Background and objective Recent longitudinal studies have suggested an association of high serum parathyroid hormone levels (PTH) with elevated cardiovascular risk in the general population. This study presents analyses of the prognostic value of baseline PTH for subsequent cardiovascular events and all-cause mortality in a high-risk population with stable coronary heart disease.
Methods Based on measurements of PTH levels in 1133 patients recruited at two German rehabilitation clinics and followed over 8 years, multivariate Cox regression analysis was performed to estimate the risk of secondary cardiovascular events (including myocardial infarction, stroke and death due to cardiovascular diseases) and all-cause-mortality according to PTH quartiles (Q1–Q4) and continuous PTH concentrations.
Results During follow-up, 153 cardiovascular events and 124 deaths occurred. Age and sex-adjusted Cox regression analysis yielded statistically significant positive associations of PTH with both cardiovascular event incidence and all-cause mortality (HR (95% CI) per SD increase of PTH: 1.35 (1.21–1.51) and 1.25 (1.11–1.42), respectively). Associations remained essentially unchanged after additional adjustment for multiple cardiovascular risk factors. More detailed dose–response analyses showed strong risk elevation for above-normal levels of PTH (>95th percentile), with essentially no association at lower levels.
Conclusion The results of this first detailed study in a cohort of patients with stable coronary heart disease suggest an independent predictive value of above-normal PTH for the prognosis in patients with stable coronary heart disease.
- cardiovascular diseases
- cohort studies
- coronary artery disease (CAD)
- parathyroid hormone
- risk stratification
Statistics from Altmetric.com
- cardiovascular diseases
- cohort studies
- coronary artery disease (CAD)
- parathyroid hormone
- risk stratification
Parathyroid hormone (PTH) is a peptide-hormone in bone metabolism contributing to calcium homeostasis. It is secreted by the parathyroid glands upon stimulation by decreased extracellular calcium levels, in order to maintain the balance by mobilising calcium from bone and by increasing intestinal absorption and renal reabsorption of calcium.1 PTH has been suggested to play a role in the development of cardiovascular diseases as well,2 although a possible direct influence of PTH on the cardiovascular system remains unclear. While PTH might be inhibitory to vascular calcification through osteoblast-like cells in the vessel wall,3 serum PTH was found to be positively associated with cardiovascular disease.4 This finding might, at least in part, be based on indirect mechanisms: primary hyperparathyroidism, which results from a disorder of the parathyroid gland and goes along with excessive secretion of PTH2 5 is associated with known cardiovascular risk factors like dyslipidaemia,6 7 left ventricular hypertrophy,8 9 hypertension10 11 and insulin resistance.12 13
Several epidemiological studies including patients with hyperparathyroidism confirmed an elevated all-cause and cardiovascular mortality risk compared with healthy subjects.14 15 However, very little is known about associations of PTH with cardiovascular risk in the general population or in subjects with pre-existing cardiovascular disease. The few population-based studies reported to date, including two longitudinal analyses, and a very recent prospective analysis in a cohort of patients undergoing coronary angiography support a positive association between PTH level and cardiovascular disease.16–19
Evidence concerning the prognostic value of PTH in coronary heart disease patients is lacking. Such knowledge would be highly relevant due to the high prevalence of coronary heart disease worldwide and the availability of various pertinent treatment options.20 In this article, we present results of a prospective study conducted in a cohort of initially 1206 patients with stable coronary heart disease from Germany, followed over 8 years.
In the prospective KAROLA study (‘Langzeiterfolge der KARdiOLogischen Anschlussheilbehandlung’) in-patients of cardiac rehabilitation after an acute coronary event (myocardial infarction (MI)/cardiac surgery for coronary heart disease/acute coronary syndrome) were enrolled. In Germany, in-patient rehabilitation programmes in specialised centres are accessible to every post-acute coronary syndrome patient. Between January 1999 and May 2000, 1206 such patients were consecutively recruited for the KAROLA study at the beginning of the 3-week rehabilitation programme in two cooperating centres (Klinik am Südpark, Bad Nauheim and Schwabenland-Klinik, Isny-Neutrauchburg). The included patients had a main diagnosis of coronary heart disease (ICD-9 pos. 410–414), were 30 to 70 years old and entered rehabilitation within 3 months from the acute event. Further details of recruitment and data collection have previously been described.21–23
The study was conducted according to the Declaration of Helsinki and was approved by the Ethics Boards of the medical faculties of the universities of Ulm and Heidelberg and of the Physicians' Chambers of the German federal states of Baden-Württemberg and Hessen. All patients gave written informed consent.
Baseline information was retrieved from standardised questionnaires, which were completed by the patients at the beginning and end of rehabilitation and covered medical history, behavioural and sociodemographic factors. Furthermore, medical records from the acute care clinic and the rehabilitation clinic were retrieved.
Follow-up information about potential secondary events was obtained at 1, 3, 4.5, 6 and 8 years after rehabilitation discharge by mailed standardised questionnaires from patients and their primary care physicians. Secondary cardiovascular events were defined as physician-reported non-fatal MI or ischaemic stroke or cardiovascular death, the latter being defined according to the major cause of death reported on death certificates obtained from public health authorities (ICD-9 pos. 390–459; ICD-10 pos. I0–I99 and R57.0).
A blood sample, taken from the fasting patient in a sitting position at the end of the rehabilitation programme, was stored at −80°C until analysis of the following laboratory markers.
Serum lipids, high-sensitive C-reactive protein (hs-CRP), lipoprotein-associated phospholipase A2, fasting glucose, creatinine and creatinine clearance as well as N-terminal pro-B-natriuretic peptide (NT-proBNP) were determined as previously described.21–23 Serum calcium, inorganic phosphate and albumin were measured photometrically using an Olympus AU 2700 device. The interassay variation coefficients provided by the manufacturer ranged from 0.94% to 0.96%, from 1.23% to 1.55%, and from 1.55% to 2.62%, respectively. Serum 25-hydroxyvitamin D (25-OH-D), osteocalcin and intact PTH were quantified by electrochemical luminescence immunoassay method on a modular analytic system (cobas; Roche Diagnostics, Mannheim, Germany). The corresponding interassay variation coefficients provided by the manufacturer ranged from 4.4% to 7.1% for 25-OH-D, from 1.1% to 1.6% for osteocalcin and from 1.6% to 3.4% for PTH.
The study population was first described according to baseline characteristics. Mean and SD of PTH levels are provided for each stratum of the categorical covariables. Differences between the strata were evaluated for statistical significance by Satterthwaite t-test or Welch analysis of variance. Correlations of continuous variables with PTH levels were examined by Spearman rank correlation coefficients.
Continuous covariables (categorised for descriptive purposes as appropriate) included age, body mass index (BMI), triglycerides, high-density lipoprotein (HDL), total cholesterol, hs-CRP, NT-proBNP, calcium, phosphate, albumin and 25-OH-D. Gender, season (blood sampling from March–May, June–August, September–November or December–February), smoking status (never, ex or current smoker), number of affected vessels (≤1, 2 or 3–4), history of MI, treatment with β-blockers, lipid-lowering drugs, ACE inhibitors, calcium antagonists or diuretics, creatinine clearance (≤60 and >60 ml/min) and history of diabetes or hypertension were entered as categorical variables.
For reasons of comparability, the analytical strategy closely followed the lines of previous publications.16 Initially introducing all aforementioned variables, we selected independent predictors of PTH levels on a two-sided α-level of 0.05 in linear regression analysis, using a backward elimination method.
After visualisation of the association of PTH quartiles with occurrence of secondary cardiovascular events and all-cause mortality using log(-log) survivor functions and Kaplan–Meier plots, we employed multivariable Cox proportional hazards modelling to analyse the prognostic value of PTH, categorised into quartiles and as a continuous variable (divided by its SD, for both outcomes). We assessed the proportional hazards assumption by means of Schoenfeld tests using the SAS macro SCHOEN24 in the fully adjusted models (using continuous PTH). Apart from age and sex-adjusted models (model 1), we adjusted for different sets of confounders: in model 2 (fully adjusted) we introduced established cardiovascular risk factors (age, gender, hypertension, diabetes, smoking, BMI, total cholesterol, HDL-cholesterol, triglycerides, use of diuretics, ACE-inhibitors, β-blockers, calcium antagonists, lipid-lowering drugs, history of MI and number of affected vessels). In addition, we separately tested interaction terms for every covariable with PTH. Models 3 to 5 consisted of model 2 extended by factors associated with mineral metabolism (model 3: model 2 + calcium, albumin, phosphate, 25-OH-D, creatinine clearance and season of blood draw), or by factors reflecting potential pathways of the link between PTH and the outcomes like inflammation (model 4: model 2 + hs-CRP) and ventricular dysfunction (model 5: model 2 + NT-proBNP). Due to some detected non-linearity for the unadjusted association of CRP and NT-proBNP with the outcomes examined, we additionally introduced a quadratic term for those variables in models 4 and 5.
In order to determine the discriminatory abilities of PTH, the net reclassification improvement (NRI) and integrated discrimination improvement (IDI)25 by PTH as a continuous variable were calculated for the fully adjusted models for both outcomes.
Additional exploratory analyses included dose-response modelling with fractional polynomials26 and subgroup analyses with further stratification of subjects in the upper quartile (>95th percentile, 75–95th percentile). Statistical tests were two-sided and p values <0.05 were considered significant. All statistical analyses were performed with the SAS statistical software package (SAS V.9.2).
PTH could be measured in 1133 patients (93.9% of initially included KAROLA participants). Of these, 1039 patients (91.7%) had physician-based follow-up information about secondary cardiovascular events, and 1133 (100%) patients had follow-up information about survival status.
The distribution of PTH was modestly right skewed with a median (IQR) of 34 (26–44) pg/ml and a mean (SD) of 37.3 (18.6) pg/ml; 91.5% of the values were situated in the reference range (15–65 pg/ml) given by the assay manufacturer (3.4% <15, 5.0% >65 pg/ml). As shown in table 1, PTH tended to be higher in older subjects, patients with a history of hypertension or with an impaired ventricular function as well as in participants treated with diuretics, whereas lower PTH levels were observed among patients treated with aspirin or lipid-lowering drugs. Furthermore, PTH varied significantly with the season, with lower values in autumn (mean 30.3 pg/ml) compared with the other seasons (means 36.3–40.4 pg/ml). Although a history of diabetes did not seem to be statistically associated with PTH, further stratification of patients with diabetes according to treatment revealed that patients with diabetes being treated with oral antidiabetic drugs showed the lowest mean PTH level (34.5 (15.3)) compared with those not being treated (37.0 (16.0)) and those receiving an insulin therapy (41.8 (30.4)). Among continuous covariables, significant positive correlations with PTH were found for cystatin C, NT-proBNP, lipoprotein-associated phospholipase A2 (Lp-PLA2), hs-CRP and osteocalcin. In contrast, PTH was negatively associated with creatinine clearance, 25-OH-D, calcium and phosphate (table 2).
The final PTH linear regression model resulting from backward elimination yielded independent associations of high PTH with treatment with diuretics, low creatinine clearance, high BMI, high triglycerides, high NT-proBNP and high HDL-cholesterol, whereas low PTH was independently associated with blood drawing in autumn, a history of diabetes, treatment with lipid-lowering drugs, high phosphate, high total cholesterol and high 25-OH-D (see supplementary table 1, available online only).
During a median (IQR) follow-up time of 8.1 (6.1–8.2) years for cardiovascular events and 8.2 (8.1–8.4) years for all-cause mortality, 153 cardiovascular events and 124 deaths occurred. The Kaplan–Meier plots for PTH quartiles and cardiovascular event incidence and all-cause mortality shown in figure 1 and 2 tentatively suggest moderate positive associations (log rank p=0.053 and 0.13, respectively).
In age and sex-adjusted Cox regression models, a non-significant risk increase was observed in the highest compared with the lowest quartile for both cardiovascular events (HRQ4vs.1=1.55 (0.98–2.45)) and all-cause mortality (HRQ4vs.1=1.73 (0.98–3.04)) (table 3).
As can be seen in table 3, when including PTH as a continuous variable, age and sex-adjusted models as well as models adjusted for established cardiovascular risk factors (‘fully adjusted models’) showed a significant risk increase with higher PTH for both outcomes (secondary cardiovascular events: fully adjusted HR=1.30 (1.15–1.48) per SD of PTH; all-cause mortality: fully adjusted HR=1.21 (1.05–1.40) per SD of PTH). Note, that for cardiovascular event incidence the scaled Schoenfeld residuals were positively correlated with time to event (r=0.178, p=0.04), suggesting that the overall HR estimated in this model might be an overestimate during early and an underestimate during later follow-up.
NRI measures suggested no significant overall improvement by the addition of PTH to the fully-adjusted models for either outcome (secondary event incidence: NRI=0.121 (95% CI: −0.058 to 0.301), p=0.19; all-cause mortality: NRI=−0.052 (−0.247–0.144), p=0.61). Similarly, IDI was very small for both outcomes, though a p value <0.05 was observed for the secondary event model (secondary event incidence: IDI=0.018 (95% CI: 0.001 to 0.035), p=0.04; all-cause mortality: IDI=0.01 (−0.003–0.023), p=0.14). Further adjustment for factors associated with mineral metabolism or CRP had little impact on the observed associations. However, associations were attenuated (and became non-significant in case of all-cause mortality) after additional adjustment for NT-proBNP (model 5), suggesting that the association between PTH and the outcomes may at least partly be mediated through ventricular dysfunction.
Further stratification of participants in the 4th quartile of PTH (data not shown) suggested elevated risks for both outcomes with PTH levels above the 95th percentile compared to levels in the 1st quartile in fully-adjusted models (secondary cardiovascular events: HR=2.62 (1.39–4.94), all-cause mortality: HR=2.60 (1.18–5.74)), while a risk increase could not be observed for PTH levels in the 75th–95th percentile (secondary cardiovascular events: HR=1.12 (0.65–1.91), all-cause mortality: HR=1.25 (0.64–2.42)).
The evaluation of fractional polynomials (based on the fully adjusted models) suggested no significant improvement in model fit with non-linear transformations of PTH, both with respect to secondary cardiovascular events and all-cause mortality (data not shown). Furthermore, no interaction term of PTH with any covariable in the fully adjusted model for secondary cardiovascular events was found to be significant. In contrast, for all-cause mortality, the interaction term of PTH with diuretics was significant at α=0.05 (p=0.03) and suggested a HR of 1.34 (1.14–1.57) per SD of PTH in the treated versus 0.90 (0.64–1.25) in the non-treated participants.
The results from this prospective cohort study including patients with stable coronary heart disease at baseline followed over 8 years support an independent association of baseline PTH with secondary cardiovascular event incidence and all-cause mortality. Notably, strongly increased risks were observed among patients with PTH over 66 pg/ml.
Our results are in line with those reported from previous population-based cohorts of elderly community-dwelling16 or institutionalised subjects,19 and from one very recent analysis in a cohort of patients who underwent coronary angiography18 in which increased cardiovascular and all-cause mortality was found in subjects with higher PTH. Comparison of the results of Hagstrom et al16 in a community-based sample with our findings reveals a remarkable similarity, with an adjusted HR of 1.38 (1.18–1.60) per SD of PTH. Both a slightly different outcome definition (cardiovascular mortality) and study population might have contributed to their higher HR in the fully adjusted quartile model (HRQ4vs.1=1.83 (1.10–3.04)). However, the use of different cutpoints in the two studies probably also played a role, since applying the quartile cutpoints of Hagstrom et al16 (28, 38 and 50 pg/ml) to our data likewise resulted in higher HR for cardiovascular event incidence (model 2: HRQ4vs.Q1=1.99 (1.22–3.25)).
The question of similarity of pathogenetic mechanisms involved in the development of primary and secondary cardiovascular events, however, cannot be answered at this point. Similar to previous reports,16 18 we observed a reduced risk estimate after adjustment for NT-proBNP, an indicator of ventricular dysfunction, whereas including hs-CRP did not change the main results.16 In addition, like other studies in diseased and healthy populations,16 18 19 our data suggested a prognostic value of PTH independent of factors of mineral metabolism, especially 25-OH-D, which likewise has been linked to various morbidities including cardiovascular diseases.27 Notably, further analyses (data not shown) could not identify an interaction between these two markers regarding cardiovascular and mortality outcomes. However, analyses in the general population and in patients undergoing coronary angiography still supported a prognostic value of PTH in the normal range for cardiovascular and all-cause mortality,16 18 an observation not confirmed in our study of coronary heart disease patients. Although a connection of PTH levels with cardiovascular risk factors appears to be established through basic research and epidemiological studies,2 4 the results of our backward linear regression analysis revealed some inconsistencies regarding the relationship of PTH with cardiovascular risk profiles in this special population since high PTH levels appeared associated with both favourable (eg, high HDL level, absence of diabetes) and adverse components (eg, high BMI, high triglyceride level). Nevertheless, and although we cannot entirely rule out the possibility of residual confounding for example by vitamin D-related factors, such as outdoor physical activity levels or vitamin D supplementation throughout follow-up, our data indicate that an independent influence of PTH on the cardiovascular system may exist besides indirect mechanisms through known, albeit only partially understood, connections with other cardiovascular risk factors contributing to the pathogenesis of recurrent events. On a cellular level, vascular smooth muscle cells that are responsible for the production of extracellular matrix in the vessel wall after being activated through mechanisms resembling bone metabolism3 have been shown to express PTH-receptors28 through which PTH might directly play a role in the development of vascular calcification, which, however, is not fully understood and therefore needs further elucidation.3 29
Some limitations of our study have to be taken into account. Even though in Germany all patients after an acute event are entitled to participate in a rehabilitation programme, rehabilitation and study participation was voluntary. Therefore, our study population might not be entirely representative of all such patients, with more severe cases probably being less likely to participate. The overall slightly lower PTH values in our cohort in comparison with the distribution in the general population as reported by Hagstrom et al16 should not be prominent enough to suggest that pronounced survivor selection bias had taken place, which would theoretically be possible through increased acute mortality in high PTH subjects. Heterogeneous PTH assay performances might more plausibly account for these differences.30 Note, in this regard, that a possible interaction of the Roche electrochemiluminescence assay for 25-OH-D with smoking has recently been described.31 The results laid out in the present manuscript, however, did not change to any relevant degree when excluding the low number of current smokers (data not shown). The low percentage of women (15.5%) in our cohort as often seen in cardiological rehabilitation programmes32 unfortunately prevented us from calculating sex-specific estimates, a limitation that might be overcome in future larger studies. Finally, our study examined the prognostic value of a single baseline measurement of PTH for secondary events. To gain more definite insights into the question of causality, repeated PTH measurements throughout long-term follow-up clearly would be advantageous.
Whereas our analysis concentrated on the prognostic value of PTH predominantly in the ‘normal range’, additional analyses in our data with further stratification of patients in the upper quartile of the PTH range suggested a strong association of PTH with both outcomes in patients with pathologically high PTH, along with the absence of such associations with lower PTH levels. While these patterns, as well as the fact that PTH levels have also been found to be positively associated with other cardiovascular (and non-cardiovascular) health risks like for example adverse outcomes in heart failure,33 might suggest potential benefits of interventions targeted at lowering high PTH levels, pertinent recommendations would be premature in the absence of data from intervention studies. Depending on the underlying cause, which could not be determined unequivocally for the patients in our sample, above-normal PTH can be lowered for example by parathyroidectomy,5 calcimimetics,34 treatment with calcium/vitamin D supplements,35 36 or vitamin D receptor activators.37 Interventions are commonly undertaken in manifest hyperparathyroidism, and some evidence has been presented previously that such interventions could successfully reduce PTH secretion and adverse cardiovascular outcomes in patients with elevated PTH levels,5 supporting the hypothesis of a causal relationship between PTH levels and the outcomes investigated. While for the time being we cannot derive evidence for therapeutic benefit from reducing pathological PTH levels in patients with stable coronary heart disease from our observational study, this question would appear worthwhile to be followed up in future clinical research.
In our analysis, PTH levels were independently associated with secondary cardiovascular events and all-cause mortality in a population of stable coronary heart disease patients during long-term follow-up. In particular, strongly increased risks were observed in subjects with above-normal PTH. Due to the lack of longitudinal data on PTH, both in the general population and in cardiovascular patients, and given the high prevalence of cardiovascular diseases worldwide, this subject warrants further investigation, especially since interventions aiming at PTH reduction are feasible.
The authors would like to thank Claudia El Idrissi-Lamghari from the Division of Clinical Epidemiology and Aging Research (German Cancer Research Center, Heidelberg, Germany) and the co-workers in the two rehabilitation clinics for their kind support in data collection and data management. They are grateful to Synlab Heidelberg, Eppelheim for offering the possibility to measure levels of PTH and other markers of bone metabolism in their laboratory.
Funding This study was funded by the Willy Robert Pitzer Foundation (Zanderstrasse 28, 61231 Bad Nauheim, Germany), Association of German Pension Fund Agencies (Ruhrstrasse 2, 10709 Berlin, Germany), German Ministry of Education and Research (Hannoversche Strasse 28–30, 10115 Berlin, Germany).
Competing interests None declared.
Patient consent Obtained.
Ethics approval This study was conducted with the approval of the Ethic Boards of the medical faculties of the universities of Ulm and Heidelberg and of the Physicians' Chamber of the federal states of Baden-Württemberg and Hessen.
Provenance and peer review Not commissioned; externally peer reviewed.
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.