Tammo Brunekreef

Systemic Lupus Erythematosus (SLE) is a rare systemic autoimmune disease with a prevalence of 43.7 per 100.000 people. Symptoms range from relatively mild complaints, such as muscle and joint pains or a characteristic skin rash, to disabling or potentially life-threatening symptoms, such as psychosis and inflammation of internal organs. Because SLE can cause a wide variety of symptoms, establishing the diagnosis is often difficult. Many symptoms caused by SLE are not specific to the disease and have many other, more common possible causes. This often leads to a diagnostic delay, where the time between the onset of first symptoms and the diagnosis regularly spans months to several years. During this period, untreated disease can cause organ damage and imposes a high disease burden on patients. It is therefore important to improve the diagnostic process for SLE and to reduce this diagnostic delay.

Part 1
Because SLE is an uncommon condition, collecting high-quality data for research is challenging. Setting up structured trials to gather such data is time consuming and expensive. Therefore, other data sources are also considered for research purposes, such as the Electronic Health Records (EHR). One of the EHR’s main functions is to record information relevant to managing patients’ disease, including symptoms and diagnoses. As a result, the EHR contains large amounts of data that may be highly relevant for answering research questions. A major challenge when using EHR data for research is that it contains a lot of free text or unstructured data, which is difficult to process.
In most cases, relevant data are manually extracted from the EHR by conducting chart reviews, which is time consuming and subject to substantial inter- and intraobserver variation. In Chapter 2 we describe how we developed an algorithm to filter relevant information from EHR data. For our research this focused mainly on whether or not a diagnosis of SLE was present. With our algorithm we show that unstructured EHR data can be used to obtain relevant information in an automated and thus reproducible way that can be reused for research. The algorithm searches for occurrences of specific keywords and evaluates the context of those keywords to recognize, for example, negations. Tools like our algorithm can help facilitate large-scale chart review.

Part 2
The diagnosis of SLE is a clinical diagnosis made on the basis of a combination of clinical and immunological features. One of the most important immunological features of SLE is the presence of autoantibodies. Over the years more than 100 different autoantibodies associated with SLE have been discovered. Many of these autoantibodies target antigens from the cell nucleus and thus fall into the group of antinuclear antibodies (ANAs). These autoantibodies play an important role in the pathophysiology of SLE. Some autoantibodies are known to appear in the blood of future patients years before the diagnosis of SLE, which offers potential for using autoantibodies as biomarkers for early diagnosis. Many of these antibodies, however, also occur in a small proportion of the healthy population, which limits the potential for a screening test in the general population. The presence of these autoantibodies in healthy individuals could be seen as a form of benign autoimmunity that does not lead to clinical autoimmune disease.
In clinical practice it is common for patients to have a positive ANA test result without ultimately receiving a diagnosis of an autoimmune disease. In Chapter 5 we investigated whether these positive tests should be interpreted as a sign of a possible future autoimmune disease or as clinically irrelevant. We show that only a small proportion of patients with complaints that could be related to an autoimmune disease eventually receive such a diagnosis. A positive ANA test by itself was a poor predictor of later receiving a diagnosis of an autoimmune disease. Having both symptoms and the presence of specific ANAs related to the suspected autoimmune disease was a stronger predictor of progression to disease, although this still occurred infrequently.
Although the predictive value of individual autoantibodies appears limited, in Chapter 4 we demonstrate that constructing autoantibody profiles offers additional opportunities. In this study we used a large cohort of SLE patients and multiple relevant control groups to create autoantibody profiles. We then show that future SLE patients often already exhibit an autoantibody profile before diagnosis that can be used to identify patients at high risk of developing SLE at an early stage.
In Chapter 3 we show that antibodies directed against CpG oligodeoxynucleotides (ODNs) have a high prevalence in the blood of SLE patients and therefore make an important contribution to forming the autoantibody profiles. Previous studies have shown that CpG ODNs can stimulate plasmacytoid dendritic cells (pDCs) via TLR9. pDCs are important producers of inflammatory type I interferons (IFN) and are likely largely responsible for the increase of the so-called IFN signature in SLE patients. In Chapter 6 we further investigated the relationship between autoantibodies and the IFN signature. We show that antibodies directed against EphB2, RipP1, and RNP70 are associated with a high IFN score.
In Chapter 7 we provide an overview of recent developments in diagnostics for SLE. We considered both autoantibodies and other biomarkers, such as chemokines, cytokines, RNAs, and gene expression. Although this review did not identify a perfect biomarker, it does provide greater insight into the pathophysiology of SLE. We argue that the quality of diagnostic studies could be significantly improved by establishing large population-based cohorts. Such cohorts can yield valuable insights for many diseases and, specifically for SLE, enable research into immunological changes in patients prior to diagnosis.

Conclusion
In this thesis I demonstrate how we used several innovative approaches to collect both clinical and immunological data. We show that technical tools can be designed to unlock additional data sources for research. We explored the presence of various autoantibodies and their relationship with other immunological phenomena, and we show that autoantibody profiles can be used to recognize future SLE patients at an early stage, with the aim of reducing diagnostic delay.