Background

Approximately, 65,000 end-stage renal disease (ESRD) patients start chronic hemodialysis (HD) every year, in Europe [1]. These patients are at a high risk of early and midterm death, the 1-year and 3-year mortality ranging between 10 and 17% and 32–45%, respectively [1,2,3,4,5,6,7,8,9].

Early and mid-term death after initiating dialysis is therefore a relevant problem that needs to be accounted for before renal replacement therapy becomes necessary and discussed with patients and families.

In the last 20 years, there have been significant advances in technology and the quality of chronic dialysis procedures and improvements in managing anemia, arterial hypertension, vascular access, and uremia/replacement therapy-related complications, such as osteodystrophy and cardiovascular diseases [10, 11].

At the same time, the proportion of incident patients aged over 65 years in Europe increased from 22% to 1980 to 55% in 2005, according to the European Renal Association Dialysis and Transplantation (ERA-EDTA) [12, 13], with an increase in the patients with age-related vascular nephropathies and diabetic nephropathy [14].

Interestingly, in the last few decades, a moderate progressive increase of short-term survival in incident HD patients has been observed over time in some studies [1,2,3,4,5,6,7] but not in others [4, 8, 9]. However, when present, the increase of survival was independent of the age of the patients’ population [1,2,3,4,5,6,7]. In a study conducted, using the Lazio Regional Dialysis and Transplant Registry (RRDTL) data, on a 1995–2006 cohort (about 9,000 subjects), the overall mortality rate remained essentially unchanged over the 12-year period, despite a 5-year increase in the median age, without statistically significant differences over time in annual mortality rates by cause of death [4].

In addition, a reduction over time of in excess mortality for atheromatous cardiovascular disease and infections has been reported, possibly due to the better management of dialysis patients with myocardial infarction or stroke in the last decades and better prevention of infections [1].

In this study, using the RRDTL, trends over time in 1-year and 3-year mortality were examined in ESRD patients starting chronic HD in Lazio, a region in central Italy that includes Rome, and has 5,5 million residents.

Methods

Aims

The present study aimed to analyse mortality within 1 year and 3 years from incidence in chronic HD among patients residing in Lazio region.

In detail objectives were:

  1. 1.

    To calculate the annual 1-year and 3-year mortality rate over a period of 9 years overall and by gender and age classes;

  2. 2.

    To analyse the association between period of incidence in HD and 1-year and 3-year mortality;

  3. 3.

    To identify factors associated with 1-year and 3-year mortality.

Source of the data

The source of the data is the RRDTL a population-based registry established in 1994. It collects detailed information on socio-demographic status, clinical characteristics, dialysis treatments, and drug therapy for all patients undergoing dialysis in all public and private accredited dialysis centers of the Lazio Region in Italy. All dialysis units have to register the information on their patients and to update information every 6-months. The dialysis units are required to communicate the date and the reason of dialysis treatment’s termination (kidney transplant, renal recovery, transfer to another dialysis centers, death) [15, 16]. Details on the RRDTL are reported elsewhere [4, 17,18,19,20]. The Lazio regional mortality registry was used to confirm the date of death registered in the RRDTL [21].

Study population

This cohort study includes ESRD patients who started HD between January 2008 and December 2016 and were recorded in the RRDTL at incident date.

Exclusion criteria

Patients less than 18 years old and with a follow-up period shorter than 91 days were excluded to avoid possible acute dialysis cases, which may have been incorrectly recorded as chronic at the RRDTL.

Outcome

The outcomes are death within 1 year of dialysis and death within 3 years of dialysis.

Follow-up periods

Each patient with at least 91 days of follow-up after the date of first dialysis treatment (index date) was followed from index date up to the first of the following: death, change in prognosis (transplantation, renal function recovery), loss to follow-up and the end of the study. The end of the study was 1 year and 3 years from the index date for 1 and 3-year outcomes respectively. Change in prognosis or loss to follow-up were reasons for censuring.

Exposure (aim2)

Periods of incidence in HD: 2008–2010, 2011–2013, 2014–2016.

Co-variates or potential risk factors

Variables recorded in the RRDTL were evaluate at the index date and were considered as co-variates for the aim 2 and as potential risk factors for the aim 3.

Socio-demographic and clinical variables

The following socio-demographic variables were considered: gender, age classes (< 64, ≥ 65 years), place of birth (Italy, other countries), residence (Rome Municipality, Rome Province, Other Lazio’s Municipalities) and educational level (Up to Middle School, High School or more). Furthermore, were considered both clinical variables such as body mass index (BMI < 18.0, 18.0 ≤ BMI < 25.0, 25.0 ≤ BMI < 30.0, BMI ≥ 30.0), self-sufficiency (Total autonomy, Autonomy in some activities, Not self -sufficient), cause of ESRD (renal vascular disease, diabetic nephropathy, glomerulonephritis, cystic renal disease and familial nephropathy, interstitial and toxic nephritis/pyelonephritis, systemic disease, renal malformation, other nephropathies, unknown) and the following comorbidities: severe hypertension, heart disease, diabetes, peripheral vascular diseases, cerebrovascular disease, chronic obstructive pulmonary disease (COPD), cancer, thyroid disease, lipid metabolism’s alteration, motor deficit, liver disease, extrauremic anemia, dementia and psychiatric disease, malnutrition, obesity: (yes vs. no).

Treatment variables

Use of drugs for diabetes, hypertension, other cardiovascular disease and anemia, and type of vascular access (catheter-CVC, fistula-FAV) evaluated at the index date.

Care-related variables

Type of dialysis centre (public, private), pre-dialysis counselling (patients followed by nephrodialytic clinic in the 6 months prior to the start of dialysis; yes vs. no).

Laboratory findings

The laboratory findings collected included: haemoglobin Hb (g/dL), calcium (mg/dL), serum creatinine (mg/dL), serum phosphate level (mg/dL), serum albumin (mg/dL) evaluated at the index date.

Statistical analysis

Socio-demographic, clinical, treatment, care-related variables and laboratory findings were presented as percentage or mean [IQR], according to periods of incidence in dialysis and to vital status within 1 year and 3 years from index date. The relationship between categorical variables and period or vital status were tested by Chi-square test. The association between normally distributed continuous variables and period or vital status were tested by ANOVA or T-test respectively; the association with the other continuous variables were tested by non-parametric tests, namely Kruskal-Wallis or Wilcoxon rank sum tests.

Crude Mortality Rates per 100 Person Years (CMR*100PY) were calculate at 1 year and 3 years by calendar year of incidence (from 2008 to 2016), as the ratio of the number of deaths over the person time at risk. Plots of mortality rates across years were presented stratified by gender and age classes. For both outcomes JoinPoint models were performed to evaluate time trends in mortality rates overall and stratified by gender and age classes [22].

Cumulative survival estimates at 1 year and 3 years since index date were presented as Kaplan-Meier curves for the three periods and compared using the log-rank test. Kaplan-Meier curves were also presented according to age classes.

Furthermore, Cox univariate proportional hazards regression analyses were performed to explore the association between potential confounders and mortality within 1 year and 3 years since index date, and a multivariable Cox model was used to explore the association between periods of incidence and mortality in HD and the two outcomes adjusted for potential confounders.

Finally, Cox multivariable proportional hazards regression analyses were performed to identify factors associated with 1-year and 3-year mortality.

Variables significantly associated with the outcomes in the univariate analysis were introduced in the multivariate Cox models by a stepwise procedure. Comparative risk estimates were expressed as hazard ratios (HR) and 95% Confidence Interval (CI). Statistical significance was set at a 2-tailed p-value of 0.05.

A further analysis of the mortality outcome occurring after the first year from the index date within the third year was performed to evaluate the determinants of mortality in this specific period.

All data were analysed using SAS version 9.4 (SAS Institute, Cary, NC, USA).

Results

In the cohort of 6,997 incident chronic HD patients during the period 2008–2016 the majority was male (64.5%), over 65 years old (66.1%), born in Italy (91.7%) and residing in the Municipality of Rome (50.2%).

Characteristics of the study population by periods.

The baseline characteristics of the population at the index date according to periods of incidence in HD are shown in Table 1.

Table 1 Demographic and clinical characteristics of HD incident patients, by period of incidence in HD
Full size table

Statistically significant differences between periods were found for site of birth, educational level and self-sufficiency, respectively with a higher percentage of patients born abroad of Italy (9.8% vs. 7.0%), with a higher level of education (36.9% vs. 30.0%) and totally autonomous patients (56.5% vs. 36.2%), in the third period respect to the first one. The prevalence of comorbidities such as severe hypertension, cancer, thyroid disease and obesity increased in a statistically significant way across periods (Table 1). A lower percentage of FAV (47.3% vs. 57.4%) and of patients receiving pre-dialysis counselling (71.7% vs. 78.0%) was found in the third period compared to the first one. A higher percentage of patients undergoing dialysis in private clinic (48.5% vs. 46.6%) was found in the third period compared to the first one. Finally, statistically significant differences between periods were found for all laboratory findings.

Characteristics of the study population by vital status

Table 2 shows characteristics of the study population by vital status at 1 year and 3 years and their corresponding unadjusted cause-specific HRs along with 95%CI.

Table 2 Demographic and clinical characteristics of HD patients by vital status for 1 and 3-year mortality and unadjusted cause-specific HR.
Full size table

1-year mortality

923 out of 6,997 died within 1 year of follow-up; the percentage of censored subjects during the first year of dialysis was 3% of which 1% refers to patients who received renal transplantation; those percentages were similar during the three periods of study. One-year crude mortality rates per 100 Person Years in the three periods were 2008–2010: 13.5 (95%CI: 12.1–15.2); 2011–2013: 14.2 (95%CI: 12.7–15.9); 2014–2016: 14.5 (95%CI: 13.0-16.2).

One-year mortality was higher for subjects over 65 vs. under 65 years (HR: 2.78, 95%CI: 2.33–3.31), born in Italy vs. born in other countries (HR: 3.75, 95%CI: 2.48–5.66), resident in Rome Province and in Other Lazio’s municipalities vs. resident in Rome municipality (HR: 1.27, 95%CI: 1.09–1.49 and 1.25, 95%CI: 1.07–1.45, respectively), with a low level of education vs. high (HR: 1.45, 95%CI: 1.25–1.68), BMI < 18.0 kg/m2 vs. normal weight (HR: 1.76, 95%CI: 1.39–2.21), not self-sufficient vs. total autonomy (HR: 5.82, 95%CI: 4.89–6.92). The nephropathy with the highest 1-year mortality was the systemic one vs. unknown nephropathy (HR: 1.67 95%CI: 1.22–2.28), and the HRs for comorbidities significantly associated with 1-year mortality ranged from 0.62 for severe hypertension to 2.28 for malnutrition. Finally, 1-year mortality was higher for users of diabetes drugs (HR: 1.20, 95%CI: 1.04–1.38), for users of anti-anaemic drugs (HR: 1.50, 95%CI: 1.18–1.91) and for subjects with a CVC as vascular access vs. FAV (HR: 3.05, 95%CI: 2.64–3.52).

3-year mortality

2,253 out of 6,997 died within 3 years of follow-up; the percentage of censored subjects during the 3 years of dialysis was 9% of which 5% refer to patients who received renal transplantation; those percentages were similar during the three periods of study. Three-year crude mortality rates per 100 Person Years in the three periods were 2008–2010: 13.7 (95% CI: 12.8–14.7); 2011–2013: 13.9(95% CI: 12.9–14.9); 2014–2016: 13.5(95%CI: 12.6–14.5). Three-year mortality was higher for subjects older than 65 years vs. younger (HR: 2.95, 95%CI: 2.64–3.30), born in Italy vs. born in other countries (HR: 3.06, 95%CI:2.42–3.85), resident in Rome Province and in Other Lazio’s municipalities vs. resident in Rome municipality (HR: 1.22, 95%CI: 1.10–1.35 and 1.20, 95%CI: 1.08–1.32, respectively) with a low level of education vs. high (HR: 1.39, 95%CI: 1.26–1.52), BMI < 18.0 kg/m2 vs. normal weight (HR: 1.44, 95%CI: 1.22–1.70), and not self-sufficient vs. total autonomy (HR: 4.36, 95%CI: 3.93–4.84). The nephropathy with the highest 3-year mortality was the systemic one vs. unknown nephropathy (HR: 1.34, 95%CI: 1.07–1.67), and the HRs for the comorbidities significantly associated with 3-year mortality ranged from 0.69 for severe hypertension to 2.02 for dementia and psychiatric disease. Finally, 3-year mortality was higher for users of antidiabetic drugs (HR: 1.23, 95%CI: 1.12–1.34), for users of other cardiovascular drugs (HR: 1.25, 95%CI: 1.14–1.37) and for users of anti-anaemic drugs (HR: 1.24, 95%CI: 1.07–1.43), and for subjects with a CVC as vascular access vs. FAV (HR: 2.09, 95%CI: 1.92–2.27).

Mortality annual trend

Figure 1 shows the trend of the crude annual 1-year and 3-year mortality rate by year of incidence, gender and age classes. The crude mortality rate remained largely unchanged over the years with small annual fluctuations in both 1-year and 3-year mortality. Even after stratification by gender and age classes no significant changes emerged over the years. The JoinPoint analysis confirmed the absence of any trends in annual rates for the two outcomes, both with the assumption of zero joinpoint and with the assumption of one joinpoint. (Supplementary Figure S1)

Figure 1
figure 1

Crude 1-year and 3-year mortality rates by calendar years of incidence in HD

Full size image

Association between periods of incidence in HD and 1- year and 3-year mortality

Kaplan-Meir curves show the absence of a statistically significant differences in survival at 1 year (log-rank test p = 0.666) and 3 years (log-rank test p = 0.882) since HD incidence across the three considered periods. The same results were confirmed when stratifying by age groups (survival at 1 year < 65: log-rank test p = 0.705, ≥ 65: log-rank test p = 0. 563; survival at 3 years < 65: log-rank test p = 0.955, ≥ 65: log-rank test p = 0.990). The 1-year survival probability was: 0.94 (95%CI: 0.93–0.95) in age class < 65 years and 0.83 (95%CI: 0.82–0.84) in age class ≥ 65 years, this difference increased considering 3-year survival: 0.83 (95%CI: 0.81–0.85) and 0.58 (95%CI: 0.57–0.59) respectively. Kaplan-Meir curves are shown in Supplementary Figure S2.

Variables found to be significantly associated with mortality at 1 year and 3 years and included in the multivariate analysis were: age at incidence, place of birth, residence, BMI, self-sufficiency, cause of ESRD, severe hypertension, heart disease, cancer, peripheral vascular diseases, liver disease, dementia and psychiatric disease, type of vascular access, HB, serum creatinine, serum albumin. For mortality at 3 years COPD and lipid metabolism’s alteration were also found.

In the adjusted Cox regression model for 1-year mortality HRs of 1.10 95%CI:0.93–1.29 and 1.06 95%CI:0.90–1.25 were found, respectively in periods 2011–2013 and 2014–2016 compared to the reference period 2008–2010.

Similar results were obtained for 3-year mortality with HRs of 1.05 95%CI:0.94–1.17 and 1.03 95%CI: 0.93–1.14, respectively in periods 2011–2013 and 2014–2016 compared to the reference period 2008–2010.

Factors associated with 1- year and 3-year mortality

After stepwise regression factors associated with 1-year and 3-year mortality were: age at incidence, place of birth, residence, BMI, self-sufficiency, cause of ESRD, severe hypertension, heart disease, peripheral vascular diseases, cancer, liver disease, dementia and psychiatric disease, type of vascular access, haemoglobin, serum creatinine, serum albumin. COPD and lipid metabolism’s alteration were included as covariates only in the 3-year mortality Cox model. HR and 95% CIs are reported in Fig. 2.

Figure 2
figure 2

Factors associated within 1 and 3 years from incidence in HD. Multivariable Cox regression analysis

Full size image

The results for the mortality outcome occurring after the first year from the index date within the third year were presented in Supplementary Table S1.

Discussion

The present study showed that the 1-year and 3-year mortality in incident HD patients was stable over a period of 9 years, since 2008 to 2016 in Lazio, a central region of Italy including the city of Rome. This observation is confirmed both by Kaplan-Meier survival curves, also after stratification by age and gender, and by raw and adjusted Cox regression models. In the light of an increasing complexity of patients starting HD over time, this is a positive signal, which confirms the consistent quality of HD assistance in Lazio.

Nevertheless, about 14% and 35% of incident HD patients died within the first and third year since the start of HD, in the three calendar periods, respectively, and the hypothesis of increasing survival as a consequence of advances in clinical care and specifically of improvements of dialysis techniques and materials, new medications, increase of life expectancy, better management of comorbidities and malnutrition, better prevention of infections and an uptake of clinical practice guidelines, occurring over time, was not met.

Using a different analytic approach, two Registry studies have demonstrated, recently, that mortality rates decreased significantly over calendar periods in patients initiating HD. Foster et al. [7] have demonstrated that among almost 2 million individuals initiating ESRD care in the United States from 1995 to 2013, the excess risk of ESRD-related compared relative to the general population decreased significantly with advancing calendar time. Similarly, data from the ERA-EDTA Registry show that, between 2007 and 2015, in patients initiating HD, the excess mortality rate decreased from 178 extra deaths per 1000 person-year to 154 extra deaths per 1,000 person-years [1]. In the UK Registry essentially no change was observed in the death rate from 2013 to 2014 on a background of a declining trend in the death rate overall and over the past decade [9].

The lack of improvement in 1-year and 3-year mortality in the period 2008–2016 in the Lazio incident HD patients is difficult to explain, taking into consideration that the survival had an about 1-year improvement in the general population in the same period, either in Lazio and in Italy [23] and considering that the dialysis centres of Lazio included in the RRDTL follow the international and national guidelines for the treatment of ESRD patients and offer every advance in patient care. Moreover, unvaried mortality trends were observed in all ages (< 65 years, ≥ 65 years), which is in disagreements with findings from other studies. Foster et al. [7] reported that younger incident HD patients showed significant larger relative improvements in excess mortality than older people and Boenink et al. [1] demonstrated that the strongest decrease in excess mortality was observed in patients ≥ 65 years, with smaller reductions among the youngest patients. Accordingly, in the UK registry, the death rate per 1,000 patient years in the first year of starting HD from 2003 to 2012 showed a declining trend with a larger rate of decline in patients aging ≥ 65 years [9]. We underline that the purpose of our study is to assess the trend and determinants of mortality in HD, not to make comparisons between regions. However, comparisons with mortality rates observed from other registries should be made with caution, as there may be a different case mix of people on HD. In particular, the lower renal transplant and peritoneal dialysis supply rates may result in younger, healthier individuals among HD patients in Lazio. Furthermore, the rates of transplantation and peritoneal dialysis incidence in Lazio did not change substantially during the period of the study, with annul standardised kidney transplant rates per 100,000 ranging from 2.4 to 3.6 in the period 2008–2017 [24] and the proportion of patients treated with peritoneal dialysis among dialysis patients range from 8 to 13% [16]. Therefore, it is unlikely that the potential selection bias could have affected the analyses and the conclusions of our study.

Another relevant finding of the present study is that the factors associated with the risk of dying were almost the same at 1 and 3 years. As expected, mortality was higher for patients over 65 years of age, underweight according to BMI, not self-sufficient, and with a catheter as first vascular access. Moreover, mortality was lower in patients born in countries other than Italy probably due to the fact that these patients where younger than Italians (mean age 52 vs. 69 years).

In addition, for both 1-year and 3-year mortality, the comorbidity factors of greatest impact were, dementia or psychiatric disorders, heart disease, peripheral vascular disease, COPD, cancer, and liver disease. The observation that patient characteristics at the time of dialysis entry associated with mortality were almost the same for both 1-year mortality and 3-year mortality may suggest a longer-term effect of patient conditions at baseline.

Finally, variables associated with 1-year and 3-year mortality did not differ significantly among the three periods of incidence. Interestingly, in the study by Boenink et al. [1], ESRD patients starting HD showed a decrease in excess mortality for all causes of death with advancing calendar time, especially for atheromatous cardiovascular disease and infections. The RRDTL makes it possible to collect a lot of information about the incident dialysis patients that allows a precise description of individual socio-demographic characteristics and clinical conditions. These characteristics were considered in the models to obtain an adjusted estimate of the 1-year and 3-year mortality. However, residual confounding due to factors not retrievable in the RRDTL survey might play a role. One strength of this study is the inclusion of a large population with a long-term follow-up. To the best of our knowledge, this is one of the few studies that have recently evaluated early survival in a European population of dialysis patients. A further strength is the availability of a regional dialysis registry (RRDTL), which provides detailed data of all dialysis patients of Lazio region, that can be integrated by data from other regional healthcare databases, making it a powerful epidemiological tool to monitor survival rates.

Conclusions

In summary, the present study showed that 1- and 3-year mortality in patients with end-stage renal disease starting HD in the Lazio region, was stable over 9 years. Investigation on longer term mortality (e.g., 5-year) may add information helpful to interpret the phenomenon. In the meantime, evidence from previous studies in the same context could be considered for potential improvements, such as an increased use of FAV which has been associated with better outcomes [17].