Effects of Anabolic-Androgenic Steroids on Muscular Strength Janet D. Elashoff, PhD; Alan D. Jacknow, MD; Sara G. Shain, MS; and Glenn D. Braunstein, MD
• Objective: To assess the effects of anabolicandrogenic steroids on human muscle strength. • Data Sources: A MEDLINE search for the period from January 1966 to April 1990, supplemented by manual searches of previous reviews, produced 30 studies in which subjects received more than one dose of the study steroid and in which changes in muscular strength were measured. • Study Selection: Of the 30 studies, 14 were not included in the detailed data summary because they did not use a placebo control, did not randomize subjects to groups, or did not make objective strength measurements, or because percent change in strength data could not be abstracted. • Data Extraction: Details of study design, reporting of results, and the adequacy and correctness of statistical methods were tabulated. Percent improvement in strength for the largest muscle group studied was computed, using the difference between results for the placebo and for the steroid-treated groups. • Data Synthesis: Previously trained athletes show slightly greater improvements in strength in the anabolic-androgenic steroid-treated group than in the placebo group, with a median difference of 5% across the nine studies (range, 1.2% to 18.7%). A meta-analysis of the three studies with enough information to compute effect size showed a mean difference of 1.0 standard deviations (95% CI, 0.49 to 1.5). However, the poor overall quality of the studies in terms of design, sample size, and analysis; the lack of a dose-response effect across the narrow range of dosages tested; and the tendency for differences to be smaller in the larger studies throw these results into question. No evidence was found to support enhanced muscle strength with steroid use in eight studies in untrained normal volunteers. • Conclusions: Anabolic steroids may slightly enhance muscle strength in previously trained athletes. No firm conclusion is possible concerning the efficacy of anabolic steroids in enhancing overall athletic performance. Results for the low steroid dosages studied in the published reports cannot be generalized to steroid-using athletes taking megadose regimens.
1 he anabolic effects of androgens were first shown in the 1930s by Kochakian and Murlin (1,2). Later studies showed dissociation between anabolic and androgenic effects of synthetic and naturally occurring androgens (3). In the early 1950s, performances of several weightlifting teams improved dramatically, and rumors suggested this was caused by the use of androgens as ergogenic aids. The supposed success of these drugs led many more competitors in the weightlifting and throwing sports to use them in athletic events, including the 1956 Olympic Games in Melbourne, Australia (4). At that time the scientific evidence supporting their use was minimal, and elite athletes used anecdotal reports to support their continued use. By 1976 a number of studies had been conducted, some of which concluded that there was a positive effect and others, that there was no positive effect of steroids on strength in trained athletes. In view of conflicting evidence of their effectiveness, reports of side effects, ethical considerations, and the development of
Table 1. Studies Not Included in the Data Summary and the Reasons for Exclusion Study (Year)
Reference
Study design inadequate Keul et al. (1976) Johnson and O'Shea (1969) O'Shea and Winkler (1970) Tahmindjuis (1976)
12 13 14 15
Methods unclear, no randomization Volunteers received steroid No blinding, three same-order crossovers No controls
Did not measure strength objectively Ariel and Saville (1972) 16 Knee jerk reflex times only Crist et al. (1983) 17 Subjective ratings of strength only Freed et al. (1975) 18 Subject reports of initial strength O'Shea (1970) 19 Swimming speed only Percent improvements could Bowers and Reardon (1972) Hervey et al. (1976) Fowler et al. (1965)
not be computed 20 Strength data not reported 21 22
Johnson et al. (1972)
23
Hervey (1975)
24
Hervey et al. (1981)
25
Annals of Internal Medicine. 1991;115:387-393. From Cedars-Sinai Medical Center-UCLA School of Medicine, Los Angeles, California. For current author addresses, see end of text.
Reasons
Strength data not reported Percent change could not be computed Percent change could not be computed Data for first period only not obtainable Data for first period only not obtainable
©1991 American College of Physicians
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
387
methods to screen for their use, the International Olympic Committee banned anabolic steroids for the first time for the 1976 Olympics in Montreal, Canada (5). In 1977, the American College of Sports Medicine (ACSM) issued a position statement on anabolic steroids based on a survey of the literature. They concluded: "The administration of anabolic-androgenic steroids to healthy humans below age 50 in medically approved therapeutic doses often does not of itself bring about any significant improvements in strength, aerobic endurance, lean body mass, or body weight. There is no conclusive scientific evidence that extremely large doses of anabolic steroids either aid or hinder athletic performance'' (6). This view was not totally accepted by elite athletes, and research on anabolic steroid effects continued into the early 1980s. In 1984, the ACSM revised its previous position and, based on another review of the literature, concluded. 'The gains in muscular strength achieved through high intensity exercise and proper diet can be increased by the use of anabolic-androgenic steroids in some individuals" (7). The paper went on to deplore the use of steroids because of the "ethical principles of athletic competition." During the 1980s, four reviews (8-11) were published; all pointed out that about half the studies of the effect of anabolic steroids on muscular
strength had concluded that there was an effect whereas about half found no significant effects. Three of the reviews (9-11) pointed out many problems in study design, analysis, and reporting. Haupt and Rovere (10) reviewed in detail 24 studies of the effects of anabolic steroids on human strength and numerous studies and case reports of side effects, concluding that "anabolic steroids have their most pronounced effects in those athletes who have trained to the point that they are in a chronic catabolic state." Wilson (11) reviewed in detail 16 studies of muscular strength that appeared "to be adequately designed and blinded" and also reviewed data on side effects; he concluded that no "improvement in strength can be demonstrated consistently when androgens are administered double blind to athletes." All previous reviews have dichotomized studies into those claiming statistical significance and those not claiming statistical significance rather than address the question of size and consistency of observed steroid effects. In view of the importance of the question of steroid effects on muscle strength, we carefully reviewed the available studies and abstracted mean percent improvement in strength for placebo and steroidtreated groups to assess the existence and size of anabolic-androgenic steroid effects on the strength of trained athletes and of untrained healthy young men.
Table 2. Study Characteristics (Studies Ordered by Total Dosage of Anabolic Steroid) Study (Year)
Studies using trained athletes Weiss and Muller (1968) Ariel (1973) Ward (1973) 0'Shea(1971) Loughton and Ruling (1977) Stamford and Moffatt (1974) Ariel (1974) Golding et al. (1974) O'Shea (1974) Studies using untrained subjects Loughton and Ruling (1977) Johnson et al. (1975) Win-May and Mya-Tu (1975) Casner(1971) Fahey and Brown (1973) Samuels et al. (1942) Munson (1970) Stromme et al. (1974)
Refer- Number ence of Subjects
Steroid
Mg per Weeks Randomization Day and Blinding
Training Measurement
Protein Used
2.6 4 4 4
"Random" Double-blind Single-blind* Double-blind*
Yes No Yes Yes
Yes Yes No Yes
No No No Yes
6
Double-blind
Yes
Yes
Yes
4§ 4 9 5
Single-blind|| Double-blind Double-blind*§ "Random"
No Yes No Yes
Yes Yes Yes Yes
Yes No Yes Yes
10/5* 6
6 3
Double-blind Double-blind
Yes Yes
Yes Yes
Yes Yes
5 6
13 6
Double-blind Double-blind
No No
Yes Yes
No No
8 Nandrolone decanoate 11 Methyl testosterone 50/200** 4 4 20 Oxandrolone Mesterolone 75/150tt 8
Double-blind Single-blind Double-blind Double-blind
Yes Yes Yes Yes
Yes Yes Yes Yes
No No No Yes
26 27 28 29
32 6 16 20
Methandrostenolone Methandrostenolone Methandrostenolone Methandrostenolone
10 10 10 10
30t
6
Methandrostenolone
10/5$
31 32 33 34
12 10 40 18
Methandrostenolone Methandrostenolone Methandrostenolone Stanozolol
20 15 10 8
30 35
6 13
Methandrostenolone Stanozolol
36 37
31 27
Methandrostenolone Stanozolol
38 39 40 41
28 4 39 21
* Steroid- and placebo-treated subjects were paired. t Athletes were from a wrestling team. $ Subjects received 10 mg/d for the first 3 weeks and 5 mg/d for the last 3 weeks. § Steroid was given from Monday to Friday only. || Assignment to groups was based on matching rather than on randomization. H The total dosage was 3 mg/kg body weight. ** Subjects received 50 mg/d for the first 3 weeks and 200 mg/d for the last week. tt Subjects received 75 mg/d for the first 4 weeks and 150 mg/d for the last 4 weeks. 388
Standardization
1 September 1991 • Annals of Internal Medicine • Volume 115 • Number 5
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
Table 3. Study Results (Studies Ordered by Total Dosage of Anabolic Steroid) Study (Year)
Studies using trained athletes Weiss and Muller (1968) Ariel (1973) Ward (1973) 0'Shea(1971) Loughton and Ruling (1977) Stamford and Moffat (1974) Ariel (1974) Golding et al. (1974) O'Shea (1974) Studies using untrained subjects Loughton and Ruling (1977) Johnson et al. (1975) Win-May and Mya-Tu (1975) Casner(1971) Fahey and Brown (1973) Samuels et al. (1942) Munson (1970) Strommeetal. (1974)
Reference Number of Subjects
Report of Results
Statistical Methods
Test of Muscle Group
Total Dosage*
Percent Mean Strength Increase Placebo Steroid Difference Group Group
By subject Figure Means ± SE By subject
Incorrect Incorrect Incorrect Correctt
309 476 476 476
Other Bench press Bench press Bench press
0.2 1.4 4.3 8.3
1.5 20.1 13.2 16.9
1.3 18.7 8.9 8.6
6
Means only
Unclear
536
Bench press
23.3
26.6
3.3
31 32 33 34
12 10 40 16
Means only Figure Means only Mean ± SE
Not stated Incorrect Not stated Correctt
3.1 5.1 8.1 6.7
5.5 18.5 9.3 11.8
2.4 13.4 1.2 5.0
30 35
6 13
Means only Means only
Unclear Correct
536 630
11.2 2.0
20.5 2.9
9.3 0.9
36 37
31 27
Means only Means only
Unclear Correct
774 1260
Hand grip Squat
3.7 2.5
6.9 -11.9
3.2 -14.4
38 39 40 41
28 4 32 21
Means only Figure By subject Figure
Incorrect Not done Correct Not stated
2571$ 2818 6328 7245
Bench press Hand grip Squat Squat
11.3 9.2 15.8 10.7
17.3 22.2 5.8 14.2
6.0 12.9 -10.0 3.5
26 27 28 29
32 6 16 18
30
680 714 1071 1400
Bench Bench Bench Bench
press press press press
Bench press Knee extension
* Total anabolic steroid dosage = daily steroid dose times number of days times steroid factor (steroid factors are as follows: methandrostenolone, 1.7; stanozolol, 5; oxandrolone, 11.3; methyl testosterone, 1.15; nandrolone decanoate, 11.3; mesterolone, 1.15). t Although the methods appear to be correct, the calculations do not appear to be correct. t Computed by multiplying the total dosage of 3 mg/kg times an average body weight of 75.8 kg.
Methods Selection of Studies A MEDLINE search was conducted for the period from January 1966 to April 1990 and was supplemented by manual searches of previously published reviews and position papers. We searched for studies that evaluated the use of anabolic steroids in ergogenic activity in which subjects received more than a single administration of the steroid in question and in which muscular strength was measured quantitatively. We included 30 studies of the effect of anabolic steroid use on strength performance in our report (12-41). We eliminated four studies (12-15) because they either did not provide a placebo control or did not randomize subjects to groups (Table 1). We placed no limitations, however, on study size, time interval, dosage, or choice of steroid.
Measures of Steroid Effect Because different exercises were used in the various studies, a kinesiologist determined the dominant muscle groups tested by each exercise. Because different end points and measurements were used, a percent improvement was calculated for each study in the following exercises: bench press, hand grip, squat, knee extension, other. If results were available on more than one of these exercises, results for the exercise listed first above was used; that is, bench press results were used if available; if not, hand grip results were used, and so on. Four studies (16-19) were excluded from analysis because they did not objectively measure strength. Results in each study were expressed by computing percent improvement in strength for the placebo and for the steroidtreated groups and by taking the difference. The percent improvement was calculated by subtracting the mean initial
strength measurement from the mean final measurement, dividing this result by the mean initial measurement, and multiplying the quotient by 100. In the three studies in which the data for each subject were reported, the percent improvement was calculated in this same way for comparability of results. In four studies, results were given only in graphic form and had to be estimated (27, 32, 39, 41). Four studies were excluded because strength data were not reported (20, 21) or because percent change could not be computed (22, 23). In crossover studies, only the data from the first period were used. Data from the second period in a crossover study are not valid unless responses return to baseline promptly after the end of the first period and there is no carryover effect. Two studies (24, 25) could not be included in the summary because percent improvement could not be computed for the first period. Steroid Dosage Studies varied in the anabolic steroid used, the dosage, and the study length; in order to evaluate dosage effects, we calculated a total anabolic dosage by multiplying the daily dose of the drug times the length of the study (in days) and the anabolic potency compared with testosterone in bioassay. The relative potency factors used were computed as the mean of the values reported by Potts and colleagues (3) and Weller (42): methyltestosterone and mesterolone = 1.15; methandrostenolone = 1.7; stanozolol = 5; oxandrolone = 11.3; nandrolone decanoate = 11.3. Study Characteristics Table 2 shows the design characteristics of each study; the first section of the table details the nine studies (26-34) in which trained athletes were studied, and the second section details eight studies (30, 35-41) using untrained subjects. (One
1 September 1991 • Annals of Internal Medicine • Volume 115 • Number 5
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
389
typically administered for 4 to 6 weeks; one study gave steroids for less than 3 weeks and one study administered steroids for 13 weeks. The methods of randomization and blinding are summarized in Table 2. In two studies, the authors reported only that subjects had been "randomly" assigned and that one group had received "placebo"; only one study (39) reported a test for the success of subject blinding, and only one study assessed subject compliance because the steroid was given by injection (38). Standardized weight-training methods were used in 12 studies, and standardized methods of measuring subjects' gains in strength were used in 16. Eight studies reported the use of controlled protein supplementation; in the other studies, protein supplementation was not mentioned. Even for the studies in trained athletes (usually weightlifters), which used methandrostenolone in all but one case, there were marked differences between studies in the age of subjects, the degree of previous training, the exercise and testing protocols used, and dietary and nutritional status before and during the study. Results
Figure 1. Percent mean improvement of the steroid group compared with that of the placebo group for trained and untrained subjects. Results for each study are plotted using a letter denoting the steroid used. D = methandrostenolone, S = stanozolol, O = oxandrolone, T = methyl testosterone, N = nandrolone decanoate, M = mesterolone. article [30] was treated as two studies in our review because it included studies of both trained and untrained athletes.) The studies are ordered by total anabolic dosage (see Table 3). All subjects studied were men under 40 years old; all but two of the studies using trained athletes studied weightlifters. All the studies are small, the smallest studying two subjects in each group and the largest, 20 subjects in each group; none of the studies justified their sample size choice with power computations or confidence intervals. Only one of the studies mentioned dropouts. All but one of the studies using trained athletes used methandrostenalone, whereas the studies of untrained athletes used six different steroids. Steroids were 390
Table 3 shows the results for each of the studies with the number of subjects completing the study. Results had to be read from figures in four of the studies (27, 32, 39, 41). Of the 16 reports, no statistical tests were used in one study (39); statistical methods were not stated in three studies (31, 33, 41); and the statistical methods used were not clearly described in two (30, 36). Statistical methods appeared to be incorrect in five studies (26-28, 32, 38). Weiss and Muller (26) apparently used the independent samples f-test instead of the paired r-test to compare pretest and post-test results. We obtained results that differed from those of the authors. Ariel (27, 32) compared the regression slopes across weekly testing sessions during control and experimental conditions, but the degrees of freedom reported indicate that week rather than subject was used as the unit of analysis. Ward (28) and Fahey and Brown (38) did not use initial pairing or stratification of subjects in their analyses. Of the five studies in which the statistical methods appear to be correct, discrepancies in the calculations were noted in the two papers by O'Shea (29, 34). The total anabolic steroid dosage ranged from 409 to 7245 mg in testosterone units. Eight of the studies using trained athletes provided data on the bench press, whereas the muscle group for which data were available was not uniform in the studies of untrained athletes. For trained athletes, the percent mean strength increase in the placebo group ranged from 0.2% to 23.3% (median, 5.1%), and the increase in the steroid group ranged from 1.5% to 26.6% (median, 13.2%); differences ranged from 1.2% to 18.7% (median, 5.0%). For untrained athletes, the percent mean strength increase in the placebo group ranged from 2.0% to 15.8% (median, 10.0%), and the increase in the steroid group ranged from - 11.9% to 22.2% (median, 10.6%); differences ranged from - 14.4% to 12.9% (median, 3.4%). Side effects or biochemical evaluations other than those addressing the issue of weight gain were only discussed in seven studies. Summaries across Studies Figure 1 shows the percent mean improvement of the drug-treated group (top) compared with the percent
1 September 1991 • Annals of Internal Medicine • Volume 115 • Number 5
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
Figure 2. Relation of steroid effect to total anabolic dosage in trained athletes. For nine studies in trained athletes, percent mean improvement of the steroid-treated subjects minus percent mean percent improvement of placebo-treated subjects is plotted against the log of the total anabolic dosage calculated (described in text).
mean improvement of the placebo-treated group {bottom) in the largest muscle group evaluated. The dotted line drawn in the separate graphs for trained and untrained athletes represents the line at which the percent improvements in the treated and placebo group are equal; the letter symbol denotes the steroid used. All nine of the studies using trained athletes are above the line of identity that would support the conclusion that there is a significant effect of steroid based on the sign test. This is not the case, however, with the studies of untrained subjects. Because of the diversity in steroids used and in the muscle groups tested as well as the lack of consistency in results, no conclusions about steroid effectiveness in untrained subjects can be reached; the data do not suggest a positive effect. To evaluate further the studies in trained athletes, we plotted the difference in percent improvement between anabolic steroid-treated subjects and placebo-treated subjects against the log of total anabolic dosage (Figure 2). There is no evidence of a positive relation between the total steroid dosage and the effect on strength, although the dosage range is relatively narrow. Figure 3 shows the difference in percent improvement plotted against the number of subjects completing the study in each group; the largest apparent effects of steroid are seen in the smaller studies.
of the changes in each group. This figure is computed using the actual outcome measurements and not percent improvement. Two of the nine studies provided individual data from which the effect sizes could be computed (26, 29); two others reported standard errors of change within each group (28, 34); however, in one of these studies (28), the subjects were paired and the appropriate standard deviation for the effect size (the standard deviation [SD] of differences in paired changes) was not reported. Thus only three of the nine studies of trained athletes provided adequate information to compute effect sizes for a meta-analysis. The estimated effect sizes in SD units were 0.22 for one study (26), 2.3 for another (29), and 1.9 for the third (34). These three studies are heterogeneous, reflecting use of two different steroids and two different muscle groups. We assumed that the estimated effect sizes had an asymptotic normal distribution, and we computed confidence intervals using the formula of Hedges and Olkin (43) for estimation of the asymptotic variance. Figure 4 shows the 95% confidence intervals based on large sample normal theory for the effect sizes of three individual studies and for the three studies combined. The effects for the individual studies were combined using weights inversely proportional to the variance in each study (43); the estimated combined effect was 1.0 (95% CI, 0.49 to 1.5). This analysis indicated that two of the studies showed a significant steroid effect and that the combined results of the three studies were significant at the 0.05 level. Discussion The nine studies of trained athletes, taken at face value, support the claim that anabolic steroids improve
Meta-analysis of Effect Sizes To compare studies with different designs, different outcome measures, and different sample sizes, we computed "effect sizes." In this case, the effect size is the difference between the mean improvement in the steroid group and the mean improvement in the placebo group divided by a pooled estimate of the standard deviation
Figure 3. Relation of steroid effect to study sample size in trained athletes. For nine studies in trained athletes, percent mean improvement of the steroid-treated subjects minus percent mean improvement of placebo-treated subjects is plotted against the total number of subjects in both groups completing the study.
1 September 1991 • Annals of Internal Medicine • Volume 115 • Number 5
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
391
Figure 4. Confidence intervals for study effect sizes. Ninety-five percent confidence intervals based on large sample theory for the three studies in trained athletes that provided enough information to compute individual effect size and the effect size of the three studies combined. The effect size is the difference between the mean improvement in the steroid group and the mean improvement in the placebo group divided by a pooled estimate of the standard deviation of the changes in each group; effect sizes were computed using the actual outcome measurements and not percent improvement. Any confidence interval including zero (denoted by the dotted line) indicates a nonsignificant result.
the strength of trained athletes because slightly greater improvements in strength were found in the anabolicandrogenic steroid-treated group compared with the placebo group in all nine studies. It must be noted, however, that these studies are heterogeneous in design and that our conclusions are restricted to the effects of methandrostenolone on bench press results (with one exception for each). G0tzsche (44) discussed sources of bias in the reporting of double-blind drug trials. In the context of doubleblind, randomized, controlled trials of nonsteroidal antiinflammatory drugs in rheumatoid arthritis, he found considerable evidence of publication bias and of preferential publication of ''positive'' over "negative" results. In reviews, he found reference bias. He also discussed sources of bias in the quality of study design and in the quality of statistical analysis. Many of the sources of bias outlined by G0tzsche (44) are problems in the papers that we reviewed. Although the quality of these studies of anabolic steroids is somewhat hard to assess because of incomplete reporting, there is room for possible bias. In particular, the success of blinding is questionable in a situation in which effects on weight gain, acne, beard growth, and psychologic status make blinding difficult (45, 46). Only two studies included in our summary reported on success of blinding or subject compliance. Improvements were ob392
served in all the placebo groups, an effect previously documented by Ariel and Saville (47). No measures were taken in any study to ascertain whether placebogroup or steroid-group members took black-market steroids during the studies. Previous reviews have divided studies into those yielding statistically significant results compared with those not yielding statistically significant results. Such dichotomies may reflect only minor variations in sample size among studies rather than reflect the probable size of the differences between steroid and placebo. It is important to estimate the size and precision of observed differences and to evaluate the relation to dose and other study characteristics. In addition, basing assessments on the reported "significance" of results is questionable given the frequency of inappropriate statistical methods and computational mistakes. The lack of any apparent dose-response relation brings into question whether steroids have any marked effect in this dosage range. The absence of a doseresponse relation may reflect the narrow dose range and the relatively small total dosages given to the subjects as well as the short duration (generally 6 weeks or less) of the studies. Larger dosages of the steroids or quantities given for a longer period may have resulted in a more easily demonstrable effect. Forbes (48), in an evaluation of the effect of anabolic steroids on lean body mass, found a sharp increase in response at a total anabolic dosage of 113 000 (a bodybuilder taking a total dosage of 10 000 mg of oxandrolone). The maximum total anabolic dosage given trained athletes in the reviewed studies was 7245 mg. It is interesting to note that on the Forbes dose-response curve for lean body mass, this dosage is at the start of the upslope, suggesting that the dosages of steroids used in the published studies were too low to see a dose-response relation, if one does exist. It is questionable how applicable the studies are to the actual practice of drug use in athletes and body builders. Often, several different steroids are taken in amounts that range from 10 to more than 100 times greater than the therapeutic dosages recommended for treatment of medical conditions or that are used in the published studies. Anabolic steroids may also be used in conjunction with other ergogenic aids such as human chorionic gonadotropin, growth hormone, and central nervous system stimulants or depressants. The drugs may be given together or in a staggered regimen and may be administered on a daily basis for a long period or given in a cyclical fashion. Under these circumstances, it is impossible to "dissect out" any results that can be attributed to anabolic steroids from the anecdotal usage reports that purport to show a positive effect of such steroid use. On the basis of these considerations, we conclude that although the data suggest that well-trained athletes may have a greater strength gain while taking anabolic steroids compared with placebos, the data are insufficient to allow any firm conclusion about the efficacy of anabolic steroids in enhancing overall athletic performance. Acknowledgments: The authors thank Frank Shellock, PhD, for help in determining the dominant muscle groups tested by each exercise.
1 September 1991 • Annals of Internal Medicine • Volume 115 • Number 5
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
Requests for Reprints: Glenn D. Braunstein, MD, Department of Medicine, Room Bl 18, Cedars-Sinai Medical Center, 8700 Beverly Boulevard, Los Angeles, CA 90048. Current Author Addresses: Drs. Braunstein and Elashoflfand Ms. Shain: Cedars-Sinai Medical Center, 8700 Beverly Boulevard, Los Angeles, CA 90048. Dr. Jacknow: Southern California Permanente Medical Group, 13652 Cantara Street, Panorama City, CA 91402.
24. 25.
26.
References 1. Kochakian CD, Murlin JR. Effect of male hormone on protein and energy metabolism of castrate dogs. J Nutr. 1935; 10:437-59. 2. Kochakian CD, Murlin JR. The relationship of the synthetic male hormone, androstenedione, to the protein and energy metabolism of castrate dogs and the protein metabolism of a normal dog. Am J Physiol. 1936;117:642-57. 3. Potts GO, Arnold A, Beyler AL. Dissociation of the androgenic and other hormonal activities from the protein anabolic effects of steroids. In: Kochakian CD, ed. Anabolic-Androgenic Steroids. Berlin: Springer-Verlag; 1976:361-406. 4. Bergman R, Leach RE. The use and abuse of anabolic steroids in Olympic-caliber athletes. Clin Orthop. 1985;198:169-72. 5. Beckett AH, Cowan DA. Misuse of drugs in sport. Br J Sports Med. 1979;12:185-94. 6. American College of Sports Medicine. Position statement on the use and abuse of anabolic-androgenic steroids in sports. Med Sci Sports. 1977;9(4):xi-xii. 7. American College of Sports Medicine position stand on the use of anabolic-androgenic steroids in sports. Am J Sports Med. 1984; 12: 13-8. 8. Ryan AJ. Anabolic steroids are fool's gold. Fed Proc. 1981 ;40: 2682-8. 9. Lamb DR. Anabolic steroids in athletics: how well do they work and how dangerous are they? Am J Sports Med. 1984;12:31-8. 10. Haupt HA, Rovere GD. Anabolic steroids: a review of the literature. Am J Sports Med. 1984;12:469-84. 11. Wilson JD. Androgen abuse by athletes. Endocr Rev. 1988;9:181-99. 12. Keul J, Deus W, Kindermann W. Anabole Hormone: Schadigung, Leistungfahigkeit und Stoffwechsel. Med Klin. 1976;71:497-503. 13. Johnson LC, O'Shea JP. Anabolic steroid: effects on strength development. Science. 1969;164:957-9. 14. O'Shea JP, Winkler W. Biochemical and physical effects of an anabolic steroid in competitive swimmers and weightlifters. Nutrition Report International. 1970;2:351-62. 15. Tahmindjis AJ. The use of anabolic steroids by athletes to increase body weight and strength. Med J Aust. 1976;1:991-3. 16. Ariel G, Saville W. Effects of anabolic steroids on reflex components. J Appl Physiol. 1972;2:795-7. 17. Crist DM, Stackpole PJ, Peake GT. Effects of androgenic-anabolic steroids on neuromuscular power and body composition. J Appl Physiol. 1983;54:366-70. 18. Freed DL, Banks AJ, Longson D, Burley DM. Anabolic steroids in athletics: crossover double-blind trial on weightlifters. BMJ. 1975; 2:471-3. 19. O'Shea, JP. Anabolic steroid: effect on competitive swimmers. Nutrition Report International. 1970;1:337-43. 20. Bowers R, Reardon J. Effects of methandrostenolone (Dianabol) on strength development and aerobic capacity. Med Sci Sports. 1972; 4:54. 21. Hervey GR, Hutchinson I, Knibbs AV, Burkingshaw L, Jones PR, Norgan NG, et al. "Anabolic" effects of methandione in men undergoing athletic training. Lancet. 1976;ii:699-701. 22. Fowler WM, Gardner GW, Egstrom GH. Effect of an anabolic steroid on physical performance on young men. J Appl Physiol. 1965;20:1038-40. 23. Johnson LC, Fisher G, Silvester LJ, Hofheins CC. Anabolic steroid:
27. 28. 29.
30. 31.
32. 33.
34.
35.
36. 37.
38.
39.
40.
41.
42.
43. 44. 45. 46. 47. 48.
effect on strength, body weight, oxygen uptake and spermatogenesis upon mature males. Med Sci Sports. 1972;4:43-5. Hervey GR. Are athletes wrong about anabolic steroids? Br J Sports Med. 1975;9:74-7. Hervey GR, Knibbs AV, Burkinshaw L, Morgan DB, Jones PR, Chettle DR, et al. Effects of methandienone on the performance and body composition of men undergoing athletic training. Clin Sci. 1981;60:457-61. Weiss U, Muller H. Zer Frage der Beeinflussung des Krafttrainings Duech Anabole Hormone. Schweizerische Zeitschrift fiir Sportmedizin. 1968;16:79-89. Ariel A. The effect of anabolic steroid upon skeletal muscle contractile force. J Sports Med Phys Fitness. 1973;13:187-90. Ward, P. The effect of an anabolic steroid on strength and lean body mass. Med Sci Sports. 1973;5:277-82. O'Shea JP. The effects of an anabolic steroid on the dynamic strength level of weightlifters. Nutrition Report International. 1971; 4:363-70. Loughton SJ, Ruhling RO.tness. 1977;17:285-96. Stamford BA, Moffatt R. Anabolic steroid: effectiveness as an ergogenic aid to experienced weight trainers. J Sports Med Phys Fitness. 1974;14:191-7. Ariel G. Residual effect of an anabolic steroid upon isotonic muscular force. J Sports Med Phys Fitness. 1974;14:103-11. Golding LA, Freydinger JE, Fishel SS. Weight, size and strength unchanged with steroids. The Physician and Sports Medicine. 1974; 2:39-43. O'Shea JP. A biochemical evaluation of the effects of stanozolol on adrenal, liver and muscle function in humans. Nutr Rep Int. 1974; 10:381-8. Johnson LC, Roundy ES, Allsen PE, Fischer AG, Silvester LF. Effect of anabolic steroid treatment on endurance. Med Sci Sports. 1975; 7:287-9. Win-May M, Mya-Tu M. Effects of anabolic steroids on physical fitness. J Sports Med Phys Fitness. 1975;15:266-71. Casner SW Jr, Early RG, Carlson RB. Anabolic steroid effects on body composition in normal young men. J Sports Med Phys Fitness. 1971;11:98-103. Fahey TD, Brown CH. The effects of an anabolic steroid on the strength, body composition, and endurance of college males when accompanied by a weight training program. Med Sci Sports. 1973; 5:272-6. Samuels LT, Henschel AF, Keys A. Influence of methyl testosterone on muscular work and creatinine metabolism in normal young men. J Clin Endocrinol Metab. 1942;2:649-54. Munson AR. Some effects of anaboic steroids during weight training [Thesis]. School of Education, University of Southern California; 1970. Stromme SB, Meen HD, Aakvaag A. Effects of an androgenicanabolic steroid on strength development and plasma testosterone levels in normal males. Med Sci Sports. 1974;6:203-8. Weller, O. Die androgene Wirkung von la-Methyl-5ot-androstan17/3-ol-3-on im Vergleich zum 17a-Methyl-A4-androsten-17/3-ol-3on. Arzneismittelforschung. 1966;16:465-6. Hedges LV, Olkin I. Statistical Methods for Meta-analysis. Orlando, Florida: Academic Press, Inc.; 1985. Gotzsche PC. Bias in double-blind trials. Dan Med Bull. 1990;37: 329-36. Freinhar JP, Alvarez W. Androgen-induced hypomania. J Clin Psychiatry. 1985;46:354-5. Pope HG Jr, Katz DL. Bodybuilder's psychosis [Letter]. Lancet. 1987;1:863. Ariel G, Saville W. Anabolic steroids: The physiologic effects of placebos. Med Sci Sports. 1972;4:124-6. Forbes GB. The effect of anabolic steroids on lean body mass: the dose response curve. Metabolism, 1985;34:571-3.
1 September 1991 • Annals of Internal Medicine • Volume 115 • Number 5
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
393
• Objective: To assess the effects of anabolicandrogenic steroids on human muscle strength. • Data Sources: A MEDLINE search for the period from January 1966 to April 1990, supplemented by manual searches of previous reviews, produced 30 studies in which subjects received more than one dose of the study steroid and in which changes in muscular strength were measured. • Study Selection: Of the 30 studies, 14 were not included in the detailed data summary because they did not use a placebo control, did not randomize subjects to groups, or did not make objective strength measurements, or because percent change in strength data could not be abstracted. • Data Extraction: Details of study design, reporting of results, and the adequacy and correctness of statistical methods were tabulated. Percent improvement in strength for the largest muscle group studied was computed, using the difference between results for the placebo and for the steroid-treated groups. • Data Synthesis: Previously trained athletes show slightly greater improvements in strength in the anabolic-androgenic steroid-treated group than in the placebo group, with a median difference of 5% across the nine studies (range, 1.2% to 18.7%). A meta-analysis of the three studies with enough information to compute effect size showed a mean difference of 1.0 standard deviations (95% CI, 0.49 to 1.5). However, the poor overall quality of the studies in terms of design, sample size, and analysis; the lack of a dose-response effect across the narrow range of dosages tested; and the tendency for differences to be smaller in the larger studies throw these results into question. No evidence was found to support enhanced muscle strength with steroid use in eight studies in untrained normal volunteers. • Conclusions: Anabolic steroids may slightly enhance muscle strength in previously trained athletes. No firm conclusion is possible concerning the efficacy of anabolic steroids in enhancing overall athletic performance. Results for the low steroid dosages studied in the published reports cannot be generalized to steroid-using athletes taking megadose regimens.
1 he anabolic effects of androgens were first shown in the 1930s by Kochakian and Murlin (1,2). Later studies showed dissociation between anabolic and androgenic effects of synthetic and naturally occurring androgens (3). In the early 1950s, performances of several weightlifting teams improved dramatically, and rumors suggested this was caused by the use of androgens as ergogenic aids. The supposed success of these drugs led many more competitors in the weightlifting and throwing sports to use them in athletic events, including the 1956 Olympic Games in Melbourne, Australia (4). At that time the scientific evidence supporting their use was minimal, and elite athletes used anecdotal reports to support their continued use. By 1976 a number of studies had been conducted, some of which concluded that there was a positive effect and others, that there was no positive effect of steroids on strength in trained athletes. In view of conflicting evidence of their effectiveness, reports of side effects, ethical considerations, and the development of
Table 1. Studies Not Included in the Data Summary and the Reasons for Exclusion Study (Year)
Reference
Study design inadequate Keul et al. (1976) Johnson and O'Shea (1969) O'Shea and Winkler (1970) Tahmindjuis (1976)
12 13 14 15
Methods unclear, no randomization Volunteers received steroid No blinding, three same-order crossovers No controls
Did not measure strength objectively Ariel and Saville (1972) 16 Knee jerk reflex times only Crist et al. (1983) 17 Subjective ratings of strength only Freed et al. (1975) 18 Subject reports of initial strength O'Shea (1970) 19 Swimming speed only Percent improvements could Bowers and Reardon (1972) Hervey et al. (1976) Fowler et al. (1965)
not be computed 20 Strength data not reported 21 22
Johnson et al. (1972)
23
Hervey (1975)
24
Hervey et al. (1981)
25
Annals of Internal Medicine. 1991;115:387-393. From Cedars-Sinai Medical Center-UCLA School of Medicine, Los Angeles, California. For current author addresses, see end of text.
Reasons
Strength data not reported Percent change could not be computed Percent change could not be computed Data for first period only not obtainable Data for first period only not obtainable
©1991 American College of Physicians
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
387
methods to screen for their use, the International Olympic Committee banned anabolic steroids for the first time for the 1976 Olympics in Montreal, Canada (5). In 1977, the American College of Sports Medicine (ACSM) issued a position statement on anabolic steroids based on a survey of the literature. They concluded: "The administration of anabolic-androgenic steroids to healthy humans below age 50 in medically approved therapeutic doses often does not of itself bring about any significant improvements in strength, aerobic endurance, lean body mass, or body weight. There is no conclusive scientific evidence that extremely large doses of anabolic steroids either aid or hinder athletic performance'' (6). This view was not totally accepted by elite athletes, and research on anabolic steroid effects continued into the early 1980s. In 1984, the ACSM revised its previous position and, based on another review of the literature, concluded. 'The gains in muscular strength achieved through high intensity exercise and proper diet can be increased by the use of anabolic-androgenic steroids in some individuals" (7). The paper went on to deplore the use of steroids because of the "ethical principles of athletic competition." During the 1980s, four reviews (8-11) were published; all pointed out that about half the studies of the effect of anabolic steroids on muscular
strength had concluded that there was an effect whereas about half found no significant effects. Three of the reviews (9-11) pointed out many problems in study design, analysis, and reporting. Haupt and Rovere (10) reviewed in detail 24 studies of the effects of anabolic steroids on human strength and numerous studies and case reports of side effects, concluding that "anabolic steroids have their most pronounced effects in those athletes who have trained to the point that they are in a chronic catabolic state." Wilson (11) reviewed in detail 16 studies of muscular strength that appeared "to be adequately designed and blinded" and also reviewed data on side effects; he concluded that no "improvement in strength can be demonstrated consistently when androgens are administered double blind to athletes." All previous reviews have dichotomized studies into those claiming statistical significance and those not claiming statistical significance rather than address the question of size and consistency of observed steroid effects. In view of the importance of the question of steroid effects on muscle strength, we carefully reviewed the available studies and abstracted mean percent improvement in strength for placebo and steroidtreated groups to assess the existence and size of anabolic-androgenic steroid effects on the strength of trained athletes and of untrained healthy young men.
Table 2. Study Characteristics (Studies Ordered by Total Dosage of Anabolic Steroid) Study (Year)
Studies using trained athletes Weiss and Muller (1968) Ariel (1973) Ward (1973) 0'Shea(1971) Loughton and Ruling (1977) Stamford and Moffatt (1974) Ariel (1974) Golding et al. (1974) O'Shea (1974) Studies using untrained subjects Loughton and Ruling (1977) Johnson et al. (1975) Win-May and Mya-Tu (1975) Casner(1971) Fahey and Brown (1973) Samuels et al. (1942) Munson (1970) Stromme et al. (1974)
Refer- Number ence of Subjects
Steroid
Mg per Weeks Randomization Day and Blinding
Training Measurement
Protein Used
2.6 4 4 4
"Random" Double-blind Single-blind* Double-blind*
Yes No Yes Yes
Yes Yes No Yes
No No No Yes
6
Double-blind
Yes
Yes
Yes
4§ 4 9 5
Single-blind|| Double-blind Double-blind*§ "Random"
No Yes No Yes
Yes Yes Yes Yes
Yes No Yes Yes
10/5* 6
6 3
Double-blind Double-blind
Yes Yes
Yes Yes
Yes Yes
5 6
13 6
Double-blind Double-blind
No No
Yes Yes
No No
8 Nandrolone decanoate 11 Methyl testosterone 50/200** 4 4 20 Oxandrolone Mesterolone 75/150tt 8
Double-blind Single-blind Double-blind Double-blind
Yes Yes Yes Yes
Yes Yes Yes Yes
No No No Yes
26 27 28 29
32 6 16 20
Methandrostenolone Methandrostenolone Methandrostenolone Methandrostenolone
10 10 10 10
30t
6
Methandrostenolone
10/5$
31 32 33 34
12 10 40 18
Methandrostenolone Methandrostenolone Methandrostenolone Stanozolol
20 15 10 8
30 35
6 13
Methandrostenolone Stanozolol
36 37
31 27
Methandrostenolone Stanozolol
38 39 40 41
28 4 39 21
* Steroid- and placebo-treated subjects were paired. t Athletes were from a wrestling team. $ Subjects received 10 mg/d for the first 3 weeks and 5 mg/d for the last 3 weeks. § Steroid was given from Monday to Friday only. || Assignment to groups was based on matching rather than on randomization. H The total dosage was 3 mg/kg body weight. ** Subjects received 50 mg/d for the first 3 weeks and 200 mg/d for the last week. tt Subjects received 75 mg/d for the first 4 weeks and 150 mg/d for the last 4 weeks. 388
Standardization
1 September 1991 • Annals of Internal Medicine • Volume 115 • Number 5
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
Table 3. Study Results (Studies Ordered by Total Dosage of Anabolic Steroid) Study (Year)
Studies using trained athletes Weiss and Muller (1968) Ariel (1973) Ward (1973) 0'Shea(1971) Loughton and Ruling (1977) Stamford and Moffat (1974) Ariel (1974) Golding et al. (1974) O'Shea (1974) Studies using untrained subjects Loughton and Ruling (1977) Johnson et al. (1975) Win-May and Mya-Tu (1975) Casner(1971) Fahey and Brown (1973) Samuels et al. (1942) Munson (1970) Strommeetal. (1974)
Reference Number of Subjects
Report of Results
Statistical Methods
Test of Muscle Group
Total Dosage*
Percent Mean Strength Increase Placebo Steroid Difference Group Group
By subject Figure Means ± SE By subject
Incorrect Incorrect Incorrect Correctt
309 476 476 476
Other Bench press Bench press Bench press
0.2 1.4 4.3 8.3
1.5 20.1 13.2 16.9
1.3 18.7 8.9 8.6
6
Means only
Unclear
536
Bench press
23.3
26.6
3.3
31 32 33 34
12 10 40 16
Means only Figure Means only Mean ± SE
Not stated Incorrect Not stated Correctt
3.1 5.1 8.1 6.7
5.5 18.5 9.3 11.8
2.4 13.4 1.2 5.0
30 35
6 13
Means only Means only
Unclear Correct
536 630
11.2 2.0
20.5 2.9
9.3 0.9
36 37
31 27
Means only Means only
Unclear Correct
774 1260
Hand grip Squat
3.7 2.5
6.9 -11.9
3.2 -14.4
38 39 40 41
28 4 32 21
Means only Figure By subject Figure
Incorrect Not done Correct Not stated
2571$ 2818 6328 7245
Bench press Hand grip Squat Squat
11.3 9.2 15.8 10.7
17.3 22.2 5.8 14.2
6.0 12.9 -10.0 3.5
26 27 28 29
32 6 16 18
30
680 714 1071 1400
Bench Bench Bench Bench
press press press press
Bench press Knee extension
* Total anabolic steroid dosage = daily steroid dose times number of days times steroid factor (steroid factors are as follows: methandrostenolone, 1.7; stanozolol, 5; oxandrolone, 11.3; methyl testosterone, 1.15; nandrolone decanoate, 11.3; mesterolone, 1.15). t Although the methods appear to be correct, the calculations do not appear to be correct. t Computed by multiplying the total dosage of 3 mg/kg times an average body weight of 75.8 kg.
Methods Selection of Studies A MEDLINE search was conducted for the period from January 1966 to April 1990 and was supplemented by manual searches of previously published reviews and position papers. We searched for studies that evaluated the use of anabolic steroids in ergogenic activity in which subjects received more than a single administration of the steroid in question and in which muscular strength was measured quantitatively. We included 30 studies of the effect of anabolic steroid use on strength performance in our report (12-41). We eliminated four studies (12-15) because they either did not provide a placebo control or did not randomize subjects to groups (Table 1). We placed no limitations, however, on study size, time interval, dosage, or choice of steroid.
Measures of Steroid Effect Because different exercises were used in the various studies, a kinesiologist determined the dominant muscle groups tested by each exercise. Because different end points and measurements were used, a percent improvement was calculated for each study in the following exercises: bench press, hand grip, squat, knee extension, other. If results were available on more than one of these exercises, results for the exercise listed first above was used; that is, bench press results were used if available; if not, hand grip results were used, and so on. Four studies (16-19) were excluded from analysis because they did not objectively measure strength. Results in each study were expressed by computing percent improvement in strength for the placebo and for the steroidtreated groups and by taking the difference. The percent improvement was calculated by subtracting the mean initial
strength measurement from the mean final measurement, dividing this result by the mean initial measurement, and multiplying the quotient by 100. In the three studies in which the data for each subject were reported, the percent improvement was calculated in this same way for comparability of results. In four studies, results were given only in graphic form and had to be estimated (27, 32, 39, 41). Four studies were excluded because strength data were not reported (20, 21) or because percent change could not be computed (22, 23). In crossover studies, only the data from the first period were used. Data from the second period in a crossover study are not valid unless responses return to baseline promptly after the end of the first period and there is no carryover effect. Two studies (24, 25) could not be included in the summary because percent improvement could not be computed for the first period. Steroid Dosage Studies varied in the anabolic steroid used, the dosage, and the study length; in order to evaluate dosage effects, we calculated a total anabolic dosage by multiplying the daily dose of the drug times the length of the study (in days) and the anabolic potency compared with testosterone in bioassay. The relative potency factors used were computed as the mean of the values reported by Potts and colleagues (3) and Weller (42): methyltestosterone and mesterolone = 1.15; methandrostenolone = 1.7; stanozolol = 5; oxandrolone = 11.3; nandrolone decanoate = 11.3. Study Characteristics Table 2 shows the design characteristics of each study; the first section of the table details the nine studies (26-34) in which trained athletes were studied, and the second section details eight studies (30, 35-41) using untrained subjects. (One
1 September 1991 • Annals of Internal Medicine • Volume 115 • Number 5
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
389
typically administered for 4 to 6 weeks; one study gave steroids for less than 3 weeks and one study administered steroids for 13 weeks. The methods of randomization and blinding are summarized in Table 2. In two studies, the authors reported only that subjects had been "randomly" assigned and that one group had received "placebo"; only one study (39) reported a test for the success of subject blinding, and only one study assessed subject compliance because the steroid was given by injection (38). Standardized weight-training methods were used in 12 studies, and standardized methods of measuring subjects' gains in strength were used in 16. Eight studies reported the use of controlled protein supplementation; in the other studies, protein supplementation was not mentioned. Even for the studies in trained athletes (usually weightlifters), which used methandrostenolone in all but one case, there were marked differences between studies in the age of subjects, the degree of previous training, the exercise and testing protocols used, and dietary and nutritional status before and during the study. Results
Figure 1. Percent mean improvement of the steroid group compared with that of the placebo group for trained and untrained subjects. Results for each study are plotted using a letter denoting the steroid used. D = methandrostenolone, S = stanozolol, O = oxandrolone, T = methyl testosterone, N = nandrolone decanoate, M = mesterolone. article [30] was treated as two studies in our review because it included studies of both trained and untrained athletes.) The studies are ordered by total anabolic dosage (see Table 3). All subjects studied were men under 40 years old; all but two of the studies using trained athletes studied weightlifters. All the studies are small, the smallest studying two subjects in each group and the largest, 20 subjects in each group; none of the studies justified their sample size choice with power computations or confidence intervals. Only one of the studies mentioned dropouts. All but one of the studies using trained athletes used methandrostenalone, whereas the studies of untrained athletes used six different steroids. Steroids were 390
Table 3 shows the results for each of the studies with the number of subjects completing the study. Results had to be read from figures in four of the studies (27, 32, 39, 41). Of the 16 reports, no statistical tests were used in one study (39); statistical methods were not stated in three studies (31, 33, 41); and the statistical methods used were not clearly described in two (30, 36). Statistical methods appeared to be incorrect in five studies (26-28, 32, 38). Weiss and Muller (26) apparently used the independent samples f-test instead of the paired r-test to compare pretest and post-test results. We obtained results that differed from those of the authors. Ariel (27, 32) compared the regression slopes across weekly testing sessions during control and experimental conditions, but the degrees of freedom reported indicate that week rather than subject was used as the unit of analysis. Ward (28) and Fahey and Brown (38) did not use initial pairing or stratification of subjects in their analyses. Of the five studies in which the statistical methods appear to be correct, discrepancies in the calculations were noted in the two papers by O'Shea (29, 34). The total anabolic steroid dosage ranged from 409 to 7245 mg in testosterone units. Eight of the studies using trained athletes provided data on the bench press, whereas the muscle group for which data were available was not uniform in the studies of untrained athletes. For trained athletes, the percent mean strength increase in the placebo group ranged from 0.2% to 23.3% (median, 5.1%), and the increase in the steroid group ranged from 1.5% to 26.6% (median, 13.2%); differences ranged from 1.2% to 18.7% (median, 5.0%). For untrained athletes, the percent mean strength increase in the placebo group ranged from 2.0% to 15.8% (median, 10.0%), and the increase in the steroid group ranged from - 11.9% to 22.2% (median, 10.6%); differences ranged from - 14.4% to 12.9% (median, 3.4%). Side effects or biochemical evaluations other than those addressing the issue of weight gain were only discussed in seven studies. Summaries across Studies Figure 1 shows the percent mean improvement of the drug-treated group (top) compared with the percent
1 September 1991 • Annals of Internal Medicine • Volume 115 • Number 5
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
Figure 2. Relation of steroid effect to total anabolic dosage in trained athletes. For nine studies in trained athletes, percent mean improvement of the steroid-treated subjects minus percent mean percent improvement of placebo-treated subjects is plotted against the log of the total anabolic dosage calculated (described in text).
mean improvement of the placebo-treated group {bottom) in the largest muscle group evaluated. The dotted line drawn in the separate graphs for trained and untrained athletes represents the line at which the percent improvements in the treated and placebo group are equal; the letter symbol denotes the steroid used. All nine of the studies using trained athletes are above the line of identity that would support the conclusion that there is a significant effect of steroid based on the sign test. This is not the case, however, with the studies of untrained subjects. Because of the diversity in steroids used and in the muscle groups tested as well as the lack of consistency in results, no conclusions about steroid effectiveness in untrained subjects can be reached; the data do not suggest a positive effect. To evaluate further the studies in trained athletes, we plotted the difference in percent improvement between anabolic steroid-treated subjects and placebo-treated subjects against the log of total anabolic dosage (Figure 2). There is no evidence of a positive relation between the total steroid dosage and the effect on strength, although the dosage range is relatively narrow. Figure 3 shows the difference in percent improvement plotted against the number of subjects completing the study in each group; the largest apparent effects of steroid are seen in the smaller studies.
of the changes in each group. This figure is computed using the actual outcome measurements and not percent improvement. Two of the nine studies provided individual data from which the effect sizes could be computed (26, 29); two others reported standard errors of change within each group (28, 34); however, in one of these studies (28), the subjects were paired and the appropriate standard deviation for the effect size (the standard deviation [SD] of differences in paired changes) was not reported. Thus only three of the nine studies of trained athletes provided adequate information to compute effect sizes for a meta-analysis. The estimated effect sizes in SD units were 0.22 for one study (26), 2.3 for another (29), and 1.9 for the third (34). These three studies are heterogeneous, reflecting use of two different steroids and two different muscle groups. We assumed that the estimated effect sizes had an asymptotic normal distribution, and we computed confidence intervals using the formula of Hedges and Olkin (43) for estimation of the asymptotic variance. Figure 4 shows the 95% confidence intervals based on large sample normal theory for the effect sizes of three individual studies and for the three studies combined. The effects for the individual studies were combined using weights inversely proportional to the variance in each study (43); the estimated combined effect was 1.0 (95% CI, 0.49 to 1.5). This analysis indicated that two of the studies showed a significant steroid effect and that the combined results of the three studies were significant at the 0.05 level. Discussion The nine studies of trained athletes, taken at face value, support the claim that anabolic steroids improve
Meta-analysis of Effect Sizes To compare studies with different designs, different outcome measures, and different sample sizes, we computed "effect sizes." In this case, the effect size is the difference between the mean improvement in the steroid group and the mean improvement in the placebo group divided by a pooled estimate of the standard deviation
Figure 3. Relation of steroid effect to study sample size in trained athletes. For nine studies in trained athletes, percent mean improvement of the steroid-treated subjects minus percent mean improvement of placebo-treated subjects is plotted against the total number of subjects in both groups completing the study.
1 September 1991 • Annals of Internal Medicine • Volume 115 • Number 5
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
391
Figure 4. Confidence intervals for study effect sizes. Ninety-five percent confidence intervals based on large sample theory for the three studies in trained athletes that provided enough information to compute individual effect size and the effect size of the three studies combined. The effect size is the difference between the mean improvement in the steroid group and the mean improvement in the placebo group divided by a pooled estimate of the standard deviation of the changes in each group; effect sizes were computed using the actual outcome measurements and not percent improvement. Any confidence interval including zero (denoted by the dotted line) indicates a nonsignificant result.
the strength of trained athletes because slightly greater improvements in strength were found in the anabolicandrogenic steroid-treated group compared with the placebo group in all nine studies. It must be noted, however, that these studies are heterogeneous in design and that our conclusions are restricted to the effects of methandrostenolone on bench press results (with one exception for each). G0tzsche (44) discussed sources of bias in the reporting of double-blind drug trials. In the context of doubleblind, randomized, controlled trials of nonsteroidal antiinflammatory drugs in rheumatoid arthritis, he found considerable evidence of publication bias and of preferential publication of ''positive'' over "negative" results. In reviews, he found reference bias. He also discussed sources of bias in the quality of study design and in the quality of statistical analysis. Many of the sources of bias outlined by G0tzsche (44) are problems in the papers that we reviewed. Although the quality of these studies of anabolic steroids is somewhat hard to assess because of incomplete reporting, there is room for possible bias. In particular, the success of blinding is questionable in a situation in which effects on weight gain, acne, beard growth, and psychologic status make blinding difficult (45, 46). Only two studies included in our summary reported on success of blinding or subject compliance. Improvements were ob392
served in all the placebo groups, an effect previously documented by Ariel and Saville (47). No measures were taken in any study to ascertain whether placebogroup or steroid-group members took black-market steroids during the studies. Previous reviews have divided studies into those yielding statistically significant results compared with those not yielding statistically significant results. Such dichotomies may reflect only minor variations in sample size among studies rather than reflect the probable size of the differences between steroid and placebo. It is important to estimate the size and precision of observed differences and to evaluate the relation to dose and other study characteristics. In addition, basing assessments on the reported "significance" of results is questionable given the frequency of inappropriate statistical methods and computational mistakes. The lack of any apparent dose-response relation brings into question whether steroids have any marked effect in this dosage range. The absence of a doseresponse relation may reflect the narrow dose range and the relatively small total dosages given to the subjects as well as the short duration (generally 6 weeks or less) of the studies. Larger dosages of the steroids or quantities given for a longer period may have resulted in a more easily demonstrable effect. Forbes (48), in an evaluation of the effect of anabolic steroids on lean body mass, found a sharp increase in response at a total anabolic dosage of 113 000 (a bodybuilder taking a total dosage of 10 000 mg of oxandrolone). The maximum total anabolic dosage given trained athletes in the reviewed studies was 7245 mg. It is interesting to note that on the Forbes dose-response curve for lean body mass, this dosage is at the start of the upslope, suggesting that the dosages of steroids used in the published studies were too low to see a dose-response relation, if one does exist. It is questionable how applicable the studies are to the actual practice of drug use in athletes and body builders. Often, several different steroids are taken in amounts that range from 10 to more than 100 times greater than the therapeutic dosages recommended for treatment of medical conditions or that are used in the published studies. Anabolic steroids may also be used in conjunction with other ergogenic aids such as human chorionic gonadotropin, growth hormone, and central nervous system stimulants or depressants. The drugs may be given together or in a staggered regimen and may be administered on a daily basis for a long period or given in a cyclical fashion. Under these circumstances, it is impossible to "dissect out" any results that can be attributed to anabolic steroids from the anecdotal usage reports that purport to show a positive effect of such steroid use. On the basis of these considerations, we conclude that although the data suggest that well-trained athletes may have a greater strength gain while taking anabolic steroids compared with placebos, the data are insufficient to allow any firm conclusion about the efficacy of anabolic steroids in enhancing overall athletic performance. Acknowledgments: The authors thank Frank Shellock, PhD, for help in determining the dominant muscle groups tested by each exercise.
1 September 1991 • Annals of Internal Medicine • Volume 115 • Number 5
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
Requests for Reprints: Glenn D. Braunstein, MD, Department of Medicine, Room Bl 18, Cedars-Sinai Medical Center, 8700 Beverly Boulevard, Los Angeles, CA 90048. Current Author Addresses: Drs. Braunstein and Elashoflfand Ms. Shain: Cedars-Sinai Medical Center, 8700 Beverly Boulevard, Los Angeles, CA 90048. Dr. Jacknow: Southern California Permanente Medical Group, 13652 Cantara Street, Panorama City, CA 91402.
24. 25.
26.
References 1. Kochakian CD, Murlin JR. Effect of male hormone on protein and energy metabolism of castrate dogs. J Nutr. 1935; 10:437-59. 2. Kochakian CD, Murlin JR. The relationship of the synthetic male hormone, androstenedione, to the protein and energy metabolism of castrate dogs and the protein metabolism of a normal dog. Am J Physiol. 1936;117:642-57. 3. Potts GO, Arnold A, Beyler AL. Dissociation of the androgenic and other hormonal activities from the protein anabolic effects of steroids. In: Kochakian CD, ed. Anabolic-Androgenic Steroids. Berlin: Springer-Verlag; 1976:361-406. 4. Bergman R, Leach RE. The use and abuse of anabolic steroids in Olympic-caliber athletes. Clin Orthop. 1985;198:169-72. 5. Beckett AH, Cowan DA. Misuse of drugs in sport. Br J Sports Med. 1979;12:185-94. 6. American College of Sports Medicine. Position statement on the use and abuse of anabolic-androgenic steroids in sports. Med Sci Sports. 1977;9(4):xi-xii. 7. American College of Sports Medicine position stand on the use of anabolic-androgenic steroids in sports. Am J Sports Med. 1984; 12: 13-8. 8. Ryan AJ. Anabolic steroids are fool's gold. Fed Proc. 1981 ;40: 2682-8. 9. Lamb DR. Anabolic steroids in athletics: how well do they work and how dangerous are they? Am J Sports Med. 1984;12:31-8. 10. Haupt HA, Rovere GD. Anabolic steroids: a review of the literature. Am J Sports Med. 1984;12:469-84. 11. Wilson JD. Androgen abuse by athletes. Endocr Rev. 1988;9:181-99. 12. Keul J, Deus W, Kindermann W. Anabole Hormone: Schadigung, Leistungfahigkeit und Stoffwechsel. Med Klin. 1976;71:497-503. 13. Johnson LC, O'Shea JP. Anabolic steroid: effects on strength development. Science. 1969;164:957-9. 14. O'Shea JP, Winkler W. Biochemical and physical effects of an anabolic steroid in competitive swimmers and weightlifters. Nutrition Report International. 1970;2:351-62. 15. Tahmindjis AJ. The use of anabolic steroids by athletes to increase body weight and strength. Med J Aust. 1976;1:991-3. 16. Ariel G, Saville W. Effects of anabolic steroids on reflex components. J Appl Physiol. 1972;2:795-7. 17. Crist DM, Stackpole PJ, Peake GT. Effects of androgenic-anabolic steroids on neuromuscular power and body composition. J Appl Physiol. 1983;54:366-70. 18. Freed DL, Banks AJ, Longson D, Burley DM. Anabolic steroids in athletics: crossover double-blind trial on weightlifters. BMJ. 1975; 2:471-3. 19. O'Shea, JP. Anabolic steroid: effect on competitive swimmers. Nutrition Report International. 1970;1:337-43. 20. Bowers R, Reardon J. Effects of methandrostenolone (Dianabol) on strength development and aerobic capacity. Med Sci Sports. 1972; 4:54. 21. Hervey GR, Hutchinson I, Knibbs AV, Burkingshaw L, Jones PR, Norgan NG, et al. "Anabolic" effects of methandione in men undergoing athletic training. Lancet. 1976;ii:699-701. 22. Fowler WM, Gardner GW, Egstrom GH. Effect of an anabolic steroid on physical performance on young men. J Appl Physiol. 1965;20:1038-40. 23. Johnson LC, Fisher G, Silvester LJ, Hofheins CC. Anabolic steroid:
27. 28. 29.
30. 31.
32. 33.
34.
35.
36. 37.
38.
39.
40.
41.
42.
43. 44. 45. 46. 47. 48.
effect on strength, body weight, oxygen uptake and spermatogenesis upon mature males. Med Sci Sports. 1972;4:43-5. Hervey GR. Are athletes wrong about anabolic steroids? Br J Sports Med. 1975;9:74-7. Hervey GR, Knibbs AV, Burkinshaw L, Morgan DB, Jones PR, Chettle DR, et al. Effects of methandienone on the performance and body composition of men undergoing athletic training. Clin Sci. 1981;60:457-61. Weiss U, Muller H. Zer Frage der Beeinflussung des Krafttrainings Duech Anabole Hormone. Schweizerische Zeitschrift fiir Sportmedizin. 1968;16:79-89. Ariel A. The effect of anabolic steroid upon skeletal muscle contractile force. J Sports Med Phys Fitness. 1973;13:187-90. Ward, P. The effect of an anabolic steroid on strength and lean body mass. Med Sci Sports. 1973;5:277-82. O'Shea JP. The effects of an anabolic steroid on the dynamic strength level of weightlifters. Nutrition Report International. 1971; 4:363-70. Loughton SJ, Ruhling RO.tness. 1977;17:285-96. Stamford BA, Moffatt R. Anabolic steroid: effectiveness as an ergogenic aid to experienced weight trainers. J Sports Med Phys Fitness. 1974;14:191-7. Ariel G. Residual effect of an anabolic steroid upon isotonic muscular force. J Sports Med Phys Fitness. 1974;14:103-11. Golding LA, Freydinger JE, Fishel SS. Weight, size and strength unchanged with steroids. The Physician and Sports Medicine. 1974; 2:39-43. O'Shea JP. A biochemical evaluation of the effects of stanozolol on adrenal, liver and muscle function in humans. Nutr Rep Int. 1974; 10:381-8. Johnson LC, Roundy ES, Allsen PE, Fischer AG, Silvester LF. Effect of anabolic steroid treatment on endurance. Med Sci Sports. 1975; 7:287-9. Win-May M, Mya-Tu M. Effects of anabolic steroids on physical fitness. J Sports Med Phys Fitness. 1975;15:266-71. Casner SW Jr, Early RG, Carlson RB. Anabolic steroid effects on body composition in normal young men. J Sports Med Phys Fitness. 1971;11:98-103. Fahey TD, Brown CH. The effects of an anabolic steroid on the strength, body composition, and endurance of college males when accompanied by a weight training program. Med Sci Sports. 1973; 5:272-6. Samuels LT, Henschel AF, Keys A. Influence of methyl testosterone on muscular work and creatinine metabolism in normal young men. J Clin Endocrinol Metab. 1942;2:649-54. Munson AR. Some effects of anaboic steroids during weight training [Thesis]. School of Education, University of Southern California; 1970. Stromme SB, Meen HD, Aakvaag A. Effects of an androgenicanabolic steroid on strength development and plasma testosterone levels in normal males. Med Sci Sports. 1974;6:203-8. Weller, O. Die androgene Wirkung von la-Methyl-5ot-androstan17/3-ol-3-on im Vergleich zum 17a-Methyl-A4-androsten-17/3-ol-3on. Arzneismittelforschung. 1966;16:465-6. Hedges LV, Olkin I. Statistical Methods for Meta-analysis. Orlando, Florida: Academic Press, Inc.; 1985. Gotzsche PC. Bias in double-blind trials. Dan Med Bull. 1990;37: 329-36. Freinhar JP, Alvarez W. Androgen-induced hypomania. J Clin Psychiatry. 1985;46:354-5. Pope HG Jr, Katz DL. Bodybuilder's psychosis [Letter]. Lancet. 1987;1:863. Ariel G, Saville W. Anabolic steroids: The physiologic effects of placebos. Med Sci Sports. 1972;4:124-6. Forbes GB. The effect of anabolic steroids on lean body mass: the dose response curve. Metabolism, 1985;34:571-3.
1 September 1991 • Annals of Internal Medicine • Volume 115 • Number 5
Downloaded From: http://annals.org/ by a University of California San Diego User on 12/15/2016
393
