Article Text

Concise report
Generalisability of clinical registers used for drug safety and comparative effectiveness research: coverage of the Swedish Biologics Register
  1. M Neovius1,
  2. JF Simard1,
  3. A Sundström1,
  4. L Jacobsson2,
  5. P Geborek2,
  6. T Saxne2,
  7. N Feltelius3,
  8. L Klareskog4,
  9. J Askling1,4,
  10. for the ARTIS Study Group
  1. 1Clinical Epidemiology Unit, Department of Medicine, Karolinska Institutet, Stockholm, Sweden
  2. 2Department of Rheumatology, Lund University, Lund, Sweden
  3. 3Medical Products Agency, Uppsala, Sweden
  4. 4Rheumatology Unit, Department of Medicine, Karolinska Institutet, Stockholm, Sweden
  1. Correspondence to Martin Neovius, Clinical Epidemiology Unit (T2), Department of Medicine, Karolinska Institutet, SE-171 76 Stockholm, Sweden; martin.neovius{at}


Objective To determine coverage and generalisability of data in the Swedish Biologics Register ARTIS.

Methods Patients with adult onset rheumatoid arthritis (RA) were identified in the National Patient Register and the Swedish Rheumatology Quality Register, including the ARTIS cohort of patients exposed to biological agents. Exposure to etanercept and adalimumab between 2006 and 2008 was determined by register linkage to the Prescribed Drug Register which contains patient-level data on >99% of all etanercept and adalimumab use in Sweden.

Results Of 62 897 patients with RA, 6510 had received treatment with etanercept or adalimumab according to the Prescribed Drug Register. Of these, 5673 were also registered in ARTIS, resulting in a national coverage of 87%. The regional variation was small with >85% coverage in 18 of 21 counties. In multivariable analysis, ARTIS-registered and non-registered patients did not differ by age (p=0.62), sex (p=0.84) or education level (p=0.24).

Conclusion Nationwide drug dispensing and demographic data may function as quality metrics for coverage and generalisability assessments. Using such data, the coverage of ARTIS was estimated at 87% with no indications of compromised external generalisability regarding demography.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Clinical registers, claims data and electronic medical records are increasingly used and proposed as data sources for drug safety and comparative effectiveness research.1,,6 For example, clinical registers for biological agents have aided post-marketing pharmacovigilance studies of new and often complex drugs guiding both regulatory and clinical decision-making.2 3 7,,10 In Sweden, long-term surveillance of biological treatments for rheumatoid arthritis (RA) is conducted using a national population-based clinical register (ARTIS) which is regularly enriched with cancer, death and productivity outcome data by linkages to other national registers.1 Through large-scale longitudinal data collection in routine care, register linkages using ARTIS as a data backbone enable assessments of patient subgroups, time frames and outcomes unavailable within randomised trials such as uncommon adverse events, effects on hard outcomes (as opposed to surrogate markers)11,,13 and head-to-head drug comparisons.4

Increasing reliance on register data raises demand for methods to assess internal and external validity. The extent of source population coverage in registers is often unknown and raises concerns about generalisability. Such concerns limit the usefulness of any findings obtained from register data. Previously used means to investigate coverage of treatment registers have included comparisons of register data against aggregated sales data.14

This study investigated an alternative approach to assess coverage and generalisability of clinical register data using ARTIS as an example. Taking advantage of the opportunity to perform register linkages and the existence of governmental individual patient data registers in Sweden, we explored the possibility of quantifying the coverage and generalisability of ARTIS regarding patients with RA receiving biological treatment.


Setting: The Swedish National Health Service

Sweden has a population of 9.3 million (31 May 2009, and consists of 21 counties. The healthcare system is tax funded and offers universal access. Prescription drugs are provided free of charge above a threshold of SEK1800 annually (approximately €180), and all prescription drug use is registered in the Prescribed Drug Register. The register is complete, although missing data on the personal identification number are present in approximately 0.3% of records. However, only ambulatory care use is possible to link on the patient level, while hospital use is not.

Register linkage procedure

Based on each Swedish resident's unique personal identification number, two national and virtually complete registers operated by the National Board of Health and Welfare (the National Patient Register and the Prescribed Drug Register) were linked with the Swedish Rheumatology Quality Register of which ARTIS forms one part (figure 1; described in more detail in online supplement). Patients with adult onset RA were identified via the National Patient Register (inpatient and non-primary care outpatient visits) and the Swedish Rheumatology Quality Register.15 Thereafter, data on use of etanercept, adalimumab, infliximab, anakinra, abatacept and rituximab were retrieved from the Prescribed Drug Register and ARTIS.

Figure 1

Venn diagram illustrating the overlap in number of patients identified from each source (total n=62 897). SRQ consists of patients with early rheumatoid arthritis (RA) and ARTIS patients (patients treated with biological agents); PDR allows only ambulatory care use of biological agents to be linked on a patient level; ARTIS includes both ambulatory care and hospital use of biological agents. Biological agents are: etanercept (nPDR=4617; nARTIS=3966), adalimumab (nPDR=2596; nARTIS=2131), infliximab (nPDR=520; nARTIS=2178), abatacept (nPDR=51; nARTIS=157), anakinra (nPDR=147; nARTIS=86) and rituximab (nPDR=73; nARTIS=717) use in the period 1 January 2006 to 31 December 2008.

Outcome measure

Coverage was calculated for use of etanercept and/or adalimumab for which >99% of all use is in ambulatory care.16 Hence the Prescribed Drug Register allows for patient-level linkage of practically all patients receiving these drugs in Sweden. Coverage was defined as the number of patients with RA exposed to etanercept or adalimumab in both ARTIS and the Prescribed Drug Register divided by the number of exposed patients with RA in the Prescribed Drug Register.

Census and education data

Data were retrieved from Statistics Sweden to ascertain whether patients were registered in Sweden and alive on 31 December in 2006, 2007 and 2008. Education level was also retrieved from Statistics Sweden and categorised into ≤9, 10–12 and >12 years.

Statistical analysis

The data were analysed using SAS Version 9 (SAS Institute Inc, Cary, North Carolina, USA). Overall and county-specific coverage estimates were cross-tabulated. To assess generalisability of the ARTIS data, differences between etanercept/adalimumab exposed patients registered and not registered in ARTIS regarding age, sex and education level were investigated using logistic regression.


In the outpatient, inpatient and Swedish Rheumatology Quality Register, a total of 62 897 patients were identified alive and registered in Sweden during the study period (figure 1). In the Prescribed Drug Register, 7014 patients had registered ambulatory care dispensings with etanercept, adalimumab, infliximab, anakinra, abatacept and/or rituximab. Based on both ambulatory care and hospital use, 8493 patients were registered in ARTIS for use of biological agents. While ARTIS had more patients identified with biological agents primarily used in the hospital (infliximab, rituximab, abatacept), not all patients on the biological agents used primarily in ambulatory care were registered in ARTIS for any of the drugs (etanercept, adalimumab, anakinra; footnote in figure 1).

Of the 7014 patients exposed to biological agents identified in the Prescribed Drug Register, 6510 patients had dispensings of etanercept or adalimumab registered (table 1). In ARTIS, 5673 of these 6510 patients were registered with use of either etanercept or adalimumab, resulting in a national coverage of 87%. Geographical variation across counties was small, with 8 of 21 counties having a coverage of >90% and 18 with a coverage of >85% (figure 2). Of all patients exposed to biological agents identified in the Prescribed Drug Register, 92% were also identified in ARTIS (6480/7014; including etanercept, adalimumab, infliximab, anakinra, abatacept and rituximab).

Figure 2

County variation in coverage for biological treatment with etanercept or adalimumab in the Swedish Biologics Register ARTIS in patients with rheumatoid arthritis (compared with the Prescribed Drug Register, 2006–8).

Table 1

Characteristics of patients receiving etanercept and/or adalimumab

Etanercept/adalimumab-exposed patients registered in ARTIS (n=5673) did not differ from those not registered in ARTIS but identified with treatment in the Prescribed Drug Register (n=837) by age (55.4 vs 56.0 years; p=0.16), sex (77% vs 77% women; p=0.85) or education level (≤9 years: 26% vs 28%; 10–12 years: 45% vs 45%; and >12 years: 29% vs 26%; p=0.16). In multivariable analysis, neither age (p=0.62), sex (p=0.84) nor education level (p=0.24) were associated with non-registration in ARTIS.


ARTIS covered 87% of Swedish patients with RA prescribed etanercept or adalimumab between 2006 and 2008. The remaining 13% did not differ from the rest by age, sex or education level. Together these findings indicate that (1) data from ARTIS are generalisable to the Swedish population receiving biological agents; (2) additional efforts can identify exposed patients not in ARTIS; and (3) linkages to governmental and virtually complete registers provide a powerful yet simple means to assess coverage and generalisability of clinical registers. Generic by nature, this algorithm can be used in other settings.

Register studies are commonly conducted without formal assessments of the completeness of the registered data—that is, of the potential for selection bias and limited generalisability of results. No previous study has quantified the coverage in ARTIS for patients exposed to biological agents although a comparison of official aggregated sales in 2001–3 and register-based estimates of treatment costs in southern Sweden indicated a >90% coverage of the regional component (SSATG) of ARTIS used in this area.14

The high national coverage and generalisability found in this study supports the use of ARTIS data for research, regulatory and clinical purposes. In Sweden, data from clinical quality registers such as ARTIS are reported annually to the healthcare authorities for evaluation of regional differences, potentially influencing policy. It is therefore crucial to quantify coverage to determine whether variations in drug provision or other quality indicators reflect actual differences or only regional variations in coverage.16

In addition to providing information about coverage and generalisability, patients exposed to the drug but not present in ARTIS were identified. Including exposed but previously unknown patients in analyses using register-based outcomes increases power, reduces selection bias and decreases exposure misclassification when making comparisons with control patients sampled from the RA population.

The main strength of this study was the availability of >99% complete data for use of biological agents for nearly the whole Swedish RA population. One limitation is that indication for use is not registered in the Prescribed Drug Register. Some patients with RA receiving biological agents may therefore have been missed. To capture the RA population we included patients with a diagnosis of RA in inpatient or non-primary outpatient care. Although some of these patients may be misclassified, >90% of RA diagnoses in inpatient17 and 85% in outpatient care18 are patients fulfilling the ACR classification criteria. Furthermore, the approach is ideally suited for drugs used only in ambulatory care and dispensed directly to patients from pharmacies as hospital drug use cannot currently be linked on the patient level. In Sweden, hospital use dominates for infliximab, abatacept and rituximab.16 We therefore restricted the analysis to etanercept and adalimumab. However, although infliximab has a large share of overall use of biological agents, abatacept and rituximab constitute only 4% of the biological treatments in ARTIS (31 May 2008) in which both hospital and ambulatory use are registered.

In summary, ARTIS has high coverage for patients with RA exposed to biological agents, and patients not registered do not differ demographically from registered patients. This increases confidence in the generalisability of ARTIS-based safety and comparative effectiveness findings. The approach can be applied to other clinical registers to assess generalisability and reduce exposure misclassification.


The authors thank the following rheumatologists for contributing data to the Swedish Rheumatology Quality Registers. Yngve Adolfsson, Sunderby Hospital, Luleå; Ewa Berglin, Norrland's University Hospital, Umeå; Torgny Smedby, Östersund County Hospital; Rüdi Götze, Sundsvall County Hospital; Anna-Carin Holmqvist, Hudiksvall's Hospital; Sven Tegmark, Gävle County Hospital; Jörgen Lysholm, Falu lasarett, Falun; Solveig Gustafsson, Karlstad's Central Hospital; Eva Baecklund, Akademiska Hospital, Uppsala; Rolf Oding, Västerås Hospital; Per Salomonsson, University Hospital, Örebro; Birgitta Nordmark, Karolinska University Hospital, Solna; Ingiäld Hafström, Karolinska University Hospital, Huddinge; Göran Lindahl, Danderyd's Hospital, Stockholm; Gun Sandahl, Queen Sophia Hospital, Stockholm; Martin Mousa, Visby lasarett, Visby; Anders Lindblad, Visby Privat, Visby; Åke Thörner, Mälarsjukhuset, Eskilstuna; Lars Cöster, University Hospital, Linköping; Sören Transö, County Hospital Ryhov, Jönköping; Olle Svernell, Västervik's Hospital; Claudia Jacobs, Oskarshamn's Hospital; Bengt Lindell, Kalmar County Hospital; Maria Söderlin, Växjö's Central Hospital; Olof Börjesson, Växjö Privat; Göran Kvist, Centrallasarettet Borås; Karin Svensson, Kärnsjukhuset, Skövde; Tomas Torstenson, Uddevalla Hospital; Ingeli Andreasson, Göteborg privat; Lennart Bertilsson, Sahlgrenska University Hospital, Gothenburg; Tore Saxne, University Hospital in Lund; Miriam Karlsson, Lasarettet Trelleborg; Annika Teleman, Spenshult, Oskarström; Catharina Keller, Helsingborg's Lasarett; Astrid Schröder, Ängelholm's Hospital; Jan Theander, Kristianstad's Central Hospital; Christina Book, MAS University Hospital, Malmö. They also thank Elizabeth Arkema for language revision and Maud Rütting at the Medical Products Agency.


Supplementary materials

  • Web only data

    Files in this Data Supplement:


  • Competing interests The ARTIS Study Group conducts scientific analyses using data from the Swedish Biologics Register ARTIS run by the Swedish Society for Rheumatology. For the maintenance of this register, the Swedish Society for Rheumatology has received funding independent of the conduct of these scientific analyses, from Schering-Plough, BMS, Wyeth, Abbott Laboratories and Roche.

  • Ethics approval Ethical approval was granted by the regional ethics committee, Karolinska Institutet, Sweden.

  • Provenance and peer review Commissioned; externally peer-reviewed.