Diagnosed CD

Although several studies indicate that diagnosed CD is associated with increased mortality, data are contradictory with regards to the extent of such excess mortality. The first half of this article reviews the largest and most recent mortality studies, all of which included at least 1000 patients with CD.

The first such study was an Italian multicenter study of 1072 individuals with CD. In this study, adults with biopsy-proven CD from 11 gastroenterology units were identified ( Table 1 ), and the investigators found a 2-fold increased risk of death in CD (standardized mortality ratio [SMR] 2.0; 95% confidence interval [CI] = 1.5–2.7).

The most important finding of that study was the association between severe clinical presentation and higher mortality (SMR = 2.5; 95% CI = 1.8–3.4). In contrast, no association was found in patients who had only mild CD (SMR = 1.1) or asymptomatic CD (SMR = 1.2). The lack of excess mortality in these 2 groups indicates that patients with CD with mild symptoms have the same mortality risk as the general population.

A second finding of the Italian study was the relation between diagnostic delay and mortality. Patients with more than 10 years of diagnostic delay were at a higher risk of death than those with less than 1 year of diagnostic delay. Although at first this finding seems reasonable, the increased risk of death in those with diagnostic delay is difficult to reconcile with the low SMR in those diagnosed in the oldest age group, in which diagnostic delay ought to have been longest. One interpretation of the lower excess mortality in older patients with CD is that the relative importance of CD as a cause of death diminishes when more people (with or without CD) have other diseases.

In a second study, Peters and colleagues investigated mortality in 10,032 Swedish patients with CD. The strength of this study is its statistical power, which allowed the investigators to calculate narrow CIs and to explore cause-specific death. Consistent with the Italian study, the Swedish study found the lowest SMRs in the oldest age group (SMR = 1.7), whereas adolescents with CD and young adults were at a more than 3-fold increased risk of death. Two factors may have contributed to the high SMR for death in this study. First, the study was based on inpatients with CD, although modern CD rarely requires hospital admission (although many patients with CD in Sweden in the 1960s to 1980s were admitted as part of the small intestinal biopsy procedure). Inpatients may suffer from more advanced CD, with more symptoms, or more comorbid conditions. All these factors could have contributed to the 2-fold increased risk of death seen in the study by Peters and colleagues. When Peters and colleagues restricted their dataset to patients who had only a diagnosis of CD and no other recorded comorbidity, the SMR decreased to 1.4. In all patients in this study, CD was diagnosed before 1994. The introduction of CD serology has since meant that CD is diagnosed in more patients with mild disease. Drawing from the evidence of Corrao and colleagues that patients with asymptomatic CD had a lower mortality risk, it is reasonable to conclude that recently diagnosed patients have lower mortality risk (although contradicting data have also been reported).

In contrast to the studies by Corrao and colleagues and Peters and colleagues, more recent studies have found only modestly increased risks of death in CD. From 4732 patients with CD, West and colleagues reported an overall mortality hazard ratio (HR) of 1.31, a figure that decreased to 1.17 after exclusion of the first year of follow-up. This study was able to adjust for smoking and body mass index (BMI, calculated as weight in kilograms divided by the square of height in meters) in a subset of patients (with no change of risk estimates) and estimate the absolute excess mortality (see Table 1 ). Taking advantage of long-term follow-up of a regional cohort, Grainge and colleagues later examined the mortality in 1092 British patients from southern Derbyshire. That study found a 1.37-fold increased risk of death in CD. The study by Grainge and colleagues focused on mortality change over time and confirmed Swedish data that mortality in CD has not changed substantially since the 1980s.

The largest mortality study to date used histopathology data from Sweden’s 28 pathology departments to identify more than 46,000 individuals with either CD (n = 29,096) or CD-related diseases (see later discussion). The HR for death in biopsy-verified CD was 1.39 (95% CI = 1.33–1.45), decreasing to 1.26 when the first year of follow-up was excluded. After more than 5 years of follow-up, the HR remained slightly more than 1. Several findings in this study are worth emphasizing. First, in contrast to a recent British study, the Swedish study found similar HRs for death in men (1.28) and women (1.24). Second, as in several other studies, excess mortality was lowest among those in whom CD was diagnosed at an older age. If we accept the assumption that CD often begins at a younger age, the lack of excess mortality in those whose CD is diagnosed at an older age (supported by recent US findings ) could mean that long-standing untreated CD may have only a minimal effect on mortality.

Undiagnosed CD

Whereas 4 studies of undiagnosed CD have shown a similar mortality in patients and controls, 2 studies have shown an increased risk of death in patients with undiagnosed CD. There are 2 main reasons for a potentially increased risk of death in patients with undiagnosed CD: (1) CD may confer an inherent excess mortality independently of whether it is being diagnosed or not; and (2) undiagnosed CD, as opposed to diagnosed CD, remains untreated, with ongoing exposure to gluten.

In a German study, 15 of 63 individuals (24%) with positive antibodies to tissue transglutaminase (tTG) died during follow-up. This finding corresponded to an HR of 2.53. Given that the investigators used only 1 serologic marker for their CD diagnosis (positive tTG), this HR is surprisingly high. Without confirming the tTG with either endomysial antibody (EMA) or small intestinal biopsy, a significant proportion of individuals should have false-positive results. The lack of confirmation of CD might then lead to dilution of data and a lower risk estimate for death. One potential reason for the high mortality HR in this study is that some patients with tTG positivity may not have had CD but rather had liver or heart disease, which could also cause tTG positivity and increase mortality.

The only other study showing substantial excess mortality in CD is an American study that used stored blood samples from young men at a US Air Force base (HR = 3.9). Several facts are worth noting in this study by Rubio-Tapia and colleagues. First, the follow-up period was longer than in any of the other studies. The excess mortality became apparent only after 15 to 20 years of follow-up. Potentially, the other studies with shorter follow-up may have failed to recognize an excess mortality that is evident only after a certain number of years with CD. Second, the number of individuals with CD was small (n = 14) and the CI was wide (95% CI = 2.0–7.5). It cannot be ruled out that the high HR was a chance finding and that the true HR for death in this study was just more than 2 rather than about 4 (still within the 95% CI). Third, because of the historical character of the study, the investigators were unable to verify the diagnosis of CD through biopsy, although all tTG-positive individuals required confirmation with EMA. Fourth, among those who had positive CD serology, there may have been individuals with other (diagnosed or undiagnosed) diseases, such as certain liver and heart disease, which per se are associated with an increased risk of death. Because health care has vastly improved in the past 50 years, the HR for death of 3.9 in young men with undiagnosed CD in the 1940s and 1950s does not automatically mean that undiagnosed (and untreated) CD today is associated with a 4-fold increased risk of death. Nevertheless, the study by Rubio-Tapia and colleagues constitutes one of the strongest arguments so far that undiagnosed CD may have severe adverse health effects.

The study by Godfrey and colleagues found no increased risk of death in older patients (aged ≥50 years) with undiagnosed CD (HR = 0.80; 95% CI = 0.45–1.41). Of some 16,000 patients tested for CD, 0.8% had positive or borderline tTG and positive EMA. However, this study had a significantly shorter follow-up than that of Rubio-Tapia and colleagues. In addition, it is possible that the influence of CD (diagnosed or undiagnosed) diminishes at older ages, when the reference individuals acquire other diseases. In accordance with the lack of excess mortality, the investigators also found that seropositive and seronegative individuals had a similar BMI, and with the exception of slightly lower cholesterol and ferritin levels, they also had similar biochemical data (hemoglobin, vitamin B ₁₂ , folic acid, and albumin levels), suggesting that they did not have more than marginal malnutrition. The fact that CD was clinically detected in 15% of those with undiagnosed CD during follow-up (and presumably treated with a GFD) may also have affected the HR, but cannot explain the lack of excess mortality, because studies of diagnosed CD show a positive association with death.

Finnish and British data indicate that patients with undiagnosed CD in these countries are not at increased risk of mortality. Both studies took advantage of centralized systems to identify recorded death dates, but the studies also adjusted for smoking. The Finnish study used data from an earlier screening study (Mini-Finland) and identified 74 EMA-positive individuals and 204 Celikey tTG-positive individuals (Pharmacia Diagnostics, Uppsala) ; in addition, these individuals were all positive for a second type of tTG test. The adjusted HR for death was 0.78 in EMA-positive individuals and 1.19 in Celikey tTG-positive individuals. None of these risk estimates was statistically significant.

A British study examined 7527 individuals with blood samples from 2001. Of these individuals, 87 were regarded as positive for CD (EMA-positive, but not confirmed through biopsy). This study points to the necessity to adjust for risk factors when estimating risk of death. In the original unadjusted analyses, the HR for death was 0.73, but after adjustment for age, sex, socioeconomic group, and smoking status, this figure rose to 0.98 (95% CI = 0.57–1.69). This finding is likely because of the low prevalence of smoking among British patients with CD ; however, this characteristic has not been confirmed in Swedish or American data.

Cause-Specific Mortality

The excess risk of death in patients with CD was long believed to be caused by malignancies. This belief is only partly true. The commonest cause of death in patients with CD in our population-based Swedish cohort was cardiovascular death (1007 deaths during follow-up, compared with 773 deaths from malignancy). That the relative risk of death from malignancy (HR = 1.55) is higher than that of cardiovascular disease (HR = 1.19) does not mean that malignancy causes more deaths than cardiovascular disease in patients with CD. Recent data indicate that patients with CD are at increased risk of ischemic heart disease, stroke, atrial fibrillation, and potentially also autoimmune heart disease. In the recent study by Canavan and colleagues of undiagnosed CD (EMA-positivity), the adjusted risk of cardiovascular death (HR = 1.39, although nonsignificant: 95% CI = 0.66–2.92) was higher than that of death from malignancy (HR = 1.27; 95% CI = 0.57–2.85).

Several studies have reported an increased risk of respiratory death in CD (with Lohi and colleagues showing an increased risk of respiratory death in Celikey tTG-positive patients, whereas the risk was not statistically significantly increased in EMA-positive patients). This risk increase could be of multiple origins. Previous research has shown that patients with CD are at increased risk of pneumococcal infection, which is expected given their predisposition to hyposplenism. Other potential causes of respiratory death in patients with CD include pulmonary embolism, tuberculosis, and influenza.

Although suicide is a rare condition, it constitutes an important cause of death in younger individuals in Western countries. It has also been suggested that some excess mortality noted in type 1 diabetes mellitus, another autoimmune disorder, is caused by suicide and external causes of death. Psychiatric disease is more common in CD (although probably not schizophrenia), and psychiatric disease is a risk factor for suicide in patients with CD. One smaller British study has also reported an increased risk of suicide in children with CD. We therefore explored the risk of suicide in our cohort of biopsy-verified patients with CD and found a small but statistically significantly increased risk of suicide (HR = 1.55; 95% CI = 1.15–2.10; based on 54 completed suicides).

Mortality in Children with CD

Few studies have investigated mortality in children with CD. Solaymani-Dodaran and colleagues stratified their data set according to age at diagnosis and found a 2.6-fold increased risk of death in patients diagnosed with CD in childhood. What was especially concerning with this study was that deaths caused by accidents, suicide, and violence were increased (based on 7 deaths) and that excess mortality increased with longer follow-up (SMR was 3.32 after more than 5 years after diagnosis).

Swedish data have since confirmed that childhood CD is associated with excess mortality (excluding the first year of follow-up: HR = 1.78) and remains increased after more than 5 years of follow-up (HR = 1.90). Because of the concerns raised by Solaymani-Dodaran and colleagues about deaths from external causes, we specifically looked at suicide rates in children and adolescents with CD in another population-based study. In this study, we found no particular increase in suicide rates in patients diagnosed with CD before the age of 20 years (HR = 1.42). However because HRs were similar in all age strata in our cohorts, we concluded that an early diagnosis of CD is unlikely to eliminate the increased suicide risk seen in CD.

Individuals Without VA

In our paper on biopsy-verified CD and future mortality, we also investigated mortality in patients undergoing small intestinal biopsy but who did not have CD. A notable finding was that the mortality excess risk was higher in patients with inflammation (Marsh stage 1–2, intraepithelial lymphocytosis: HR = 1.72) than in those with VA (Marsh stage 3, CD: HR = 1.39). There are several explanations for this finding. One is that patients without VA have traditionally not adopted a GFD. Hence, their inflammation is likely to remain active, whereas patients with VA (CD) would have received a GFD with a high chance of long-term mucosal healing (assuming that the GFD is protective against mortality). Another possible explanation is that not all duodenal inflammation is caused by CD. Whereas the heredity for CD was 12% in patients with VA in our study, it was lower in those with inflammation (5%), perhaps reflecting a more heterogeneous origin of this condition. Mahadeva and colleagues have reported that 40% of individuals with intraepithelial lymphocytosis develop CD (here defined as VA), but that most individuals do not. Still, when 2 researchers (independently) manually examined more than 1500 biopsy reports with VA or inflammation, few patients with inflammation had a non-CD diagnosis recorded in the biopsy register (inflammatory bowel disease was the most common other diagnosis and mentioned in 1.6% of the biopsy reports with Marsh 1–2). It is therefore difficult to draw any conclusions regarding mortality in CD from our findings of an increased risk of death in patients with Marsh stage 1 to 2.

Underlying Mechanism for the Excess Mortality in CD

The dominant causes of death in patients with CD seem to be cardiovascular death and death from malignancy. Because patients with CD often lack traditional cardiovascular risk factors (they have less hypertension or no difference in blood pressure, more beneficial lipid profile, lower BMI, [and most, but not all, studies suggest an inverse relation with smoking]), an explanation for the increased risk of cardiovascular death must be sought elsewhere. Some studies have found no or even a negative association between CD and cardiovascular disease, whereas the largest study so far (991 patients with CD developed ischemic heart disease during follow-up) found that patients with CD have about a 20% increased risk of cardiovascular disease.

The most plausible link between CD and cardiovascular disease (as well as with death from malignancy) is the increase in inflammation, both before and after CD diagnosis. Although Lewis and colleagues did not compare C-reactive protein (CRP) values in patients with CD and controls, the average CRP level at diagnosis (based on 94 patients with CD) suggests that patients with CD have a normal CRP. However, other research has shown that patients with CD may have increased levels of other inflammatory markers (such as interleukin 4, 6, and 10, and tumor necrosis factor α) that could play a detrimental role for the risk of cardiovascular disease and malignancy.

It is also probable that CD protects against certain types of disease because patients often have a slightly lower BMI and smoke less often. Among diseases that may be less common in CD is breast cancer, which is a common cause of death in women.

GFD and Mortality

A GFD is the only treatment of CD. Numerous reports confirm that treatment of CD results in better growth and overall health. The effect of GFD on mortality is less certain, and no prospective study on mortality has compared patients who adhere to a GFD with those who are nonadherent. Data on the role of GFD for mortality risk are therefore limited to retrospective data (often medical record reviews), which carry a strong risk of bias. In the study by Corrao and colleagues, patients likely to have been on a GFD were at a nonsignificantly lower risk of death compared with the general population (SMR = 0.5; 95% CI = 0.2–1.1). Those with uncertain dietary adherence had a 2-fold increased risk of dying, whereas those who were unlikely to have followed the diet were at a 6-fold increased risk of dying (95% CI = 4.0–8.8). However, a problem with these calculations is that most of the data in patients who died during the follow-up were obtained from relatives or friends, whereas individuals who were alive at the end of the follow-up were asked to grade their dietary adherence themselves. Thus, it cannot be ruled out that death per se influenced the relatives’ responses and that patients who were alive overestimated their self-reported dietary adherence.

It could be argued that the higher mortality in patients diagnosed with CD in the 1970s and 1980s when strict adherence to GFD may not have been as widespread as it is currently is proof that a GFD is important to minimize excess mortality. Meanwhile, the lack of a major difference in mortality between individuals with VA (HR = 1.39) and those with normal mucosa but positive CD serology (HR = 1.35) suggests that a GFD may not have such a strong effect on mortality. The main effect of a GFD is to heal the mucosa, but even individuals with normal mucosa seem to be at increased risk of death in this study. However, a GFD may have other beneficial effects on the health of patients with CD. Some investigators have directly or indirectly contended that a delay in the institution of a GFD increases the risk of other autoimmune comorbidities (with detrimental effects on mortality) or affects the mortality risk per se. One of our studies showed increased risk of death in children with biopsy-verified CD, suggesting that short diagnostic delay in diagnosis may not prevent excess mortality.

Mortality: Implications for the Nondietary Management of CD

Given that cardiovascular disease is a common cause of death in patients with CD, general preventive measures are also relevant in patients with CD. Such measures include cessation of smoking, treatment of hypertension and hyperlipidemia, encouraging physical activity, and avoiding overweight. Although many of these risk factors are inversely related to CD, they are still highly prevalent in adults with CD.

A second theme of prevention deals with infectious and respiratory disease. Given the increased risk of both pneumococcal infections and influenza, immunization against these infections may be considered in patients with CD.

Diagnosed CD

Although several studies indicate that diagnosed CD is associated with increased mortality, data are contradictory with regards to the extent of such excess mortality. The first half of this article reviews the largest and most recent mortality studies, all of which included at least 1000 patients with CD.

The first such study was an Italian multicenter study of 1072 individuals with CD. In this study, adults with biopsy-proven CD from 11 gastroenterology units were identified ( Table 1 ), and the investigators found a 2-fold increased risk of death in CD (standardized mortality ratio [SMR] 2.0; 95% confidence interval [CI] = 1.5–2.7).

The most important finding of that study was the association between severe clinical presentation and higher mortality (SMR = 2.5; 95% CI = 1.8–3.4). In contrast, no association was found in patients who had only mild CD (SMR = 1.1) or asymptomatic CD (SMR = 1.2). The lack of excess mortality in these 2 groups indicates that patients with CD with mild symptoms have the same mortality risk as the general population.

A second finding of the Italian study was the relation between diagnostic delay and mortality. Patients with more than 10 years of diagnostic delay were at a higher risk of death than those with less than 1 year of diagnostic delay. Although at first this finding seems reasonable, the increased risk of death in those with diagnostic delay is difficult to reconcile with the low SMR in those diagnosed in the oldest age group, in which diagnostic delay ought to have been longest. One interpretation of the lower excess mortality in older patients with CD is that the relative importance of CD as a cause of death diminishes when more people (with or without CD) have other diseases.

In a second study, Peters and colleagues investigated mortality in 10,032 Swedish patients with CD. The strength of this study is its statistical power, which allowed the investigators to calculate narrow CIs and to explore cause-specific death. Consistent with the Italian study, the Swedish study found the lowest SMRs in the oldest age group (SMR = 1.7), whereas adolescents with CD and young adults were at a more than 3-fold increased risk of death. Two factors may have contributed to the high SMR for death in this study. First, the study was based on inpatients with CD, although modern CD rarely requires hospital admission (although many patients with CD in Sweden in the 1960s to 1980s were admitted as part of the small intestinal biopsy procedure). Inpatients may suffer from more advanced CD, with more symptoms, or more comorbid conditions. All these factors could have contributed to the 2-fold increased risk of death seen in the study by Peters and colleagues. When Peters and colleagues restricted their dataset to patients who had only a diagnosis of CD and no other recorded comorbidity, the SMR decreased to 1.4. In all patients in this study, CD was diagnosed before 1994. The introduction of CD serology has since meant that CD is diagnosed in more patients with mild disease. Drawing from the evidence of Corrao and colleagues that patients with asymptomatic CD had a lower mortality risk, it is reasonable to conclude that recently diagnosed patients have lower mortality risk (although contradicting data have also been reported).

In contrast to the studies by Corrao and colleagues and Peters and colleagues, more recent studies have found only modestly increased risks of death in CD. From 4732 patients with CD, West and colleagues reported an overall mortality hazard ratio (HR) of 1.31, a figure that decreased to 1.17 after exclusion of the first year of follow-up. This study was able to adjust for smoking and body mass index (BMI, calculated as weight in kilograms divided by the square of height in meters) in a subset of patients (with no change of risk estimates) and estimate the absolute excess mortality (see Table 1 ). Taking advantage of long-term follow-up of a regional cohort, Grainge and colleagues later examined the mortality in 1092 British patients from southern Derbyshire. That study found a 1.37-fold increased risk of death in CD. The study by Grainge and colleagues focused on mortality change over time and confirmed Swedish data that mortality in CD has not changed substantially since the 1980s.

The largest mortality study to date used histopathology data from Sweden’s 28 pathology departments to identify more than 46,000 individuals with either CD (n = 29,096) or CD-related diseases (see later discussion). The HR for death in biopsy-verified CD was 1.39 (95% CI = 1.33–1.45), decreasing to 1.26 when the first year of follow-up was excluded. After more than 5 years of follow-up, the HR remained slightly more than 1. Several findings in this study are worth emphasizing. First, in contrast to a recent British study, the Swedish study found similar HRs for death in men (1.28) and women (1.24). Second, as in several other studies, excess mortality was lowest among those in whom CD was diagnosed at an older age. If we accept the assumption that CD often begins at a younger age, the lack of excess mortality in those whose CD is diagnosed at an older age (supported by recent US findings ) could mean that long-standing untreated CD may have only a minimal effect on mortality.

Undiagnosed CD

Whereas 4 studies of undiagnosed CD have shown a similar mortality in patients and controls, 2 studies have shown an increased risk of death in patients with undiagnosed CD. There are 2 main reasons for a potentially increased risk of death in patients with undiagnosed CD: (1) CD may confer an inherent excess mortality independently of whether it is being diagnosed or not; and (2) undiagnosed CD, as opposed to diagnosed CD, remains untreated, with ongoing exposure to gluten.

In a German study, 15 of 63 individuals (24%) with positive antibodies to tissue transglutaminase (tTG) died during follow-up. This finding corresponded to an HR of 2.53. Given that the investigators used only 1 serologic marker for their CD diagnosis (positive tTG), this HR is surprisingly high. Without confirming the tTG with either endomysial antibody (EMA) or small intestinal biopsy, a significant proportion of individuals should have false-positive results. The lack of confirmation of CD might then lead to dilution of data and a lower risk estimate for death. One potential reason for the high mortality HR in this study is that some patients with tTG positivity may not have had CD but rather had liver or heart disease, which could also cause tTG positivity and increase mortality.

The only other study showing substantial excess mortality in CD is an American study that used stored blood samples from young men at a US Air Force base (HR = 3.9). Several facts are worth noting in this study by Rubio-Tapia and colleagues. First, the follow-up period was longer than in any of the other studies. The excess mortality became apparent only after 15 to 20 years of follow-up. Potentially, the other studies with shorter follow-up may have failed to recognize an excess mortality that is evident only after a certain number of years with CD. Second, the number of individuals with CD was small (n = 14) and the CI was wide (95% CI = 2.0–7.5). It cannot be ruled out that the high HR was a chance finding and that the true HR for death in this study was just more than 2 rather than about 4 (still within the 95% CI). Third, because of the historical character of the study, the investigators were unable to verify the diagnosis of CD through biopsy, although all tTG-positive individuals required confirmation with EMA. Fourth, among those who had positive CD serology, there may have been individuals with other (diagnosed or undiagnosed) diseases, such as certain liver and heart disease, which per se are associated with an increased risk of death. Because health care has vastly improved in the past 50 years, the HR for death of 3.9 in young men with undiagnosed CD in the 1940s and 1950s does not automatically mean that undiagnosed (and untreated) CD today is associated with a 4-fold increased risk of death. Nevertheless, the study by Rubio-Tapia and colleagues constitutes one of the strongest arguments so far that undiagnosed CD may have severe adverse health effects.

The study by Godfrey and colleagues found no increased risk of death in older patients (aged ≥50 years) with undiagnosed CD (HR = 0.80; 95% CI = 0.45–1.41). Of some 16,000 patients tested for CD, 0.8% had positive or borderline tTG and positive EMA. However, this study had a significantly shorter follow-up than that of Rubio-Tapia and colleagues. In addition, it is possible that the influence of CD (diagnosed or undiagnosed) diminishes at older ages, when the reference individuals acquire other diseases. In accordance with the lack of excess mortality, the investigators also found that seropositive and seronegative individuals had a similar BMI, and with the exception of slightly lower cholesterol and ferritin levels, they also had similar biochemical data (hemoglobin, vitamin B ₁₂ , folic acid, and albumin levels), suggesting that they did not have more than marginal malnutrition. The fact that CD was clinically detected in 15% of those with undiagnosed CD during follow-up (and presumably treated with a GFD) may also have affected the HR, but cannot explain the lack of excess mortality, because studies of diagnosed CD show a positive association with death.

Finnish and British data indicate that patients with undiagnosed CD in these countries are not at increased risk of mortality. Both studies took advantage of centralized systems to identify recorded death dates, but the studies also adjusted for smoking. The Finnish study used data from an earlier screening study (Mini-Finland) and identified 74 EMA-positive individuals and 204 Celikey tTG-positive individuals (Pharmacia Diagnostics, Uppsala) ; in addition, these individuals were all positive for a second type of tTG test. The adjusted HR for death was 0.78 in EMA-positive individuals and 1.19 in Celikey tTG-positive individuals. None of these risk estimates was statistically significant.

A British study examined 7527 individuals with blood samples from 2001. Of these individuals, 87 were regarded as positive for CD (EMA-positive, but not confirmed through biopsy). This study points to the necessity to adjust for risk factors when estimating risk of death. In the original unadjusted analyses, the HR for death was 0.73, but after adjustment for age, sex, socioeconomic group, and smoking status, this figure rose to 0.98 (95% CI = 0.57–1.69). This finding is likely because of the low prevalence of smoking among British patients with CD ; however, this characteristic has not been confirmed in Swedish or American data.

Cause-Specific Mortality

The excess risk of death in patients with CD was long believed to be caused by malignancies. This belief is only partly true. The commonest cause of death in patients with CD in our population-based Swedish cohort was cardiovascular death (1007 deaths during follow-up, compared with 773 deaths from malignancy). That the relative risk of death from malignancy (HR = 1.55) is higher than that of cardiovascular disease (HR = 1.19) does not mean that malignancy causes more deaths than cardiovascular disease in patients with CD. Recent data indicate that patients with CD are at increased risk of ischemic heart disease, stroke, atrial fibrillation, and potentially also autoimmune heart disease. In the recent study by Canavan and colleagues of undiagnosed CD (EMA-positivity), the adjusted risk of cardiovascular death (HR = 1.39, although nonsignificant: 95% CI = 0.66–2.92) was higher than that of death from malignancy (HR = 1.27; 95% CI = 0.57–2.85).

Several studies have reported an increased risk of respiratory death in CD (with Lohi and colleagues showing an increased risk of respiratory death in Celikey tTG-positive patients, whereas the risk was not statistically significantly increased in EMA-positive patients). This risk increase could be of multiple origins. Previous research has shown that patients with CD are at increased risk of pneumococcal infection, which is expected given their predisposition to hyposplenism. Other potential causes of respiratory death in patients with CD include pulmonary embolism, tuberculosis, and influenza.

Although suicide is a rare condition, it constitutes an important cause of death in younger individuals in Western countries. It has also been suggested that some excess mortality noted in type 1 diabetes mellitus, another autoimmune disorder, is caused by suicide and external causes of death. Psychiatric disease is more common in CD (although probably not schizophrenia), and psychiatric disease is a risk factor for suicide in patients with CD. One smaller British study has also reported an increased risk of suicide in children with CD. We therefore explored the risk of suicide in our cohort of biopsy-verified patients with CD and found a small but statistically significantly increased risk of suicide (HR = 1.55; 95% CI = 1.15–2.10; based on 54 completed suicides).

Mortality in Children with CD

Few studies have investigated mortality in children with CD. Solaymani-Dodaran and colleagues stratified their data set according to age at diagnosis and found a 2.6-fold increased risk of death in patients diagnosed with CD in childhood. What was especially concerning with this study was that deaths caused by accidents, suicide, and violence were increased (based on 7 deaths) and that excess mortality increased with longer follow-up (SMR was 3.32 after more than 5 years after diagnosis).

Swedish data have since confirmed that childhood CD is associated with excess mortality (excluding the first year of follow-up: HR = 1.78) and remains increased after more than 5 years of follow-up (HR = 1.90). Because of the concerns raised by Solaymani-Dodaran and colleagues about deaths from external causes, we specifically looked at suicide rates in children and adolescents with CD in another population-based study. In this study, we found no particular increase in suicide rates in patients diagnosed with CD before the age of 20 years (HR = 1.42). However because HRs were similar in all age strata in our cohorts, we concluded that an early diagnosis of CD is unlikely to eliminate the increased suicide risk seen in CD.

Individuals Without VA

In our paper on biopsy-verified CD and future mortality, we also investigated mortality in patients undergoing small intestinal biopsy but who did not have CD. A notable finding was that the mortality excess risk was higher in patients with inflammation (Marsh stage 1–2, intraepithelial lymphocytosis: HR = 1.72) than in those with VA (Marsh stage 3, CD: HR = 1.39). There are several explanations for this finding. One is that patients without VA have traditionally not adopted a GFD. Hence, their inflammation is likely to remain active, whereas patients with VA (CD) would have received a GFD with a high chance of long-term mucosal healing (assuming that the GFD is protective against mortality). Another possible explanation is that not all duodenal inflammation is caused by CD. Whereas the heredity for CD was 12% in patients with VA in our study, it was lower in those with inflammation (5%), perhaps reflecting a more heterogeneous origin of this condition. Mahadeva and colleagues have reported that 40% of individuals with intraepithelial lymphocytosis develop CD (here defined as VA), but that most individuals do not. Still, when 2 researchers (independently) manually examined more than 1500 biopsy reports with VA or inflammation, few patients with inflammation had a non-CD diagnosis recorded in the biopsy register (inflammatory bowel disease was the most common other diagnosis and mentioned in 1.6% of the biopsy reports with Marsh 1–2). It is therefore difficult to draw any conclusions regarding mortality in CD from our findings of an increased risk of death in patients with Marsh stage 1 to 2.

Underlying Mechanism for the Excess Mortality in CD

The dominant causes of death in patients with CD seem to be cardiovascular death and death from malignancy. Because patients with CD often lack traditional cardiovascular risk factors (they have less hypertension or no difference in blood pressure, more beneficial lipid profile, lower BMI, [and most, but not all, studies suggest an inverse relation with smoking]), an explanation for the increased risk of cardiovascular death must be sought elsewhere. Some studies have found no or even a negative association between CD and cardiovascular disease, whereas the largest study so far (991 patients with CD developed ischemic heart disease during follow-up) found that patients with CD have about a 20% increased risk of cardiovascular disease.

The most plausible link between CD and cardiovascular disease (as well as with death from malignancy) is the increase in inflammation, both before and after CD diagnosis. Although Lewis and colleagues did not compare C-reactive protein (CRP) values in patients with CD and controls, the average CRP level at diagnosis (based on 94 patients with CD) suggests that patients with CD have a normal CRP. However, other research has shown that patients with CD may have increased levels of other inflammatory markers (such as interleukin 4, 6, and 10, and tumor necrosis factor α) that could play a detrimental role for the risk of cardiovascular disease and malignancy.

It is also probable that CD protects against certain types of disease because patients often have a slightly lower BMI and smoke less often. Among diseases that may be less common in CD is breast cancer, which is a common cause of death in women.

GFD and Mortality

A GFD is the only treatment of CD. Numerous reports confirm that treatment of CD results in better growth and overall health. The effect of GFD on mortality is less certain, and no prospective study on mortality has compared patients who adhere to a GFD with those who are nonadherent. Data on the role of GFD for mortality risk are therefore limited to retrospective data (often medical record reviews), which carry a strong risk of bias. In the study by Corrao and colleagues, patients likely to have been on a GFD were at a nonsignificantly lower risk of death compared with the general population (SMR = 0.5; 95% CI = 0.2–1.1). Those with uncertain dietary adherence had a 2-fold increased risk of dying, whereas those who were unlikely to have followed the diet were at a 6-fold increased risk of dying (95% CI = 4.0–8.8). However, a problem with these calculations is that most of the data in patients who died during the follow-up were obtained from relatives or friends, whereas individuals who were alive at the end of the follow-up were asked to grade their dietary adherence themselves. Thus, it cannot be ruled out that death per se influenced the relatives’ responses and that patients who were alive overestimated their self-reported dietary adherence.

It could be argued that the higher mortality in patients diagnosed with CD in the 1970s and 1980s when strict adherence to GFD may not have been as widespread as it is currently is proof that a GFD is important to minimize excess mortality. Meanwhile, the lack of a major difference in mortality between individuals with VA (HR = 1.39) and those with normal mucosa but positive CD serology (HR = 1.35) suggests that a GFD may not have such a strong effect on mortality. The main effect of a GFD is to heal the mucosa, but even individuals with normal mucosa seem to be at increased risk of death in this study. However, a GFD may have other beneficial effects on the health of patients with CD. Some investigators have directly or indirectly contended that a delay in the institution of a GFD increases the risk of other autoimmune comorbidities (with detrimental effects on mortality) or affects the mortality risk per se. One of our studies showed increased risk of death in children with biopsy-verified CD, suggesting that short diagnostic delay in diagnosis may not prevent excess mortality.

Mortality: Implications for the Nondietary Management of CD

Given that cardiovascular disease is a common cause of death in patients with CD, general preventive measures are also relevant in patients with CD. Such measures include cessation of smoking, treatment of hypertension and hyperlipidemia, encouraging physical activity, and avoiding overweight. Although many of these risk factors are inversely related to CD, they are still highly prevalent in adults with CD.

A second theme of prevention deals with infectious and respiratory disease. Given the increased risk of both pneumococcal infections and influenza, immunization against these infections may be considered in patients with CD.