Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=igas20
Scandinavian Journal of Gastroenterology
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/igas20
Capturing biologic treatment for IBD in the
Swedish Prescribed Drug Register and the Swedish
National Patient Register – a validation study
Gabriella Bröms, Jonas Söderling, Michael C. Sachs, Jonas Halfvarson, the
SWIBREG study group, Par Myrelid, Jonas F. Ludvigsson, Åsa H. Everhov &
Ola Olén
To cite this article: Gabriella Bröms, Jonas Söderling, Michael C. Sachs, Jonas Halfvarson, the SWIBREG study group, Par Myrelid, Jonas F. Ludvigsson, Åsa H. Everhov & Ola Olén (2021) Capturing biologic treatment for IBD in the Swedish Prescribed Drug Register and the Swedish National Patient Register – a validation study, Scandinavian Journal of Gastroenterology, 56:4, 410-421, DOI: 10.1080/00365521.2021.1884894
To link to this article: https://doi.org/10.1080/00365521.2021.1884894
© 2021 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
View supplementary material
Published online: 25 Feb 2021. Submit your article to this journal
Article views: 342 View related articles
ORIGINAL ARTICLE
Capturing biologic treatment for IBD in the Swedish Prescribed Drug Register
and the Swedish National Patient Register
– a validation study
Gabriella Br€omsa,b , Jonas S€oderlinga,c
, Michael C. Sachsc, Jonas Halfvarsond, the SWIBREG study group, Par Myrelide,f , Jonas F. Ludvigssonc,g , Åsa H. Everhova,hand Ola Olena,h,i
a
Clinical Epidemiology Division, Department of Medicine Solna, Karolinska Institutet, Stockholm, Sweden;bDepartment of Internal Medicine, Danderyds Hospital, Stockholm, Sweden;cDepartment of Medical Epidemiology and Biostatistics, Karolinska Institutet, Solna, Sweden;
d
Department of Gastroenterology, Faculty of Medicine and Health, €Orebro University, €Orebro, Sweden;eDepartment of Surgery, Link€oping University Hospital, Link€oping, Sweden;fDepartment of Clinical and Experimental Medicine, Link€oping University, Link€oping, Sweden;
g
Department of Pediatrics, €Orebro University Hospital, €Orebro, Sweden;hSachs’ Children and Youth Hospital, Stockholm South General Hospital, Stockholm, Sweden;iDepartment of Clinical Science and Education S€odersjukhuset, Karolinska Institutet, Stockholm, Sweden
ABSTRACT
Background: It is not known to what extent biologic treatment for IBD is captured in the Swedish Prescribed Drug Register (PDR) and the National Patient Register (NPR).
Methods: A cross-sectional study from July 2005 until 2017, comparing data on biologic treatment in the PDR and the NPR with medical records. We assessed the proportion of started treatment episodes in the medical records that were found in the PDR/NPR ever, withinþ/ one year and within þ/ three months; for any biologic drug, per specific drug (infliximab, adalimumab, golimumab, vedolizu-mab, ustekinumab), by calendar period (2005–2008, 2009–2012, and 2013–2017) and by study center. For adalimumab, we assessed the validity of end of treatment episodes.
Results: Medical records of 1361 patients and 2323 treatment episodes with any biologic were reviewed and 80.1% (95% CI: 78.4–81.7) were ever captured in the PDR/NPR in. A time window of þ/ one year or þ/ three months reduced the sensitivity to 63.3% (95% CI: 61.3–65.3) and 52.6% (95% CI: 50.5–54.6), respectively. The sensitivity was high (>85%) for the prescribed injection drugs adalimumab, golimumab, and ustekinumab for all time windows and for adalimumab end of treat-ment, while considerably lower for the infusion drugs infliximab and vedolizumab.
Conclusions: The PDR and the NPR are reliable data sources on treatment with injection biologics in patients with IBD in Sweden. Infliximab and vedolizumab are poorly captured, why PDR/NPR data should only be used after careful consideration of their limitations or complemented by other data sources, e.g., the disease-specific quality register SWIBREG.
ARTICLE HISTORY
Received 21 August 2020 Revised 25 January 2021 Accepted 30 January 2021
KEYWORDS
IBD; Crohn’s disease; ulcerative colitis; IBD unclassified; biologics; anti-TNF; validation
Background
Biologic treatment has become a mainstay option in the treatment of inflammatory bowel disease (IBD). The anti-tumor necrosis factor agents (anti-TNF); infliximab, adalimu-mab and golimuadalimu-mab, the anti-integrin vedolizuadalimu-mab, and the anti-interleukin 12/23 ustekinumab are increasingly used to treat Crohn’s disease (CD) and ulcerative colitis (UC) [1]. Since infliximab was approved for CD in Sweden in 1999, other biologic agents have followed and treatment indica-tions have expanded (Supplementary Appendix Table 2). The increasing use of biologic agents is motivated primarily by evidence from randomized control trials (RCTs) [2–9]. In gen-eral, the generalizability of RCTs is limited by narrow eligibil-ity criteria, which restrict the study population to certain ages and disease characteristics. Moreover, follow-up time is
limited. Real-world evidence of effectiveness and long-term safety, as well as data applicable to a wider population, are needed to assess treatment in the clinical setting.
The Prescribed Drug Register (PDR) and the National Patient Register (NPR) are national health registers and con-tain data that, because of mandatory reporting, exhibit close to complete coverage of individuals in Sweden [10,11]. In the PDR, prescribed drug treatment in outpatient care is cap-tured by electronic records of filled prescriptions. However, biologic agents administered by infusion, in particular inflixi-mab and vedolizuinflixi-mab, are not expected to be completely captured in the PDR. As they are administered in-hospital, the majority of infusion drugs are acquired via stock orders by the hospital and cannot be traced to any individual patient. Some hospital administrations may instead be identi-fied from infusion procedure codes in combination with
CONTACTGabriella Br€oms gabriella.broms@ki.se Clinical Epidemiology Division, Karolinska Institutet, T2 Karolinska University Hospital, Solna, Stockholm, SE-171 76, Sweden
Supplemental data for this article can be accessedhere.
ß 2021 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.
2021, VOL. 56, NO. 4, 410–421
specified drugs according to the Anatomical Therapeutic Chemical Classification System (ATC) in the NPR, but the val-idity of this register data is unknown.
The PDR and the NPR have been used in observational studies of IBD [12,13], but it is not known to what extent biologic treatment for IBD is captured in these registers. We therefore performed a comprehensive medical record review in a sample of biologically treated pediatric and adult IBD patients across Sweden and compared the treatment data found in the medical records with data found in the PDR and the NPR.
Method Study design
We performed a cross-sectional study, comparing data on biologic treatment from the NPR and the PDR with informa-tion in patient records as our gold standard (Figure 1).
Setting and data sources
In Sweden, healthcare is accessible to all and free of charge. Reporting to the national health registers, which are adminis-tered by the National Board of Health and Welfare, is manda-tory, resulting in close to complete coverage. The Personal Identity Number (PIN) assigned to all residents at birth or upon immigration allows for the individual linkage of data across registers [14].
The NPR was set up in some regions in 1964, initially including only inpatient visits, and reached virtually complete
geographical coverage in 1987 [11]. From 2001, all out-patient visits to non-primary care are also included. Patients with IBD are generally treated by gastroenterologists in Sweden. Records include date of visit and main and con-tributory diagnostic codes according to the Swedish version of the International Classification of Diseases (ICD). It is also possible to add codes of performed procedures and codes for administered drugs according to the Anatomical Therapeutic Chemical Classification System (ATC), although the National Board of Health and Welfare acknowledges that data on ATC codes are not complete [15].
The PDR records all filled prescriptions since July 2005, including their ATC code, date of dispensing, strength and number of tablets/syringes [10]. ATC-codes for the included biologics are listed in Supplementary Appendix Table 3. Biosimilars share the ATC-code of the reference product.
The ESPRESSO (Epidemiology Strengthened by histoPathology Reports in Sweden) database includes data on gastrointestinal biopsies from all 28 pathology depart-ments in Sweden from January 1965 to April 2017, including date of biopsy, anatomic location and morphology according to the Systematized Nomenclature of Medicine system (SNOMED) [16].
Study population
Individuals of all ages, both pediatric and adult patients, with IBD were defined by either at least two separate visits (out-patient visits or hospitalizations) with IBD diagnostic codes in the NPR, or one visit supported by a colorectal biopsy with a
Figure 1. Flowchart of patient selection. Capturing biological treatment for IBD in the Swedish Prescribed Drug Register and the Swedish Patient Register– a valid-ation study.
SNOMED code indicating IBD in the ESPRESSO register. IBD subtypes were classified as CD, UC or IBD unclassified based on the two first diagnostic listings of IBD (Supplementary Appendix Table 4). We required patients to be alive on 1 July 2005 for the possibility of identifying their treatment in the PDR.
Medical record review
Patients with known biologic treatment according to medical records were identified at each participating center, which included the University Hospital of Umeå, €Ornsk€oldsvik Hospital, Karlstad Hospital, €Orebro University Hospital, Karolinska University Hospital, Link€oping University Hospital, Norrk€oping Hospital and Skåne University Hospital in Lund and Malm€o. These centers correspond to an uptake area cov-ering close to one third of the Swedish population. At smaller centers (e.g., Umeå), medical records of all patients with known biologic treatment were reviewed. At large cen-ters (e.g., Lund and Malm€o), patients were listed and their medical records reviewed consecutively according to this list until the end of data collection time, resulting in a total coverage of above 80% among patients with known biologic treatment at the included centers. Data were collected by one or two medical students at each center primarily between January and April 2016, while data collection was carried out in 2017 at one center. The students were super-vised by experienced gastroenterologists. At the time of medical record review, physicians responsible for the record review had no information on whether or not a patient’s use of biologics was captured in any way in PDR or the NPR. We used a standardized case report form (CRF) including the PIN, date of IBD diagnosis, and extent/location of inflamma-tion. Data on biologic treatment between 1 July 2005 and the date of the record review in 2016 or 2017 included name of drug, dose, interval, start and end date of treatment and reason for starting and ending treatment. The included biologic drugs were infliximab, adalimumab, golimumab, vedolizumab, and ustekinumab. Treatment that was noted as initiated before the introduction of the drug as a treatment option was assumed to be erroneously recorded and omitted in the final analysis, while treatment initiated after introduc-tion, but before specific approval to treat either CD or UC, was included.
Statistical analysis
Descriptive statistics for the study population including sex, age at diagnosis, calendar year of diagnosis, type of IBD, and country of birth were tabulated by the availability of infor-mation – for the patients whose medical records were reviewed, for patients with biologic treatment according to the PDR and the NPR, and for the entire population of patients with IBD in Sweden.
The agreement between data from the medical records and the combined information from the PDR and the NPR was assessed by counting a) individual patients and b) treat-ment episodes in the medical records that corresponded
with a record in the national registers. Confidence intervals were calculated using the Wilson method [17]. For treatment episodes, the start in the registers was defined as either a filled prescription in the PDR or an ATC code for biologics from a visit in the NPR and validated according to their date of appearance in the registers ever, within ± one year and within ± three months of treatment start in the medical records. An interval of 90 days for procedures and 180 days for prescriptions between register records of treat-ment, or a switch to another biologic was considered the start of a new treatment episode. We required an ATC code for NPR records, instead of merely a procedure code, for example ‘infusion’ or ‘injection’. Although likely that these procedure codes do pertain to biologic treatment when used in combination with a diagnostic code of CD or UC, they may also relate to other administered substances. We report the procedure codes that were found in connection with ATC codes for biologics inSupplementary Appendix Table 5.
We assessed the sensitivity for any biologic drug and per specific drug overall for individuals and treatment episodes, and for episodes also stratified by calendar period (2005–2008, 2009–2012, and 2013–2017) and by study center for infliximab, adalimumab and vedolizumab. We investi-gated type of biologic drug used as first, second, and third line treatment, overall and over calendar periods. We also reported the source of the register information, the PDR, the NPR or both.
The validity of register data for end of treatment episodes was considered for adalimumab only, and among episodes with a verified treatment start within three months in the PDR. We did not find it meaningful to study this aspect for infusion drugs because of the observed low validity of treat-ment start, or for the other injection drugs because of the small numbers. Ongoing treatment episodes were assigned an end date corresponding to the date of review for medical records. The number of defined daily doses collected at the last filled prescription for adalimumab was added to the treatment episode according to the registers. The analysis of end of treatment for adalimumab was supplemented by sen-sitivity analysis allowing 30 days before and after stop in registers, respectively.
To assess the significance of imperfect sensitivity for bio-logic treatment in the PDR/NPR, the impact of varying levels of misclassification on a hypothetical assessed relative risk was illustrated in simulation graphs. We plotted graphs given a ratio of treated vs untreated of 1:10, 1:2 and 1:1, where misclassified treated patients were assumed to be untreated.
Results
We identified 103,247 patients with incident or prevalent IBD according to the NPR and ESPRESSO (Table 1). Of these, the medical records of 1361 patients with known use of biologics were reviewed. The patients whose records were reviewed were younger (median age 26 years) and evenly distributed between CD/UC (50.1%/47.8%, with 2.1% with IBD-U as their first diagnosis), while in the entire population median age was 39 years and the majority (63.1%) had UC as first IBD
diagnosis. Characteristics of patients whose records were reviewed were similar to patients with a record of biologic treatment in the PDR or the NPR. However, while the propor-tion of infliximab/adalimumab as first biologic drug in the reviewed records was 68.5%/30.1%, the distribution of users was the opposite in patients with biologic treatment accord-ing to PDR/NPR (29.6%/66.9%). Only a few percent had goli-mumab, vedolizumab or ustekinumab as their first biologic drug. The majority of reviewed records came from university hospitals (>70%;Supplementary Appendix Table 6).
Considering the source (PDR or NPR), 95% of the individu-als with biologic treatment according to medical records and
also captured in the registers, were identified in the PDR, and 13% in the NPR (Supplementary Appendix Table 7). For injection drugs adalimumab, golimumab and ustekinumab, 100% were found in the PDR. Of the 9700 individuals with biologic treatment according to the registers, 90% were identified from the PDR. Of the 15,363 treatment episodes found in the registers, 83% were identified from the PDR, with only a small proportion (1%) recorded in both sources (Supplementary Appendix Table 8). Considering separate treatment occasions (filled prescriptions in the PDR or drug administrations in the NPR), 94% were identified from the PDR and 6% from the NPR.
Table 1. Characteristics of the study population. Characteristics
Medical record review patients
Patients with biologic treatment in the PDR/NPR
Overall patients with IBD (NPRþ ESPRESSO)
N 1361 9700 103,247
Sex
Female 592 (43.5%) 4666 (48.1%) 50,811 (49.2%)
Male 769 (56.5%) 5034 (51.9%) 52,436 (50.8%)
Age at IBD diagnosis
Mean (SD) 29.3 (13.8) 31.8 (15.3) 41.9 (19.1)
Median (IQR) 25.8 (18.7–36.9) 28.3 (19.9–41.4) 39.3 (26.3–56.3)
Range, min–max 4.5–86.1 0.9–85.5 0.0–101.4
Categories, n (%) 0–<18 y 295 (21.7%) 1816 (18.7%) 9374 (9.1%) 18–<30 y 559 (41.1%) 3432 (35.4%) 24,414 (23.6%) 30–<40 y 219 (16.1%) 1803 (18.6%) 19,045 (18.4%) 40–<50 y 142 (10.4%) 1199 (12.4%) 15,670 (15.2%) 50–<60 y 97 (7.1%) 881 (9.1%) 13,910 (13.5%) 60–<70 y 40 (2.9%) 438 (4.5%) 10,996 (10.7%) 70 y 9 (0.7%) 131 (1.4%) 9838 (9.5%)
Year of IBD diagnosis
1969–1986 80 (5.9%) 488 (5.0%) 9724 (9.4%) 1987–1996 117 (8.6%) 797 (8.2%) 1112 (9.8%) 1997–2000 119 (8.7%) 728 (7.5%) 8054 (7.8%) 2001–2005 398 (29.2%) 2168 (22.4%) 24,971 (24.2%) 2006–2010 408 (30.0%) 2398 (24.7%) 20,262 (19.6%) 2011–2017 239 (17.6%) 3121 (32.2%) 30,124 (29.2%)
Type of first IBD diagnosis
CD 682 (50.1%) 5018 (51.7%) 34,548 (33.5%) UC 650 (47.8%) 4389 (45.2%) 65,168 (63.1%) IBD-U 29 (2.1%) 293 (3.0%) 3531 (3.4%) Country of birth Nordic country 1207 (88.7%) 8700 (89.7%) 94,971 (92.0%) Non-Nordic country 154 (11.3%) 999 (10.3%) 8271 (8.0%) Missing 0 1 (0.0%) 5 (0.0%)
First biologic drug
Infliximab 932 (68.5%) 2869 (29.6%) 2869 (2.8%)
Adalimumab 409 (30.1%) 6493 (66.9%) 6493 (6.3%)
Golimumab 9 (0.7%) 242 (2.5%) 242 (0.2%)
Vedolizumab 11 (0.8%) 60 (0.6%) 60 (0.1%)
Ustekinumab 0 36 (0.4%) 36 (0.0%)
Age at first biologic treatment start
Mean (SD) 36.3 (14.8) 38.7 (16.1) –
Median (IQR) 33.7 (24.1–46.3) 37.0 (25.3–50.9) –
Range, min–max 6.1–87.5 0.8–87.4 –
Categories, n (%) 0–<18 y 77 (5.7%) 746 (7.7%) – 18–<30 y 490 (36.0%) 2721 (28.1%) – 30–<40 y 298 (21.9%) 1956 (20.2%) – 40–<50 y 223 (16.4%) 1716 (17.7%) – 50–<60 y 164 (12.0%) 1379 (14.2%) – 60–<70 y 87 (6.4%) 902 (9.3%) – 70 y 22 (1.6%) 280 (2.9%) –
Year of first biologic treatment start
2005–2008 436 (32.0%) 1803 (18.6%) –
2009–2012 503 (37.0%) 2751 (28.4%) –
Types of treatment episodes in medical records
Among the patients whose medical records were reviewed, 2323 treatment episodes were identified in the medical records (Table 2). The majority (54.5%) of all treatment epi-sodes were with infliximab, followed by adalimumab (36.7%). Infliximab and adalimumab dominated as first line biologic treatment (68.5% and 30.1% respectively), as well as second line (38.2% and 51.2% respectively). In the most recent calen-dar period (2013–2017), the use of vedolizumab increased as second (19.0%) and third (37.8%) line treatment.
Sensitivity for individuals with treatment with biologics ever and for the start of treatment episodes
Considering all included biologic drugs, 74.4% (95% CI: 72.0–76.6) of all individuals with biologic treatment accord-ing to the medical records had at least one record of this in the PDR or NPR (Table 3). Looking instead at treatment epi-sodes, 80.1% (95% CI: 78.4–81.7) were ever captured in PDR or NPR. Limiting the accepted time window to ± one year or three months reduced the sensitivity to 63.3% (95% CI: 61.3–65.3) and 52.6% (95% CI: 50.5–54.6), respectively. The proportion of treatment episodes that was captured in PDR/ NPR decreased across calendar periods from 85.6% (95% CI: 82.5–88.3) in 2005–2008 to 76.8% (95% CI: 74.1–79.3) in 2013–2017.
Of all individual drugs, the sensitivity was high for adali-mumab, goliadali-mumab, and ustekinumab (88.6% [95% CI: 86.3–90.6], 90.0% [95% CI: 74.4–96.5], and 87.5% [95% CI: 52.9–97.8], respectively), even with the strictest time window of three months. The high sensitivity was consistent across calendar periods for adalimumab, while the assessment over calendar periods was not possible for golimumab and
ustekinumab, given their novelty. For infliximab, 40.6% (95% CI: 38.0–43.4) of treatment episodes in the medical records were ever captured by records in PDR/NPR and 18.5% (95% CI: 13.3–25.0) of vedolizumab treatment episodes were cap-tured. Limiting the time window to three months, the sensi-tivity of PDR/NPR records was 28.1% (95% CI: 25.7–30.7) for infliximab and 10.7% (95% CI: 6.9–16.3) for vedolizumab. The sensitivity for infliximab being ever captured in PDR/NPR decreased over time (51.2% [95% CI: 45.9–56.3] in 2005–2008 and 29.2% [25.2–33.6] in 2013–2017). The sensitivity was even lower in 2013–2017 (14.7% [95% CI: 11.8–18.3]) if the stricter time-window of three months was used. The impact of varying levels of misclassification is illustrated in Supplementary Appendix Figures 1 and 2.
We found no obvious differences by the type of hospital, i.e., university hospitals vs regional hospitals, but large differ-ences in infliximab sensitivity across individual centers (5.8–98.8%;Supplementary Appendix Table 9).
Timing of treatment episodes in PDR/NPR vs medical records
For the absolute majority of treatment episodes that did have a corresponding record in PDR/NPR, the treatment start according to PDR/NPR was dated close to the date of start in the medical records (within a month’s time, Figure 2(a–e)). For adalimumab and golimumab, some 10% of the prescrip-tions were filled around a month after treatment start according to the medical record.
Sensitivity of end of treatment episodes for adalimumab
Among the adalimumab treatment episodes with a corre-sponding start date in the PDR within three months, 54% also had an end date within one month of the end of treat-ment according to the medical records, and 85% within the range 3 months to þ3 months (Figure 3). The correspond-ing numbers in the sensitivity analyses shiftcorrespond-ing the time win-dow 30 days and þ30 days respectively, were 55% and 45%, and 84% and 82% (Supplementary Appendix Figures 3 and 4).
Discussion
In this validation study of >1300 IBD patients, PDR/NPR records were able to capture three out of four individuals and four out of five treatment episodes in patients ever treated with biologics. The PDR/NPR records were very good at capturing injection drugs such as adalimumab, golimu-mab, and ustekinugolimu-mab, but poor at capturing infusion drugs such as infliximab and vedolizumab. For injection biologics, start dates and end dates according to the doctor’s intention in medical records corresponded well with PDR/NPR records.
The sensitivity of PDR/NPR records for identifying patients ever treated with biologics in general was high, but decreased gradually when restricting the grace period to one year, or to three months. The impact of this misclassification
Table 2. Characteristics of validated treatment episodes from July, 2005, in patients with inflammatory bowel disease (patientsN ¼ 1361).
Drug
Episodes, N (%)
Total 1st line 2nd line 3rd line
Total 2323 1361 630 334 Infliximab 1265 (54.5%) 932 (68.5%) 240 (38.2%) 93 (27.9%) Adalimumab 852 (36.7%) 409 (30.1%) 322 (51.2%) 121 (36.3%) Golimumab 30 (1.3%) 9 (0.7%) 7 (1.1%) 14 (4.2%) Vedolizumab 168 (7.2%) 11 (0.8%) 59 (9.4%) 98 (29.4%) Ustekinumab 8 (0.3%) 0 1 (0.2%) 7 (2.1%) Year 2005–2008 550 436 100 14 Infliximab 350 (63.6%) 318 (72.9%) 25 (25.0%) 7 (50.0%) Adalimumab 200 (36.4%) 118 (27.1%) 75 (75.0%) 7 (50.0%) Golimumab 0 0 0 0 Vedolizumab 0 0 0 0 Ustekinumab 0 0 0 0 Year 2009–2012 781 503 218 60 Infliximab 467 (59.8%) 351 (69.8%) 90 (41.3%) 26 (43.3%) Adalimumab 311 (39.8%) 151 (30.0%) 126 (57.8%) 34 (56.7%) Golimumab 3 (0.4%) 1 (0.2%) 2 (0.9%) 0 Vedolizumab 0 0 0 0 Ustekinumab 0 0 0 0 Year 2013–2017 992 422 311 259 Infliximab 448 (45.2%) 263 (62.3%) 125 (40.2%) 60 (23.2%) Adalimumab 341 (34.4%) 140 (33.2%) 121 (38.9%) 80 (30.9%) Golimumab 27 (2.7%) 8 (1.9%) 5 (1.6%) 14 (5.4%) Vedolizumab 168 (16.9%) 11 (2.6%) 59 (19.0%) 98 (37.8%) Ustekinumab 8 (0.8%) 0 1 (0.3%) 7 (2.7%)
on an assessed relative risk depends on the sample size ratio of the comparison groups and the effect size of the studied association. A 20% misclassification of the treated group would have little impact even with an observed relative risk of 4, if the comparison group is large (ratio 1:10). Accordingly, data from the PDR/NPR may be adequate in observational studies in which exposure is defined by any treatment period with biologics. For example, since biologics are indicated for a certain disease phenotype, the data may be used to define patients that have ever experienced severe disease [18,19]. This level of accuracy in timing may also be sufficient to study long-term effects such as the risk of devel-oping malignant disease. On the other hand, it may not be appropriate to use this data alone in observational studies on short-term safety, such as the risk of infections after exposure to biologics.
It was clear that the mode of administration was key for the sensitivity of the PDR/NPR data. The injection drugs, ada-limumab, goada-limumab, and ustekinumab, are well captured in the PDR/NPR data. The register data was also very accurate
concerning the timing of treatment start and end. For adali-mumab and goliadali-mumab, it was common to start the treat-ment episode according to the registers about a month after the start according to the medical records, which may reflect a practice of administering the first injections at the hospital, before initiating a prescription. For ustekinumab, common practice is to administer the first dose as an infusion. This did not seem to affect the results of sensitivity according to different time windows, however there were few ustekinu-mab treatment episodes assessed in this study.
The infusion drugs infliximab and vedolizumab were poorly captured, except at €Ornsk€oldsvik Hospital. It is evident that the use of prescriptions and procedure codes for these drugs differ across hospitals and regions in Sweden, as can be inferred from assessments of national data previously conducted by the National Board of Health and Welfare [20,21]. The major determining factors of whether infusions are recorded in the registers are 1) for the PDR, whether drugs are individually prescribed or stock ordered, and 2) in the NPR, whether ATC codes for procedures are manually
Table 3. (A) Proportion of individual patients (N ¼ 1361) with biologic treatment according to the medical records and captured in PDR or NPR.
Drug Total N Captured N % (95% CI)
Any biologic drug 1361 1012 74.4% (72.0–76.6)
Infliximab 1014 408 40.2% (37.3–43.3)
Adalimumab 743 731 98.4% (97.2–99.1)
Golimumab 30 29 96.7% (83.3–99.4)
Vedolizumab 154 20 13.0% (8.6–19.2)
Ustekinumab 7 6 85.7% (48.7–97.4)
(B) Proportion of treatment episodes (N ¼ 2323) captured in PDR or NPR: ever, with a corresponding start date ± one year, or a corresponding start date ± 3 months, overall and by calendar period.
Ever Start dateþ/ 1 year Start dateþ/ 3 months
Drug Total N Cap N % (95% CI) Cap N % (95% CI) Cap N % (95% CI)
Overall
Any biologic drug 2323 1861 80.1% (78.4–81.7) 1471 63.3% (61.3–65.3) 1221 52.6% (50.5–54.6) Infliximab 1265 514 40.6% (38.0–43.4) 412 32.6% (30.0–35.2) 356 28.1% (25.7–30.7) Adalimumab 852 839 98.5% (97.4–99.1) 813 95.4% (93.8–96.6) 755 88.6% (86.3–90.6) Golimumab 30 29 96.7% (83.3–99.4) 29 96.7% (83.3–99.4) 27 90.0% (74.4–96.5) Vedolizumab 168 31 18.5% (13.3–25.0) 27 16.1% (11.3–22.4) 18 10.7% (6.9–16.3) Ustekinumab 8 7 87.5% (52.9–97.8) 7 87.5% (52.9–97.8) 7 87.5% (52.9–97.8) By calendar period
Any biologic drug
2005–2008 550 471 85.6% (82.5–88.3) 371 67.5% (63.4–71.2) 310 56.4% (52.2–60.5) 2009–2012 781 628 80.4% (77.5–83.0) 518 66.3% (62.9–69.6) 455 58.3% (54.8–61.7) 2013–2017 992 762 76.8% (74.1–79.3) 582 58.7% (55.6–61.7) 456 46.0% (42.9–49.1) Infliximab 2005–2008 350 179 51.1% (45.9–56.3) 151 43.1% (38.1–48.4) 129 36.9% (32.0–42.0) 2009–2012 467 204 43.7% (39.3–48.2) 172 36.8% (32.6–41.3) 161 34.5% (30.3–38.9) 2013–2017 448 131 29.2% (25.2–33.6) 89 19.9% (16.4–23.8) 66 14.7% (11.8–18.3) Adalimumab 2005–2008 200 196 98.0% (95.0–99.2) 191 95.5% (91.7–97.6) 174 87.0% (81.6–91.0) 2009–2012 311 306 98.4% (96.3–99.3) 300 96.5% (93.8–98.0) 276 88.7% (84.8–91.8) 2013–2017 341 337 98.8% (97.0–99.5) 322 94.4% (91.5–96.4) 305 89.4% (85.7–92.3) Golimumab 2005–2008 0 – – – – – – 2009–2012 3 3 100% (43.9–100) 3 100% (43.9–100) 2 66.7% (20.8–93.9) 2013–2017 27 26 96.3% (81.7–99.3) 26 96.3% (81.7–99.3) 25 92.6% (76.6–97.9) Vedolizumab 2005–2008 0 – – – – – – 2009–2012 0 – – – – – – 2013–2017 168 31 18.5% (13.3–25.0) 27 16.1% (11.3–22.4) 18 10.7% (6.9–16.3) Ustekinumab 2005–2008 0 – – – – – – 2009–2012 0 – – – – – – 2013–2017 8 7 87.5% (52.9–97.8) 7 87.5% (52.9–97.8) 7 87.5% (52.9–97.8) Cap N¼ captured N.
Figure 2. (a ) Inf lix im ab tr e at me nt st ar t in rev ie we d m ed ic al re co rd s vs re g is te rs (n ¼ 1 2 6 5 ). Tr ea tm en t ep is o d e s in th e m e d ic al re co rd s w it h o u t a co rr es p o nd in g reg is te r re co rd are sh o w n in g re y, at th e sm al le st va lu e o f th e x-sc al e. (b ) A da lim u m ab tr e at me nt st ar t in re vie w e d m ed ic al re co rd s vs re g is te rs (n ¼ 8 5 2 ). Tr ea tm en t e p is o d e s in th e me d ic al re co rd s w it h o u t a co rr es p o nd in g reg is te r re co rd ar e sh o w n in g re y, at th e sm al le st va lu e o f th e x-sc al e.
Figure 2. (c) Golimumab treatment start in reviewed medical records vs registers (n ¼ 30). Treatment episodes in the medical records without a corresponding register record are shown in grey, at the smallest value of the x-scale. (d) Vedolizumab treatment start in medical records reviewed vs registers (n ¼ 168). Treatment episodes in the medical records without a corresponding register record are shown in grey, at the smallest value of the x-scale.
Figure 2. (e) Ustekinumab treatment start in medical records reviewed vs registers (n ¼ 8). Treatment episodes in the medical records without a corresponding register record are shown in grey, at the smallest value of the x-scale.
recorded, which in turn is dependent on how well the soft-ware for medical documentation aligns with reporting to the register, and how much reporting is encouraged and imple-mented (A. Jacobsson, personal communication 23 November 2020). Although the reporting of procedure codes (e.g., intravenous drug administration) may be important for economic compensation, to specify ATC codes does not add anything. This makes the impact of this motivating factor dif-ficult to assess for our study, in which we required full ATC codes to define administration in the NPR. Also, visits
administered by nurses, such as visits for infusion treatment, may not be reported to the registers at all.
Apart from regional differences, routines have also changed through the years; fewer infliximab treatment epi-sodes were recorded in the registers in 2013–2017 as com-pared to 2005–2008. Instead of distribution by individual prescriptions, regional care-givers may opt to provide infu-sion drugs via stock orders to hospitals, which facilitates the clinical logistics and may allow the regions to negotiate a more favorable price, although all costs are ultimately
Figure 3. Adalimumab treatment stop in medical records reviewed vs registers for patients with a corresponding treatment start within 3 months (n ¼ 751; [4 patients were found in the NPR only]). Patients with ongoing treatment in journals reviewed after validation performed were assigned date of validation, also for patients with ongoing treatment in the registers.
covered by the regional care-giver regardless. Over the years, stock ordering has become more rational as biologics have become more commonly used, and prices have decreased. On a separate note, the majority of treatment episodes iden-tified were sourced from the PDR, rather than from proced-ure codes in the NPR. The NPR data on biologics may be used to gain as much information about infusion biologics as possible, but does not add much to PDR data for injection biologics. To gain more information about biologic treat-ment, in particular infusion treattreat-ment, among IBD patients in Sweden in the future, the reporting of ATC codes to the NPR should be further encouraged. In the meantime, another potential data source may be the disease-specific Swedish Inflammatory Bowel Disease Registry (SWIBREG). Although reporting to SWIBREG is not mandatory like to the national registers, coverage is continuously increasing and particular efforts have been made to include all patients with biologic treatment [1].
Infliximab has remained the most common first line treat-ment, although its use has decreased over calendar periods. Therefore, the impact of insufficient information on inflixi-mab on total biologics use may decrease over time. However, the alternative that increased the most in use over time was vedolizumab, also an infusion drug. Consequently, the PDR/NPR data in most counties/hospitals are not suffi-cient when studying treatment lines or switches between biologic drugs. It should be noted that a formulation of vedolizumab for subcutaneous injection was recently approved by the European Commission, which is likely to change the sensitivity of vedolizumab data in the registers in the future.
Strengths and limitations
We conducted a comprehensive medical record review of biologic treatment in IBD patients. Thanks to the PIN, we were able to link this detailed data to the nationwide health registers PDR and NPR and to assess the validity of biologic treatment in these data sources. A well-structured CRF aided medical students in extracting information in a standardized manner under the supervision of experienced gastroenterolo-gists. Although we did not include all centers in Sweden, the included patients represent IBD patients from different regions and hospital levels covering around one third of the Swedish population. Moreover, the population selected for record review had similar characteristics, and seem to well represent the entire group of patients defined as treated with biologics according to PDR/NPR records as seen in Table 1. As for all studies using the PDR, we were not able to ascertain actual administration of the drugs. However, the use of filled prescriptions, as opposed to merely prescribed drugs, increases the likelihood that the drugs were taken according to the doctor’s intent. We were also, with the data at hand, not able to study and assess the validity of the register information for changes in dose.
Conclusions
The PDR and the NPR are excellent sources of information on treatment with injection biologics such as adalimumab, golimumab, and ustekinumab in patients with IBD in Sweden. Use of infliximab and infusion vedolizumab is poorly captured, why these data should only be used after careful consideration of their limitations or complemented by other data sources, such as the disease-specific quality regis-ter SWIBREG.
Acknowledgements
The authors acknowledge the reviewers of medical records; Anna Fr€oborg, Emelie Mårdberg, Helena K Ohlsson, Ida Holm, Johan Weiber, Johanna Engstr€om, Lars Becker, Natalia Mouratidou and Per Olsson. The authors also acknowledge Anders Jacobsson, statistician at the National Board of Health and Welfare for valuable input through personal communication.
Collaborators as part of the SWIBREG study group Henrik Hjortswang
Jan Bj€ork Olof Grip Martin Rejler
Ulrika Lorentzon Fagerberg Caroline Nordenvall Hans Strid Michael Eberhardson Susanna J€aghult Marie Andersson Pontus Karling Malin Olsson Per Hellstr€om Jonas Bengtson Ethical approval
The study has been approved by the Ethical Review Board in Stockholm (registration number: 2016/191-31/2).
Author contributions
Guarantor: Br€oms, S€oderling, and Olen had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis; study concept: Olen; design: all co-authors; acquisition of data: Olen; analysis: Br€oms, S€oderling, and Sachs; drafting of the manuscript: Br€oms, S€oderling, Everhov, and Olen; critical revision of the manuscript for important intellectual content: All co-authors; supervision: Olen.
Disclosure statement
Dr. Br€oms has been involved in post-marketing pregnancy safety studies at Karolinska Institutet partly financed by Janssen and Pfizer.
Dr. Halfvarson served as speaker and/or advisory board member for AbbVie, Celgene, Celltrion, Dr Falk Pharma and the Falk Foundation, Ferring, Hospira, Janssen, MEDA, Medivir, MSD, Novartis, Olink Proteomics, Pfizer, Prometheus Laboratories, Sandoz, Shire, Takeda, Thermo Fisher Scientific, Tillotts Pharma, Vifor Pharma, UCB and received grant support from Janssen, MSD and Takeda.
Dr. Myrelid has been involved in projects partly financed by investi-gator-initiated grants from Janssen, Ferring and Pfizer. None of those studies have any relation to the present study. He has also participated in advisory boards or lecturing for Janssen, Ferring, Takeda, and Pfizer
regarding topics not related to the present study. Dr Myrelid is chairman of the national Swedish IBD register (SWIBREG).
Dr. Everhov has participated in projects at Karolinska Institutet partly financed by investigator-initiated grants from Janssen and Ferring.
Dr. Olen has been PI on projects at Karolinska Institutet partly financed by investigator-initiated grants from Janssen and Ferring, and also reports a grant from Pfizer in the context of a national safety moni-toring program. None of those studies have any relation to the present study. Karolinska Institutet also has received fees for Olen’s lectures and participation on advisory boards from Janssen, Ferring, Takeda, and Pfizer regarding topics not related to the present study.
Dr. Br€oms was supported by Stockholm Region (postdoctoral appointment). Dr. Myrelid was supported by Regional Agreement on Medical Training and Clinical Research between €Osterg€otland Region and Link€oping University (ALF). Dr. Everhov was supported by the Bengt Ihre Research Fellowship. Dr. Olen was supported by grants from the Swedish Medical Society, Karolinska Institutet Foundations, Young Scholar Award from the Strategic Research Area Epidemiology program at Karolinska Institutet, and Regional Agreement on Medical Training and Clinical Research between Stockholm Region and Karolinska Institutet (ALF).
None of the funding organizations has had any role in the design and conduct of the study; in the collection, management, and analysis of the data; or in the preparation, review, and approval of the manuscript.
ORCID
Gabriella Br€oms http://orcid.org/0000-0002-2423-1968
Par Myrelid http://orcid.org/0000-0001-7518-9213
Jonas F. Ludvigsson http://orcid.org/0000-0003-1024-5602
References
[1] SWIBREG årsrapport 2019/SWIBREG annual report 2019 [Internet]. 2020 [cited 2020 Dec 1]. Available from:https://www.swibreg.se/ wp-content/uploads/2020/09/SWIBREG_Arsrapport_2019.pdf
[2] Targan SR, Hanauer SB, van Deventer SJ, et al. A short-term study of chimeric monoclonal antibody cA2 to tumor necrosis factor alpha for Crohn’s disease. Crohn’s Disease cA2 Study Group. N Engl J Med. 1997;337:1029–1035.
[3] Hanauer SB, Feagan BG, Lichtenstein GR, et al. Maintenance inflix-imab for Crohn’s disease: the ACCENT I randomised trial. Lancet. 2002;359:1541–1549.
[4] Colombel JF, Sandborn WJ, Rutgeerts P, et al. Adalimumab for maintenance of clinical response and remission in patients with Crohn’s disease: the CHARM trial. Gastroenterology. 2007;132: 52–65.
[5] Rutgeerts P, Sandborn WJ, Feagan BG, et al. Infliximab for induc-tion and maintenance therapy for ulcerative colitis. N Engl J Med. 2005;353:2462–2476.
[6] Sandborn WJ, van Assche G, Reinisch W, et al. Adalimumab indu-ces and maintains clinical remission in patients with
moderate-to-severe ulcerative colitis. Gastroenterology. 2012;142: 257–65.e1–3.
[7] Sandborn WJ, Feagan BG, Marano C, et al., PURSUIT-SC Study Group. Subcutaneous golimumab induces clinical response and remission in patients with moderate-to-severe ulcerative colitis. Gastroenterology. 2014;146:85–95. quiz e14–5.
[8] Feagan BG, Rutgeerts P, Sands BE, et al. Vedolizumab as induc-tion and maintenance therapy for ulcerative colitis. N Engl J Med. 2013;369:699–710.
[9] Feagan BG, Sandborn WJ, Gasink C, et al. Ustekinumab as induc-tion and maintenance therapy for Crohn’s disease. N Engl J Med. 2016;375:1946–1960.
[10] Wallerstedt SM, Wettermark B, Hoffmann M. The First Decade with the Swedish Prescribed Drug Register– a systematic review of the output in the scientific literature. Basic Clin Pharmacol Toxicol. 2016;119:464–469.
[11] Ludvigsson JF, Andersson E, Ekbom A, et al. External review and validation of the Swedish national inpatient register. BMC Public Health. 2011;11:450.
[12] Olen O, Askling J, Sachs MC, et al. Mortality in adult-onset and elderly-onset IBD: a nationwide register-based cohort study 1964-2014. Gut. 2020;69:453–461.
[13] Eberhardson M, Soderling JK, Neovius M, et al. Anti-TNF treat-ment in Crohn’s disease and risk of bowel resection-a population based cohort study. Aliment Pharmacol Ther. 2017;46:589–598. [14] Ludvigsson JF, Otterblad-Olausson P, Pettersson BU, et al. The
Swedish personal identity number: possibilities and pitfalls in healthcare and medical research. Eur J Epidemiol. 2009;24: 659–667.
[15] Bortfall och kvalitet i patientregistret: Socialstyrelsen [Quality of the National Patient Register]. National Board of Health and Welfare; 2018 [cited 2020 Mar 06]. Available from: https://www. socialstyrelsen.se/statistik-och-data/register/alla-register/patientre-gistret/bortfall-och-kvalitet/
[16] Ludvigsson JF, Lashkariani M. Cohort profile: ESPRESSO (Epidemiology Strengthened by histoPathology Reports in Sweden). Clin Epidemiol. 2019;11:101–114.
[17] Wilson EB. Probable inference, the law of succession, and statis-tical inference. J Am Stat Assoc. 1927;22:209–212.
[18] Torres J, Bonovas S, Doherty G, et al. ECCO guidelines on thera-peutics in Crohn’s disease: medical treatment. J Crohns Colitis. 2020;14:4–22.
[19] Harbord M, Eliakim R, Bettenworth D, et al., European Crohn’s and Colitis Organisation [ECCO]. Third European Evidence-based Consensus on Diagnosis and Management of Ulcerative Colitis. Part 2: current management. J Crohns Colitis. 2017;11:769–784. [20] Kodningskvalitet i patientregistret– Ett nytt verktyg f€or att m€ata
kvalitet. Socialstyrelsen. National Board of Health and Welfare; 2013. [p. 14–15, Diagrams 3 and 4]. Available from:https://www. socialstyrelsen.se/globalassets/sharepoint-dokument/artikelkata-log/klassifikationer-och-koder/2013-3-10.pdf
[21] F€orstudie kring individdata f€or rekvisitionsl€akemedel.: Socialstyrelsen. National Board of Health and Welfare; 2012 [p. 11, Figure 2]. Available from: https://www.socialstyrelsen.se/glob-alassets/sharepoint-dokument/artikelkatalog/ovrigt/2012-4-21.pdf
