Bernd Grosche Federal Office for Radiation Protection, Germany

Transcription

1 Bernd Grosche Federal Office for Radiation Protection, Germany The work was carried with the support from the Grant agreement no () with the EU under the EURATOM FP7 Programme

2 from Grosche et al, CAJGH (in press)

3 NIIRME, Semey, Kazakhstan NNC, Kurchatov, Kazakhstan NIRS, Chiba, Japan IARC, Lyon, France BfS, Neuherberg, Germany and REA, Tokyo, Japan

4 Assumption: 2 cohorts historical cohort, set up by NIIRME new cohort, established by NNC with financial support from the Japanese Radiation Effects Association (REA). The data from the latter are owned by REA and held by NIRS.

5 from Bauer et al, Rad Res 2005

6

7 Settlement Settlement Settlement Settlement Settlement Dolon Irtyshsk Semenovka Lokot others Znamenka Shcherbakty Sosnovka Novopokrovka unknown Kainar Bolshaya Vladimirovka Izvestka Akbulak Karaul Mostik Besterek Aigyrjal (Kaskabulak) Kenes Cheremushki Sarzhal Tailan (Kundyzdy) Kenjekol Budene Semipalatinsk Bobrovka Jana-aul Malaya Vladimirovka Kurchatov Isa Kachiry Kanonerka Pavlodar Tarkhanka Zhelezinka Semiyarka

8 NIIRME not only holds the historical cohort but a registry of 145,306 individuals (with personal identifiers) considered as being affected by the atomic bomb testing, according to Kazakh governmental legislation

9

10 register includes individuals being considered as affected by the testing since 1949 as well as their offspring. Follow-up of individuals included in the register is conducted on an annual basis mortality since 2002 also incidence data

11

12 Evaluation of individual identifiers Testing possibilities of record linkage between the two cohorts Evaluation of uncertainties in individual record linkage Establishing SOP document for future merging of the two cohorts Decision on data ownership, data access and data storage

13 NIIRME (file: All_people.xls) NNC (file: person_.xls) Variable Name Description Variable Name Description sysid Random numbers identifying cohort members person_id Random numbers identifying cohort members firstname Individual's first name name Individual's first name lastname Individual's last, or family name. Multiple records exist for individuals who experienced name changes, especially women fam1-fam4 Individual's last, or family name. Multiple records exist for individuals who experienced name changes, especially women midlname Individual's patronymic name otch Individual's patronymic name Gender Specifies an individual's sex as male or sex Specifies an individual's sex as male or female female birthdt Date of birth for each individual birth_dt Date of birth for each individual deaddt Date of death death_dt Date of death icd9 Cause of death from international classification of disease coding, revision 9 PlaceName Individual's settlement of residence local_no A number indexing an individual's location (settlement) during a specified time period: this number must be linked to a second file to obtain the name of the location (localtable_.xls) StartDate Start date of period that an individual lived in mother_id person_id of individual's mother, if available a settlement EndDate End date of period that an individual lived in a settlement father_id person_id of individual's father, if available

14 linkage exercise based on the two large data sets Random sample of 3,000 individuals from each of the data sets Using RecLink3 software

15 The structure of individuals is different in the two registries The NIIRME registry has a person-period structure (multiple records per person for each period of observation) while NNC has a person-level structure (one record per person with multiple time periods)

16 Variable Type Reliability individual ID display n/a First name name 5 Last name name 2 Patronymic name name 3 sex number 6 birth year code 6 birth day display n/a birth month display n/a

17 This criteria yields 45 matches among the 3,000 total records. Based on these records, we conclude the following: matches that receive a probability above 90% appear to be true matches; matches between 75%-89% are possible matches; matches below 75% seem unreliable

18 To check whether matches between 75%-89% probably are true matches, it would be necessary to conduct inspection of each individual record. Further information that could aid in this task are second or third last name records, as well as settlement records. However, settlement is time varying, and substantial work would be required to go through the datasets to identify an individual s location at the time of relevant bomb testing.

19 Because of the different structures of the data bases, one would be required to link across multiple datasets and identify the time period at which an individual was at a settlement of interest. A complete linkage of both data bases is possible, but will be very work intensive. It is estimated that 1 person-year would be needed for a complete linkage.

20 NIIRME data set belongs to NIIRME, is continuously updated, and is open for Kazakh researchers on demand. NIIRME wishes to make this data set also available to a broader scientific community include the respective information into the STORE platform ( This work is in progress. REA confirmed that the data from the new cohort may be used in the frame of related research

21 The NIIRME register is a good resource for future cohort studies. Focus on individuals from the exposed settlements (zones of extraordinary and maximal radiation risk) Information from historical cohort: Number of study subjects, duration of follow-up and dose range (0 to > 600 mgy) sufficient for risk analyses. Mean dose <100 mgy makes the population of high interest for today s research.

22 More information is needed to better understand the construction of the NIIRME register. Population from any comparison area should not be used previous analyses showed a large difference in the baseline risks between the respective areas (Bauer et al., 2005, for cancer and Grosche et al., 2011, for cardiovascular diseases, respectively). Using NNC data might help in case missing information in the NIIRME data base needs to be filled. Links to other activities related to SNTS should be tightened.

23 23,960 families in the registry 13,692 families with information on 2nd generation 10,268 families with information on 3rd generation

24