Validation of disease registration in pregnant women in the Medical Birth Registry of Norway

Acta Obstetricia et Gynecologica. 2009; 88: 10831089 ORIGINAL ARTICLE Validation of disease registration in pregnant women in the Medical Birth Registry of Norway ANDERS ENGELAND 1,2, TONE BJØRGE 2,3, ANNE KJERSTI DALTVEIT 2,3, STEIN EMIL VOLLSET 2,3 & KARI FURU 1 1 Department of Pharmacoepidemiology, Division of Epidemiology, Norwegian Institute of Public Health, N-0403 Oslo, Norway, 2 Department of Public Health and Primary Health Care, University of Bergen, N-5018 Bergen, Norway, and 3 Medical Birth Registry of Norway, Division of Epidemiology, Norwegian Institute of Public Health, N-5018 Bergen, Norway Abstract Objective. To evaluate the reliability of maternal disease registration in the Medical Birth Registry of Norway (MBRN). Study design. Validation study. Setting. The two nationwide population-based registries, the Norwegian Prescription Database (NorPD) and the MBRN, were linked using a unique personal identification number. Population. Pregnant women (n108,489), pregnancies (lasting more than 22 weeks) starting at March 30, 2004 onwards and ending before January 1, 2007, registered in MBRN. Main outcome measures. Registrations of maternal diagnoses (diabetes, asthma, and epilepsy) in MBRN compared to their dispensal of prescribed medicines as an indication of disease. The sensitivity, specificity, and positive and negative predictive values of diabetes, epilepsy, and asthma in the MBRN were calculated using information from NorPD as the gold standard. Methods. Validation study comparing maternal diagnoses in MBRN with prescriptions of drugs and reimbursement codes registered in NorPD. Results. The sensitivity for the diagnosis of diabetes (any type) and diabetes type 1 in MBRN were estimated to be 72% (95% confidence interval: 6974) and 90% (8693%). For epilepsy and asthma, the sensitivity of MBRN was estimated to be 74% (6978) and 51% (4952), respectively. For asthma, the sensitivity increased to over 70% when it was restricted to individuals with more serious disease. Conclusions. The sensitivity of the registration of type 1 diabetes was 90%, while the sensitivity for any type of diabetes was lower and similar to the sensitivity of the registration of epilepsy. The registration of asthma overall was 51%, but considerably higher for more serious asthma. Except for diabetes type 1, the medical disease diagnoses in the MBRN should be dealt with carefully. Key words: Registration of maternal diseases, diabetes, epilepsy and asthma, registry Introduction The major objective of establishing the Medical Birth Registry of Norway (MBRN) in 1967 was to detect any increased rates for birth defects or other adverse pregnancy outcomes, and to establish a basis for epidemiological research. Some maternal diseases may increase the risk of congenital malformations as well as other perinatal health problems (1,2). Therefore, information on maternal diseases before and during pregnancy has been recorded in the registry. Particular attention has been paid to the course and outcome of pregnancy in women with diabetes (3,4), epilepsy (5,6), and hypertension (7). The validity of the information on maternal diseases in the MBRN is not well documented. However, the validity of diabetes mellitus diagnoses in the database has recently been explored (8). This was done using two sets of data; medical records (9,10) and the Norwegian Childhood Diabetes Registry (11). One of the conclusions from that evaluation was that the positive predictive value (PPV) for pre-gestational diabetes (80%) in 1998 was less than optimal, but acceptable. For gestational diabetes, the PPV was 90%. Furthermore, Hoff et al. (12) found a PPVof at least 88% for myasthenia gravis registered in MBRN during 19672004. Skomsvoll et al. (13) found a Correspondence: Anders Engeland, Department of Pharmacoepidemiology, Division of Epidemiology, Norwegian Institute of Public Health, P.O. Box 4404 Nydalen, N-0403 Oslo, Norway. E-mail: anders.engeland@isf.uib.no (Received 12 January 2009; accepted 18 June 2009) ISSN 0001-6349 print/issn 1600-0412 online # 2009 Informa UK Ltd. (Informa Healthcare, Taylor & Francis AS) DOI: 10.1080/00016340903128454

1084 A. Engeland et al. sensitivity of 88% for maternal rheumatic disease in the MBRN before 1996. Most of these studies were based on data prior to 1998, when a new notification form was introduced in the MBRN. For some diseases (e.g. diabetes, epilepsy, asthma) specific medications are required, and are also used during pregnancy. The establishment of the Norwegian Prescription Database (NorPD) in 2004 made it possible to link data on maternal diseases before and during pregnancy from the MBRN with the use of medication (14). The aim of our study was to validate recorded disease diagnoses in MBRN using prescribed medicines and reimbursement codes in the NorPD as the standard. Material and methods Data sources NorPD (15) is a research database which captures all dispensed prescriptions in Norway from January 1, 2004, and covers the entire population of Norway (4.8 million). NorPD contains information on all prescribed drugs, reimbursed or not, dispensed at pharmacies to individual patients treated in ambulatory care. Data on use in institutionalized patients in nursing homes and hospitals are also collected, but these figures are only registered at an institutional and not at the individual level. Therefore, drugs dispensed to institutions are not included. For each prescription, the sex and age of the patient, demographic information, dispensal date, and detailed drug information are registered. The indication for prescribing is not recorded. However, the code of reimbursement is recorded (for all reimbursed drugs), and this may in some cases function as a proxy for diagnosis. In Norway, pharmaceutical treatment is reimbursed for a number of diseases or conditions (16). Reimbursement is given only when the patient has a chronic disease, for which long-term medication (more than three months) is necessary (16). Classification of drugs is based on the Anatomical Therapeutic Chemical (ATC) classification system as of 2007 (17). MBRN is a population-based registry containing information on all births in Norway since 1967 (more than 2.3 million) (18). MBRN is based on compulsory notification of every birth or late abortion from 12 completed weeks of gestation onwards, and includes identification of the parents by their national identification numbers, demographic information of the parents, maternal diagnoses before and during pregnancy, complications during pregnancy and delivery, length of pregnancy, as well as information on the infant, including birth defects and other perinatal problems (18). A standard antenatal form is completed at visits to a general practitioner, midwife, or obstetrician during pregnancy and is brought by the mother to the place of birth. The midwife enters additional data recorded at the time of birth. Followup data are added to the form until discharge of hospital births. Since 1998, data obtained in neonatal wards on congenital conditions for infants transferred to such units after birth have also been included. In our study, only pregnancies lasting more than 22 completed weeks were included. From late 1998, a new notification form for births was introduced. This form aimed to improve the registrations in MBRN. In this form there are precoded fields for disease in mothers prior to pregnancy, such as type 1 and 2 diabetes, asthma and epilepsy, and gestational diabetes. In addition, there are open text fields for maternal diagnoses before and during pregnancy. The use of drugs in pregnancy is also recorded. Textual information on maternal diseases is coded by the staff at the MBRN using International Classification of Diseases (ICD)-10 codes. Some women with polycystic ovary syndrome (PCOS) are prescribed metformin as treatment for infertility. PCOS is also associated with type 2 diabetes (19). We cannot, however, differentiate between diabetic and PCOS patients who are prescribed metformin. Hence, analyses were performed excluding prescriptions on metformin to women registered with PCOS in MBRN while studying diabetes overall and diabetes type 2. Study subjects Data in the MBRN were linked to NorPD using the unique 11-digit identification number, assigned to all individuals living in Norway after 1960. In our study, women with pregnancies registered in MBRN were included, beginning at March 30, 2004 or later and ending before January 1, 2007. Only the first pregnancy during the study period was included for each woman. Prescription data from NorPD from January 2004 until December 31, 2006 were used. March 30, 2004 was defined as the starting date of pregnancies to ensure that information on all prescriptions dispensed from three months before conception until the end of pregnancy was included. Methodology Length of pregnancy was usually determined by routine ultrasound around week 18 (97%) and the date of conception estimated. Each pregnancy was

Registration of maternal diseases 1085 divided into trimesters (week 012, week 1326, and week 27 onwards). Prescriptions during the three last months before conception and during pregnancy were included from NorPD. Prescriptions of specific drugs, combined with reimbursement codes, were used as proxies of disease:. Dispensed drugs used to treat diabetes (ATC code A10) were used as proxy for any type of diabetes. Dispensed insulin (A10A) could more specifically also be used as an indicator of type 1 or gestational diabetes. To distinguish between patients with type 1 and gestational diabetes, insulin prescriptions dispensed during the three months before conception were used.. Dispensed selective beta-2-adrenoreceptor agonists as inhalants (ATC code R03AC), fixed dose combination of inhaled beta-2-adrenoreceptor agonists and long-acting glucocorticoids (R03AK), inhalants of glucocorticoids and anticholinergics (R03B), or other systemic drugs for obstructive airway diseases (R03D) with specific reimbursement codes were used as a proxy for asthma. According to guidelines, medical treatment of asthma should follow a stepwise approach until control of symptoms is achieved using the lowest possible dose (20). Each step could be indicated by the type of drugs prescribed: (1) Inhaled selective beta-2-adrenoreceptor agonists (R03AC). (2) Inhaled glucocorticoids (R03BA). (3) Both R03AC and R03BA or adrenergics, and other drugs for obstructive airway diseases (R03AK). (4) Other systemic drugs for obstructive airway diseases (R03D).. Dispensed antiepileptics (ATC code N03A) with a specific reimbursement code were used as a proxy for epilepsy. Using registration in NorPD as the gold standard, we calculated the sensitivity, specificity, and positive and negative predictive values:. the sensitivity of diagnoses in MBRN was calculated as the proportion of those diagnosed in NorPD who were diagnosed in MBRN;. the specificity was calculated as the proportion of those not diagnosed in NorPD who were not diagnosed in MBRN either;. the PPV was calculated as the proportion diagnosed in MBRN who were also diagnosed in NorPD; and. the negative predictive value (NPV) was calculated as the proportion not diagnosed in MBRN who were not diagnosed in NorPD either. For sensitivity, specificity, PPV, and NPV, 95% confidence intervals (95% CI) were calculated using continuity corrected score intervals (21). Results A total of 108,489 pregnancies were included in our study, 376 (0.3%) of the mothers were registered with PCOS. According to data from MBRN, a total of 1,929 (1.8%) mothers were registered with diabetes, 855 (0.8%) with epilepsy, and 4,549 (4.2%) with asthma (Table I). For all diagnoses studied, the estimated specificity was close to 100%. Calculations Table I. Characteristics of the study population. Data from pregnancies in 20042006 (n108,489), Medical Birth Registry of Norway (MBRN). Total number of women Percentage (%) Age (years) B20 2,381 2.2 2029 49,460 45.6 3039 53,817 49.6 40 2,831 2.6 Birth year B1960 47 0.04 19601969 14,043 12.9 19701979 71,030 65.5 1980 23,369 21.5 Length of pregnancy (completed weeks) 2324 178 0.2 2536 6,039 5.6 3739 43,820 40.4 4041 50,371 46.4 4245 8,072 7.4 46 9 0.008 Diabetes in MBRN Yes 1,929 1.8 Epilepsy in MBRN Yes 855 0.8 Asthma in MBRN Yes 4,549 4.2 Pregnancy outcome Abortion 34 0.03 Stillbirth 367 0.3 Livebirth 108,088 99.6

1086 A. Engeland et al. using six months instead of three months prior to conception, gave similar results for diabetes, epilepsy, and asthma. Diabetes Of 1,929 women registered with diabetes in MBRN before or during pregnancy, 926 (48%) were dispensed with at least one drug used in the treatment of diabetes. In addition, 363 women who were not registered with diabetes in the MBRN were dispensed drugs used to treat diabetes. Among the women who received such drugs, 926 (72%) were registered in the MBRN as having diabetes, i.e. a sensitivity of 72% (Table II). The PPV of diabetes registered in MBRN was 48%. Of the 1,289 women dispensed drugs used to treat diabetes, most received fast (56%) and intermediate-acting (50%) insulin and analogs for injection (ATC codes A10AB and A10AC, Table III). While excluding metformin-prescriptions in women with PCOS, 899 of 1,198 women who were dispensed drugs used to treat diabetes, were registered with diabetes in MBRN, i.e. a sensitivity of 75%. A total of 1,060 women were registered with gestational diabetes in MBRN. Among these, 32 (3%) received drugs used to treat diabetes during the three months prior to conception, seven of these women had PCOS and received metformin. Among women registered with gestational diabetes not receiving drugs prior to conception, 25% received drugs during pregnancy. According to MBRN, 520 women had type 1 diabetes. Of these, 307 (59%) received drugs used to treat diabetes during the three months before conception. Excluding metformin-prescriptions in women with PCOS, one of these women was removed. Of the women registered with type 1 diabetes, a total of 289 (56%) were dispensed insulins during the three months before conception. Of those not registered in MBRN as having type 1 diabetes, 33 (0.03%) were dispensed insulin. A total of 461 (89%) women registered with type 1 diabetes were dispensed insulin during the three months before conception and during pregnancy. In addition, 409 women not registered as having type 1 diabetes also received insulin. Of these, 281 were registered as having gestational diabetes. Of 221 women registered with type 2 diabetes in MBRN, 99 (45%) received drugs used to treat diabetes. While excluding metformin-prescriptions in women with PCOS, 97 of these women were dispensed drugs used to treat diabetes. Epilepsy A total of 314 (37%) women registered with epilepsy in MBRN received drugs for epilepsy. The sensitivity for epilepsy of the registry was estimated to be 74%. In women who were not registered in MBRN as having epilepsy, 112 used epilepsy drugs and 74 of these were dispensed epilepsy drugs during pregnancy. Asthma Of women registered with asthma in MBRN, 2,098 (46%) were dispensed drugs for asthma. In women not registered in the MBRN as having asthma, 2,043 (2%) used asthma drugs. Of all women who were dispensed drugs for asthma, about half (51%) were registered with a diagnosis of asthma in MBRN. The sensitivity of registration of asthma in the MBRN increased according to the seriousness of asthma Table II. Registered diseases in the Medical Birth Registry of Norway (MBRN) in women (n108,489) before and during pregnancy vs. diseases diagnosed using prescriptions and reimbursement codes registered in the Norwegian Prescription Database (NorPD). Disease in both registries (A) Disease only in MBRN (B) Disease only in NorPD (C) Sensitivity a (95% CI) Positive predictive value (PPV) b (95% CI) Specificity c (95% CI) Diabetes (any type) 926 1,003 363 72 (6974) 48 (4650) 99 (9999) Type 1 diabetes d 289 231 33 90 (8693) 56 (5160) 100 (100100) Epilepsy 314 541 112 74 (6978) 37 (3440) 100 (99100) Asthma e 2,098 2,451 2,043 51 (4952) 46 (4548) 98 (9898) Step I 636 3,913 955 40 (3842) 14 (1315) 96 (9696) Step IIIII 1,255 3,294 901 58 (5660) 28 (2629) 97 (9797) Step IV 128 4,421 47 73 (6679) 3 (23) 96 (9696) a SensitivityA/(AC). b PPVA/(AB). c SpecificityD/(BD). d Diagnosis in NorPD based on dispensed insulins (ATC: A10A) during the three months prior to conception. e Based on guidelines for medical treatment of asthma, asthmatic women were divided into groups according to assumed seriousness of their asthma (20).

Table III. Dispensed drugs for diabetes, epilepsy, and asthma from three months prior to conception until birth for pregnant women (n108,489). Data from the Norwegian Prescription Database (NorPD). Number of women with]1 prescription Disease Dispensed drugs (ATC code) Total Three months before conception First trimester Second trimester Third trimester Diabetes Insulins and analogs for injection, fast-acting (A10AB) 725 268 373 474 573 Insulins and analogs for injection, intermediate-acting (A10AC) 647 198 282 368 479 Insulins and analogs for injection, intermediate-acting 112 20 33 54 87 combined with fast-acting (A10AD) Insulins and analogs for injection, fast-acting (A10AE) 46 32 34 27 27 Biguanides (A10BA) 479 407 270 67 43 Sulfonamides, urea derivatives (A10BB) 18 13 10 1 1 Thiazolidineones (A10BG) 5 5 3 1 0 Epilepsy Phenobarbital (N03AA02) 10 5 5 5 6 Phenotyin (N03AB02) 6 5 4 4 5 Phensuximide (N03AD01) 4 3 3 2 2 Clonazepam (N03AE01) 21 16 15 12 13 Carbamazepine (N03AF01) 98 67 66 64 65 Oxcarbazepine (N03AF02) 21 13 16 19 19 Valproic acid (N03AG01) 59 35 29 31 35 Lamotrigine (N03AX09) 211 148 149 138 136 Topiramate (N03AX11) 24 19 15 10 11 Gabapentin (N03AX12) 16 15 4 4 3 Levetiracetam (N03AX14) 33 25 23 27 27 (N03AX16) 2 2 1 1 1 Asthma Selective beta-2 adrenoreceptor antagonists (R03AC) 3,150 1,299 1,181 1,274 1,212 Adrenergics and other drugs for obstructive airway 1,593 891 622 576 542 diseases (R03AK) Glucocortisoids (R03BA) 966 345 338 397 360 Anticholinergics (R03BB) 70 38 31 27 21 Antiallergic agents, excl. corticosteroids (R03BC) 58 10 21 25 19 Xanthines (R03DA) 23 12 12 11 11 Leukotriene receptor antagonists (R03DC) 149 109 52 21 14 Registration of maternal diseases 1087

1088 A. Engeland et al. (see Methods ): Step I, sensitivity 40%; Step IIIII, sensitivity 58%; Step IV, sensitivity 73%. Discussion In this population-based study, we validated the registration of diseases in women in the MBRN using prescribed medicines and reimbursement codes registered in NorPD as the gold standard. We observed a sensitivity of the registration of type 1 diabetes of 90%, while the sensitivity for any type of diabetes was lower and similar to the sensitivity of the registration of epilepsy (72 and 74%). The sensitivity of the registration of asthma was about 51%, but considerably higher for those with more serious disease. A major strength of our study was the use of two nationwide registries and the unique 11-digit identification number, which ensured valid data linkage between the two registries. Another advantage was that the NorPD includes prescriptions from all practicing physicians prescribing to patients living outside institutions in Norway, both general practitioners and specialists. In addition, NorPD has not previously been used in studies assessing the validity of maternal disease diagnoses in MBRN. Diseases included in our study were those which were easily identifiable through the prescription of drugs, i.e. diseases where the use of specific medicines, combined with reimbursement codes, implies the presence of the disease. However, the absence of prescriptions does not necessarily imply absence of the disease. Moreover, women who receive medication as inpatients in hospitals were not included in NorPD. For this reason, we were only able to give estimates of the minimum PPV in MBRN on the assumption that those who were dispensed the disease-specific drugs (for epilepsy and asthma also with the relevant reimbursement codes) had the disease. If we also assumed that among those not being dispensed drugs, the proportion of true diagnoses registered in MBRN was equal to or less than among those dispensed drugs, the maximum sensitivity in MBRN could be calculated. Since most pregnant women do not have any of the diseases we are studying, the specificity for all these diseases will be close to 100%. In addition, women with the diseases do not necessarily need medication during the period studied here. Data on drugs prescribed, rather than dispensed, can lead to the misclassification of drug exposure because of primary non-compliance (22). Our data were based on dispensed rather than prescribed drugs, and this may increase the usefulness of the data. Assuming that those who received drugs used to treat diabetes were correctly diagnosed with diabetes, the sensitivity of diabetes in MBRN was at most 72%. Not all patients with diabetes use these drugs. We do not know whether women not receiving drugs for diabetes in connection with pregnancy were diabetics or not. However, in the USA, 16% of adults with diagnosed diabetes (type 1 or type 2), do not take insulin or oral medication (23). If we base estimates on dispensed drugs in the three months before conception, at least 6% of the mothers registered with gestational diabetes in the MBRN had diabetes before pregnancy. Using dispensed insulin as a measure, the sensitivity of type 1 diabetes in MBRN was at most 90%. This is in agreement with a recent study by Stene et al. (8), comparing registrations in the MBRN with registrations in the Norwegian Childhood Diabetes Registry for births in 19992004. In that study, a sensitivity of 88% was observed for registration of type 1 diabetes in MBRN. For this disease most patients will need prescriptions of insulin regularly. Assuming that those receiving drugs for epilepsy had a correct diagnosis (epilepsy), the sensitivity was at most 74%. Among those not diagnosed in MBRN, a large proportion was dispensed anti-epileptics during pregnancy. This diagnosis has been used in a recent study looking at the adult intelligence in offspring of mothers with epilepsy (6). In that study, Øyen et al. found a prevalence of maternal epilepsy of 4.1/1,000, using births from 1967 to 1979. In our study, the prevalence of maternal epilepsy registered in MBRN was 7.9/1,000. Assuming that individuals who received drugs for asthma actually had asthma, the sensitivity of the asthma diagnosis in MBRN was, at most, 51%. However, the sensitivity of registration of asthma in MBRN seemed to increase according to the seriousness of the disease. Due to the low sensitivity, analyses including this diagnosis should be interpreted with caution. In conclusion, the registration of diabetes (any type) and epilepsy in MBRN was estimated to have sensitivity of 72 and 74%, respectively. The sensitivity of the registration of type 1 diabetes was higher (90%), while the sensitivity of the registration of asthma, in general, was only 51%. However, a higher sensitivity of asthma was achieved when the analysis was restricted to individuals presumed to have more serious disease. Except for diabetes type 1, the

Registration of maternal diseases 1089 diagnoses in MBRN explored in this study should be dealt with very carefully. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper. References 1. Holmes LB, Harvey EA, Coull BA, Huntington KB, Khoshbin S, Hayes AM, et al. The teratogenicity zof anticonvulsant drugs. N Engl J Med. 2001;/344:/11328. 2. Macintosh MC, Fleming KM, Bailey JA, Doyle P, Modder J, Acolet D, et al. Perinatal mortality and congenital anomalies in babies of women with type 1 or type 2 diabetes in England, Wales, and Northern Ireland: population based study. BMJ. 2006;333:171180. 3. Egeland GM, Skjærven R, Irgens LM. Birth characteristics of women who develop gestational diabetes: population based study. BMJ. 2000;/321:/5467. 4. Egeland GM, Skjærven R, Irgens LM. The reproductive health of daughters of pregestational diabetic women: Medical Birth Registry of Norway. Paediatr Perinat Epidemiol. 2002;/ 16:/2906. 5. King PB, Lie RT, Irgens LM. Spina bifida and cleft lip among newborns of Norwegian women with epilepsy: changes related to the use of anticonvulsants. Am J Public Health. 1996;/86:/ 14546. 6. Øyen N, Vollset SE, Eide MG, Bjerkedal T, Skjærven R. Maternal epilepsy and offsprings adult intelligence: a population-based study from Norway. Epilepsia. 2007;/48:/17318. 7. Rasmussen S, Irgens LM. History of fetal growth restriction is more strongly associated with severe rather than milder pregnancy-induced hypertension. Hypertension. 2008;/51:/ 12318. 8. Stene LC, Eidem I, Vangen S, Joner G, Irgens LM, Moe N. The validity of the diabetes mellitus diagnosis in the Medical Birth Registry of Norway. Nor J Epidemiol. 2007;/17:/16574. 9. Hawthorne G, Irgens LM, Lie RT. Outcome of pregnancy in diabetic women in northeast England and in Norway, 19947. BMJ. 2000;/321:/7301. 10. Hawthorne G, Irgens LM, Lie RT, Moe N, Jervell J. Retraction of paper on maternal diabetes. BMJ. 2003; 327:929. 11. Aamodt G, Stene LC, Njølstad PR, Søvik O, Joner G. Spatiotemporal trends and age-period-cohort modeling of the incidence of type 1 diabetes among children aged B15 years in Norway 19731982 and 19892003. Diabetes Care. 2007;/30:/8849. 12. Hoff JM, Daltveit AK, Gilhus NE. Myasthenia gravis in pregnancy and birth: identifying risk factors, optimising care. Eur J Neurol. 2007;/14:/3843. 13. Skomsvoll JF, Østensen M, Baste V, Irgens LM. Validity of a rheumatic disease diagnosis in the Medical Birth Registry of Norway. Acta Obstet Gynecol Scand. 2002;/81:/8314. 14. Engeland A, Bramness JG, Daltveit AK, Rønning M, Skurtveit S, Furu K. Prescription drug use among fathers and mothers before and during pregnancy. A populationbased cohort study of 106,000 pregnancies in Norway 200406. Br J Clin Pharmacol. 2008;/69:/65360. 15. Furu K. Establishment of the nationwide Norwegian Prescription Database (NorPD) new opportunities for research in pharmacoepidemiology in Norway. Nor J Epidemiol. 2008;/ 18:/12936. 16. Hågå A, Sverre J. Pricing and reimbursement of pharmaceuticals in Norway. Eur J Health Econ. 2002;/3:/21520. 17. WHO Collaborating Centre for Drug Statistics Methodology. ATC classification index with DDDs 2007. Oslo, Norway: WHO Collaborating Centre for Drug Statistics Methodology; 2006. 18. Irgens LM. The Medical Birth Registry of Norway. Epidemiological research and surveillance throughout 30 years. Acta Obstet Gynecol Scand. 2000;/79:/4359. 19. Glueck CJ, Goldenberg N, Sieve L, Wang P. An observational study of reduction of insulin resistance and prevention of development of type 2 diabetes mellitus in women with polycystic ovary syndrome treated with metformin and diet. Metabolism. 2008;/57:/95460. 20. Global Initiative for Asthma (GINA). Global strategy for asthma management and prevention. 2008. Available online at: http://www.ginasthma.org (Assessed April 15, 2009). 21. Vollset SE. Confidence intervals for a binomial proportion. Stat Med. 1993;/12:/80924. 22. Beardon PH, McGilchrist MM, McKendrick AD, McDevitt DG, MacDonald TM. Primary non-compliance with prescribed medication in primary care. BMJ. 1993;/307:/8468. 23. Centers for Disease Control and Prevention.National diabetes fact sheet: general information and national estimates on diabetes in the United States, 2007. Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention; 2008.