Glomerular Function Time Trends in Long-Term Survivors of Childhood Cancer: A Longitudinal Study

Because of the advances in the treatment of childhood cancer, most patients are expected to become long term survivors (1, 2). However, the improved prognosis has been accompanied by late, treatment related complications. As a result, childhood cancer survivors (CCS) are a growing group of individuals who are at risk of premature morbidity and mortality (3, 4).

Nephrotoxicity is one of the late effects identified in CCS, presenting either as a decline in glomerular filtration rate (GFR), deterioration of tubular function and/or proteinuria. It has been suggested that CCS treated with ifosfamide, cisplatin, carboplatin, high-dose (HD-)methotrexate, HD-cyclophosphamide, radiotherapy to the kidney region, and/or (partial) nephrectomy are at risk for nephrotoxicity (5–17). Recently, we showed in a large cross-sectional cohort study, including 1,442 CCS with a median followup of 12 years after diagnosis, that 28% had ≥1 renal late effect(s), of whom, 62 (4.6%) developed glomerular dysfunction (GFR <90 mL/min/1.73 m²). Higher ifosfamide, cisplatin, and carboplatin dose, high-dose (HD-)cyclophosphamide, nephrectomy, and longer follow-up were identified as risk factors for glomerular dysfunction (18).

It is, however, not clear whether CCS will show recovery, deterioration or no change in glomerular function, and what the effects of specific cancer treatments are on the pattern over time. A few smaller cohort studies with limited followup durations have evaluated the change in renal function over time, but results have been inconclusive (7–11, 19).

Chronic kidney disease may affect quality of life and can lead to premature co-morbidity such as cardiovascular disease, anemia, bone disease, and malnutrition (20). To establish adequate followup protocols for CCS, it is important to know their pattern of glomerular function over time. Therefore, the aim of this study was to investigate time trends and predictors of glomerular function in a large cohort of long term CCS.

In 1996, the Late Effects Outpatient Clinic for CCS was established in the Emma Children's Hospital/Academic Medical Center (EKZ/AMC). Here, we offer medical followup to CCS for the assessment of late effects. All five-year CCS were identified using the hospital's Childhood Cancer Registry (established in 1966), which contains data on all patients treated for childhood cancer in the EKZ/AMC regarding diagnosis, treatment, and follow-up. All CCS surviving ≥5 years after diagnosis and still alive at the start of the outpatient clinic were invited to participate and were subsequently enrolled into prospective followup protocols tailored to previous diagnosis and treatment. To be eligible for this study, patients had to meet the following criteria: 1) diagnosed between January 1966 and January 2003 and treated for a primary malignancy; 2) aged <18 years at diagnosis; 3) treated primarily in the EKZ/AMC; 4) survived ≥5 years after diagnosis; and 5) aged 18 years or older at glomerular function testing.

All CCS who visited the outpatient clinic between January 1996 and July 2010 received a full medical assessment, consisting of a medical history, physical examination, and additional risk-based diagnostic tests and counseling. CCS were seen at regular intervals (every 1, 2, or 5 years), depending on previous cancer treatment and assumed risk for late effects. For example, low-risk childhood cancer survivors (e.g. survivors treated with surgery only) were invited every 5 years, while high-risk survivors (e.g. survivors treated with radiotherapy and/or chemotherapy and surgery) were invited every year. For each CCS, we recorded patient and treatment characteristics and serum creatinine measurements from each outpatient clinic visit.

The EKZ/AMC institutional review board reviewed and approved the collection of data used for the analyses presented. Written informed consent was obtained from all participants.

To assess glomerular function, we measured serum creatinine using an enzymatic IDMS-traceable method (Roche P800, Roche Diagnostics). We calculated the GFR using the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) formula for adult CCS (21, 22). We defined glomerular dysfunction as a GFR <90 mL/min/1.73 m² as recommended by the Kidney Disease Outcome Quality Initiative guidelines (23).

To evaluate the patterns of glomerular function over time, we fitted multivariable linear random effects models for GFR as a continuous outcome variable, and multivariable logistic regression models for GFR as a dichotomous outcome variable.

In the random effects models, estimates were obtained via restricted maximum likelihood. In the logistic regression models, estimates were obtained using generalized estimating equations (GEE) with a compound symmetry working correlation matrix. The random effects models and GEE adjust for the correlation between repeated observations measured in the same subject and can handle longitudinal data on subjects with varying number of and unequally spaced observations. All performed GFR measurements were taken into account. Followup time was included as a continuous variable.

We first fitted models that only included followup time and had received any type of potentially nephrotoxic therapy, allowing for an interaction between the two. Next, we included all following potentially nephrotoxic therapies in the prediction models: administration and cumulative doses of ifosfamide, cisplatin, and carboplatin, administration of HD-methotrexate (≥1 g/m²/course) and HD-cyclophosphamide (≥1 g/m²/course or a total cumulative dose of ≥10 g/m²), radiotherapy to the kidney region, and (partial) nephrectomy. In addition, sex and age at diagnosis were included in the models. We allowed for interactions between age at diagnosis and nephrectomy, and all individual treatments (including cumulative ifosfamide, cisplatin, and carboplatin dose) and followup time. We modeled the effects of followup time and age at diagnosis via restricted cubic splines with 4 knots to allow for a non-linear relationship with the outcome variables. Because parameter estimates of individual spline components are hard to interpret, we presented results graphically using predicted values from the models. Plots were stratified by important determinants of glomerular function. To graphically present the effects of cumulative dose and age at diagnosis (all continuous variables), we divided their distribution into tertiles. Variable reference values and P values are presented in the figure legends.

We carried out analyses using R version 2.14 (R Foundation for Statistical Computing). All statistical tests were two sided and P values < 0.01 were considered statistically significant.

The study population consisted of 1,251 out of 1,502 eligible adult CCS, of whom 1,122 (89.7%) underwent glomerular function testing (Supplementary Fig. S1). Clinical characteristics of CCS are presented in Tables 1 and 2, and Supplementary Table S1. In total, 6,412 GFR measurements have been carried out in 1,122 CCS, of whom 920 (82.0%) had repeated observations. Of these 920 CCS, the median followup years from first until last glomerular function test was 7.3 years (range 0.8–14.3), with a median of 6 measurements per subject (range 2–15). The screening frequency was comparable between CCS treated with and without nephrotoxic therapy (0.96 and 0.95 per year, respectively), and between CCS with a normal and an abnormal GFR during the course of followup (0.95 and 0.94 per year, respectively).

In Fig. 1, the patterns of GFR over time for CCS treated with and without potentially nephrotoxic therapy are presented. GFR declined in both groups during followup (P < 0.001). At 5 years after diagnosis, the mean GFR for CCS treated with and without potentially nephrotoxic therapy was 132.1 mL/min/1.73 m² (95% CI, 130.5–133.6) and 139.0 mL/min/1.73 m² (95% CI, 137.0–141.1), respectively. At 35 years after diagnosis, the mean GFR for CCS treated with and without potentially nephrotoxic therapy was 95.2 mL/min/1.73 m² (95% CI, 92.2–97.9) and 100.2 mL/min/1.73 m² (95% CI, 98.1–102.3), respectively. The differences in GFR between both groups were highly significant (P < 0.001), but the differences in time trends were not (P = 0.04).

In the multivariable model, the GFR deteriorated over time in all CCS (P < 0.001). Figure 2 shows the GFR time trends for the separate potentially nephrotoxic therapies. For ifosfamide, cisplatin, and carboplatin, curves are shown for a cumulative dose of 30 g/m² ifosfamide, 320 mg/m² cisplatin, and 2,000 mg/m² carboplatin. In Supplementary Fig. S2, the GFR time trends for different cumulative doses of ifosfamide and cisplatin are presented.

Overall, ifosfamide was associated with a lower GFR (P < 0.001). No ifosfamide by time interaction was found (P = 0.08), meaning that the pattern of GFR over time was not significantly different for CCS treated with and without ifosfamide (Fig. 2A). As shown in Supplementary Fig. S2, A–C, a higher cumulative ifosfamide dose was associated with a lower GFR (P < 0.001). The time trends did not significantly differ by whether lower or higher doses of ifosfamide were given (P = 0.09).

Cisplatin had a significant overall effect on GFR (P <0.001). This effect seems to be related to cumulative cisplatin dose (P < 0.001), with higher dose (especially ≥500 mg/m² cisplatin) related to a lower GFR. In addition, a cisplatin by-time (P = 0.002) and a cisplatin dose-by-time interaction (P = 0.004) was shown, indicating that CCS treated with higher doses of cisplatin showed different GFR time trends as compared with CCS treated without and with lower doses of cisplatin (Fig. 2B and Supplementary Fig. S2, D–F). The deterioration rate was higher in CCS treated with higher doses of cisplatin versus lower doses upto 25 years after childhood cancer diagnosis.

Overall, carboplatin was associated with a lower GFR (P = 0.006). No significant effect of cumulative carboplatin dose was observed (P = 0.07). The time trends did not significantly differ by whether carboplatin had been given and by cumulative carboplatin dose (P = 0.24; Fig. 2C).

Overall, HD-cyclophosphamide was associated with a lower GFR (P = 0.005). A significant interaction with time was shown (P = 0.006), indicating that CCS treated with and without HD-cyclophosphamide showed different GFR time trends, although differences were small (Fig. 2D).

HD-methotrexate (P = 0.07), and radiotherapy without chemotherapy (P = 0.012) were overall not significantly associated with GFR. No significant interactions with time for HD-methotrexate (P = 0.17) and radiotherapy (P = 0.04) were found (Fig. 2E and F).

Nephrectomy was associated with a lower GFR (P < 0.001). The GFR time trend was not significantly different for CCS treated with and without nephrectomy (P = 0.26; Fig. 2G). In addition, older age at childhood cancer diagnosis was associated with a lower GFR (P < 0.001). An interaction effect between nephrectomy and age at diagnosis was shown (P = 0.002), meaning that the effect of nephrectomy on GFR was different for CCS nephrectomized at an older age versus CCS nephrectomized at a younger age (Supplementary Fig. S3). CCS nephrectomized at an older age showed a faster decline in GFR as compared with CCS nephrectomized at a younger age.

No significant effect of sex was shown (P = 0.34).

To confirm the clinical relevance of the results on continuous GFR, we present the time trends in glomerular dysfunction probability (GFR <90 mL/min/1.73 m²) for CCS treated with and without potentially nephrotoxic therapy in Fig. 3. The glomerular dysfunction probability increased in both groups (P < 0.001). At 15 years after diagnosis, the mean glomerular dysfunction probability for CCS treated with and without potentially nephrotoxic therapy was 5.4% (95% CI, 4.0–7.4) and 1.7% (95% CI, 0.1–5.2), respectively. At 35 years after diagnosis, the mean glomerular dysfunction probability for CCS treated with and without potentially nephrotoxic therapy was 26.4% (95% CI, 20.6–33.0) and 6.6% (95% CI, 4.4–9.6), respectively. These differences were highly significant (P < 0.001), but there were no differences in time trends between the two groups (P = 0.11).

The glomerular dysfunction probability time trends for the separate potentially nephrotoxic therapies and for different cumulative doses of ifosfamide and cisplatin are shown in Fig. 4 and Supplementary Fig. S4. As in line with the results on continuous GFR, the glomerular dysfunction probability increased over time in all CSS (P < 0.001). In addition, the analyses confirmed that ifosfamide (P < 0.001), higher cumulative ifosfamide dose (P < 0.001), cisplatin (P < 0.001), higher cumulative cisplatin dose (effect most profound in the highest dose group of ≥500 mg/m²; P < 0.001), and nephrectomy (P < 0.001) were associated with a higher glomerular dysfunction probability (Fig. 4A, B and G and Supplementary Fig. S4, A–F). In addition, the glomerular dysfunction probability time trend was influenced by whether cisplatin was given and by cumulative cisplatin dose (P = 0.005). The glomerular dysfunction probability increased faster in CCS treated with higher doses of cisplatin versus lower doses upto 35 years after childhood cancer diagnosis.

In contrast with the results on continuous GFR, HD-cyclophosphamide (P = 0.095) was not significantly associated with glomerular dysfunction probability (Fig. 4D). In addition, a significant interaction with time for nephrectomy was observed (P = 0.002), indicating that the pattern of glomerular dysfunction probability over time was different for CCS treated with, versus without nephrectomy (Fig. 4G). Moreover, no interaction between nephrectomy and age at diagnosis was observed (P = 0.29).

This is the first large longitudinal cohort study reliably investigating time trends and predictors of glomerular function over time in adult CCS, taking into account the correlation between repeated observations measured within the same subject. This study shows that CCS treated with nephrotoxic therapy have a worse glomerular function upto 35 years after cancer diagnosis compared with CCS treated without nephrotoxic therapy. Especially ifosfamide, higher doses of cisplatin, and nephrectomy are associated with lower GFR and with glomerular dysfunction (GFR <90 mL/min/1.73 m²), which persists during the entire followup period. These CCS experience an early decline in glomerular function that does not recover. In all CCS, irrespective of treatment, glomerular function deteriorates over time. This decline starts at least as early as 5 years after cancer diagnosis and persists over the following 30 years. There is a suggestion that the glomerular function deterioration rate is higher in CCS treated with higher doses of cisplatin as compared with other CCS. Strengths of this study are the long and nearly complete followup, and the multivariable longitudinal analyses including precise information on treatment-related predictors.

To date, a few small studies have addressed glomerular function over time in CCS. Two studies investigated glomerular function in 63 CCS treated with cisplatin and/or carboplatin and 25 with ifosfamide, 1 and 10 years after completion of treatment (7, 8). Patients showed either recovery or deterioration in glomerular function, but the time trends were not significant. Two studies did also not show significant differences in GFR patterns in 23 CCS treated with carboplatin and 14 with cisplatin after 2 and 6 years, respectively (9, 10), whereas another study showed a significant recovery of GFR 4 years after end of treatment among 40 CCS treated with cisplatin (19). The results of these studies must, however, be interpreted with caution, because sample sizes were small, followup was relatively short, and only few followup measurements were taken into account.

In the general population, the decline in glomerular function is a normal and expected phenomenon. The decline in creatinine clearance is a physiologic and pathologic consequence of aging (24–26). A cross-sectional study among 174,448 subjects aged ≥18 years showed that the age-related decline amounts 8–11 mL/min/1.73 m² per 10 years (25). However, as longitudinal data on glomerular function within the normal population is not available, we do not know if the decline in CCS is worse than the decline in the general population. We used CCS treated without potentially nephrotoxic therapy as an internal control group. Although this may have led to an underestimation of the effect sizes in the risk factor analyses (e.g. due to antibiotic treatment for neutropenia in the control group which might increase the risk of renal disease), we had the advantage that longitudinal data was available for both groups and that the GFR assessment method was identical. It is essential to realize that our population is still young (median age 28 years) and has a long life expectancy. As glomerular function continues to deteriorate, CCS will be at increased risk for premature chronic renal failure.

Thus far, no studies have evaluated predictors of glomerular function patterns over time in CCS. We showed that especially CCS treated with cisplatin seem to have different glomerular function patterns over time as compared with CCS treated without cisplatin, particularly the higher cisplatin doses. There is an indication that higher doses of cisplatin are associated with a faster increase in glomerular dysfunction probability over time.

As in line with our results, previous cross-sectional studies showed that higher ifosfamide and cisplatin doses, and nephrectomy were associated with diminished GFR (12–14, 18). Moreover, we observed that HD-cyclophosphamide was related with a lower GFR, but not with a GFR <90 mL/min/1.73 m². The effect of HD-cyclophosphamide seems more a temporary than a longterm effect, because the association decreases over time. Our results on HD-cyclophosphamide are not in concordance with the current literature which suggests no pathophysiologic evidence for glomerular toxicity (5).

We observed that older age at diagnosis is associated with a decline in GFR over time, especially in CCS treated with nephrectomy. Previous studies in Wilms' tumor and neuroblastoma survivors did not find an influence of age at nephrectomy (27–29). However, some other cross-sectional studies investigating nephrotoxicity of ifosfamide and platinum derivates did report correlations between GFR and older age at treatment (7, 12). A possible explanation for the effect of older age at diagnosis in nephrectomized CCS may be that a younger kidney has a higher functional-reserve capacity after a nephrectomy than an older kidney.

Although the study population was representative and consisted of 83% of the original cohort of adult CCS, selection bias cannot be completely ruled out. There were relatively less CCS treated with ifosfamide and carboplatin in the glomerular-function testing group as compared with CCS not included in the study group. Because ifosfamide and carboplatin are nephrotoxic, this may have underestimated our observed effects on glomerular function. In addition, a larger proportion of CCS who were not included in the study group had suffered a recurrence and had died during followup, which in turn may have underestimated the results if these CCS were in worse health than those still alive and without a recurrence. Other biases may have also been present in this study as the frequency of followup visits for CCS with more severe health problems may be higher than for healthy CCS. The GEE approach assumes data to be missing completely at random. This means that it is assumed that the availability of measurements in an individual does not depend on the actual value of the outcome, conditionally on the co-variables in the model and that might not be true. However, we have no indication that CCS treated with nephrotoxic therapy and with an abnormal GFR had more frequent screening than other CCS. Moreover, it is likely that non renal health problems were the main cause for variations in followup visit frequency. Therefore, the risk of bias due to differential frequency of measurement in our study is most likely limited.

The use of the CKD-EPI formula to calculate estimated GFR might be another drawback of this study. It is well established that no formula is as reliable as a gold standard clearance measurement such as the ⁵¹Cr-EDTA clearance (30). GFR estimating formulae have an increasing error in estimating the GFR when the value is above 60 mL/min/1.93 m² (21, 31). However, clearance measurement is a cumbersome, invasive test, not suitable for screening purposes. Earley and colleagues found that the CKD-EPI formula was the estimation method of preference for public health and general clinical practice usage when compared with the MDRD-formula (22). Furthermore, a low GFR (<60 mL/min/1.93 m²), as calculated by estimation formulae, was associated with all-cause and cardiovascular mortality in the general population in a large meta-analysis by Matsushita and colleagues (32), indicating the clinical relevance of using the estimated GFR.

Initially we included all CCS, both adults and children, in our study. For CCS aged <18 years, we calculated the GFR by the Schwartz formula (33). However, there are no data showing the relationship between GFR values estimated for children using the Schwartz formula and the GFR values estimated for adults using the CKD-EPI formula. When analyzing children and adults together, we found out that it was not reliable to compare the GFR values of both groups (Supplementary Fig. S5). We observed that at the time CCS shifted from the Schwartz to the CKD-EPI formula, the GFR value increased considerably, biasing the results. Therefore, we decided to conduct separate analyses for children and adults. Due to the low power, i.e., small number of followup measurements in CCS aged <18 years, we were only able to evaluate the glomerular function time trends in adult CCS aged ≥18 years. Future studies investigating glomerular function in both children and adults should always be aware of the discordance between these formulae.

In conclusion, CCS treated with nephrotoxic therapy, especially ifosfamide, high doses of cisplatin and nephrectomy, have a worse glomerular function compared with other CCS and this does not recover. We have an indication that CCS treated with higher doses of cisplatin experience faster decline in glomerular function than other CCS. Moreover, as the other treatments could not explain the differences in deterioration rates, there might be a suggestion that the first glomerular function assessment may be a more sufficient predictor for the pattern of glomerular function over time instead of nephrotoxic treatment. As glomerular function continues to deteriorate, CCS are at risk for premature chronic renal failure. For CCS treated with ifosfamide, higher doses of cisplatin and nephrectomy, special attention is warranted as deterioration occurs earlier which poses this group at additional premature risk. Future studies should focus on the predictive value of the first glomerular function assessment and on the longitudinal evaluation of other outcomes related to glomerular function, including proteinuria and hypertension.

No potential conflicts of interest were disclosed.

Conception and design: R. L. Mulder, S.L. Knijnenburg, E.C. van Dalen, H.J.H. van der Pal, A.H. Bouts, H.N. Caron, L.C.M. Kremer

Development of methodology: R.L. Mulder, S.L. Knijnenburg, R.B. Geskus, L.C.M. Kremer

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H.J.H. van der Pal

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. L. Mulder, S.L. Knijnenburg, R.B. Geskus, E.C. van Dalen, H.J.H. van der Pal, L.C.M. Kremer

Writing, review, and/or revision of the manuscript: R.L. Mulder, S.L. Knijnenburg, R.B. Geskus, E.C. van Dalen, H.J.H. van der Pal, C.C.E. Koning, A.H. Bouts, H.N. Caron, L.C.M. Kremer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R.L. Mulder, S.L. Knijnenburg, H.J.H. van der Pal

Study supervision: L.C.M. Kremer

This study was supported by the Tom Voûte Fund, Amsterdam, the Netherlands. The funding organization had no involvement in the design and conduct of the study; in the collection, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.