Article Text

Extended report
Association between testosterone levels and risk of future rheumatoid arthritis in men: a population-based case–control study
  1. Mitra Pikwer1,
  2. Aleksander Giwercman2,
  3. Ulf Bergström1,
  4. Jan-Åke Nilsson1,
  5. Lennart T H Jacobsson1,3,
  6. Carl Turesson1
  1. 1Section of Rheumatology, Department of Clinical Sciences, Lund University, Malmö, Sweden
  2. 2Reproductive Medicine Centre (RMC), Skåne University Hospital, Lund University, Malmö, Sweden
  3. 3Department of Rheumatology and Inflammation Research, The Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden
  1. Correspondence to Dr Mitra Pikwer, Department of Rheumatology, Lund University, Skåne University Hospital, Malmö SE-205 02, Sweden; Mitra.pikwer{at}


Objectives Rheumatoid arthritis (RA) is less common among men than women, and sex hormones have been suggested to play a part in the pathogenesis. Lower levels of testosterone have been demonstrated in men with RA, but it is not known if these changes precede the disease.

Methods In a nested case–control study, using information and blood samples from a population-based health survey, we identified incident cases of RA by linking the cohort to local and national RA registers. Two controls for each validated case, matched for age, sex and year of screening, were selected from the health survey. Using stored blood samples, collected between 08:00 and 10:00 am after an overnight fast, we analysed levels of testosterone and other reproductive hormones.

Results Serum was available from 104 cases (median time from screening to RA diagnosis 12.7 years (range 1–28); 73% rheumatoid factor (RF) positive at diagnosis or later) and 174 matched controls. In conditional logistic regression models, adjusted for smoking and body mass index, lower levels of testosterone were associated with subsequent development of RF-negative RA (OR 0.31 per SD, 95% CI 0.12 to 0.85), with a weaker association with RF-positive RA (OR 0.87 per SD; 95% CI 0.53 to 1.43). Levels of follicle-stimulating hormone were significantly increased in pre-RF-negative RA (p=0.02), but decreased in pre-RF-positive RA (p=0.02).

Conclusions Lower levels of testosterone were predictive of RF-negative RA, suggesting that hormonal changes precede the onset of RA and affect the disease phenotype.

  • Rheumatoid Arthritis
  • Epidemiology
  • Treatment
View Full Text

Statistics from


Risk factors for rheumatoid arthritis (RA) include genetic,1 environmental2 and hormonal factors.3 Autoimmunity affects men to a lesser extent than women,4 and, during the fertile years, RA has a female/male incidence ratio of 4–6 : 1.5 With increasing age, the sex difference in incidence narrows.6 In cross-sectional studies, lower levels of serum testosterone have been found in both male and female patients with RA compared with healthy controls,7–10 and low levels of testosterone have been observed in synovial fluids in both female and male patients with active disease.11 Proinflammatory cytokines are known to stimulate the hypothalamic–pituitary–adrenal axis but suppress the hypothalamic–pituitary–gonadal (HPG) axis, suggesting that low testosterone levels might be a consequence of the inflammatory disease.12 Alternatively, the measured low testosterone levels may reflect a role for androgens in the pathogenesis of RA, as suggested by the age-related sex differences in the epidemiology of the disease.

A large prospective study of women did not find any association between androgen levels measured at a single time point or polymorphism in the sex hormone receptors and risk of RA. The only prospective study on men, based on 32 incident male RA cases from Finland, did not show any significant differences in testosterone levels before RA onset compared with controls.13 However, this study did not adjust for potential confounders and had limited power for stratification by different phenotypes of RA.13 Taken together, there are limited data on the importance of androgen levels for the development of RA, in particular in men. Based on the literature, testosterone is of particular interest, but other hormones such as follicle-stimulating hormone (FSH) and luteinising hormone (LH) are necessary to interpret the origin of differences in testosterone. The interplay of these hormones in the HPG axis is illustrated in figure 1. Data on sex hormone-binding globulin (SHBG) are needed to calculate free testosterone levels.

Figure 1

A schematic overview of the hypothalamus–pituitary–gonadal (HPG) axis and the impact of testosterone on immune cells. FSH, follicle-stimulating hormone; GnRH, gonadotropin releasing hormone, LH, luteinising hormone.

The aim of this study was to measure testosterone and other sex hormones in a larger sample of men who subsequently developed RA, to investigate if differences in hormone concentrations from matched controls could be detected years before diagnosis and if such patterns differed between subtypes of RA.

Patients and methods

Source population: the Malmö Preventive Medicine Program

Between 1974 and 1992, the Malmö Preventive Medicine Program (MPMP), a preventive case-finding programme, was conducted in Malmö, Sweden (population 235 000 in 1974).14 The programme included a total of 22 444 males born between 1949 and 1921 and 10 902 females born between 1938 and 1925. The aim of the health survey was to screen large strata of the adult population in order to identify individuals for preventive intervention. The overall attendance rate was 71.2%. The vast majority of participants were Caucasians of Scandinavian origin. The subjects underwent physical examination including height and weight measurements and laboratory tests, and completed a self-administered questionnaire on health and lifestyle factors.15 The subjects were invited to leave blood samples in the morning, between 08:00 and 10:00 am, after an overnight fast. The samples were stored at −20°C.16

Selection of cases and controls

In a previous survey,17 we identified individuals who developed RA after inclusion in this cohort and up to 31 December 2004, by linking the MPMP register to a community-based RA register,18 ,19 the local outpatient clinic administrative register for Malmö University Hospital, the National Hospital Discharge Register and the National Cause of Death Register.17 The community-based RA register has been shown to include more than 90% of patients in the catchment area.19 The Swedish national inpatient register includes more than 99% of all hospital discharges and has a high validity for RA and many other diagnoses.20

In a structured review of all medical records, possible cases were validated and classified according to the 1987 American College of Rheumatology criteria for RA.21 Four controls for each validated case, matched for sex, year of birth and year of screening, who were alive and free from RA when the index person was diagnosed with RA, were selected from the MPMP cohort. Vital status and information on emigration were retrieved from the national census, and controls who were not alive or living in Sweden through the index date were excluded.

For the present study, serum samples from two controls per case were collected from the MPMP bio bank. For various reasons, samples were missing from a subset of cases as well as controls. For cases with available serum, the retrieval of control samples was extended to include the two remaining matched controls, when such were available.

This study was approved by the regional research ethics committee for southern Sweden. All participants gave their informed consent to be included in the MPMP and the Malmö RA register, respectively. No informed consent was obtained specifically for the present study.

Socioeconomic background and comorbidities

Data on socioeconomic status were derived from self-reported job titles in the Swedish national censuses, as previously described.17 Briefly, occupations were coded and converted into standardised social class categories, and subjects were classified as ‘blue-collar workers’ (manual workers, both skilled and unskilled), ‘white-collar workers’ (non-manual employees and self-employed professionals) and ‘others’. Housewives, students and unemployed without any other self-reported job title during the study period were excluded from this classification.22 Smoking was identified as current smoking versus current non-smoking (ie, never smoked or past smoking).

Data on self-reported overall health and self-reported cancer, diabetes and cardiovascular disease (the latter classified as self-report of hospitalisation for stroke, physician diagnosis of angina pectoris, or current use of heart medication) at baseline were extracted from the self-administered questionnaire.

Laboratory tests

Serum total testosterone, SHBG, LH and FSH concentrations were quantified by ElectroChemiLuminiscence Immunoassay based on a ruthenium derivative according to routine methods used at the Department of Laboratory Medicine, Skåne University Hospital. Free testosterone was calculated from total testosterone and SHBG levels using the Vermeulen formula.23 The detection limits for testosterone, SHBG, LH and FSH were 0.0087 nmol/l, 0.35 nmol/l, 0.10 IU/l and 0.10 IU/l, respectively. Imprecision levels for low and high levels were 2.4% and 4.0% for testosterone, 1.0% and 1.1% for SHBG, 2.0% and 2.2% for LH, and 3.3% and 2.2% for FSH, respectively. The erythrocyte sedimentation rate (ESR) was measured at screening according to the standard Westergren method.

Data on rheumatoid factor (RF) tests were collected from the databases of the two clinical immunology laboratories in the area.

Statistical analysis

The impact of baseline hormone levels on the risk of RA was examined by bivariate conditional logistic regression analysis, taking into account the matched design of the study. For comparability of the impact of hormones with different concentrations, ORs for RA were calculated per SD of testosterone, FSH, LH and SHBG. Potential confounders were examined in a similar manner. Correlations between body mass index (BMI) and testosterone were examined using Pearson's test. Multivariate logistic regression analysis was used to adjust for potential confounders. Analyses were stratified by RF status at diagnosis or later (ever positive vs negative) and also by time from screening to RA diagnosis (above vs below the median). Statistical significance was set at p<0.05 (two-sided test).


Cases and controls

As previously reported, 151 male cases of incident RA were identified.17 For the present study, stored serum was available from 104 male subjects who subsequently developed RA and 174 matched controls. RF status at diagnosis or later was available for 83 patients. Age at screening (mean age 45 vs 46 years), age at RA diagnosis (mean 59 years in both groups) and the proportion of blue-collar workers (52% vs 54%) were similar in this subset with known RF status and in the entire population of incident male RA cases, respectively. Characteristics of pre-RA cases and controls are described in table 1. Patients who developed RF-negative RA were older at screening (mean 48 vs 45 years) and at RA diagnosis (mean 64 vs 58 years) than those who developed RF-positive RA. There were no substantial differences in ESR between cases and controls (table 1). Comorbidities and self-reported health status were similar in cases and controls (table 1). Comparison between those with and without available sera showed similar frequencies of full self-reported health among both cases (76% vs 70%) and controls (70% vs 74%). Age at screening in cases and controls and age at diagnosis in cases were similar in those with and without serum available (data not shown).

Table 1

Baseline characteristics of pre-RA cases and controls, stratified by RF status at diagnosis or later in the cases

Confounders: smoking and BMI

Mean BMI in pre-RA cases was lower than in controls (table 1), and there was a negative correlation between BMI and testosterone levels (r=−0.46; p<0.001) as well as free testosterone levels (r=−0.36; p<0.001), but not between BMI and other hormones (SHBG, LH and FSH). As previously reported,17 smoking was associated with RA (table 1), and smokers had higher levels of all measured hormones (testosterone: mean 23.0 vs 20.2 nmol/l, p=0.001; free testosterone: mean 0.44 vs 0.41 nmol/l, p=0.01; FSH: mean 7.10 vs 5.86 IU/l, p=0.06; LH: mean 7.72 vs 5.06 IU/l, p=0.05; SHBG: mean 38.4 vs 32.6 nmol/l, p<0.01; for levels in all cases and controls, see table 2). In bivariate logistic regression analysis, smoking was associated with increased risk of RF-positive RA, and lower BMI tended to be associated with, in particular, RF-negative RA (table 3). Blue-collar worker status was also a potential confounder, being associated with both RA (table 1), as previously reported,17 and a tendency towards higher serum levels of testosterone (mean 22.2 vs 20.9 nmol/l in white-collar workers; p=0.17).

Table 2

Hormone levels in all cases and controls, stratified by RF status in the case at diagnosis or later, and by median time from inclusion to RA diagnosis

Table 3

Unadjusted associations between levels of hormones, smoking, BMI and the risk of RA

Hormone levels and the risk of RA

Cases tended to have lower mean concentrations of all measured hormones except SHBG compared with controls, with the main differences observed among those with a longer time from screening to diagnosis (table 2). Individuals who developed RA >12.7 years after screening were older (mean 65 vs 54 years) and more likely to be RF negative (65% vs 35%) at diagnosis.

In bivariate analyses, without adjustment for potential confounders, there was a trend towards a negative association between total and free testosterone levels and RF-negative RA. There was a statistically significant positive association between FSH levels and risk of RF-negative RA (table 3).

In multivariate logistic regression analysis, adjusted for BMI and smoking, there was a trend towards a negative association between testosterone levels (both total and free testosterone) and subsequent development of RA, which was statistically significant for RF-negative cases (table 4).

Table 4

Associations between levels of measured hormones and risk of RA adjusted for smoking and/or BMI

Serum levels of FSH were positively associated with future development of RF-negative RA and negatively associated with future RF-positive RA, in both bivariate (table 3) and multivariate analysis, adjusted for smoking (table 4). When stratified by smoking or median BMI, similar associations between hormone levels and RA were seen in all strata (data not shown).

To exclude nascent inflammation in pre-RA cases as an explanation for the differences in testosterone levels, we performed an additional analysis, limited to the cases in which RA diagnosis was >5 years after screening (n=89; 21 RF negative; 51 RF positive; 17 unknown) and their controls, with similar results for testosterone (adjusted ORs with 95% CIs for all cases 0.82 (0.55 to 1.21), for RF-negative RA 0.31 (0.11 to 0.84), and for RF-positive RA 0.89 (0.53 to 1.50)) and for the other hormones (data not shown).

When further adjustment for socioeconomic status was performed, point estimates were similar, with statistically significant negative associations between RF-negative RA and total or free testosterone (OR (95% CI) for total testosterone for all cases 0.75 (0.51 to 1.10), for RF-negative RA 0.28 (0.10 to 0.78), and for RF-positive RA 0.84 (0.51 to 1.39); OR (95% CI) for free testosterone for all cases 0.75 (0.52 to 1.10), for RF-negative RA 0.33 (0.12 to 0.88), and for RF-positive RA 0.82 (0.50 to 1.33)).


This study demonstrated a negative association between levels of testosterone and free testosterone and subsequent development of RF-negative RA in men. There were also differences in gonadotropin levels between pre-RA cases and controls, with distinct and different patterns for pre-RF-negative (higher FSH) and pre-RF-positive RA (lower FSH).

With the exception of a small study from Finland, by Heikkila et al,13 this is the first study to explore hormone levels in men before RA diagnosis.

Low testosterone could be a consequence of primary testicular dysfunction, a primary dysfunction in the hypothalamus–pituitary part of the HPG axis, but also a result of inflammation, which may act on both the central and peripheral portion of the HPG axis to reduce testosterone production.24 However, the low ESR levels in the present sample, and the lack of differences in ESR or self-reported health status between pre-RA cases and controls, do not indicate inflammation due to early arthritis as the main explanation for the observed differences. Furthermore, when men who developed RA within 5 years of screening were excluded, similar results were obtained.

The associations between smoking, BMI and hormone levels in the present study are supported by previous findings.25 ,26 Smoking is a well-known risk factor for RA,27 and a recent case–control study demonstrated an association between low BMI and seronegative RA in men.28 The rationale for adjusting for smoking and BMI in the multivariate analysis is therefore not only based on data from the present sample, but is also compatible with the literature.

Testosterone has been proposed to have anti-inflammatory functions, by suppressing both the cellular and humoral immune system (figure 1).29 Male sex has been found to be an independent predictor of remission in early RA.30 This has led to the hypothesis that androgen supplementation may be useful in the management of RA. Testosterone as treatment for RA has, however, so far been investigated in limited samples, with different results.31 ,32

It has been suggested that patients with RA have an altered hypothalamic–pituitary–adrenal axis with inappropriately low cortisol and gonadotropin levels compared with that expected given the state of inflammation.33 A Swedish case–control study of men with early RA diagnosis identified lower testosterone levels in cases, and noted that older men with RA (>50 years old) had low LH levels despite low testosterone, indicating a central genesis to the deficiency.34 Furthermore, they showed a negative correlation between testosterone levels and disease activity.35 In our study, LH levels were also lower in pre-RA cases than controls despite lower testosterone, although the differences for LH did not reach statistical significance. We also identified discrepancies in FSH levels between RF-negative and RF-positive RA, indicating different patho-mechanisms behind the two subtypes of the disease. FSH levels were higher before onset of RF-negative RA, indicating an increased hypothalamus–pituitary response to testicular dysfunction. In contrast, lower FSH levels were found before onset of RF-positive RA, implicating an impaired hypothalamus–pituitary function as a possible factor in the pathogenesis of RF-positive RA. It has previously been suggested that chronic stress may contribute to the pathogenesis of RA through hormonal and neuroimmunomodulatory mechanisms.11 ,36 However, in a previous study of women, we did not find self-reported stress to be a predictor of RA.3 This area should be further studied.

RA is a heterogeneous disease, which may be subdivided into different phenotypes, of which RF status is the most widely recognised and considered a stable phenotype based on follow-up studies.37 Predictors may differ between the sexes, geographical area38 and phenotypes. For instance, smoking is a predictor of RF-positive disease, with a stronger impact in men,27 whereas in women, age at menopause and breast-feeding history predominantly affect the risk of RF-negative disease with onset after age 45.3 ,39 In the present study, the average age at diagnosis of RF-negative RA was substantially higher than of RF-positive disease. Taken together, these results suggest that changes in sex hormones mainly influence onset of RF-negative RA in older individuals of both sexes. Lower levels of testosterone in men, and early menopause in women, may be signs of premature aging,3 putting them at greater risk of developing not only RA but also other diseases, as reduced levels of adrenal hormones have been seen in older patients with polymyalgia rheumatica.33 This is of particular interest because of the clinical overlap between polymyalgia rheumatica and elderly-onset seronegative RA.

This is a unique material of fasting morning blood samples obtained from a relatively large number of men who subsequently developed RA. Limitations include the fact that samples were stored for a number of years before analysis. However, this is unlikely to have had a systematic effect on our comparison, since cases and controls were matched for year of screening. Furthermore, owing to missing blood samples and missing data on RF status in a subset of the cases, statistical power is limited for some of the subanalyses.

Strengths include the population-based approach, and the fact that we used a validated local register, together with the Swedish national inpatient register. Previously reported incidence estimates indicate that we captured virtually all incident RA cases during the study period,17 suggesting that our cases are representative of RA cases in the community. On the other hand, the cases were mainly Caucasians of Scandinavian heritage, and the results may not apply to other ethnic groups or other geographic settings.

A further strength is the availability of data on confounders and comorbidities from the health survey. Our analyses indicate that differences in comorbidities and socioeconomic status did not explain the associations between baseline hormone levels and future risk of RA. However, unmeasured confounding by other exposures, such as alcohol use or diet, cannot be excluded.

In conclusion, we report a negative association of testosterone and free testosterone levels with the risk of developing RF-negative RA in men. Since this is the first major study of testosterone and related hormones in the preclinical phase of RA, our findings should be verified in other populations.


The authors would like to thank Göran Berglund and collaborators in the MPMP steering committee for excellent advice on the study, Anders Dahlin and Håkan Andersson for excellent help with data extraction from the MPMP database, Yvonne Bengtsson, Lisbet Andreassen and Charlotte Becker from the Department of Laboratory Medicine, Skåne University Hospital, for their work with analyses of testosterone, FSH, LH and SHBG, Yvonne Giwercman for helpful comments on the study, and Gunnel Henriksson and Lennart Truedsson for providing information from the clinical immunology databases on RF tests.


View Abstract

Supplementary materials

  • Press release

    Press release

    Files in this Data Supplement:

  • Press release

    Press release

    Files in this Data Supplement:


  • Handling editor Tore K Kvien

  • Contributors All authors contributed to the conception, design, analysis and interpretation of data. All authors participated in drafting the manuscript and revising it critically for important intellectual content. The final version of the manuscript was approved by all authors.

  • Funding This study (R-157811) was supported by the Swedish Research Council (2010–2891), the Swedish Rheumatism Association, Lund University and the County of Skåne.

  • Ethics approval Regional research ethics committee for southern Sweden approved this study.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.