Calciphylaxis epidemiology, risk factors, treatment and survival among French chronic kidney disease patients: a case-control study

Background Calcific Uremic Arteriolopathy (CUA) is a rare disease, causing painful skin ulcers in patients with end stage renal disease. Recommendations for CUA management and treatment are lacking. Methods We conducted a retrospective cohort study on CUA cases identified in western France, in order to describe its management and outcome in average clinical practices. Selection was based on the Hayashi diagnosis criteria (2013) extended to patients with eGFR < 30 mL/min/1.73m2. Dialyzed CUA cases were compared with 2 controls, matched for age, gender, region of treatment and time period. Results Eighty-nine CUA cases were identified between 2006 and 2016, including 19 non dialyzed and 70 dialyzed patients. Females with obesity (55.1%) were predominant. Bone mineral disease abnormalities, inflammation and malnutrition (weight loss, serum albumin decrease) preceded CUA onset for 6 months. The multimodal treatment strategy included wound care (98.9%), antibiotherapy (77.5%), discontinuation of Vitamin K antagonists (VKA) (70.8%) and intravenous sodium thiosulfate (65.2%). 40.4% of the patients died within the year after lesion onset, mainly under palliative care. Surgical debridement, distal CUA, localization to the lower limbs and non calcium-based phosphate binders were associated with better survival. Risks factors of developing CUA among dialysis patients were obesity, VKA, weight loss, serum albumin decrease or high serum phosphate in the 6 months before lesion onset. Conclusion CUA involved mainly obese patients under VKA. Malnutrition and inflammation preceded the onset of skin lesions and could be warning signs among dialysis patients at risk. Trial registration ClinicalTrials.gov identifier NCT02854046, registered August 3, 2016.


Background
Calcific uremic arteriolopathy (CUA), also called calciphylaxis, is a rare but devastating disease involving patients with end stage renal disease (ESRD). CUA causes painful skin lesions that evolve to ulcerative lesions at risk of superinfection and sepsis [1], with a poor prognosis. One year survival rates vary between 45 to 55% [2][3][4]. CUA management lacks strong recommendations [5] and therefore is heterogeneous. Reported risk factors of CUA are female sex, obesity, diabetes mellitus, vitamin K antagonists (VKA) and ESRD [6]. Dysregulation of calcium-phosphate metabolism also participates to its development. Histopathological findings of skin lesions mostly associate thromboses and vessel calcifications [7]. We decided to conduct the first study on CUA in the French population. Our main objective was to describe diagnosis management, treatment and outcome of CUA in ESRD and stage 4-5 CKD patients. Secondary objectives were to analyze risk factors of developing calciphylaxis and influencing patient survival in the dialyzed cohort.

Study patients
We first conducted a retrospective cohort study and secondly a case control analysis. The nephrologists from Western France were asked to report their patients diagnosed for CUA. Patients were also identified by searching the MEDIAL dialysis regional data base. Inclusion and non-inclusion criteria checking and collection of the data in the medical records were performed by the investigators in the center of care of each case.
Hayashi [8] criteria were used for diagnosis: chronic hemodialysis or estimated Glomerular Filtration Rate (eGFR) below 15 mL/min/1.73m 2 , more than two painful non-treatable skin ulcers with concomitant painful purpura and localization of skin ulcers on the trunk, extremities or penis with concomitant painful purpura. Typical histopathological findings (necrosis and ulceration of the skin with calcification of the tunica media and internal elastic membrane of small to medium-sized arterioles of dermis and subcutaneous fat) can replace a clinical feature.
Inclusion criteria were: CUA according to Hayashi criteria, onset of cutaneous lesions between 1st January 2006 and 31th December 2016, patients > 18 yo. Patients with eGFR between 15 and 30 mL/min/1.73m 2 (CKD EPI formula) (the serum creatinine at onset of CUA was considered) were also included if all other inclusion criteria were met. CUA was eliminated if a differential diagnosis seemed more likely or was confirmed by skin biopsy, or in case of severe atherosclerotic vascular disease in the wound area. CUA patients under hemodialysis or peritoneal dialysis at onset of CUA lesions were assigned to the dialysis group, while the others to the non dialysis group.

Controls selection
In order to explore risk factors of CUA among dialyzed patients, each CUA dialysis patient was matched with two controls identified in the REIN registry of treated ESRD in France. Matching criteria were: gender, age (± 2 years), treatment by hemodialysis in the same geographical area and at the same time of CUA diagnosis in the case. Among the potential controls for each case, two were randomly and anonymously selected. After checking the absence of diagnosis of CUA, collection of the data in the medical records of the control patients was performed by the investigator directly in the center of care of the patient.

Study data
The analyzed data were demographic, history of kidney disease, cardiovascular comorbidities and other previously reported risk factors of CUA. Thrombophilia was defined as acquired or congenital antithrombin/protein C/protein S deficiency, antiphospholipid syndrome, activated protein C resistance, prothrombin mutation. "Onset date" of CUA was the date where typical skin lesions were mentioned in medical records. Laboratory data at onset were recorded as well as the worst values within the 6 months before diagnosis. As intact Parathyroid hormone (iPTH) measurement kits were different between laboratories, we normalized iPTH with the upper limit of the normal range for each laboratory. Medications and dialysis parameters were recorded. For each CUA case, clinical presentation with lesion distribution defined as proximal (extremities proximal to knees and elbows, trunk, breast and penis) and/or distal (extremities distal to knees and elbows), evolution, diagnosis and treatment methods and outcomes were collected.
A written consent form was given to each patient, except for deceased patients and loss of follow-up patients. The study was approved by the ethics committee of the Nantes University Hospital. All data collected were de-identified.

Statistical analyses
Frequency of categorical variables, median and interquartile range (IQR) for non-normally distributed variables were reported. Survival curves were determined using the Kaplan-Meier method. Survival analysis using Cox models were used to determine survival predictors. In the group of CUA patients treated by dialysis, univariate conditional logistic regression analysis on matched case-controls was performed. Variables with p < 0.20 were included in the multivariate logistic regression analysis. Non dialyzed CUA patients were not included in the risk factor analysis because of the lack of controls for those patients.
All analyses were performed using the SAS program (version 9.4) (SAS Institute Inc., Cary, NC). Statistical Fig. 1 Flow chart of selection of Calcific Uremic Arteriolopathy cases significance was set as p < 0.05. The study protocol was pre-registered on clinicaltrials.gov under the number NCT02854046.

Results
Two hundred fourteen eligible cases were identified (Fig. 1). Ninety-five patients with incomplete diagnosis criteria and 22 with a differential diagnosis were excluded (Table S1). Eighty-nine cases of CUA were finally included, 74% diagnosed between 2013 and 2016.

Laboratory findings
Adjusted serum calcium, serum phosphate and normalized iPTH were significantly higher in dialyzed CUA patients than in matched dialyzed controls at lesion onset and in the six preceding months ( Table 2). Malnutrition preceded CUA onset, with a median albumin decrease of 2.7 g/L within the 6 months before onset and Creactive protein (CRP) was high at both times.

Medications at CUA lesion onset
Active and native vitamin D were not significantly more prescribed in CUA patients, nor statin and cinacalcet (Table 3). Calcium-based phosphate binders were more frequently administered in CUA patients; 71.9% were under VKA. The median time between VKA introduction and onset of CUA was 3.2 years (IQR 1.8-6.6) and was shorter for dialyzed vs non dialyzed patients (2.6     Thirty-six patients (40.5%) suffered from a proximaltype CUA, while 26 (29.2%) had a distal-type, and 27 (30.3%) both proximal and distal. Lower limbs were involved in most of the patients (86.5%), especially under the knees (34.8%), while trunk lesions were found in 50.6%, mainly in the abdomen (27.0%). Upper limb lesions were present in 22.5%. A median of 5 lesions (IQR 3-6) per patient were found and were mostly ulcerative (95.5%).

CUA treatments
Treatments used for CUA were wound care, intravenous Sodium Thiosulfate (STS), renal replacement therapy modification and nutritional support therapy (Table 4). Other treatments were discontinued, notably VKA, calcium supply and vitamin D. Median dosing of intravenous STS was 75 g per week, and the cumulative dose was higher for dialysis patients. Some treatments were scarcely used, and included, intra-lesional STS (1.1%),  (Table 5), risk factors independently associated CUA in dialyzed patients were obesity, coronary artery disease, weight loss over the last 6 months, serum phosphate increase within 6 months before diagnosis and VKA therapy. As lower odds were associated with serum albumin increase within the 6 months before onset of lesions (OR 0.2), serum albumin decrease was also a risk factor of CUA among the dialysis cases.

CUA outcome
40.4% of deaths due to calciphylaxis occurred during the first year after diagnosis and 56.2% after 5 years ( Table 6). Mortality of dialysis cases was significantly higher than paired hemodialysis controls (Hazard Ratio (HR) 3.4; 95% Confidence Interval (CI) 2.2-5.2; p < 0.001) (Fig. 2). The median delay between the onset of lesions and death was 4.1 months (IQR 2.2-14.2). The main circumstance of death was in palliative care for CUA patients (49.1%).

Factors predictive of survival
Higher BMI (HR 0.79; p = 0.004), higher serum albumin at onset of lesions (HR 0.70; p < 0.001) and VKA discontinuation (HR 0.41; p = 0.01) were associated with better survival in univariate analysis only (Table  S3). After removal of patients treated less than 2 weeks or with a cumulative dose below 150 g of STS, a trend to better survival was associated with STS cumulated dose and duration (HR 0.87; CI 0.77-0.97; p = 0.02). By multivariate analysis, items with p < 0.2 in the univariate analysis were included, except sevelamer and lanthanum carbonate treatment at diagnosis, normalized PTH, eKt/V, number of skin biopsies, STS duration, VKA discontinuation and CUA recurrence, because of missing data. Adjusted serum calcium and calcium phosphate product were removed because of linkage to serum calcium and phosphate. Thus 192 patients were included in the multivariate analysis. Finally, factors independently associated with survival were surgical debridement (HR 0.11) and antibiotherapy (HR 0.25) (Table 7), whereas parathyroidectomy increased the risk of death (HR 29.5).

Discussion
The 89 CUA patients of our study were typically 70-year old overweight diabetic females under VKA therapy (72%). In these cases, bone mineral disease parameters were out of the recommended ranges and were associated with inflammation and malnutrition.
These patients were identified from all over western France and the informations provided by our study are the results of average clinical management of these patients and not the experience of one expert care center. By following the Hayashi criteria [8], a skin biopsy was not a prerequisite in our study and confirmed diagnosis only in required clinical situations, in particular, to rule out a differential diagnosis. Because this can worsen lesions, skin biopsies are frequently avoided. Specificity is also questioned because of the frequence of extravascular calcifications in ESRD. Skin biopsies can confirm diagnosis by showing the combination of arteriolar media calcification and thrombosis that is associated with CUA [9]. We think that the identified CUA cases are the reflect of complexity of CUA diagnosis in clinical practice.
The demographic data and CUA predisposing factors identified in our study were consistent with other casecontrol studies [4,10]. Inflammation and bone mineral disease abnormalities, especially hyperphosphatemia and hyperparathyroidism, and malnutrition preceded CUA onset by months [2,10]. For the first time, weight loss within the 6 months before CUA onset was identified by our study as a risk factor in dialysis patients.
The association of CUA and VKA therapy has already been reported [4,8,11]. By decreasing carboxylated matrix Gla Protein (cMGP), VKA are suspected to contribute to vascular calcification and therefore promote calciphylaxis. A low level of cMGP have also been highlighted in CUA cases associated with Vitamin K deficiency [12]. Besides, Warfarin could paradoxically favor thrombosis locally, by blocking protein S endothelial secretion in response to stress [13]. Thrombophilia is also a known risk factor of calciphylaxis [14,15] and we assume that the low prevalence of thrombophilia in our study might be due to lack of systematic screening. As Direct oral anticoagulants have no pro-thrombotic effect and given that vessel thrombosis may play a key role in calciphylaxis, they have been used to replace VKA once diagnosis of CUA is confirmed [16,17]. Two retrospective studies [16,17] have assessed the safety of Apixaban in CUA patients: 4 bleeding events in 20 dialysis patients were found and a lower mortality rate was demonstrated. Additional comparative studies are of course necessary.
We were surprised to identify so many non-dialyzed patients with calciphylaxis. The frontier between uremic calciphylaxis and non-uremic calciphylaxis is difficult to define. Studies on non-uremic calciphylaxis are mainly case reports and case series. Interestingly, a review on Non Nephrogenic Calciphylaxis (NNC), defined as calciphylaxis occurring in patients without impaired renal function (eGFR > 60 mL/min/1.73m 2 ), showed that VKA and obesity were the two main conditions associated with NNC [18]. Mean blood mineral parameters were normal. Calciphylaxis seems to be the conjunction of multiple conditions (obesity, VKA, bone mineral disease abnormalities, uremia, inflammation) with a broad spectrum of variations. In our study, non-dialyzed CUA patients were more obese and inflammation prior CUA onset was more severe than in dialyzed CUA patients. A Hazard ratio (HR) with 95% confidence interval link between arteriolar calcification and adipocytes could explain the preferential localization of calciphylaxis in adipose tissue areas and the increased risk of calciphylaxis associated with obesity [4,19]. Childhood obesity is already known to increase coronary artery calcification by middle age [20]. An increase in NNC reported cases [18,19] might not only be secondary to increased medical awareness of this disease, but also might be linked to the epidemic of obesity. In our study, non-dialyzed cases had the same survival as dialysis cases, whereas some studies have reported a better prognosis of NNC [21]. The mortality rate was particularly high in our study. Ulcerative lesions and proximal-type CUA, known as poor prognosis factors [2,3], were predominant. The prolonged diagnosis delay (46 vs 28 days in the German registry [22]) could be explained by the use of ulcerative skin lesions as inclusion criteria, because non-ulcerative lesions (plaques, nodules) precede the ulcerative lesions by several days [3]. This diagnosis delay could also be linked to a lack of acknowledgment of early CUA lesions, as seen in the Japanese case control study [10]. The dialysis vintage, shorter among the dialyzed cases than the matched controls (1.4 vs 2.2 years) could not explain the excess of mortality in dialyzed cases. So calciphylaxis does not necessarily occur after a long dialysis vintage. This is consistent with the important number of CUA cases identified among patients with CKD stage 4-5.
The main circumstance of death in our study was palliative care (49%), in the absence of efficient therapy to treat CUA. Due to its promising treatment properties [23,24], STS was frequently administered, but was only associated with a trend toward better prognosis after exclusion of patients treated less than 2 weeks. Trials are in progress to assess the real benefit of STS in CUA. As demonstrated by other studies, surgical debridement had a net impact on survival [25][26][27] and should be proposed more widely. Contrary to other studies [26], parathyroidectomy was associated with a poorer prognosis, but given that only 5 parathyroidectomies were performed the conclusions are limited.

Conclusions
Our study confirms the data reported by others on CUA but showed for the first time the contribution of significant unintentional weight loss. Few therapeutic measures seem efficient. Among them, STS is commonly used, but its benefit has still to be proved. The ongoing trials are of major interest.
Additional file 1: Table S1. Differential diagnosis idendified among eligible patients. Table S2. Univariate logistic regression analysis of risk factors of CUA in dialysis cases compared to matched dialysis controls. Table S3. Univariate conditional logistic regression analysis of survival predictors among Calcific Uremic Arteriolopathy cases.