Estimated Number of Patients with Chronic Renal Failure but not with End-Stage Renal Disease in Japan : Comparisons Between Two Estimation Methods

We estimated the number of patients with chronic renal failure (CRF) but not with end-stage renal disease (ESRD) in Japan by two methods: a nationwide survey and an estimation using incidence rates of ESRD among CRF patients as well as the number of incident ESRD. The former estimated the number of patients with CRF but not with ESRD as 154,000 (95% confidence interval 139,000-168,000), while so did the latter as 142,000 (133,000-150,000). It follows that the number of CRF patients including ESRD could amount to more than 300,000, which would give a substantial impact on public health in Japan. Despite some own methodological issues in estimation, the two estimates were reasonably in good agreement: supporting their methodological validity in estimation, since sources of errors appeared to be quite independent of each other. J Epidemiol, 1998 ; 8 : 278-284.

Patients with end-stage renal disease (ESRD) have been rapidly increasing in Japan.Currently, 167,000 patients undergo long-term renal replacement therapy, and another 29,000 have begun long-term renal replacement therapy, such as hemodialysis (HD) or continuous ambulatory peritoneal dialysis (CAPD), in 1996 1).
Features of patients with ESRD in Japan have been extensively described clinically and somewhat epidemiologically by the nationwide annual survey conducted by the Japanese Society for Dialysis Therapy 2,3).Little is known, however, on epidemiological features of patients with chronic renal failure (CRF) but not with ESRD.To disclose the impact of CRF on public health in Japan, we estimated the number of CRF patients without ESRD by two methods, that is, a nationwide survey, and an estimation using the incidence rates of ESRD among CRF patients as well as the number of incident ESRD cases.

MATERIALS AND METHODS
The subjects surveyed were the patients with CRF but not with ESRD.Serum creatinine concentration of 2.0 mg/dl or higher was defined as CRF.We will denote these subjects simply by "CRF patients" hereafter in our text and tables.

Nationwide survey
The nationwide survey was conducted according to a modified method of nationwide epidemiological surveys of socalled intractable diseases in Japan 4-7).The departments surveyed were randomly selected by stratified sampling from all the departments of internal medicine and urology, and other relevant sections in hospitals and clinics throughout Japan.The medical facilities were stratified according to type (clinics, general hospitals or university hospitals) as well as to number of beds (Table 1).The sampling was conducted using the registry of all the clinics and hospitals in Japan, which was obtained from the Ministry of Health and Welfare by permission.In this survey, a questionnaire was distributed to the departments selected.The questionnaire was directly mailed to the heads of 3,243 departments in October 1996, inquiring the number of patients with CRF, who satisfied the criteria (*2.0 mg/dl of serum creatinine concentration) and were regularly treated at the clinic/hospital at the survey.
The number of patients treated for CRF was estimated, based on the assumption that the response from departments is independent of the number of patients treated there 8).
The number of patients in stratum k was estimated as where sk is the estimated standard error of *k.The total number of patients a was computed as follows: where nk, Nk and Nk* denote the total number of departments, the number of responded departments, and the number of departments with i patients in stratum k, respectively.The 95% confidence interval (CI) of *k was and the 95% CI was where s is the estimated standard error of * .

Estimation using incidence rates of ESRD among CRF patients
We estimated the number of CRF patients using the data from a database of CRF patients as well as the number of incident ESRD in Japan.Serum creatinine concentration of 2.0 mg/dl or higher was also defined as CRF in this CRF database.Since December 1995, 7,893 patients with CRF were registered in the database from at least 193 clinics, 437 general hospitals and 69 university hospitals throughout Japan.A total of 377 (5.0%), 5,438 (72.8%) and 1,656 (22.2%) patients were reported from clinics, general hospitals and university hospitals, respectively (Information on medical facilities was missing for 422 patients).They were followed-up by December 1997 by means of annual questionnaire survey to the attending doctors.This database was used to estimate the incidence rates of ESRD among CRF patients.Dropouts from the follow-up, or deaths which occurred before developing ESRD were treated as censored cases.
The number of CRF patients was estimated by sex, cause of CRF and 10-year age group.When incident ESRD cases observed in the database were four or less in one stratum of CRF patients, neighboring age groups were eventually combined to include five cases or more in each stratum.ESRD was documented when a patient underwent long-term renal replacement therapy.
The annual number of incident ESRD cases in 1996 was derived from the nationwide survey on renal replacement therapy which was conducted by the Japanese Society for Dialysis Therapy in 19971).This survey covered virtually all the hospitals/clinics involved in renal replacement therapy with an extremely high response rate (95.8%).The definition of ESRD in this survey was the same as that in the CRF database.
The number of patients in stratum k was estimated as where nk, *k , Nk and ak denote the annual number of incident ESRD cases in Japan, the estimated incidence rate of ESRD among CRF patients, the observed person-years, and the annual number of incident ESRD cases in the CRF database, in stratum k, respectively.The 95% CI of ak was where sk is the estimated standard error of *k (1).
The total number of patients a was computed as follows: where s is the estimated standard error of * .
Because of small sample size, it was difficult to estimate cause-specific incidence of ESRD among CRF patients for the causes other than four primary ones (chronic glomerulonephritis, diabetic nephropathy, nephrosclerosis and polycystic kidney disease).We therefore grouped all patients with other causes of CRF, and estimated the ESRD incidence rate among them.

RESULTS
Nationwide survey Among 3,243 departments, 1,750 replied to the questionnaire; the response rate being 54.0% with 11,735 patients reported in number (Table 1).Using the formula mentioned above, the total number of CRF patients treated was estimated as 154,000 (95% CI 139,000-168,000) throughout Japan circa October 1996.

Estimation using incidence rates of ESRD among CRF patients
Table 2 summarizes the estimated number of patients with CRF caused by four primary diseases and others using data from the CRF database.In the follow-up of 8,291.7 personyears, 1,737 incident ESRD cases were observed; the incidence rate of ESRD among CRF patients being 209.5 per 1,000 person-years.
The nationwide survey by the Japanese Society for Dialysis Therapy 1) identified 28,791 incident cases of ESRD in 1996.This figure is not equal to that in the annual report of the survey 1), since additional cases have been registered after the report was published.Using the formula mentioned above, the total number of patients with CRF was estimated as 142,000 (95% CI 133,000-150,000) throughout Japan circa 1996.

DISCUSSION
We estimated the number of CRF patients without ESRD in Japan by two different approaches, and obtained similar figures of the estimated number (about 140,000 to 150,000).Thus, the number of CRF patients including ESRD could amount to more than 300,000, which would give a substantial impact on public health in Japan.In the strict sense, ESRD is a status of renal failure that The two methods we used have some own methodological issues in estimation, which should be kept in mind when interpreting the findings.Patients being treated at two or more hospitals/clinics will be associated with overestimation in the nationwide survey, while those who did not visit medical facilities would lead to underestimation.
Inaccuracy in reporting number of the patients, and relatively low response rate (54.0%) could result in both overestimation and underestimation.
On the other hand, representativeness of the patients in the CRF database and diagnostic validity of primary cause of CRF might be methodological issues when using data from the CRF database.We estimated number of CRF patients without ESRD under the assumption that the incidence rate of ESRD by sex, age and cause of CRF in the database was the same as that in all Japan.The proportion of patients treated in clinics, however, seems to be smaller in the CRF database than that in the nationwide survey.It is possible, therefore, that the prognosis among the patients in the CRF database was somewhat different from that in all Japan, though direction of this bias is not clear.If diagnostic validity of primary cause of CRF was different between the CRF database and the nationwide survey on renal replacement therapy, renal survival might be different between them: resulting in a biased estimate.CRF is clinically considered to be irreversible, and we assumed that none recovered from the disease.If some recovered, however, the incidence of ESRD among the patients in the CRF database would be underestimated, which result in overestimation of the CRF patients without ESRD.
The direction and magnitude of the above-mentioned errors should be assessed in a population-based registry of CRF patients.The sources of errors in the two estimates, however, appeared to be quite independent of each other.In general, it is unlikely by a mere coincidence that two methods with different sources of errors provided similar estimates 10).The reasonably good agreement between the two estimates, therefore, may support the methodological validity of the two different approaches in estimation.

Table 2 .
Number of patients with chronic renal failure (CRF) caused by the four primary diseases and others, estimated using the incidence rates of end-stage renal disease (ESRD) among CRF patients derived from the CRF database and number of incident ESRD in Japan.a) Rate per 1000 person-years.b) Number of deaths which occurred before developing ESRD.c) Number in 1996.There seems to be no established criteria of CRF for epidemiological studies.CRF was arbitrary defined by serum creatinine of 2.0 mg/dl or higher both in the nationwide survey and in the CRF database, and we estimated number of the patients who satisfied this criteria.A recent, large-scale epidemiological study on CRF in France 9) adopted a cutoff point of 200 µmol/l (2.26 mg/dl), which is not so much different from ours.It would be preferable, however , that the cutoff point of serum creatinine level is determined by comparison with more formal renal function tests such as creatinine clearance test.

Table 1 .
Number of the total, surveyed, responded departments and number of the reported patients with chronic renal failure in the nationwide epidemiological survey.

Table 2 .
Number of patients with chronic renal failure (CRF) caused by the four primary diseases and others, estimated using the incidence rates of end-stage renal disease (ESRD) among CRF patients derived from the CRF database and number of incident FSRD in Janan (continued.