sample of urological patients from Germany – to validate the screener.

Discussion

The AMS scale is a standardized HRQoL scale with good psychometric characteristics [1-7]. Comparison with other scales for aging males, or screeners for androgen deficiency, showed good associations, i.e., illustrated a good concurrent (criterion-oriented) validity.

Thus, the currently available methodological evidence indicates a high quality of the scale to measure and compare HRQoL over time, or before/after intervention. It suggests high reliability and high validity, as far as the process of construct validation could be completed. However, the scale was not developed or standardized as a screening instrument for androgen deficiency [1],[3]. Moreover, as far as the AMS scale per se, and its association with testosterone level is concerned, the situation is controversial: some studies could not find associations with testosterone level [15],[16], while others found close associations [11-14]. The inclusion of BMI as part of a new screener was predetermined by the data available. There might be good reason to prefer other variables describing the body shape, such as waist circumference or waist–hip ratio. Data to test this possibility, however, were not available.

Research by Kratzik et al. [11] raised the question as to whether AMS scores, in combination with age and BMI, could be used as screening scale. Based on the ANDROX study, a screening tool for low total testosterone level (< 4 ng/ml) was developed. The aim was a simple instrument for preselection of subjects that needed further medical attention (e.g., blood sampling), or for self-completion by subjects wherever complaints do not lead to a physician consultation (e.g., self-assessment of patient or subjects).

The initial exploratory analyses of AMS scores, age, and BMI in the context of total testosterone level, or low/high testosterone, led to many statistically significant associations of the individual variables with androgen level. It should be considered, however, that the application of the linear regression analysis might be problematic for associations of TT with age, because a constant or increasing trend of TT can be assumed in younger age, and a declining in later age. We abstained from age-stratified or non-linear analyses for the purpose of this very crude, descriptive orientation. It is also worth mentioning that the two samples we compared were inherently different – the ANDROX study is based on a community sample (manual workers of the Vienna municipality), and the German sample consists of selected patients who contacted urologists with complaints and any suspicion of androgen deficiency (excluding cancer or other terminal illness).

Nonetheless, the results across analyses pointed in the same direction. This encouraged us to analyse whether the combination of the three variables could be used to define groups of high and low likelihood of androgen deficiency, as measured by blood tests. For the purpose of a screening instrument – in contrast to a diagnostic measure – we deliberately defined the cut-off point of 4 ng/ml total testosterone; this might be debatable.

At screening level, we defined three levels of suspicion for AD: positive screening result or high suspicion (+) means contact a physician, equivocal result (?) could lead to the recommendation to wait, and contact a physician if complaints persist or increase, and the negative (−) screening result (= no suspicion, AD unlikely) means that no further activity is needed. These categories were associated with a different percentage of AD based on TT in blood tests, i.e., 18.7%, 40.7%, and 58.8% in the three screening categories ‘−’, ‘?’, and ‘+’. Smith et al. proposed an analogous categorization of the likelihood of androgen deficiency some years ago, for their screener for androgen deficiency [9], the results of which we found to be related, even to the AMS scale alone [10].

The basic assumption was that this kind of simple graphical solution of a screening tool could be accepted in medical practice if it proves to be valid in an independent sample. Since no other population sample with relevant data was available, we use an existing patient database. We obtained similar results when both samples (Austria and Germany) were compared using the tool.

The quality of a screening test is usually described with sensitivity and specificity. The sensitivity or correct prediction of positive subjects (= low TT value) was 48.3%, and the specificity or correct prediction of negative subjects (unsuspicious, high TT value) was 73.7% in the ANDROX sample, where the test was developed. The values in the independent ‘validation sample’ of patients were 69.9% and 64.4% for sensitivity and specificity, respectively (see Table III).

In both samples, neither sensitivity nor specificity was very high, i.e., ranging between 50% and 75%. Quite a number of subjects were left in a equivocal situation, for whom further diagnostic workup would be necessary, although they were not ‘positive’ according to the proposed screening tool. Nevertheless, the characteristics of the new composite screener (AMS + age + BMI) to detect low TT seem to be more promising than the AMS values alone, which were reported to be associated with the testosterone level in some [11-14], but not all, studies [15],[16].

The described test characteristics of the proposed composite ‘AMS screener’ will be acceptable for mass screening and preselection of subjects for further diagnostic workup. There might be also an application of this tool in the internet, for educational purposes or self-assessment of interested patients. It is likely that it is less useful to apply this screening tool as ‘diagnostic tool’ in a urological practice, as a series of blood tests for testosterone can be easily performed in this setting. It is suggested to collect own experience through self-assessment or practical application, which could further confirm or refute the described results obtained with this screener of AD, or to further define the scope of possible applications.

Conclusions

We conclude that the results of the development and initial validation of the new screener are promising. Further research and field experience are needed to confirm or refute the hypothesis that this screener is a useful tool for medical practice.