Introduction Treatment possibilities have changed in inflammatory bowel disease (IBD). We assessed changes in medical treatment and surgery over time and impact of medications on risk of surgery in a population-based cohort.
Methods 48 967 individuals were diagnosed with IBD (Crohn's disease (CD), 13 185; ulcerative colitis (UC), 35 782) during 1979–2011. Cumulative probability of receiving 5-aminosalicylic acids (5-ASA), topical, oral corticosteroids, thiopurines, and tumour necrosis factor-α (TNF-α) blockers, and of first minor or major surgery according to period of diagnosis, was estimated. Medication use and risk of surgery was examined by Cox regression.
Results 5-year cumulative probability of first major surgery decreased from 44.7% in cohort (1979–1986) to 19.6% in cohort (2003–2011) (p < 0.001) for CD, and from 11.7% in cohort (1979–1986) to 7.5% in cohort (2003–2011) (p < 0.001) for UC. Minor surgery risk decreased significantly in CD. From cohort (1995–2002) to cohort (2003–2011), a significant increase in use of thiopurines and TNF-α blockers was observed, paralleled by a significant decrease in use of 5-ASA and corticosteroids. Comparing use of azathioprine (or oral corticosteroids) to never-use, no convincing surgery-sparing effect was found. Comparing use in 3+ months of a given drug with use <3 months, only 3+ months use of oral corticosteroids reduced the risk of surgery in patients with disease duration of >1 year.
Conclusions Parallel to an increasing use of thiopurines and TNF-α blockers in IBD over time, a persistent significant decrease in surgery rates was observed along with a significant decrease in use of 5-ASA and corticosteroids. However, no convincing surgery-sparing effect of newer medications was found.
- CROHN'S DISEASE
- SURGERY FOR IBD
- ULCERATIVE COLITIS
Statistics from Altmetric.com
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.
Significance of this study
What is already known about this subject?
Treatment options in inflammatory bowel disease have changed through the last decades.
Whether these changes have had an impact on long-term need of surgery remains unknown.
What are the new findings?
We observed a significant decrease in need of surgery over the last three decades in patients with ulcerative colitis or Crohn's disease.
The decrease in need of surgery was paralleled by an increase in use of azathioprine and anti-TNF-α blockers along with a decrease in use of 5-aminosalicylates (5-ASA) and topical corticosteroids in both diseases, and a decrease in use of oral corticosteroids in Crohn's disease.
Of all medications, only long-term use of oral corticosteroids in patients with disease duration of ≥1 year, was associated with a reduced risk of surgery.
How might it impact on clinical practice in the foreseeable future?
The study provides clinicians with long-term population-based data on risk of surgery and trends in medication use in patients with inflammatory bowel disease.
The study underscores the complexity in interpreting the impact of medicine use on risk of surgery in inflammatory bowel disease.
Inflammatory bowel diseases (IBD) are multifactorial chronic diseases, where severe disease complications often result in surgery. Primary indications for surgical interventions are emergency and failure to respond to medical therapy. However, in the past 30 years, major advances in the treatment of IBD have been made, latest with the introduction of tumour necrosis factor-α (TNF-α) blockers which, in Denmark, have been in use since late 1998. Randomised controlled trials, referral centre studies, short-term follow-up studies of IBD cohorts1–,3 have indicated that introduction of immune modulators early in the disease course may reduce the need for surgery in patients with IBD. A study published in 2008 found that treatment with azathioprine in combination with TNF-α blockers early in the course of Crohn's disease (CD) was superior to current strategies with corticosteroids in inducing mucosal healing and avoiding surgery during the following 2 years.1 This is in accordance with studies from different centres suggesting that early introduction of especially thiopurines reduces the risk of surgery, particularly in CD.2 ,4–,6 Likewise, two former regional Danish population-based studies have reported a decreased need of surgery within the first year after diagnosis of IBD among patients diagnosed in the era of immune-suppressive treatment when compared to older Danish cohorts.7 ,8 However, due to short follow-up time in the mentioned studies, it remains uncertain whether the potential decrease in risk of surgery is persisting over time, or whether the need for surgery is just postponed in patients exposed to immunosuppressive and biologics.
The aim of the present study was to conduct a nationwide population-based cohort study of changes in exposure to classical and newer IBD medications as well as changes in short-term and long-term risk of major and minor surgery among Danish patients diagnosed with IBD during the last three decades. Also, we evaluated the association between exposure to 5-aminosalicylates (5-ASA), corticosteroids and thiopurines and risk of first major intestinal surgery in IBD.
We included patients 16 years or older at their first diagnosis of IBD in the Danish National Patient Register (NPR), which contains individual-level healthcare information on inpatient hospital contacts, diagnoses, surgical and other procedures performed in Danish hospitals since 1977, and in ambulatory outpatient settings since 1995.9 Linkage between registers was possible by use of the Danish Civil Registration System10 where all Danes are registered under a unique 10-digit identification number.
The cohort was defined by patients with a first diagnosis of IBD after 1978 (thereby excluding prevalent cases) using the international classification of diseases (ICD) 8th and 10th revision codes for CD (563.00–563.09 and K50) and ulcerative colitis (UC) (563.19, 569.04 and K51). The diagnoses of IBD in NPR are found to be correct and almost complete with validity estimates for registered CD and UC of, respectively, 97% and 90%, using a pathology register as reference.11 To avoid diagnostic ambiguity, patients with a diagnosis of UC and CD (n=5861) recorded in NPR were not included in this study. Patients who underwent major intestinal surgery before (median time 3.3 years) diagnosis of IBD (n=1393) were also excluded (with a similar proportion of patients excluded in each subcohort). The final study cohort consisted of 35 782 incident cases with UC and 13 185 with CD diagnosed during 1979–2011. We divided the study cohort, according to era of IBD diagnosis, into four groups; Cohort I (diagnosed 1979–1986), Cohort II (1987–1994), Cohort III (1995–2002), and Cohort IV (2003–2011) (table 1) for UC and CD, respectively.
Detailed individual-level information on medical treatment for IBD was extracted from the Danish Prescription Database (DPD),12 which contains information on all prescriptions redeemed from Danish pharmacies since 1995, including the personal identification number, dispensing date, anatomic therapeutic chemical code, number of packages, and number of defined daily doses. As IBD-relevant medications (see online supplementary table S1), we included azathioprine (the primary thiopurine used in Denmark), 5-ASA/sulfasalazine, oral- and topical corticosteroids. Treatment with TNF-α blockers was identified by combining data from the DPD; NPR (administration in inpatient and outpatient settings) and the Danish Crohn Colitis Database (DCCD).13
We defined surgical procedures for IBD as first major or first minor surgery. Major surgery included colectomies, resections and other unspecified major intestinal operations, according to the Danish Classification of Surgical Procedures and Therapies (including 1995) and, thereafter, codes from the Danish version of the Nordic Medico-Statistical Committee classification (see online supplementary table S2). Minor surgery was defined as intra-abdominal abscess-fistulas drainage and surgery for perianal complications (see online supplementary table S2).
Vital status and covariates
Information on age, gender and vital status was obtained from the Danish Civil Registration system10 which contains continuously updated information on birth, sex, place of birth, address, marital status, as well as dates of immigration, emigration, and death of citizens.
Information on comorbidities during follow-up was obtained from NPR. Comorbidities were categorised as cardiovascular disease, cerebral infarction, renal disease, chronic lung disease, liver disease, rheumatic disease, cancer, anaemia, coagulation disorders, immune disorders, human immune deficit virus infection, tuberculosis and senile dementia (see online supplementary table S3). Comorbidities were selected a priori and included in the analysis as potential confounders, and treated as 13 time-dependent binary variables.
In all analyses, including Kaplan–Meier curves, patients were followed from diagnosis of IBD until the outcome of interest, death, emigration, or end of follow-up, 31 December 2011, whichever occurred first.
Kaplan–Meier plots were used to estimate the cumulative probability of ever filling a prescription of 5-ASA, oral corticosteroid, azathioprine or TNF-α blockers or of ever undergoing first major or minor surgery according to time since diagnosis of IBD. Kaplan–Meier curves were compared by log rank test, comparing rates within the first year, 1–4 years, and 5–8 years after diagnosis. Estimating the cumulative probability without assuming independent competing risks14 gave only slightly higher estimates. To assess any association of use of medication (5-ASA, corticosteroids, azathioprine) on risk of surgery, we used Cox proportional hazards regression to estimate HRs with 95% CIs using days since IBD diagnosis as the underlying time scale. In the first type of comparison we compared the two exposure groups, current users with never users by time since diagnosis. In the second type of comparison including current users only, we compared long-term current users (3–11 months, 12+ months, 3+ months) with short-term current users (<3 months) by time since diagnosis (<1 year, ≥1 year). The HRs for medication were adjusted for gender, age, calendar period (divided into 2-year intervals), diagnosing hospital grouped in six groups according to region of Denmark (North, Middle and South Jutland, Funen, Zealand, and the capital region including Bornholm), IBD-related drugs (5-ASA, corticosteroids (oral or topical), azathioprine and anti-TNF-α blockers) and comorbidities (13 different main groups, see above). All variables were chosen a priori and were time dependent except gender and diagnosing hospital.
Medication use was treated as time-dependent variables categorised as ‘current’ users, ‘past’ users and ‘never’ users in the following manner: a patient was a current user from the date of first redeemed prescription, a gap between consecutive prescriptions of up to 100% of the duration of the preceding prescription (in defined daily doses) was accepted to leave room for variations in drug intake (using 50% or 150% gave similar results, data not shown). If two consecutive prescriptions overlapped, the overlap was disregarded and exposure time was counted from the dispensing day of the most recent prescription. Users of the drug in question who subsequently stopped treatment were recategorised to a distinct group of past user-patients and contributed person-time to this group from the day the maximum time gap after a prescription was exceeded. However, if a past user later refilled a prescription, they would contribute person-time to the current user group. Only cumulative probability for current users and never users was visualised in Kaplan Meier curves.
In an alternative analytic approach, we used stratified Cox regression with propensity score matched (1:1 with replacement) pairs as strata. We compared ever users and never users with follow-up from 1 year after diagnosis, and with medication status defined at start of follow-up. Propensity scores expressed the likelihood of being treated with a given IBD-specific medication within the first year after diagnosis. The propensity was estimated using logistic regression including the same adjustment variables described above plus use of other IBD-related medications than the drug in question within the first year after diagnosis.
Data were analysed using SAS V.9.3 (SAS Institute, Cary, North Carolina, USA).
The cohort consisted of 48 967 patients with IBD (CD, n=13 185; UC, n=35 782) diagnosed from 1979 to 2011. Cohort members were followed for up to 32 years or 134 167 person-years in CD and 364 306 person-years in UC. Patients were divided into four subcohorts according to calendar period of diagnosis as outlined in table 1.
Data on medication use was available for Cohort III (1995–2002) and Cohort IV (2003–2011).
Patients diagnosed with CD in recent years (Cohort IV) used 5-ASA to a lesser extent than patients in Cohort III, within the first year (p<0.001) and 1–4 years (p<0.001) after diagnosis (overall p<0.001). The visual difference between the two cohorts 5–8 years after diagnosis (figure 1A) was explained by the difference in the first years (table 2). For oral corticosteroids, the overall p value was 0.029, but no difference in use between Cohorts III and IV was observed within the first year after diagnosis (p=0.29), whereas a significantly lower use of oral corticosteroids in Cohort IV as compared with Cohort III was seen 1–4 year after diagnosis (p=0.006) (figure 1B, table 2). For azathioprine, patients in Cohort IV had a significantly higher use within the first year after diagnosis as compared with Cohort III (p<0.001), whereas the long-term differences seen in Kaplan–Meier plots (figure 1C) were entirely explained by the difference in the first years (1–4 years (p=.18) and 5–8 years (p=0.70))(overall p<0.001). Use of TNF-α blockers in Cohort III was only estimated for the proportion of patients diagnosed after introduction of the drug in Denmark in late 1998. Patients diagnosed in Cohort IV had a significantly higher use of TNF-α blockers within the first year (p<0.001) and 1–4 years (p<0.001) after diagnosis than patients in Cohort III, whereas the cumulative probability of use 5–8 years after diagnosis tended to equalise (p=0.35) (figure 1D, table 2) (overall p<0.001).
In patients with UC, the cumulative probability of using 5-ASA within the first (p<0.001) and 1–4 years (p<0.001) after diagnosis decreased significantly from Cohort III to Cohort IV; whereas the observed differences between curves 5–8 years (p=0.21) after diagnosis was explained by the difference in the first year (figure 1E, table 3) (overall p<0.001). Also, a significant decrease in use of topical corticosteroids was observed from Cohort III to Cohort IV within the first (p<0.001) and 1–4 years (p<0.001) after diagnosis (table 3, (not visualised in a figure)) (overall p<0.001). Contradictory to this, no difference in use of oral corticosteroids was observed from Cohort III to Cohort IV within the first year (p=0.97), 1–4 years (p=0.2) or 5–8 years (p=0.2) after diagnosis (figure 1F) (overall p=0.32). Patients in Cohort IV had a higher probability than patients in Cohort III of being exposed to azathioprine within the first year (p<0.001) and 1–4 years (p<0.001) after diagnosis, whereas the long-term difference observed (figure 1G) 5–8 years after diagnosis was explained by the initial differences in use (table 3) (overall p<0.001). Restricting Cohort III to include patients diagnosed after the introduction of TNF-α antagonist only, we observed a significantly higher probability of receiving TNF-α blockers in Cohort IV than in Cohort III within the first (p<0.001), 1–4 years (p<0.001), and 5–8 years (p<0.001) after diagnosis (overall p<0.001) (table 3, figure 1H).
Risk of surgery
A total of 4146 CD patients underwent a first major surgery during follow-up, 863 in Cohort I (1977–1986), 1148 in Cohort II (1987–1994), 1166 in Cohort III (1995–2002) and 969 in Cohort IV (2003–2011). Risk of first major surgery decreased over calendar time (comparing all four cohorts; figure 2). This was observed within the first year (p<0.001) and also 5 years (p<0.001) and 9 years (p<0.001) after diagnosis. Likewise, the cumulative probability of undergoing first major surgery was lowest in the most recent Cohorts (Cohorts III and IV) within the first year, 5 years and 9 years after diagnosis (overall p=0.004, figure 3). Restricting analysis either to intra-abdominal fistulas or perianal complications, the same significant decrease in surgery over time was observed in both groups (overall p values, p<0.001 and p<0.001). Restricting analyses to inpatients at diagnosis only, revealed nearly identical results (data not shown).
A total of 4037 UC patients underwent a first major surgery during follow-up, 943 in Cohort I (1979–1986), 1004 in Cohort II (1987–1994), 1128 in Cohort III (1995–2002) and 962 in Cohort IV (2003–2011). As for CD, risk of first major surgery decreased significantly over calendar time (comparing all four cohorts; figure 4). The decrease was found in 1, 5 and 9 years after diagnosis (all p<0.001). Restricting analysis to inpatients at diagnosis only, revealed similar results (data not shown).
Use of medications and risk of surgery
In the following, we first present the impact of current versus never use of azathioprine and oral corticosteroids on risk of surgery. Second, we present the differences in short-term and long-term use of 5-ASA, topical and oral corticosteroids, and azathioprine on risk of surgery.
Figure 5A shows the cumulative probability of first major intestinal surgery in CD according to time since diagnosis of CD (Cohorts III and IV combined) in current users and never users of azathioprine. Dividing the first year into intervals of 3 months, only in the first 3 months after diagnosis of CD a lower surgery rate was observed in current users compared to never users of azathioprine (HR, 0.50; 95% CI 0.31 to 0.82), whereas an increased surgery rate was observed in current users at all other time intervals (eg, HR 6–8 months after dx=1.95;95% CI 1.03 to 2.96; HR 12–14 months after dx=1.85;95% CI 0.97 to 3.52). Risk of surgery in current users of oral corticosteroids was increased at all times compared with non-users, figure 6A. Propensity-score matching did not change results (see online supplementary table S4).
As the higher surgery rate found in users could be due to confounding by indication, that is, the most ill patients were the ones receiving medication, we also examined rate of surgery among current drug users according to duration of use (table 4). In these analyses we looked separately at patients initiating treatment with a given drug <1 year, respectively, 1+ year after diagnosis. In patients initiating treatment with 5-ASA, topical or oral corticosteroids, or azathioprine within the first year after diagnosis, those with a current use for 3–11 months were at significantly higher risk of surgery than those with a current use of <3 months (table 4). In patients initiating treatment with 5-ASA, topical corticosteroids or oral corticosteroids, or azathioprine more than 1 year after diagnosis, only use of oral corticosteroids for 3+ months versus less than 3 months (HR, 0.42; 95% CI 0.26 to 0.69) significantly reduced the risk of surgery (table 4).
Figure 5B shows the cumulative probability of major intestinal surgery in UC according to time since diagnosis of UC (Cohort III and IV combined) in current users and never users of azathioprine. The risk of surgery, at any time, was higher in current users of azathioprine, compared to the risk in non-users (p<0.001). Likewise, current use (vs never use) of oral corticosteroids was associated with increased risk of surgery (p<0.001) (figure 6B). Again, propensity-score matching did not change results (see online supplementary table S4).
Next we analysed the risk of surgery according to duration of current drug use and disease duration. In patients initiating treatment within a year from diagnosis, a significantly increased risk of surgery was observed in those with a current use of 5-ASA or oral corticosteroids for 3–11 months (vs <3 months), and a similar tendency was observed for azathioprine (table 5). In patients with UC, initiating treatment with a given drug more than a year from diagnosis, a significantly reduced risk of surgery was observed only for those with a long-term use of oral corticosteroids (table 5).
In this nationwide population-based cohort study of 48 967 patients diagnosed with IBD over a 32-year period, we found a decrease in risk of first major intestinal surgery over calendar time, most noticeable in CD. The decrease continued long-term, suggesting that surgery is postponed not only in patients diagnosed in recent years. A significant decrease in risk of minor surgery in patients with CD was also observed in recent cohorts. Comparing the two latest cohorts, we further found a decrease in use of 5-ASA and topical corticosteroids (and oral corticosteroid in CD) paralleled by an increasing use of azathioprine and TNF-α blockers. Examining the risk of surgery according to current use of medications (compared with never users), we found, that only patients with CD using azathioprine within the first 3 months after diagnosis had a reduced risk of surgery. Considering risk of surgery among current medication users only, patients initiating treatment of 5-ASA, topical corticosteroids, oral corticosteroids, or azathioprine within the first year of diagnosis and receiving the treatment for 3–11 months had an increased risk of surgery compared to those with a current use of <3 months. In patients initiating treatment with 5-ASA, topical or oral corticosteroids, or azathioprine more than 1 year after diagnosis, only long-term use of oral corticosteroids significantly reduced the risk of surgery.
The major strength of this study was the assessment of a large national and, hence, unselected cohort of patients with IBD diagnosed over three decades. IBD diagnoses are found to be of high completeness and validity in the Danish National Patients Register.11 As access to medical services (hospitalisation, control visits, and treatment) in Denmark is independent of race, socioeconomic status and participation in health assurance programmes due to free and easy access to healthcare, the study population should represent the complete and broad spectrum of IBD. The availability of nationwide individual-level registration of all redeemed prescriptions from all pharmacies in Denmark from 1995 and onwards further strengthens this study. A potential limitation could be that redeemed prescription may not fully reflect actual drug intake, but these data are still assumed to be superior to self-reported drug intake. Another possible limitation is the use of defined daily doses and acceptance of gaps between prescriptions when estimating status of current or past use. This limitation however is judged to be minor as sensitivity analyses testing various lengths of gaps did not change results. Information on disease behaviour, disease extent at diagnosis or smoking status was not available, which limited the possibility to control for disease severity. However, from regional Danish cohort studies we know that changes in phenotypes have been minor over time.15 Furthermore, using propensity score matching as an alternative analytic approach did not change results. However, although intended to capture disease severity, propensity scores did not include direct variables of severity and behaviour, but only surrogate measures of severity. We found an increase in use of azathioprine and TNF-α blockers over time, especially in patients with CD. This is partly in accordance with observations made by Herrinton and colleagues on the potentially more selected Kaiser Permanente insurance database of 8787 patients with IBD, although TNF-α blockers had not been introduced for UC at the time of that study.16 Likewise, Kaplan and colleagues found an increase in use of immune modulators and biologicals among 775 patients with UC during 1997–2009.17
Interestingly, in the present study, the increasing use of azathioprine and TNF-α blockers was paralleled by a decrease in use of 5-ASA, topical corticosteroids and oral corticosteroids in CD. Whereas the decrease in use of 5-ASA in CD patients may be explained by changed clinical guidelines no longer supporting its use as maintenance therapy, the decreasing use of 5-ASA in patients with UC and of topical corticosteroids in both patient groups remains unexplained, although paralleled by the increase in use of immune-modulators and biological treatment. The reduced use of oral corticosteroids in patients with CD is an interesting and novel finding as it may point towards a steroid-reducing effect of newer drugs as observed in paediatric patients with CD.18
Along with the significant changes in use of classical and newer treatments for IBD, we found a persisting decrease in need of first major surgery in CD and UC, which was present in the first year after diagnosis, as shown previously,15 and was also found to persist in the long term. In UC, the 9-year risk decreased from 14.5% in Cohort I to 9.1% in Cohort IV. The first estimate is in accordance with regional Danish data from a cohort of patients with UC diagnosed In 1962–1987,19 whereas the current probability of 9.1% is in accordance with more recent results from a European cohort,20 hence suggesting that long-term colectomy rates in UC have decreased significantly over time, and that Danish rates are also approaching those of other European countries.
In CD, surgery rates also decreased significantly, even in the long term, from 50.3% 9 years after diagnosis in Cohort I to 23.3% in Cohort IV. Current surgery rates among Danish patients with CD appear to be slightly lower than those reported from Cardiff, Wales (25% at 5 years),6 and from Eastern Europe (approximately 30% at 7 years after diagnosis).5 The risk of minor surgery due to intra-abdominal fistulas or perianal complications was also significantly decreased in patients with CD, potentially due to increased use of drugs with known efficacy on penetrating disease. The decrease in need of major and minor surgery in IBD over time may also be explained by a milder disease course in patients diagnosed in recent decades,5 or by changes in smoking habits in the general population21 and, hence, potentially also in IBD. However, in Denmark, the phenotype of IBD has not been found to change markedly over time,15 nor have smoking patterns in patients with CD and UC changed significantly.8
Whether the introduction of immunosuppressive and biological drugs has had a direct effect on risk of surgery among patients with IBD remains debateable.22 We looked prospectively at the risk of first major surgery according to current duration of medicine use and time since diagnosis of IBD and observed an increased risk of surgery among users of medications prescribed within the first year of diagnosis. This is interesting, as previous studies of various methodologies2 ,5 ,6 ,23 suggest that initiating treatment early in disease course could reduce surgery rates. Ramadas et al6 found in multivariate analysis that patients with CD using thiopurines within the first year of diagnosis had a reduced risk of surgery, while Lakatos and colleagues5 observed that patients with CD using azathioprine for 6 months within the first 3 years after diagnosis had reduced risk of surgery. In a French referral centre study2 of newly diagnosed patients with CD, duration of treatment was also associated with surgery risk, whereas no such association was observed in patients with UC.24 Provided that the initiation of medical treatment per se causes delay of surgery, it is likely, that the increased risk of surgery we found within the first year after diagnosis in current users to some extent reflects failure to respond to medical treatment, as discussed by Vernier–Massouille.25 This leads to the concern that our results are reflecting confounding by indication, that is, that use of immunotherapy and oral corticosteroids reflect more severe disease. This may, to some extent, be the case despite propensity-score matching. Further, the concern raised is challenged by the reduced risk of surgery observed in patients with longer-term use of oral corticosteroids. Hence, the observed persistent decrease in surgery rates paralleled by an increase in use of immunosuppressive and biological treatment, our results do not point clearly towards a surgery-reducing effect of these treatments.
In this nationwide cohort study on patients with IBD diagnosed from 1979 to 2011 we found an increased use of azathioprine and TNF-α blockers over time parallel to a decrease in use of 5-ASA and local corticosteroids. Also, a parallel decrease in need of major and minor surgery was observed. However, a convincing surgery-sparing effect of use of newer medications was not found.
We thank Doctor Björn Pasternak for his useful comments in handling medical data.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Contributors CR: Conception and design of the study, generation and interpretation of data, drafting the manuscript and approval of the final version of the manuscript. SB: generation, analysis and interpretation of data. Revision and approval of the final version of the manuscript. MA: analysis and interpretation of data. Revision and approval of the final version of the manuscript. NMN. Generation and interpretation of data. Revision and approval of the final version of the manuscript. JW: Design of the study, analysis and interpretation of data, revision and approval of the final version of the manuscript. EL: Conception and design of the study, interpretation of data, revision and approval of the final version of the manuscript. TJ: Conception and design of the study, generation and interpretation of data, drafting and critical revision of the manuscript and approval of the final version of the manuscript.
Funding Dr Tine Jess was financially supported through a Female Research Leader grant (no. 09-066323) from the Danish Council of Independent Research and the study was further supported by the Danish Cancer Society, grant (no. R40-A1737-11-S2).
Competing interests None.
Ethics approval The study was register based and has been approved by Danish Data Protection Agency.
Provenance and peer review Not commissioned; externally peer reviewed.