Table 1 Summary of selected literature relevant to steroid hormones in all-cause mortality prediction

Mueller et al., 2014, [53]

- Cortisol

Single-center, prospective observational cohort study (6-week follow-up)/secondary chart analysis of a single-center, randomized controlled interventional study

179 prospectively recruited patients hospitalized with CAP

- In age and gender adjusted logistic regression analysis, cortisol (OR: 2.8; 95 % CI: 1.48–5.28, p = 0.002) and DHEA (OR: 2.62; 95 % CI: 1.28–5.34, p = 0.008), but not DHEA-S and none of the different ratios (cortisol/DHEA, cortisol/DHEA-s, DHEA/DHEA-S) were associated with increased 6-week all-cause mortality; the discriminatory accuracy in ROC analysis (AUC) to predict mortality was 0.74 (95 % CI: 0.60–0.88) for cortisol, 0.61 (95 % CI: 0.46–0.75) for DHEA and 0.72 (95 % CI: 0.62–0.81) for the PSI

- Single-center study

- DHEA

- Secondary chart analysis of remaining blood samples; the study had not been prospectively designed to use adrenal hormone concentrations as a primary endpoint [139]

- DHEA-S

- Cortisol, DHEA-S and DHEA levels were measured at different times during the day

Christ-Crain et al., 2007, [59]

- Free cortisol

Single-center, prospective observational cohort study (6.9 +/− 1.9-week follow-up)/secondary chart analysis of a single-center, randomized controlled interventional study

278 prospectively recruited patients presenting to the emergency department with CAP

- Initial TC and FC levels were significantly higher in non-survivors than in survivors (p < 0.001 for TC and p = 0.004 for FC); the area under the receiver operating characteristic curve (AUC) to predict death was 0.76 (95 % CI, 0.70–0.81) for TC and 0.69 (95 % CI, 0.63–0.74) for FC, whereas the AUC of the PSI was 0.76 (95 % CI, 0.70–0.81) to predict mortality

- This study had not been prospectively designed to use cortisol concentrations as a primary endpoint, and only single initial cortisol levels were measured [139]

- Total cortisol

- The prognostic accuracy of FC was not higher than the one for TC (p = 0.12)

- Cortisol levels were measured at different timepoints during the day

- No assessment of adrenal function

- TC and FC levels were independent mortality predictors in pneumonia; the prognostic accuracy of total cortisol equaled the predictive power of the PSI

Kolditz et al., 2012, [60]

Cortisol

Multicenter, prospective observational cohort study (patients recruited from CAPNETZ) (30-day follow-up)

984 hospitalized CAP patients, recruited from a multicenter national CAP-study in Germany

- After a follow-up period of 30 days initial serum cortisol levels were significantly higher in non-survivors than in survivors (p < 0.001); the AUC to predict 30-day mortality was 0.70 (cut-off serum cortisol value 795 nmol/l)

- No correction for concomitant steroid medication, due to absent data

- Predictive accuracy of the CURB-65 score alone (AUC 0.76) was significantly improved by combining with serum cortisol levels (AUC 0.81, p = 0.001)

- Since blood samples were taken at time of first contact, cortisol levels were measured at different time points during the day; however, during infectious diseases and thus increased stress levels the circadian pattern of cortisol production is often lost [140]

- Survival analysis by Kaplan-Meyer curves demonstrated a significantly different survival within cortisol-quartiles (p < 0.001), which persisted within individual CURB-65 classes (p = 0.002–0.003).

- No assessment of adrenal function

Kolditz et al., 2010, [61]

- Cortisol

Single-center, prospective observational cohort study (30-day follow-up)

59 adult patients hospitalized with CAP

- After a follow-up period of 30 days cortisol was significantly higher in non-survivors compared to survivors (p = 0.009) and was an independent predictor of short-term mortality (OR 1.002, 95 % CI 1.000–1.004, p = 0.04)

- Small sample size

- DHEA

- The prognostic accuracy of serum cortisol (AUC 0.83 (0.62–1.0); cut-off serum cortisol: 734 nmol/l) was similar to that of the PSI (AUC 0.86 [0.75–0.97]) and superior to that of the CURB-65 score (AUC 0.68, 95 % CI 0.50–0.85) in predicting 30-day mortality

- Single-center study

- DHEA-S

- No measurements of serum levels of free cortisol

- Cortisol, DHEA and DHEA-S levels were measured at different timepoints during the day, which could limit the prognostic accuracy of one single cortisol value

- Serum cortisol levels were independent 30-day mortality predictors in pneumonia, whereas DHEA and DHEA-S showed no significant difference between survivors and non-survivors (p = 0.16 and p = 0.70)

- DHEA-S shows age-dependent differences in secretion levels

Salluh et al., 2008, [62]

Total cortisol

Single-center, prospective observational cohort study (follow-up until death (in-hospital mortality) or hospital discharge)

72 patients with severe CAP admitted to the ICU

- TC levels were significantly higher in non-survivors compared to survivors (p = 0.003 for TC), whereas postcorticotropin cortisol levels and Δ-cortisol achieved no significant difference between survivors and nonsurvivors (p ≥ 0.05, exact value not shown)

- Small sample size with a mortality rate of only 16.7 % (12 patients), leading to low statistical power

- Increases in mortality rates were observed across all quartiles of baseline TC levels

- Single-center study

- The AUC to predict mortality was 0.77 (95 % CI, 0.64–0.90; p = 0.002; cut-off 25.7 μg/dl) for TC, 0.60 for delta-cortisol (95 % CI 0.42–0.78; p = 0.24), 0.58 for postcorticotropin cortisol (95 % CI 0.43–0.74; p = 0.33), 0.71 for CURB-65 (95 % CI 0.57–0.86; p = 0.01) and 0.71 for APACHE II score (95 % CI 0.56–0.86; p = 0.01), whereas baseline TC achieved a significantly higher AUC than postcorticotropin cortisol (p = 0.03)

- No data about patients with severe immunosuppression

- No measurement of free cortisol levels

- Kaplan-Meier curves demonstrated serum baseline cortisol (TC) concentration of >25.7 μg/dl on presentation to have a significantly higher risk of death (log rank test, p < 0.001) compared to baseline cortisol levels below this cut-off (p = 0.019)

- Measurement of cortisol levels at different timepoints during the day, potentially limiting the prognostic accuracy of one single cortisol value

- Potential confounding due to treatment with corticosteroids in the study population

- In univariate analysis, baseline cortisol, CURB-65 and APACHE II score were predictors of short-term mortality

Cortés-Puch et al., 2014, [63]

- Total cortisol

Experimental, prospective animal study (96-hour follow-up)

101 canine suffering from S. aureus pneumonia-induced sepsis

- At 10 h after onset of sepsis total and free cortisol levels and change in cortisol levels after exogenous ACTH administration (delta cortisol) showed no significant difference between survivors and non-survivors (p = 0.75 for TC, p = 0.80 for FC, p = 0.64 for delta cortisol after ACTH stimulation), whereas ACTH levels correlated significantly, but weakly, with mortality 10 h after onset of sepsis (p = 0.04)

- Low sample size

- Free cortisol

- Cortisol responses may partly be attributed to sedation, intubation, mechanical ventilation, and repeated ACTH stimulation testing

- At 24 h total and free cortisol, ACTH levels as well as delta cortisol after ACTH stimulation correlated significantly with

- ACTH

- Aldosterone

Ohlsson et al., 2010, [65]

- DHEA

Prospective observational population-based cohort study (mean 4.5-year follow-up)

2644 Swedish men from the Swedish multicenter Osteoporotic Fractures in Men cohort

- Low levels of DHEA-S and DHEA (quartile 1 vs. quartiles 2–4) predicted all-cause mortality (multivariate adjusted HR 1.54, 95 % CI 1.21–1.96 for DHEA-S; HR 1.48, 95 % CI 1.17–1.88 for DHEA)

- Single measurement of DHEA and DHEA-S at different timepoints during the day with subsequent possible diurnal variation in serum DHEA and DHEA-S levels

- DHEA-S

- Age-adjusted risk of all-cause mortality was increased in men within Q1 of DHEA and DHEA-S levels compared with men within the individual quartiles 2, 3 and 4 (Q2, Q3, Q4) (Q1: HR 1.00 (referent); DHEA: Q2 vs. Q1 HR 0.69, Q3 vs. Q1 HR 0.66, Q4 vs. Q1 HR 0.60; DHEA-S: Q2 vs. Q1 HR 0.69, Q3 vs. Q1 0.71, Q4 vs. Q1 0.60)

- No data about treatment with corticosteroids or other hormones, which might alter mortality risk and/or DHEA and DHEA-S levels

Hsu et al., 2012, [66]

DHEA-S

Single-center, prospective observational cohort study (mean follow-up time: 38.2 +/− 20.4 months)

200 CKD patients (men) on hemodialysis (HD) more than 6 months

- Low plasma DHEA-S levels (cut-off 790 ng/ml) showed significant association with increased all-cause mortality (HR 3.667, 95 % CI 1.710–7.909; p = 0.001) in hemodialysis men, but not in women (data not shown)

- Single-center study

- No assessmet of testosterone levels as a precursor of DHEA-S

- Multivariate Cox regression analysis adjusted for age, comorbidities such as diabetes mellitus, chronic heart failure, COPD, CRP as well as for albumin and creatinine demonstrated an independent association between low plasma DHEA-S levels and all-cause mortality in HD men (HR 2.93, 95 % CI 1.09–7.89; p = 0.033)

- Plasma DHEA-S levels were analyzed from one single pre-dialysis sample

Feng et al., 2014, [67]

- Estradiol

Single-center, prospective observational cohort study (28-day follow-up)

107 clinically diagnosed pneumonia-related septic shock patients

- Serum levels of progesterone and estradiol were significantly higher in non-survivors compared to survivors (p < 0.001), whereas testosterone levels were similar in both groups (p = 0.74)

- Single-center study

- Progesterone

- Most of the patients participating in the study were male

- Testosterone

- The discriminatory accuracy in ROC analysis (AUC) to predict 28-day mortality was 0.87 (p < 0.001) for the APACHE II score (cut-off 26 points), 0.705 (p < 0.001) for estradiol (cut-off 40 pg/ml), 0.713 (p < 0.001) for progesterone (cut-off 1.03 ng/ml) and 0.518 (p = 0.74) for testosterone (cut-off 4.4 ng/ml)

- No enrollment of premenopausal women

- Blood sample collection for hormone measurements only on the first day of septic shock (no investigation of the dynamic of serum sex hormone levels during the course of disease)

Shores et al., 2014, [68]

- Total testosterone

Prospective observational cohort study (9-year follow-up)

1032 men in the Cardiovascular Health Study (CHS)

- Adjusted for age and cardiovascular risk factors, low DHT (<25 ng/dl) and cFDHT (<0.13 ng/dl) levels showed significant associations with higher all-cause mortality, even if adjusted for cardiovascular risk factors (HR 1.31, 95 % CI 1.04–1.65 for DHT; HR 1.72, 95 % CI 1.25–2.37 for calculated free DHT; p < 0.001 for both)

- Only one single testosterone measurement

- Time of day for blood collection was not standardized; however, the effects of these varying time points of blood collection might be minimal due to less circadian fluctuation in testosterone levels in older men and no significant circadian variation of DHT levels [141]

- Calculated free testosterone

- TT and cFT as well as SHBG were not significantly associated with all-cause mortality (HR 1.05 (0.88–1.25) for testosterone; HR 1.04 (0.88–1.23) for cFT)

- DHT

- Calculated free DHT

Friedrich et al., 2012, [69]

- Total testosterone

Analysis from two German prospective cohort studies (DETECT and SHIP trial) (mean follow-up of 5.5 years)

3942 men

- Cox regression analyses adjusted for waist-to-height ratio, smoking and physical activity revealed that men with total serum testosterone levels below the 10th percentile showed an increased independent risk of all-cause mortality compared to subjects with higher hormone levels (HR 1.54, 95 % CI 1.20–1.99, p < 0.01); by additional inclusion of liver disease, increased blood pressure, diabetes and hsCRP as confounders the results were still significant for the 10th percentile as a cut-off (HR 1.38, 95 % CI 1.06–1.78, p = 0.02); using the 20th percentile as a cut-off resulted in no significant difference in all-cause morality between lower and higher testosterone levels (unadjusted model: HR 1.25, 95 % CI 0.99–1.57, p = 0.06)

- Only one single serum total testosterone measurement at baseline

- Blood collection at different timepoints during the day; however, serum samples showed only minor differences in testosterone levels collected before and after noon; therefore the effect of the diurnal variation seems to be minimal

- In univariate Kaplan-Meier survival analysis patients with lower serum total testosterone levels (cut-off: 10th percentile) showed significantly higher all-cause mortality compared to subjects with non-low hormone levels (p < 0.01)

- Use of different assay systems to determine total testosterone

- No measurement of SHBG and thus no calculation of free testosterone levels

Grossmann et al., 2014, [70]

- Total testosterone

Single-center, prospective observational cohort study (median follow-up of 8.5-years)

221 patients with CKD III-IV, undergoing dialysis or KTR

- Cox proportional hazards regression showed significant independant association between low total testosterone levels and high mortality rates (p = 0.01)

- Small female sample size

- Single-center study

- DHT

- No data about patients with CKD I-II

- Estrone

- Low testosterone concentrations were independent mortality predictors in men, whereas sex steroid levels in women showed no significant association with all-cause mortality

- Estradiol

- DHEA

Araujo et al., 2007, [71]

- Total testosterone

Prospective population-based observational cohort study (secondary chart analysis of MMAS study) (15.3-year follow-up)

1686 men aged 40 to 70 years, population-based randomly sampled

- In multivariate, age-adjusted models TT levels as well as FT, DHT and SHBG were not significantly associated with all-cause mortality

- Inclusion of mostly white men of higher socioeconomic status; thus results may not be generalizable

- Free testosterone

- Categorizing TT, FT, DHT and SHBG into 5 quintiles yielded no significant correlation with all-cause mortality either (for TT: reference group (RR = 1): TT ≥ 650 ng/dl; < 370 ng/dl: RR 1.24 (95 % CI 0.89–1.73); 370–466 ng/dl: RR 1.00 (95 % CI 0.70–1.42); 466–545 ng/dl: RR 1.05 (95 % CI 0.74–1.47); 545–650 ng/dl: RR 1.20 (95 % CI 0.86–1.69) (p for trend = 0.50))

- No data about women in this study

- DHT

- SHBG

- Lowest quintile of free testosterone level was significantly associated with decreased IHD mortality (p for trend = 0.02) and increased respiratory disease mortality (p for trend = 0.002);

- In conclusion, TT and FT as well as DHT seem to have a relatively weak or no association with all-cause mortality

De Padua Mansur et al., 2012, [72]

Estrone

Single-center, prospective observational cohort study (5.8 +/− 1.4-year follow-up)

251 postmenopausal women in the ambulatory care clinic of a tertiary cardiology hospital

- Kaplan Meier survival curve showed a significant association between higher all-cause mortality in women and low estrone levels (<15 pg/ml) (p = 0.039)

- Single-center study

- No data about male patients

- Multivariate Cox regression analysis adjusted for diabetes, body mass index, dyslipidemia, and family history showed estrone to be the only independent predictor for all-cause mortality (OR 0.45 (95 % CI 0.21–0.95); p = 0.038)

- Only 32 deaths and thus relatively low statistical power