| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
ORIGINAL ARTICLES |
From the Behavioral Endocrinology Laboratory, Department of Biobehavioral Health (D.A.G.), and Department of Sociology (A.B.), Pennsylvania State University, University Park, PA; and Department of Sociology (D.R.J.), University of Nebraska, Lincoln, NE.
Address reprint requests to: Douglas A. Granger, Department of Biobehavioral Health, 315 Health and Human Development East, Pennsylvania State University, University Park, PA 16802. Email: dag11{at}psu.edu
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSCONCLUSION AND DISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
METHODS: We investigated the models using enumerative measures of cellular and humoral immunity in a sample of 4415 men aged 30 to 48 years who were interviewed and underwent a medical examination.
RESULTS: Analysis revealed positive (and curvilinear) associations between aggressive behavior and enumerative measures of helper/inducer and suppressor/cytolytic T lymphocytes and B lymphocytes. The aggression-immunity relationship was independent of testosterone level, age, current health status, and negative health behaviors and was most pronounced for helper/inducer T cells. There was no evidence of a negative association between testosterone and any immune measure.
CONCLUSIONS: In a large sample of men, individual differences in aggressive behavior were positively associated with enumerative measures of cellular immunity.
Key Words: aggression • cellular immunity • testosterone • health behavior • health status
Abbreviations: CDC = Centers for Disease Control; DSM-III =Diagnostic and Statistical Manual of Mental Disorders,third edition; Ig = immunoglobulin; PBS = phosphate-bufferedsaline; RT = room temperature.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSCONCLUSION AND DISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
One interpretation is that the association between aggressive behavior and immune activity has important adaptive (survival) significance. Aggressiveness is vital for gaining access to females and food, protecting young, battling predators, and fighting conspecies over resources and territory. These activities increase the chances of survival and successful reproduction. On the other hand, aggressive behavior has a high likelihood of leading to trauma, wounds, and exposure to new diseases. In humans, for instance, some activities (eg, foraging, hunting, and battling conspecies) may require travel far from home. Isolated from nursing care and the protection of the home community, individuals would benefit if their immune system more effectively 1) recognizes and mobilizes its components to eliminate potential pathogens, 2) promotes efficient and complete recovery from disease, 3) facilitates repair of tissue damage from wounds, and 4) records immunologic history to be prepared to respond more efficiently if subsequent reexposure occurs.
However, some studies raise an alternative hypothesis. Noteworthy is the apparent difference between the male and female immune systems, with evidence indicating that females are immunologically stronger than males (7, 8). This gender-related immune disparity has been attributed to sociological and behavioral factors, but much of the difference has been ascribed to testosterone, the primary male sex hormone (9–13). Considerable evidence links testosterone to aggression (14, 15), dominance (16, 17), and antisocial behavior (18, 19) in humans. Perhaps individual differences in testosterone are responsible for the relationship between aggressive behavior and immunity.
These observations suggest at least two competing models for the aggressive behavior-immunity relationship. One model posits a positive association stemming from an adaptive mechanism that protects aggressive individuals from an increased risk of exposure to immune stimuli. The other suggests a negative association due to the effects of high testosterone levels. The purpose of this study was to rigorously evaluate the hypothesis that human aggression is positively associated with enumerative measures of cellular and humoral immune activity. If our basic assumption was correct, we expected part of the relationship between aggression and immune activity to be mediated by high rates of health risk behavior and circulating levels of testosterone. It should be noted that high-risk health and sensation-seeking behaviors (eg, alcohol and substance abuse, smoking, and multiple sexual partners) are only indirectly related to aggressiveness but nevertheless increase exposure to a broad range of antigenic challenges, carcinogens, and injuries that might have consequences for the immune system. We also expected that part, but not all, of the association between aggression and immunity would be accounted for by current levels of health and illness because during these episodes the immune system is activated. Therefore, we expected that the relationship between aggression and immune activity would remain after statistically controlling for high health risk behavior, current health status, and testosterone. On the other hand, if our hypothesis was incorrect, the aggression-immunity relationship would be entirely explained by these factors. To evaluate this hypothesis and the competing model, we applied logistic regression analysis to a representative sample of 4415 men with indicators of aggressive behavior (occurring from youth through adulthood) and enumerative measures of cellular and humoral immune activity.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSCONCLUSION AND DISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Comparisons of veterans who took part in the study with those who did not indicated that the men studied were representative of the entire population of veterans with respect to race, age of entry, job assignment, draft status, pay grade, and discharge status (21). The sample was also compared with the 1980 Census population data for men aged 30 to 44 years. The median number of years of education completed by men in this age category was 12.6; the median for the sample was 13.0. The racial composition of the two groups was also similar: 80% white, 9% black, and 6% Latino for the population; 82% white, 12% black, and 5% Hispanic for the sample. In another comparison of men aged 35 to 39 years in 1989, 74% of the population was married, as was 73% of this sample (22). The median age of men in the sample was 37 years. Of course, the sample is not generalizable to men who have never served in the military or who are older or younger than those in the sample or to men who did not qualify for the military for reasons of health, disability, or deferment. Only 47 cases were eliminated because of missing data.
Aggressive Behavior
The measure of aggressive behavior was derived from DSM-III symptoms assessing antisocial personality disorder. Twelve items (see Table 1) measured behavior as a youth and as an adult. The items all loaded on one factor with all coefficients meeting conventional standards for inclusion in the scale. All but two of the loadings (ie, 0.45 and 0.42) were between 0.5 and 0.6. The scale had an 
reliability coefficient of 0.75, and each item lowered the reliability of the scale when omitted. The scale was submitted to log linear transformation to correct a slight positive skew. Although the items were designed to assess aggressive-antisocial behaviors for which there are social prohibitions (eg, being arrested, fighting, using a weapon, being expelled from school, and running away), these acts in a slightly different context are similar to behaviors that have profound survival value (eg, risk-taking, such as a willingness to travel to foreign territory or to engage in conflict with others to defend or gain needed resources). Twenty-two percent reported no aggressive acts; 39%, one to two; 27%, three to five; and 12%, six or more. The median was two acts. As an indicator of validity, we drew on a measure of whether the men had problems in the military that reflected being absent without leave or confined. Confinement is one form of nonjudicial punishment that can be ordered by commanders in response to minor offenses against the Code of Military Justice or in response to violations of other military rules. As expected, this index is correlated 0.20 (p < .001) with the derived aggressive behavior scale.
|
Quality control and assurance.
The Vietnam Experience Study quality-control and -assurance procedures were rigorous and comprehensive (23). Each analytic run contained "bench" and "blind" repeated quality-control samples. Bench controls were used to monitor measurement precision within and among days. Blind repeated controls (5% of total samples) were used to monitor reproducibility within runs. All assays were conducted by certified medical technologists or medical laboratory technicians. Performance criteria for the laboratory procedures that generated the results used here were set by the CDC as interassay coefficients of variation <10% and intraassay coefficients of variation <5%.
Testosterone.
Total testosterone was measured (in ng/dl) in serum or plasma using a standard double antibody radioimmunoassay kit (Leeco Diagnostics, Southfield, MI).
Lymphocyte subsets.
Lymphocyte subsets (B cells, total T lymphocytes, helper/inducer (CD4) cells, and suppressor/cytotoxic (CD8) T cells) were measured (103 cells/mm3) by flow cytometry (24–26) from peripheral blood mononuclear cells isolated using a Ficoll-Hypaque gradient from whole blood treated with sodium heparin. Within 2 to 3 minutes of drawing blood, 5 ml of the blood was transferred into a 15-ml polypropylene tube containing 4 ml of RT RPMI and capped and mixed by inverting the tube several times. Using a bottle-top dispenser and Pasteur pipette assembly, 4 ml of RT Ficoll-Hypaque was slowly underlayered. Next, the mixture was centrifuged for 30 minutes at 450g and 24°C. Supernatant was then aspirated down to about 1 cm above the Ficoll-Hypaque-medium interface. Cells were collected at the interface with a 5-ml pipette and transferred to a separate 15-ml polypropylene tube containing 10 ml of RT RPMI (5%). The mixture was capped and gently inverted several times and then centrifuged for 15 minutes at 450g and 24°C, and the supernatant was aspirated to a final volume of 0.8 ml. Cells were resuspended and transferred to a 12- x 75-mm snap-cap polypropylene tube. One hundred twenty-five microliters of the cell suspension was transferred to a 1.5-ml polypropylene microcentrifuge tube containing 500 µl of PBS and mixed. This tube was transported on ice for automated determination of cell concentration (Coulter counter).
Specimens were stained for fluorescence-activated cell sorter analysis in groups of eight. One 96-well, V-bottom polystyrene microtiter plate (microplate) was used for every four specimens. Thirty-five microliters of cell suspension was transferred into each of five V-bottom wells in the appropriate vertical column of a microplate. The microplates and tubes were supported on a surface of ice. Each microplate was centrifuged for 5 minutes at 650g and 4°C. Plates were then inverted vigorously three to four times to remove the supernatants. Cells remained trapped by capillary forces at the bottom of the well in a volume of approximately 5 µl. Cells were resuspended by strumming the bottoms of the wells with a fingernail and placed on ice. Five microliters of 10 µg/ml IgG2a (control), OKT3, OKT4, OKT8, and CCB1 antibodies were added to wells in rows 1, 3, 5, 7, and 9, respectively. Plates were covered with a plastic lid, and cells and antibodies were mixed by gentle agitation. The microplates were incubated for 30 minutes at 4°C in the dark. Using an eight-channel pipette, cells were resuspended, one row at a time, in 200 µl of ice-cold suspension buffer by pipetting up and down multiple times. The suspensions were transferred to 500-µl polypropylene microcentrifuge tubes suspended in trays and supported on ice. Samples were stored on ice and protected from light before flow cytometric analysis.
The flow cytometer was allowed to warm up for at least 2 hours. Wattage was set at 500 MW; amperage, at 25 A; and red-green fluorescence subtraction, at 12%. The alignment standard was prepared by adding 20 µl of fluorescent microspheres to 500 µl of PBS. Alignment standards were run using the BEADS 5610 program, and adjustments were made accordingly. "Beads" were run every 20 samples to check the alignment. Using an Eppendorf repeater pipette, 20 µl of 10x propidium iodide solution was added to one microplate of cell preparations. Samples were incubated at RT for 5 minutes to permit nonviable cells to take up stain and were then returned to ice.
"Relative" results were read from the flow cytometer and were reported as percentages. Absolute values (in 103 cells/mm3) were calculated as follows: 
|
|
|
|
|
|
|
|
|
|
Immunoglobulin levels.
Immunoglobulin levels (IgG, IgA, and IgM) in serum were measured (in mg/dl) using the Beckman Immunochemistry System. A nephelometer was used to measure the rate of light scatter formation from an immunoprecipitin reaction. Anti-IgG, anti-IgA, and anti-IgM, when brought into contact with their respective antigens in solution, produce a peak rate signal proportional to the increase in light scatter produced by the antigen-antibody reaction. The optimum sample size is 250 µl. Freshly drawn serum from fasting subjects was preferred, but some samples were stored at 2 to 8°C for up to 24 hours.
The following reagents and materials were used (27): buffer, PBS with a polymer enhancer and 0.1% sodium azide as a preservative; diluent, PBS with 0.1% sodium azide as a preservative; and calibrator, 3.0 ml of processed human serum containing IgG, IgA, and IgM at concentrations near the midpoints of the measured test results (serum contained 0.1% sodium azide as a preservative; antiserum to human IgG, IgA, and IgM with 0.1% sodium azide as a preservative; and a blue dye that signals the analyzer that the antigen-antibody reaction has been started). The analyzer used a single calibrator concentration for each protein; this concentration was defined by using constants provided by an antibody card, the target value of the calibrator serum on the calibrator card, and the appropriate dilution of calibrator serum. The constants provided by the antibody card are values that represent the concentration/rate relationship throughout the measuring range. The system uses these data to characterize a curve-fitting formula that yields a mathematically unique curve for each antibody. The system is calibrated by testing a single protein concentration contained in a specific dilution of calibrator serum in duplicate. The peak rate signal obtained during calibration (the raw calibration value) is used to establish a ratio to the peak rate expected on the basis of the assigned calibrator serum target value. The calibration factor is used to adjust the analyzer electronics system gain so that the raw calibration values equal the target value specified on the calibrator card. All subsequent sample rate signals are similarly adjusted. To ensure a valid calibration, the analyzer electronics system requires that the peak rate measurement obtained during calibration be reproduced within a predefined percentage (typically ±5%) in two of two measurements or two in a series of three measurements. When the first and second peak rate values fall within the required range, the two values are averaged, and the peak rate signal is internally adjusted so that the calibrator will read at its target value. Results were reported in whole numbers (in mg/dl).
Health Behaviors and Health Status
Behaviors known to place the individual at risk of disease and injury, other than aggression itself, included number of sexual partners (
10 different partners in 1 year = 1, <10 partners in 1 year = 0), number of alcoholic drinks per day, number of cigarettes per day, and use of marijuana, cocaine, or heroin within the past 12 months. We would have preferred to have number of sexual partners as a continuous variable, but the data were originally collected using the two categories.
Five measures of health status assessed specific and general conditions and spanned both physical and mental problems. Sexually transmitted diseases measured whether the individual had contracted gonorrhea, syphilis, or both. Number of colds in 1 year assessed respiratory ailments in general. Physical trauma was captured by three items about broken bones, head injuries, and involvement in vehicular accidents. We selected these from a small pool of items that measured trauma or a source of trauma that all loaded on a single factor with coefficients of 0.5 or above. Depression was assessed using a 13-item scale derived from the DSM-III assessing affective disorders involved with appetite, sleep, fatigue, restlessness, interest in sex, feelings of worthlessness, difficulty thinking, and thoughts about death. All of these items loaded on a single factor. Ten of the 13 items had loadings of 0.5 to 0.7, and the remaining three were above 0.45. The scale had a reliability coefficient of 0.80. The final item asked respondents to rate their health as excellent, good, fair, or poor. This global question detects general self-perceptions of health not assessed by the other, more specific measures and is highly correlated with physical assessments (28).
Data Analyses
First we examined zero-order relationships. Then, using logistic regression, we estimated the percentage of change in the odds that men would be in the top quartile of cell number for each immune marker by number of aggressive acts. We estimated the percentage of change for men reporting no aggressive behavior vs. those reporting two aggressive acts, those reporting three vs. five aggressive acts, and those reporting six vs. nine aggressive acts. For each immune marker we computed two equations. The first controlled for age and testosterone and thus specified the basic relationship between aggression and each immune marker. In the second equation, we also controlled the health risk behavior and health condition variables.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSCONCLUSION AND DISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
|
|
B cell counts.
The analyses for B cells revealed a similar pattern. Two aggressive acts increased the odds of being in the top quartile by 65% over those reporting no aggressive behaviors. The odds dropped to 15% for those reporting five acts compared with three and to 8% for those reporting eight acts in comparison to six. Taking into account risk behavior and poor health reduced the odds of being in the top quartile by 40%, but the link to aggression remains strong despite this overall shift. Figure 1 depicts the curvilinear relationships between aggressive behavior and probability of being in the top quartile of CD4, CD8, and B cell counts.
|
|
CONCLUSION AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSCONCLUSION AND DISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
It may be noteworthy that the aggression-immunity association is strongest for the numbers of CD4 cells and B lymphocytes. To a large extent this subset of lymphoid cells determine the initiation, magnitude, and duration of specific cellular immune responses. For instance, the functional responsibilities of CD4 cells include the activation of B cells, secretion of proteins that regulate intercellular communications among T lymphocytes, and potentiation of the cytolytic activity of natural killer cells. CD4 cells facilitate both humoral and cellular branches of specific immunity and thus affect the immune response to both extracellular (eg, bacteria) and intracellular (eg, malignant or virally infected cells) antigens. The functional attributes of B cells include specific antigen recognition, antigen presentation, antibody production, and maintenance of immunologic memory. Although these findings focus attention on aspects of specific cell-mediated immunity (6), there is scant empirical evidence to suggest which of the many physiological pathways possibly responsible for the association should receive priority research attention. Even among the best-developed animal models, the source of the link between aggression and immunity remains unknown. Research focused on elaborating the nature and sources of the apparent aggression-lymphocyte link would seem to be the next logical step to begin the process of ruling out the many alternative hypotheses.
It is of interest that this analysis failed to reveal evidence of a negative association between testosterone and enumerative measures of immunity. Only a handful of significant relationships with circulating levels of testosterone were observed. When detected, the effect sizes were very small and in the opposite direction of that described in most of the literature (7–9). Moreover, these positive associations between testosterone and immune measures were mediated largely by the association of testosterone with risky health behaviors and practices (4). The exception is CD4 cell counts. In men with testosterone levels below 800 ng/dl, we observed a weak positive association with numbers of helper/inducer T cells. This pattern of findings seems to contradict what has been reported in the literature, but closer analysis reveals that it may be consistent with Barnard et al.’s "adaptive modulation hypothesis" (30, 31), which proposes that immunity is one fitness component among many competing for resources (eg, growth, neural integration, and sexual development). In this model, steroid hormones are mediators of the level of physiological investment in different fitness components through their effects on different biobehavioral processes. Barnard et al.’s work casts an interesting interpretation of these findings. That is, testosterone may be positively related to lymphoid cell numbers in this sample of men so additional testosterone-dependent "immunosuppression" (associated with aggression) is avoided. Additional in vivo studies are needed to clarify the complex nature of the relationship between testosterone, behavior, and immune activity in humans.
In conclusion, the results of this study demonstrate that individual differences in human aggressive behavior are related to differences in enumerative measures of cell-mediated immunity. These findings extend previous findings of work in nonhuman primates (6) and rodents (32–35) for the first time (to our knowledge) to humans. This study had noteworthy methodological strengths (eg, substantial statistical power afforded by the sample size of 4462 men) that strongly support the reliability and validity of its findings. However, these unique strengths undeniably placed practical constraints on the level of analysis that could be used to operationalize the immune variables. Although the findings raise several possible new avenues of research into the complex interactions among social behavior, evolution, and immunity in humans, interpretation of the applied and theoretical significance of these data must be qualified by the sole reliance on enumerative, as opposed to functional, measures of immunity. Also complicating the interpretation of these observations are the results of studies in nonhuman primates, which suggest that the association between aggression and immunity may be attributable to tertiary (spurious) factors such as status (ie, rank or dominance) and social context (ie, social instability), which are known to influence rates of aggressive behavior, testosterone levels, immunocompetence, and illness susceptibility (5, 6, 36, 37). Despite these limitations, we expect the present findings to serve as a useful heuristic for the design of more focused explorations of the possible link between aggression and immunity in the next generation of studies.
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSCONCLUSION AND DISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
Received for publication January 21, 1998.
Revision received January 5, 2000.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSCONCLUSION AND DISCUSSIONACKNOWLEDGMENTSREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  R. McDermott, D. Johnson, J. Cowden, and S. Rosen Testosterone and Aggression in a Simulated Crisis Game The ANNALS of the American Academy of Political and Social Science, November 1, 2007; 614(1): 15 - 33. [Abstract] [PDF]
|
|
|
|
 |
|
 J. K. Kiecolt-Glaser, L. McGuire, T. F. Robles, and R. Glaser Psychoneuroimmunology and Psychosomatic Medicine: Back to the Future Psychosom Med, January 1, 2002; 64(1): 15 - 28. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
