Risk Prediction of New Intracranial Aneurysms at Follow-Up Screening in People With a Positive Family History

Background: In first-degree relatives of patients with aneurysmal subarachnoid hemorrhage (aSAH), the risk of an intracranial aneurysm can be predicted at initial screening but not at follow-up screening. We aimed to develop a model for predicting the probability of a new intracranial aneurysm after initial screening in people with a positive family history of aSAH. Methods: In a prospective study, we obtained data from follow-up screening for aneurysms of 499 subjects with ≥2 affected first-degree relatives. Screening took place at the University Medical Center Utrecht, the Netherlands, and the University Hospital of Nantes, France. We studied associations between potential predictors and the presence of aneurysms using Cox regression analysis and the predictive performance at 5, 10, and 15 years after initial screening using C statistics and calibration plots, while correcting for overfitting. Results: In 5050 person-years of follow-up, intracranial aneurysms were found in 52 subjects. The risk of aneurysm at 5 years was 2% to 12%, at 10 years, 4% to 28%, and at 15 years, 7% to 40%. Predictors were female sex, history of intracranial aneurysms/aneurysmal subarachnoid hemorrhage, and older age. The sex, previous history of intracranial aneurysm/aSAH, older age score had a C statistic of 0.70 (95% CI, 0.61–0.78) at 5 years, 0.71 (95% CI, 0.64–0.78) at 10 years, and 0.70 (95% CI, 0.63–0.76) at 15 years and showed good calibration. Conclusions: The sex, previous history of intracranial aneurysm/aSAH, older age score provides risk estimates for finding new intracranial aneurysms at 5, 10, and 15 years after initial screening, based on 3 easily retrievable predictors; this can help to define a personalized screening strategy after initial screening in people with a positive family history for aSAH.

P eople with ≥2 first-degree relatives with aneurysmal subarachnoid hemorrhage (aSAH) are at increased risk of aSAH, with a lifetime risk of up to 25%. 1 Modeling studies have shown that for these people, preventive screening for intracranial aneurysms is cost-effective when repeated every 5 to 7 years between the ages of 20 and 70 to 80 years. 2,3 The absolute risk of finding an intracranial aneurysm is around 11% at initial screening and that of finding a new intracranial aneurysm at follow-up screening, after an initial screening, is around 7%. 4 Thus, most people undergoing screening have no aneurysms. Risk factors for finding new aneurysms at follow-up screening may help identify subgroups that have a small or high chance of a new aneurysm. The efficacy of screening may thereby be increased. Stroke Recently, we developed the number of affected relatives, age, smoking, hypertension prediction score for predicting the risk of an intracranial aneurysm at initial screening in people with ≥2 first-degree relatives with aSAH. 5 It is not known whether these same predictors can be used to predict the probability of an intracranial aneurysm at follow-up screening or whether other predictors play a role. We, therefore, aimed to develop a model for predicting the probability of finding an intracranial aneurysm during follow-up screening in people with a positive family history of aSAH.

METHODS
The data that support the findings of this study are available upon request to the corresponding author.

Study Population
We used a cohort of people with a positive family history of aSAH, defined as having ≥2 first-degree relatives who had an aSAH or 1 first-degree relative with an aSAH and ≥1 first-degree relatives with an unruptured intracranial aneurysm, screened for aneurysms in 2 hospitals: the University Medical Center Utrecht (Utrecht, the Netherlands) and Centre Hospitalier Universitaire de Nantes (Nantes, France). Both centers have a prospectively collected database with detailed information on consecutive patients screened. We used data from all the repeated screenings, following an initial screening performed between April 1993 and April 2018 (Dutch database) and between December 2012 and April 2017 (French database). Initial screening was defined as the first time imaging with magnetic resonance angiography or computed tomography (CT) angiography was performed to screen for the presence of intracranial aneurysms in a person with a positive family history for aSAH. The standard screening modality was magnetic resonance angiography; in the case of contraindications, screening was performed by CT angiography. Screening reflects clinical practice and was not according to a study protocol. Screening was usually performed from the age of 18 years until the age of ≈70 years, with the precise cutoff depending on the state of health of those screened. In the Dutch center, people were referred for screening in different ways. First, all people with aSAH who were admitted to the Neurology ward or people with an intracranial aneurysm who visited the outpatient clinic were routinely asked for details of their family history. If aSAH had occurred in their relatives, the patients were informed that their relatives were welcome to visit the outpatient clinic to be informed about screening for aneurysms. Second, people were also referred for screening by general practitioners or by neurologists and neurosurgeons from other hospitals. Third, people with aSAH and a positive family history of aSAH were advised to undergo screening for de novo intracranial aneurysms, 5 years after having had an aSAH. For the French cohort, data from the Understanding the Pathophysiology of Intracranial Aneurysm project were used. 6 This study recorded the family history of people with an intracranial aneurysm who had at least 1 firstdegree relative with an intracranial aneurysm; additional family members identified were contacted for MRI screening. Those relatives who fulfilled our criteria for a positive family history of aSAH and who had been screened for intracranial aneurysms were retrieved from this cohort. Our

Definitions
Affected first-degree relatives (parents, siblings, or children) may have a definite or probable aSAH. Definite aSAH was defined as an abrupt onset of severe headache or loss of consciousness, with or without focal neurological signs, the presence of subarachnoid blood on head CT, compatible with a ruptured intracranial aneurysm and an aneurysm on CT angiography, magnetic resonance angiography, or digital subtraction angiography. A relative with a probable aSAH was defined as a relative having a history of a stroke at an age of <70 years and a second ictus within 4 weeks followed by death for whom medical records were no longer available. 7 In affected first-degree relatives with an unruptured intracranial aneurysm, the aneurysm had to be proven by CT angiography, magnetic resonance angiography, or digital subtraction angiography. Aneurysms that, in retrospect, were visible on previous images were included at being present at the time of that images. We included patients who underwent at least 1 follow-up screening, performed no less than 3 years after the initial screening. After each screening, screenees are advised to return for further screening 5 years (with the exception of one large family in the University Medical Center Utrecht in which 3 years is advised as aneurysms developed and ruptured within the regular screening interval of 5 years in this family 8 ) later. However, this is only a recommendation, and we leave the decision with the screenees whether they want to undergo screening again. Therefore, some screenees only returned for 1 additional screening, whereas in others, multiple follow-up screenings have been performed. For the different follow-up screenings, we used all the data up to 5 years after the prior screening. We excluded people screened for intracranial aneurysms because of autosomal dominant polycystic kidney disease. The outcome of interest was detecting the presence of a new intracranial aneurysm during follow-up screening.

Model Development
Candidate predictors were preselected on the basis of literature including age, sex, smoking (former or current smoking), hypertension (defined as history of hypertension or use of antihypertensive drugs), previous intracranial aneurysm/aSAH, and the number of affected family members with aSAH or intracranial aneurysms. 4,[9][10][11][12][13] Information on all candidate predictors, including age, was collected at baseline, that is, during initial screening. The number of affected family members with aSAH or intracranial aneurysms was categorized as either 2 affected relatives or ≥3 affected relatives. Having a history of a previous intracranial aneurysm/aSAH was defined as (1) an aSAH occurring before the initial screening or (2) finding an aneurysm at a previous screening that was occluded by endovascular or microsurgical treatment.

Statistical Analysis
In the Dutch cohort, data for smoking (33%) and hypertension (39%) were missing; there were no missing data for the remaining candidate predictors in this cohort or for any predictors in the French cohort. Missing data were imputed with multiple imputation, creating 10 imputed data sets. Restricted cubic splines were used to assess whether the continuous predictor age could be analyzed as a linear term or required transformation. Age showed a linear association with the outcome. Predictors for aneurysms during screening were studied using Cox regression analysis in all 10 imputed data sets with stratification for the cohort (Dutch or French). The full model, containing all potential predictors, was simplified with backward selection, based on the Akaike Information Criterion. 14 The proportional hazards assumption was checked by visually inspecting the log minus log plot for each predictor. The model was internally validated with bootstrapping techniques, because prognostic models, derived from multivariable regression analysis, can be too optimistic and overestimate predictions when applied to a new cohort. 15,16 A shrinkage factor was estimated from the bootstrap procedure, and regression coefficients were shrunk to correct for overfitting. Discrimination of the model was examined with the C statistic, corrected for overoptimism by bootstrapping. Discrimination refers to the ability of the model to distinguish between people with and without an intracranial aneurysm. The C statistics of each imputed data set were pooled with Rubin rules. 17 Calibration of the model, which refers to the correspondence between the observed and the predicted risk, was visually inspected with 5-, 10-, and 15-year calibration plots. The regression coefficients in each imputation data set were pooled with Rubin rules. 17 To facilitate the practical application of the model, in the final model, we used the β-coefficients of the predictors to allocate points to each predictor and thus generate a risk score. 18 The score chart is accompanied by a figure that provides the 5-year absolute risk and the 10-and 15-year cumulative absolute probabilities of finding a new intracranial aneurysm. Table 1 Table S1.

RESULTS
The results of the multivariable Cox regression analysis are presented in Table 2. The following predictors of an intracranial aneurysm were identified: female sex, history of intracranial aneurysm/aSAH, and older age (sex, previous history of intracranial aneurysm/aSAH, older age [SPA] score). We combined all identified predictors in 1 model. After shrinkage, the C statistic of the model was 0.70 (95% CI, 0.61-0.78) at 5 years after initial screening, 0.71 (95% CI, 0.64-0.78) at 10 years, and 0.70 (95% CI, 0.63-0.76) at 15 years. The calibration plot showed good correspondence between predicted and observed risk ( Figure 1) with a Brier score of 0.05 and a calibration slope of 1.30 at 5 years after initial screening, a Brier score of 0.08 and a calibration slope of 1.12 at 10 years, and a Brier score of 0.08 and a calibration slope of 1.13 at 15 years. The original regression equation and baseline survival function are provided in Table S2. We translated regression coefficients into a score chart presented in Table 3. In combination with Figure S1, this score chart can be used to predict probabilities of finding an aneurysm for individual people ca. 5, 10, and 15 years after initial screening. Figure 2 shows a risk chart with estimated probabilities of finding an intracranial aneurysm ca. 5, 10, and 15 years after initial screening according to the SPA score. The probability of finding an aneurysm ranged from 2% in men aged 20 to 29 years, without a history of intracranial aneurysm/aSAH, 5 years after initial screening, up to a cumulative risk of 40% in women aged 60 to 69 years, with a history of intracranial aneurysm/aSAH, 15 years after initial screening.

DISCUSSION
We have developed the SPA score that predicts the individualized risk of a new intracranial aneurysm after initial screening in people with a positive family history of aSAH. Based on the predictors, sex, history of intracranial aneurysm/aSAH, and age, the probability of discovering a new intracranial aneurysm ranged from 2% in men aged 20 to 30 years, without a history of intracranial aneurysm/ aSAH, 5 years after initial screening, up to a cumulative risk of 40% in women aged 60 to 70 years, with a history of intracranial aneurysm/aSAH, 15 years after initial screening. Our data also show that if results of screening at 15 years are compared with those at 10 years, it is still possible to discover a new aneurysm. The chance of a new aneurysm developing between 10 and 15 years after the initial screening varies between 3% for men <40 years of age, without a history of intracranial aneurysm/ aSAH, to 12% for women and men >60 years of age.
At initial screening, predictors for the presence of intracranial aneurysms in people with a positive family history of aSAH are number of affected relatives, age, smoking, and hypertension. 5 We found that the number of affected relatives, smoking, and hypertension had no added value for the prediction of a new aneurysm at follow-up screening, while the predictors previous intracranial aneurysm/ aSAH and sex, were important. A possible explanation for sex being an important predictor is that at follow-up screening, people are older. In a meta-analysis, the prevalence of unruptured intracranial aneurysms in women was almost similar to that in men in people <50 years of age but twice as high as in men for those >50 years of age. 11 Also, the incidence of aSAH in those <50 years of age is similar in women and men but much higher in women than in men >50 years of age. 12 Thus, at increasing age with follow-up screening, the difference between women and men can become more apparent. Factors explaining the sex difference in risk of intracranial aneurysm development after the age of 50 years may be female-specific hormonal and reproductive factors. A previous systematic literature review of female risk factors for an aSAH found an increased risk for postmenopausal versus premenopausal women. The pathophysiology of this effect and its influence on the difference in incidence of aSAH between the sexes remains, however, unclear. 19 The explanation may lie in female-specific genetic factors of the X chromosome or sex-specific effects of yet unknown clinical factors, which occur more often or have a stronger effect in women.
In the prediction model for risk of aneurysms at initial screening, a history of aSAH was assessed as a predictor but had no added value when other risk factors were taken into account. 5 In the current study, for predictor history, we not only included people who had had an aSAH but also those in whom an aneurysm was found at initial screening. This combined predictor did have added value in the prediction of aneurysm risk during followup. This may be explained by the fact that the group of individuals with a positive history is now larger, leading to more power to show that this predictor contributes to the risk of de novo aneurysms. It should be noted, however, that the contribution of this predictor compared with the  other predictors seems relatively small. This is evidenced by the fact that the chances of a new aneurysm in such patients do not seem to be much increased compared with others without a previous aneurysm. Although both risk factors were identified as predictors at initial screening, our study found no added value of hypertension and smoking in the prediction of an aneurysm during follow-up. 5 In our study, we only had data on smoking and hypertension at baseline, not during followup. It is possible that at the time of the initial consultation and screening, people became more aware about risk factors for aneurysm development and may have quit smoking and had better blood pressure control. As a result, the risk factors may have had less of an influence on aneurysmal risk during follow-up compared with initial screening. In a study of people with a positive family history of aSAH, an association with current smoking was found but not former smoking-an observation that supports this hypothesis. 13 An important strength of this work is the large sample size that enabled us to study a broad range of prognostic factors. Also, the study population consisted of patients from 2 countries, which adds to the external validity of our results. Certain limitations should be considered. First, despite the prospective data collection, some data on smoking and hypertension were missing from the Dutch cohort. We used multiple imputation to predict values missing from this cohort using information from all potential predictors and outcomes. Second, we were not able to validate our model externally. We used data from 2 different countries to develop the prediction model, but the number of patients in whom new aneurysms were found during follow-up was too small to subdivide the cohort into a developmental and validation cohort. We have merged these 2 cohorts but still see that the 95% CIs of the hazard ratios of the predictors female sex and history of intracranial aneurysm/aSAH are relatively wide and not statistically significant. The data on the predictor history are consistent with an earlier study in a smaller patient population in which history of intracranial aneurysms was also found as a risk factor for finding an aneurysm at follow-up screening with a hazard ratio of 4.5 (95% CI, 1.1-18.7). 4 The fact that in this study the effect size was statistically significant may be explained by chance which is further supported by the 95% CI in that study being wider than the one found in our study (95% CI, 1.1-18.7 4 versus 0.87-2.86). Third, although calibration of our model was good at 5 years after initial screening, the risk of aneurysms was slightly underestimated at 10 years, especially in the higher risk quintile. This is, however,

60-69 4
An individual score is the sum of the points assigned to each of the predictors. aSAH indicates aneurysmal subarachnoid hemorrhage. Information on all candidate predictors, including age, was collected at baseline (ie, during initial screening). the best data we currently have to predict the probability of an aneurysm at follow-up screening. Fourth, to define a positive family history, we included family members who had a definite aSAH but also those with a probable aSAH, defined as an episode suspected to be aSAH in a person <70 years of age, such as stroke with a second ictus within 4 weeks followed by death. Because the positive predictive value of probable aSAH is 0.7, 7 family members with another type of stroke may have been misdiagnosed as having an aSAH. Fifth, information on all candidate predictors, including age, was collected during initial screening and not during the first follow-up screening, as we only have reliable information on the risk factors smoking and hypertension at baseline and not during follow-up. Finally, we did not include people screened for aneurysms who only had 1 affected first-degree relative. Consequently, our results cannot be extrapolated to this group.
For clinical practice, the SPA-prediction score can be used to personalize follow-up screening for intracranial aneurysms in people with ≥2 affected first-degree relatives. Screening may thereby become more efficient. It should be noted, however, that the SPA score has not yet been externally validated. This should be a topic for further research. Another question for further research is the influence of smoking cessation on the risk of new aneurysm formation during follow-up.

Sources of Funding
This project has received funding from the European Research Council under the European Union Horizon 2020 research and innovation program (grant agreement number 852173). Dr Zuurbier reports receiving a grant from the Remmert Adriaan Laan Foundation, during the conduct of the study. Dr Redon was supported by the French Regional Council of Pays-de-la-Loire (VaCaRMe program) and the Agence Nationale de la Recherche (ANR-15-CE17-0008-01 to Dr Rinkel). Drs Desal and Bourcier were supported by the French Ministry of Health (Clinical trial NCT02848495 to Dr Desal), the Genavie Foundation, the Société Française de Radiologie, and the Société Française de Neuroradiologie. The funding organizations were not involved in the design or conduct of the study; the collection, management, analysis, or interpretation of the data; the preparation, review, or approval of the article; or the decision to submit the manuscript for publication.

Disclosures
The authors acknowledge the support from the Netherlands Cardiovascular Research Initiative: an initiative with support of the Dutch Heart Foundation, CVON2015-08 ERASE. They are grateful to the Clinical Investigation Center (INSERM CIC1413) for its assistance in managing the Understanding the Pathophysiology of Intracranial Aneurysm biobanks.