CLINICAL ARTICLE

Unruptured intracranial aneurysm growth trajectory:

occurrence and rate of enlargement in 520 longitudinally
followed cases
Aichi Chien, PhD,1 Rashida A. Callender, MPH,2 Hajime Yokota, MD,1 Noriko Salamon, MD,1
Geoffrey P. Colby, MD, PhD,1,3 Anthony C. Wang, MD,3 Viktor Szeder, MD,1 Reza Jahan, MD,1
Satoshi Tateshima, MD,1 Juan Villablanca, MD,1 Gary Duckwiler, MD,1 Fernando Vinuela, MD,1
Yuanqing Ye, PhD,2 and Michelle A. T. Hildebrandt, PhD2
Departments of 1Radiology and 3Neurosurgery, David Geffen School of Medicine at UCLA, Los Angeles, California; and
2
Department of Epidemiology, MD Anderson Cancer Center, University of Texas, Houston, Texas

OBJECTIVE  As imaging technology has improved, more unruptured intracranial aneurysms (UIAs) are detected inci-
dentally. However, there is limited information regarding how UIAs change over time to provide stratified, patient-specific
UIA follow-up management. The authors sought to enrich understanding of the natural history of UIAs and identify basic
UIA growth trajectories, that is, the speed at which various UIAs increase in size.
METHODS  From January 2005 to December 2015, 382 patients diagnosed with UIAs (n = 520) were followed up at
UCLA Medical Center through serial imaging. UIA characteristics and patient-specific variables were studied to identify
risk factors associated with aneurysm growth and create a predicted aneurysm trajectory (PAT) model to differentiate
aneurysm growth behavior.
RESULTS  The PAT model indicated that smoking and hypothyroidism had a large effect on the growth rate of large
UIAs (≥ 7 mm), while UIAs < 7 mm were less influenced by smoking and hypothyroidism. Analysis of risk factors related
to growth showed that initial size and multiplicity were significant factors related to aneurysm growth and were consistent
across different definitions of growth. A 1.09-fold increase in risk of growth was found for every 1-mm increase in initial
size (95% CI 1.04–1.15; p = 0.001). Aneurysms in patients with multiple aneurysms were 2.43-fold more likely to grow
than those in patients with single aneurysms (95% CI 1.36–4.35; p = 0.003). The growth rate (speed) for large UIAs (≥
7 mm; 0.085 mm/month) was significantly faster than that for UIAs < 3 mm (0.030 mm/month) and for males than for
females (0.089 and 0.045 mm/month, respectively; p = 0.048).
CONCLUSIONS  Analyzing longitudinal UIA data as continuous data points can be useful to study the risk of growth and
predict the aneurysm growth trajectory. Individual patient characteristics (demographics, behavior, medical history) may
have a significant effect on the speed of UIA growth, and predictive models such as PAT may help optimize follow-up
frequency for UIA management.
https://thejns.org/doi/abs/10.3171/2018.11.JNS181814
KEYWORDS  aneurysm growth trajectory; longitudinal data; CTA follow-up; risk factors; vascular disorders

W
hen assessing unruptured intracranial aneu-
rysms (UIAs), several factors are commonly
considered, including patient life expectancy,
previous subarachnoid hemorrhage (SAH), family history,
UIA size, growth on serial imaging, and clinical symp-
toms.10 As new endovascular devices have been developed,
it has become possible to treat more UIAs, and preven-
tative treatment is increasingly common. UIA manage-
ment aims to avoid aneurysm rupture, but while 3% of the
population may harbor a UIA, only a small percentage of
these ever rupture, and despite continuing improvements,
treatment is not without risk.24,30 To further improve aneu-

ABBREVIATIONS  ICA = internal carotid artery; ISUIA = International Study on Unruptured Intracranial Aneurysms; MASAL = multivariate adaptive splines for analysis of
longitudinal data; MCA = middle cerebral artery; PAT = predicted aneurysm trajectory; SAH = subarachnoid hemorrhage; TIA = transient ischemic attack; UIA = unruptured
aneurysm.
SUBMITTED  June 24, 2018.  ACCEPTED  November 5, 2018.
INCLUDE WHEN CITING  Published online March 1, 2019; DOI: 10.3171/2018.11.JNS181814.

©AANS 2019, except where prohibited by US copyright law J Neurosurg  March 1, 2019 1
Chien et al.
rysm management, better assessment and more effective
monitoring are necessary.
The natural history of aneurysms remains poorly un-
derstood. In particular, risk factors that influence UIA
growth and the relationship of growth with rupture are
not well defined, with significant variations in findings
between past studies.2,5 Along with a lack of clear genet-
ic markers, this makes early identification of dangerous
UIAs extremely difficult. There is limited information on
which to base stratified, patient-specific UIA follow-up
management. This complicates discrimination of UIAs
that are likely to grow, becoming more likely to rupture
and more difficult to treat.
In general, previous studies of UIA growth have fo-
cused on information collected at an initial time point
versus future occurrence of growth, without investigat-
ing the speed of enlargement.3,21,22 To identify basic UIA
growth trajectories, this study instead focuses on longitu-
dinal data. It is the goal that better understanding of the
natural history of aneurysms will support more judicious
selection of UIAs for preventative treatment and enable
the development of risk-guided follow-up strategies for
UIAs. Therefore, the purpose of this study was to perform
a comprehensive analysis of UIA growth to identify risk
factors and inform the development of a predictive model
of growth rates based on patient-specific and aneurysm-
specific characteristics.
Methods
Patient Population
The study was approved by UCLA’s IRB. Procedures
followed were in accordance with institutional guidelines.
A retrospective review of aneurysm clinical imaging re-
cords from 2005 to 2015 was performed. A total of 382
patients who were diagnosed with UIAs and monitored
using clinical CTA images without treatment were includ-
ed in this study, accounting for 520 individual aneurysms
and 1636 data points. These cases generally reflect the In-
ternational Study on Unruptured Intracranial Aneurysms
(ISUIA) guidelines: the majority of monitored UIAs were
smaller than 7 mm, and larger UIAs were typically moni-
tored due to advanced patient age or the patient declining
treatment. The endpoints of the follow-up were when the
aneurysms were treated with surgical clipping or endo-
vascular therapy. Patients were lost to follow-up if they
refused further follow-up or failed to return for a follow-
up visit. The complete follow-up periods were calculated
from the date of the initial diagnostic CTA study that
identified the aneurysm to the date of the last follow-up
scan. In comparison to these cases monitored through im-
aging, during this time period, approximately 70 patients
per year underwent endovascular or surgical treatment for
intracranial aneurysms, as previously reported.22 Eight an-
eurysms that ruptured during monitoring were excluded
from the main analysis and model building, and instead
were included in the PAT testing.
Data Collection
Electronic medical records and the image database
were used to collect patient- and aneurysm-specific infor-
2 J Neurosurg  March 1, 2019
mation. Patient-specific information included demograph-
ics (age, sex), behavior (excessive alcohol use, previous or
current cigarette smoking), and medical history (coronary
artery disease, diabetes mellitus, hypertension, hypothy-
roidism, transient ischemic attack [TIA], stroke, family
history of SAH). For this study, coronary artery disease
broadly included disease related to atherosclerosis of the
arteries of the heart. The category of diabetes included
both type 1 and 2 diabetes. Aneurysm-specific informa-
tion included the number of aneurysms, aneurysm loca-
tion, and aneurysm size at the follow-up dates. Size was
recorded from diagnostic reports in which the measure-
ments were collected by board-certified neuroradiologists
following a standardized aneurysm measurement proto-
col23 and was determined according to the measurement
of the largest dimension.20,23
Statistical Analysis
The data were analyzed using the Stata statistical
package (version 14.0, StataCorp). Patient-specific and
aneurysm-specific characteristics related to growth were
analyzed using the chi-square test for categorical variables
and Student’s t-test for continuous variables. Different
definitions of growth based on change in aneurysm size
between the first and last measurements were assessed.
Thresholds of 0.5, 0.6, 0.75, and 1 mm were tested in a
sensitivity analysis to determine whether factors related to
growth changed due to the definition of growth. The odds
of growth greater than or equal to 0.6 mm as a function of
aneurysm- and patient-specific variables were determined
using univariate logistic regression with calculation of
odds ratios and corresponding 95% confidence intervals.
Multivariate logistic regression was conducted to control
for confounding of significant predictors.
Linear mixed-effects models were used to analyze
longitudinal differences in size and rate of change for a
subset of aneurysms with growth greater than or equal to
0.6 mm. A covariance matrix of the measurements over
time was used to determine the correlation structure of
the measurements within a patient. Three-level models
with random intercepts and slope using compound sym-
metric or exchangeable covariance were constructed as a
function of time and each variable. A product or interac-
tion term for time and each variable was also included to
identify differences in trajectories between subgroups. In
these models, random patient- and aneurysm-specific in-
tercepts were included to account for intraclass correlation
between multiple aneurysms from the same patient and
repeated measurements on the same aneurysm. Random
patient- and aneurysm-specific time slopes were included
in the model for patient- and aneurysm-specific variables,
respectively.
To build the predicted aneurysm trajectory (PAT)
model, multivariate adaptive splines for analysis of lon-
gitudinal data (MASAL) using the MASAL package in R
was applied to guide the development of multivariate lin-
ear mixed models for 2-level data. MASAL is a nonpara-
metric, automated model selection tool that uses methods
similar to recursive partitioning to estimate the covariance
structure and the regression function simultaneously.27–29
MASAL also uses piecewise linear splines to discover and

model potential nonlinear relationships including changes
in the rate or velocity of growth. The MASAL models
were reproduced in Stata using an unstructured covari-
ance pattern with the assumption that all aneurysms ana-
lyzed were independent. The Akaike information criterion
was used to compare the models and determine goodness
of fit. The linear growth models were applied to predict
size at the 12-month follow-up for hypothetical patients.
The rate of change and age of growth initiation were cal-
culated using the predicted values.
Results
Patient Population
The aneurysm-specific and patient-specific characteris-
tics of the 520 longitudinally followed UIAs are shown in
Table 1. On average, UIAs had an initial size of 4.8 ± 4.0
mm (± SD), were monitored by CTA for an average of 32.7
months, and the time between CTA scans was 16.1 months
with an average of 3 CTA image sets recording the UIA
progression. A majority of the patients were women (315,
82.5%) with a mean age of 61.8 years. Approximately 27%
of the patients were diagnosed with multiple aneurysms
(n = 104).
Risk Factors Associated With UIA Growth
The definition of aneurysm growth is not standardized
in the field.2,18,19 Criteria ranging from 0.5 mm to 2 mm
have been used in other studies of aneurysm growth.3,6,8,18
In medical practice, however, a cutoff of 1.0 mm is fre-
quently used.11,15 To provide clarity regarding the opti-
mal definition of UIA growth, we conducted a sensitivity
analysis to identify associations between aneurysm- and
patient-specific characteristics and 4 different definitions
of growth (0.5, 0.6, 0.75, and 1 mm). Considering that a
patient may have multiple aneurysms, patients with at
least 1 aneurysm at a cutoff were included in the respec-
tive growth subgroup. Depending on the growth threshold
definition, the percentage of aneurysms that enlarged per
year ranged from 6.14% to 2.68% (Table 1).
Initial aneurysm size was significantly associated with
growth for each definition (Table 1). Multiplicity was also
associated with growth for all definitions, with the excep-
tion of 1 mm (p = 0.22). With regard to medical history, a
diagnosis of hypertension was borderline significant with
growth for 0.6 mm (p = 0.079), 0.75 mm (p = 0.090), and 1
mm (p = 0.097). Previous history of stroke was borderline
significant for growth defined as > 1 mm only (p = 0.072).
Overall, the results of the analysis at each point were com-
parable, suggesting that risk factors impacting UIAs were
independent of the definition of growth. Because a growth
threshold of 0.6 mm had comparable results in this sensi-
tivity analysis as 0.75 mm and 1 mm, and allowed the in-
clusion of more cases to analyze growth trajectory, it was
chosen for the subsequent analyses.
Using this 0.6-mm definition as a cutoff, initial aneu-
rysm size was associated with a significant 1.09-fold in-
crease in risk of growth (95% CI 1.04–1.15; p < 0.001) for
every 1-mm increase in initial size (Table 2). Aneurysms in
patients harboring additional intracranial aneurysms were
2.43-fold more likely to grow than those in patients with
Chien et al.
single aneurysms (95% CI 1.36–4.35; p = 0.003). UIAs in
patients who had a diagnosis of hypertension were associ-
ated with a borderline significant increased risk of 1.76-
fold compared with UIAs in patients without hypertension
(95% CI 0.93–3.32; p = 0.082). Results of a multivariate
analysis of these significant and borderline significant fac-
tors are presented in Table 2. Results of the multivariate
analysis showed that initial size and multiplicity remain
significant when adjusting for the other predictors of an-
eurysm growth.
Predictors of Aneurysm Growth Behavior
Taking advantage of the availability of UIA measure-
ments over time, we investigated the predictors of UIA
growth rates (Table 3). The average growth rate for large
UIAs (≥ 7 mm) was 0.085 mm/month, which was signifi-
cantly faster than that for smaller UIAs (< 3 mm; 0.030
mm/month). Although not significant, UIAs located in the
internal carotid artery (ICA) and middle cerebral artery
(MCA) tended to grow faster than UIAs in other locations.
In the analysis of patient-specific factors, the UIA growth
rate was significantly faster for males (0.089 mm/month)
than for females (0.045 mm/month; p = 0.048). A positive
history of TIA was also associated with a faster growth
rate of 0.125 mm/month compared with a negative history
of TIA (0.044 mm/month; p = 0.016).
PAT Model
The population of patients who experienced UIA
growth ≥ 0.6 mm were used to develop prediction mod-
els for UIA growth behavior. To build the PAT model, we
first included all aneurysm- and patient-specific variables.
A stepwise procedure was then used to remove nonsig-
nificant terms. Exploratory analysis showed the predic-
tor variable initial size was significantly associated with
stroke (p = 0.015), hypertension (p = 0.025), coronary
artery disease (p = 0.004), atrial fibrillation (p = 0.007),
and diabetes mellitus (p = 0.013). Because these patient-
specific factors are collinear to initial size, or can be used
to predict initial size, they were removed during the model
building to avoid overfitting. Further refinement of the
model identified smoking history (previous or current)
and diagnosis of hypothyroidism as additional predictive
variables. The PAT model with coefficients for predicting
UIA size at follow-up is shown in Table 4. Of particular
note is the apparent strong interaction between hypothy-
roidism and smoking. Additionally, it should be noted that
the term in the PAT model, hypothyroidism × [-0.066 ×
(months of follow-up - 24)], is a spline that indicates slow-
er growth after 24 months of follow-up in hypothyroid pa-
tients, suggesting that UIAs do not grow in a linear man-
ner. Additionally, the PAT model allows different growth
patterns for UIAs with different initial sizes. When cases
were stratified by initial UIA size, removing the effects of
smoking and hypothyroidism attenuated the growth rate
prediction for large UIAs (≥ 7 mm), yet had little to no ef-
fect on the smaller UIAs. This suggests that UIAs smaller
than 7 mm were less influenced by the effects on growth
conferred by smoking and hypothyroidism. The Akaike
information criterion of the PAT model was 287.67.
J Neurosurg  March 1, 2019 3
4
Chien et al.

TABLE 1. Sensitivity analysis of patient characteristics and aneurysm growth

Aneurysm Growth ≥0.5 mm (%) Aneurysm Growth ≥0.6 mm (%) Aneurysm Growth ≥0.75 mm (%) Aneurysm Growth ≥1 mm (%)
Overall Yes No p Value Yes No p Value Yes No p Value Yes No p Value
Aneurysm-specific variables*
 Total 520 87 433 62 458 48 472 36 484

J Neurosurg  March 1, 2019

  % growth per AY  6.14  4.38  3.39  2.68
  Mean FU, mos 32.7 ± 25.4 39.9 ± 32.0 31.2 ± 23.6 0.0035 44.3 ± 34.0 31.1 ± 23.6 0.00012 46.6 ± 35.0 31.3 ± 23.8 6.3 × 10−5 51.2 ± 35.0 31.3 ± 24.0 4.98 × 10−6
  Mean initial size, 4.8 ± 4.0 5.9 ± 6.2 4.5 ± 3.4 0.0054 6.6 ± 7.1 4.5 ± 3.4 1.01 × 10−4 7.6 ± 7.7 4.5 ± 3.3 3.04 × 10−7 8.6 ± 8.6 4.5 ± 3.3 2.4 × 10−9
mm
  Initial size, mm 5.11 × 10−4 4.42 × 10−5 1.91 × 10−5 1.44 × 10−5
   ≥7 72 (13.9) 21 (24.1) 51 (11.8) 18 (29.0) 54 (11.8) 17 (35.4) 55 (11.6) 14 (38.9) 58 (12)
  5–7 87 (16.7) 16 (18.4) 71 (16.4) 11 (17.7) 76 (16.6) 9 (18.8) 78 (16.5) 8 (22.2) 79 (16.3)
  3–5 185 (35.6) 16 (18.4) 169 (39) 8 (13.0) 177 (38.6) 7 (14.6) 178 (37.7) 4 (11.1) 181 (37.4)
  <3 176 (33.8) 34 (39.1) 142 (32.8) 25 (40.3) 151 (33.0) 15 (31.3) 161 (34.1) 10 (27.8) 166 (34.3)
 Location 0.21 0.37 0.282 0.62
  ICA 273 (52.5) 43 (49.4) 230 (53.1) 33 (53.2) 240 (52.4) 25 (52.1) 248 (52.5) 22 (61.1) 251 (51.9)
  MCA 103 (19.8) 15 (17.2) 88 (20.3) 11 (17.7) 92 (20.1) 8 (16.7) 95 (20.1) 5 (13.9) 98 (20.2)
  ACA 97 (18.7) 16 (18.4) 81 (18.7) 9 (14.5) 88 (19.2) 7 (14.6) 90 (19.1) 5 (13.9) 92 (19.0)
  BAVA 47 (9) 13 (14.9) 34 (7.9) 9 (14.5) 38 (8.3) 8 (16.7) 39 (8.3) 4 (11.1) 43 (8.9)
Patient-specific variables†
 Total 382 81 301 57 325 45 337 33 349
  Mean age, yrs 61.8 ± 14.4 61.5 ± 17 61.8 ± 13.7 0.87 62.6 ± 16.5 61.6 ± 14.1 0.65 62.8 ± 17 61.6 ± 14.1 0.63 62.3 ± 18.9 61.7 ± 14.0 0.84
 Sex 0.29 0.71 0.71 0.71
  Women 315 (82.5) 70 (86.4) 245 (81.4) 48 (84.2) 267 (82.2) 38 (84.4) 277 (82.2) 28 (84.9) 287 (82.2)
  Men 67 (17.5) 11 (13.6) 56 (18.6) 9 (15.8) 58 (17.8) 7 (15.6) 60 (17.8) 5 (15.2) 62 (17.8)
  No. of aneurysms  0.0052 0.0022 0.040 0.22
  Single 278 (72.8) 49 (60.5) 229 (76.1) 32 (56.1) 246 (75.7) 27 (60) 251 (74.5) 21 (63.4) 257 (73.6)
  Multiple 104 (27.2) 32 (39.5) 72 (23.9) 25 (43.9) 79 (24.3) 18 (40) 86 (25.5 12 (36.4) 92 (26.4)
  Personal history 20 (6.8) 2 (3.3) 18 (7.7) 0.39 1 (2.4) 19 (7.5) 0.33 1 (3.3) 19 (7.2) 0.71 0 (0) 20 (7.3) 0.38
  Family history 19 (6.7) 4 (6.6) 15 (6.7) >0.99 4 (9.5) 15 (6.2) 0.50 4 (12.5) 15 (6) 0.25 3 (13.6) 16 (6.1) 0.17
 Seizure 7 (2.5) 2 (3.3) 5 (2.3) 0.65 2 (4.9) 5 (2.1) 0.28 2 (6.5) 5 (2) 0.18 2 (9.5) 5 (2) 0.091
 TIA 20 (7.1) 2 (3.4) 18 (8.1) 0.27 2 (5) 18 (7.5) 0.75 2 (6.7) 18 (7.2) >0.99 0 (0) 20 (7. 7) 0.38
 Stroke 30 (10.3) 9 (14.3) 21 (9.2) 0.24 7 (16.3) 23 (9.2) 0.17 6 (18.2) 24 (9.3) 0.13 5 (21.7) 25 (9.3) 0.072

CONTINUED ON PAGE 5 »
 » CONTINUED FROM PAGE 4
TABLE 1. Sensitivity analysis of patient characteristics and aneurysm growth
Aneurysm Growth ≥0.5 mm (%) Aneurysm Growth ≥0.6 mm (%) Aneurysm Growth ≥0.75 mm (%) Aneurysm Growth ≥1 mm (%)
Overall Yes No p Value Yes No p Value Yes No p Value Yes No p Value
Patient-specific variables† (continued)
 Hypertension 174 (52.9) 39 (55.7) 135 (52.1) 0.59 31 (64.6) 143 (50.9) 0.079 25 (65.8) 149 (51.2) 0.090 19 (67.9) 155 (51.5) 0.097
 Dyslipidemia 105 (34.0) 20 (31.3) 85 (34.7) 0.60 18 (40.0) 87 (33) 0.36 12 (34.3) 93 (33.9) 0.97 8 (33.3) 97 (34) 0.94
 CAD 36 (12.6) 9 (14.5) 27 (12.1) 0.61 8 (19.1) 28 (11.5) 0.18 6 (18.8) 30 (11.9) 0.26 3 (13.6) 33 (12.6) 0.75
  Atrial fibrillation 23 (8.2) 5 (8.3) 18 (8.1) >0.99 4 (9.8) 19 (7.9) 0.76 3 (9.7) 20 (8) 0.73 1 (4.8) 22 (8.5) >0.99
 Stenosis 44 (15.3) 5 (8.2) 39 (17.2) 0.083 5 (11.9) 39 (15.9) 0.51 4 (13.3) 40 (15.5) >0.99 3 (15) 41 (15.3) >0.99
 Arteriosclerosis 153 (53.1) 29 (47.5) 124 (54.6) 0.33 20 (47.6) 133 (54.1) 0.44 14 (46.7) 139 (53.9) 0.45 9 (45) 144 (53.7) 0.45
 DM 50 (17.1) 8 (12.9) 42 (18.2) 0.33 6 (14.3) 44 (17.5) 0.61 6 (18.8) 44 (16.9) 0.79 4 (18.2) 46 (17) 0.78
 Hypothyroidism 52 (17.8) 9 (14.5) 43 (18.6) 0.45 7 (16.7) 45 (17.9) 0.84 7 (21.9) 45 (17.2) 0.52 6 (27.3) 46 (17) 0.25
  Malignant tumor 53 (18.2) 9 (14.5) 44 (19.1) 0.40 5 (11.9) 48 (19.2) 0.26 5 (15.6) 48 (18.5) 0.70 2 (9.5) 51 (18.8) 0.39
  Past history of 175 (55) 37 (56.1) 138 (54.8) 0.85 26 (56.5) 149 (54.8) 0.83 24 (66.7) 151 (53.6) 0.14 16 (66.7) 159 (54.1) 0.23
surgery
  Excessive alcohol 6 (1.6) 0 (0) 6 (2.1) 0.35 0 (0) 6 (1.9) 0.60 0 (0) 6 (1.9) >0.99 0 (0) 6 (1.8) >0.99
use
  Cigarette smok- 102 (27.9) 20 (25.6) 82 (28.5) 0.62 12 (21.8) 90 (28.9) 0.28 10 (22.2) 92 (28.7) 0.37 6 (18.2) 96 (28.8) 0.19
ing‡
ACA = anterior cerebral artery; AY = aneurysm-year; BAVA = basilar artery–vertebral artery junction; CAD = coronary artery disease; DM = diabetes mellitus; FU = follow-up.
Boldface type indicates statistical significance. Italic type indicates a trend toward significance. Mean values are presented with SD.
* Values are presented as the number of aneurysms (%) unless stated otherwise.
† Values are presented as the number of patients (%) unless stated otherwise.
‡ Previous or current.

J Neurosurg  March 1, 2019

5
Chien et al.
Chien et al.

PAT Testing on Cases With Both Rupture and Growth TABLE 2. Risk of aneurysm growth ≥ 0.6 mm
As part of the model testing, the PAT was applied to OR 95% CI p Value
out-of-sample data from patients with aneurysms that rup-
tured during follow-up (Table 5). There were 8 patients Aneurysm-specific variables
and 11 individual aneurysms. The data set included 3 pa-   Mean initial size, mm 1.09 1.04–1.15 9.1 × 10−4
tients with 2 aneurysms, and in each patient, 1 aneurysm  Location 1.03 0.61–1.76 0.90
ruptured. The PAT model was used to predict aneurysm Patient-specific variables
size at each recorded follow-up visit. For each aneurysm,  Age 1.00 0.98–1.02 0.65
observed and predicted rates of change were calculated by
 Sex 1.16 0.54–2.49 0.71
subtracting the first and last measurements and dividing
by the total number of months of follow-up. A paired t-test  Multiplicity 2.43 1.36–4.35 0.0032
was performed to compare observed and predicted rates of   Personal history 0.30 0.04–2.30 0.25
change. The paired test suggested no significant difference   Family history 1.59 0.50–5.06 0.43
in the observed (0.041 ± 0.071 mm/month) and predicted  Seizure 2.38 0.45–12.70 0.31
(0.046 ± 0.026 mm/month) rates of change (p = 0.85). The  TIA 0.65 0.14–2.91 0.57
PAT model was also used to compute rates of change for
 Stroke 1.91 0.76–4.78 0.17
the UIAs used to develop our prediction model. The Stu-
dent t-test was used to compare predicted rates of change  Hypertension 1.76 0.93–3.32 0.082
for the ruptured and unruptured aneurysms. Although not  Dyslipidemia 1.35 0.71–2.60 0.36
significant, a higher predicted growth rate was found in  CAD 1.81 0.76–4.29 0.18
ruptured aneurysms (0.046 ± 0.026 mm/month) than in   Atrial fibrillation 1.26 0.40–3.90 0.69
unruptured aneurysms (0.038 ± 0.037 mm/month; Student  DM 0.78 0.31–1.97 0.61
t-test, p = 0.54).  Hypothyroidism 0.92 0.38–2.19 0.84
  Malignant tumor 0.57 0.21–1.52 0.26
Application of PAT to Evaluate UIAs in Real Patients and
Hypothetical Cases   Past history of surgery 1.07 0.57–2.01 0.83
To illustrate the utility of the PAT model, 3 representa-   Cigarette smoking* 0.69 0.35–1.36 0.28
tive UIAs with initial sizes of 1 mm, 4.10 mm, and 6.4 mm  Stenosis 0.72 0.27–1.94 0.51
that were not used for PAT model building were randomly  Arteriosclerosis 0.77 0.40–1.49 0.44
selected from the UIA database. PAT was then applied to Multivariate analysis of predictors
predict growth trajectory, and the findings were compared   Mean initial size, mm 1.09 1.02–1.18 0.011
to the actual growth for these UIAs (Fig. 1). These exam-  Multiplicity 2.97 1.49–5.90 0.002
ples demonstrated close agreement between predicted and
 Hypertension 1.47 0.76–2.84 0.25
observed growth.
We further applied the PAT model to a series of hypo- Boldface type indicates statistical significance. Italic type indicates a trend
thetical patients to illustrate predicted differences in pre- toward significance.
dicted size and growth rate during follow-up according to * Previous or current.
different combinations of risk factors. Figure 2 presents a
heat map representing a series of hypothetical 62-year-old
patients monitored for 5 years. This figure shows different UIA growth, most studies have been retrospective analy-
growth trajectories for patients with different combina- ses of relatively small cohorts over a relatively short period
tions of risk factors. The cumulative effects of smoking of time. With such cohort sizes, enlargement is typically
and hypothyroidism on aneurysm growth rate are much identified in fewer than 10–20 cases, increasing the poten-
greater than would be expected from a simple additive re- tial for statistical type 1 errors. As a result, meta-analyses
lationship. For patients with hypothyroidism, the growth have found the risk factors for UIA growth to be extremely
pattern is nonlinear. UIA size increases rapidly at first but heterogeneous among pooled studies.2,5
then slows down and levels off. Our examination of past studies revealed that the most
consistent predictor of UIA growth is the duration of fol-
low-up. The longer UIAs are followed, the more likely they
Discussion are to enlarge, best illustrated by the long-duration follow-
While prevention of intracranial aneurysm rupture is up of a cohort by Juvela et al.14 The study we present in-
the primary clinical interest, understanding UIA growth is cludes one of the largest reported cohorts, with an average
central to understanding the natural history of aneurysms follow-up of 2.7 years, slightly below the median of stud-
and, therefore, early detection or prevention. As new treat- ies included in meta-analyses.2 Comparing the percentage
ment devices such as flow-diverting stents have enabled of UIAs growing per aneurysm-year between studies, we
the treatment of more UIAs, with consequently fewer found that the percentage growth per aneurysm-year was
UIAs being monitored with imaging, the analysis of exist- 4.38%, which is in line with the higher previously report-
ing data sets is extremely important to understand aneu- ed values.13,18,19 Overall, it appears that as many as half of
rysm natural history. Recent meta-analyses have indicated UIAs continue to grow over time.14
that the current understanding of UIA growth risk factors Backes et al. suggested that the risk factors for aneu-
remains limited.2,5 Given the relative rarity of observing rysm growth are unlikely to be affected by the definition
6 J Neurosurg  March 1, 2019
 Chien et al.

TABLE 3. Patient- and aneurysm-specific predictors of rate (speed) of longitudinal growth

Total Growth Rate (mm/mo) 95% CI p Value
Patient-specific variables
 Age −0.00032 −0.0012 to 0.0006 0.50
 Sex
  Female 48 0.045 0.030 to 0.060 0.048
  Male 9 0.089 0.048 to 0.131 Ref
 Multiplicity
  Yes 25 0.042 0.020 to 0.064 0.31
  No 32 0.057 0.038 to 0.077 Ref
  Personal history
  Yes 1 0.008 −0.075 to 0.092 0.35
  No 41 0.049 0.033 to 0.065 Ref
  Family history
  Yes 4 0.038 −0.014 to 0.090 0.68
  No 38 0.05 0.033 to 0.066 Ref
 Seizure
  Yes 2 0.046 −0.027 to 0.120 0.93
  No 39 0.05 0.033 to 0.067 Ref
 TIA
  Yes 3 0.125 0.061 to 0.188 0.016
  No 37 0.044 0.029 to 0.059 Ref
 Stroke
  Yes 7 0.055 0.015 to 0.094 0.83
  No 36 0.05 0.032 to 0.067 Ref
 Hypertension
  Yes 32 0.051 0.035 to 0.067 0.3
  No 16 0.035 0.010 to 0.061 Ref
 Dyslipidemia
  Yes 19 0.045 0.030 to 0.069 0.78
  No 26 0.049 0.030 to 0.069 Ref
Aneurysm-specific variables
  Initial size, mm
   ≥7 18 0.085 0.062 to 0.109 6.0 × 10−5
  5–7 11 0.033 0.010 to 0.056 0.80
  3–5 8 0.04 0.001 to 0.078 0.58
  <3 25 0.03 0.011 to 0.049 Ref
 Location
  BAVA 9 0.044 0.018 to 0.071 0.49
  ACA 9 0.044 0.002 to 0.086 0.87
  MCA 11 0.055 0.025 to 0.084 0.49
  ICA 33 0.052 0.034 to 0.070 Ref
 CAD  
  Yes 8 0.055 0.016 to 0.094 0.64
  No 34 0.045 0.028 to 0.062 Ref
  Atrial fibrillation  
  Yes 4 0.03 −0.027 to 0.087 0.53
  No 37 0.049 0.032 to 0.066 Ref
 DM  
  Yes 6 0.045 0.007 to 0.083 0.91
  No 36 0.047 0.030 to 0.064 Ref
CONTINUED ON PAGE 8 »

J Neurosurg  March 1, 2019 7

Chien et al.

» CONTINUED FROM PAGE 7

TABLE 3. Patient- and aneurysm-specific predictors of rate (speed) of longitudinal growth
Total Growth Rate (mm/mo) 95% CI p Value
Aneurysm-specific variables (continued)
 Hypothyroidism  
  Yes 7 0.044 0.011 to 0.078 0.85
  No 35 0.048 0.030 to 0.066 Ref
  Malignant tumor  
  Yes 6 0.056 0.017 to 0.095 0.63
  No 36 0.046 0.029 to 0.063 Ref
  Past history of surgery  
  Yes 27 0.051 0.033 to 0.069 0.46
  No 19 0.04 0.016 to 0.063 Ref
  Cigarette smoking (previous or current)  
  Yes 12 0.083 0.042 to 0.124 0.11
  No 43 0.047 0.031 to 0.063 Ref
 Stenosis  
  Yes 6 0.072 0.029 to 0.115 0.23
  No 36 0.044 0.027 to 0.060 Ref
 Arteriosclerosis  
  Yes 20 0.047 0.023 to 0.069 0.97
  No 22 0.048 0.027 to 0.069 Ref
Boldface type indicates statistical significance.

of growth as long as the definition is consistent through-
out the study.2 Additionally, the percentage of aneurysms
that grew each year was generally inversely proportional
to the threshold used to define growth, with the additional
consideration that limitations of imaging resolution make
it difficult to identify enlargement in the shortest studies,
further decreasing this percentage. The outcome of the
sensitivity analysis in the current study was in line with
this, finding that the factors significantly related to an-
eurysm growth (particularly initial size and multiplicity)
were largely unaffected by the threshold used. In compari-
son to our previous study of small aneurysm growth, the
percentage of aneurysms that grew per aneurysm-year was
smaller in the present study, reflecting a longer average
follow-up duration and possibly less restrictive treatment
criteria in the period of 2010–2015.8
In this study, history of TIA was found to be associated
with an increased rate (speed) of UIA growth. Previously,
we found an association of TIA with aneurysm growth
in low-risk patients with multiple, small, asymptomatic
UIAs.8 A limited number of other studies have associated
UIAs with different types of cerebral ischemia; while pos-
sibly related to unusual cerebral hemodynamics, the con-
nection remains poorly defined.7,12,16 In general, risk fac-
tors relevant to UIA growth are not necessarily directly
relevant to rupture and vice versa. Some factors, such as
UIA size, appear to be shared. Others, such as hyperten-
sion, were suggestive of aneurysm growth. Finally, for oth-
ers, such as multiple aneurysms without previous SAH, the
association is only with growth. In general, while all UIAs
likely undergo growth at some point and some studies
have found ongoing growth to be associated with rupture,
8 J Neurosurg  March 1, 2019
enough other factors influence rupture that its timing is
not strongly associated with whether the UIA is detectably
growing.3,22 Additionally, some risk factors for rupture
may change during the course of growth.9
The majority of the research studies for aneurysm
growth have been focused on risk factors related to the
event of aneurysm growth defined based on changes in
size of the UIA from diagnosis to last follow-up.2,21 Few
studies have investigated longitudinal follow-up by inte-
grating patient clinical and epidemiological data to assess
changes in aneurysm size over time, and few studies have
investigated the rate (speed) of aneurysm enlargement.8 In
our previous study of a more restricted set of low-risk an-
eurysm cases (< 7-mm diameter, no history of SAH), we
found multiplicity to be a predictor of aneurysm growth
rate (speed). The current research includes a broader range
of UIA sizes, and the importance of UIA size seems to
outweigh such an effect. In that previous study, male sex
was suggestive of faster aneurysm growth, while in the
current study, this difference was significant.
In this study, the use of continuous data points al-
lowed us to create the PAT model for UIA growth trajec-
tory. During the process of building the multivariate PAT
model, variable selection was driven by a strong interac-
tion between smoking and hypothyroidism, yielding initial
size, age, sex, smoking history, and diagnosis of hypothy-
roidism as relevant parameters. As smoking has been an
important risk factor in many studies, it is our hypothesis
that because smoking is so highly correlated with risk of
cardiovascular disease, it becomes less obvious in the risk
analysis when other factors such as stroke, hypertension,
and coronary disease are included in the analysis.2,14,25,26
 Chien et al.

TABLE 4. Coefficients comprising PAT model to predict growth TABLE 5. Characteristics of aneurysms that ruptured during
trajectory of UIAs follow-up
Variable Coefficient Value
Intercept 0.23 Aneurysm-specific variables*
Initial size (mm) 1.00  Total 11
Age (yrs) −0.005   % growth per AY
Sex (0 = male, 1 = female) 0.16   Mean FU, mos 27.0 ± 21.8
Cigarette smoking (0 = no, 1 = yes) 0.14   Mean initial size, mm 6.7 ± 5.2
Hypothyroidism (0 = no, 1 = yes) −0.23   Initial size, mm
Mos FU 0.034    ≥7 2 (18.2)
Mos FU × initial size −0.005   5–7 5 (45.4)
Mos FU × (initial size – 4.7)* 0.014   3–5 1 (9.1)
Hypothyroidism × cigarette smoking 3.51   <3 3 (27.3)
Hypothyroidism × mos FU 0.041  Location
Hypothyroidism × (mos FU − 24)† −0.066   ICA 4 (36.4)
* Interaction between months follow-up and aneurysms with initial size greater   MCA 1 (9.0)
than 4.7 mm. For aneurysms with an initial size less than 4.7 mm, this term is   ACA 3 (27.3)
0. Otherwise, multiply the coefficient by months of follow-up and the difference   BAVA 3 (27.3)
between initial size (in mm) and 4.7 mm.
† Interaction between hypothyroidism and follow-up greater than 24 months. Patient-specific variables†
For follow-up periods less than 24 months, this term is 0. Otherwise, multiply  Total 8
the coefficient by diagnosis of hypothyroidism (0 = no, 1 = yes) and the differ-   Mean age, yrs 76.8 ± 5.4
ence between follow-up duration (in months) and 24 months.
 Sex
  Female 8 (100)
Thyroid disease is known to affect vessel wall function. 17   Male 0 (0)
A recent study by Atchaneeyasakul et al. suggested that   No. of aneurysms
hypothyroidism may be an independent factor associated   Single 5 (62.5)
with UIAs; we also found hypothyroidism to be an inde-   Multiple 3 (37.5)
pendent factor associated with UIA growth with an im-   Personal history 2 (25)
portant influence on growth trajectory.1 Additional studies
  Family history 0 (0)
will be important to better elucidate how hypothyroidism
impacts the natural history of UIAs.  Seizure 1 (12.5)
The PAT model predicts the rate at which a particu-  TIA 1 (12.5)
lar aneurysm may grow, not the likelihood of growth. In  Stroke 4 (50)
our application of the PAT model to hypothetical patients  Hypertension 7 (87.5)
based on a sample of patients with small aneurysms, we  Dyslipidemia 4 (50)
found that in the majority of examples, predicted UIA
 CAD 2 (25)
growth was difficult to detect with a 1-year follow-up im-
aging study (UIA growth defined by size increase greater   Atrial fibrillation 2 (25)
than 0.5–1.0 mm). For these UIAs, a follow-up > 1 year  Stenosis 1 (12.5)
may be beneficial. However there are high-risk situations,  Arteriosclerosis 8 (100)
such as when patients smoke and have hypothyroidism,  DM 3 (37.5)
in which the UIA might become > 7 mm within 1 year.  Hypothyroidism 4 (50)
This means that the initial 6-month follow-up is critical
  Malignant tumor 3 (37.5)
to identify fast-growing aneurysms, while an additional
18-month follow-up might be beneficial for most patients   Past history of surgery 7 (87.5)
based on our findings.   Excessive alcohol use 1 (12.5)
We believe that this research illustrates the value of   Cigarette smoking 2 (25)
databases of longitudinal follow-up data for UIAs for
* Values are presented as the number of aneurysms (%) unless stated
understanding UIA growth behavior. Although having otherwise.
data from a single center minimizes heterogeneity in the † Values are presented as the number of patients (%) unless stated otherwise.
medical records, validation of these findings using similar
data sets from other centers that follow UIAs would be
extremely valuable. This would also provide opportuni-
ties to further refine and strengthen the PAT model, which mapping aneurysm natural history, we may also estimate
aims to eventually provide clinicians with additional in- the point in time when aneurysm formation began. This
formation regarding UIA growth trajectory that can in- will then help early screening of high-risk aneurysms to
form treatment decisions. Moreover, by more accurately prevent sudden death.4
J Neurosurg  March 1, 2019 9
Chien et al.
FIG. 1. Comparison of observed unruptured aneurysm growth behavior
and predicted growth trajectory using the PAT model. Three UIAs from
different size groups were randomly selected from the clinical record.
PAT evaluation incorporated patient-specific and aneurysm-specific
information. The aneurysm growth observed clinically compared with
that predicted by the PAT model showed good agreement between clini-
cal reports and predictions.
FIG. 2. UIA growth predicted by the PAT model for patients with different
characteristics. Heat map showing predicted aneurysm size at follow-
up for different patient histories of smoking and hypothyroidism. The 4
quadrants of this figure correspond to different patient history combina-
tions of smoking and hypothyroidism and illustrate how these modulate
the effects of initial UIA size and follow-up duration on predicted UIA
size. Note the multiplicative effect of smoking and hypothyroidism on
growth (upper right quadrant), and the attenuation of growth after 24
months in cases with hypothyroidism (right quadrants).
10 J Neurosurg  March 1, 2019
Study Limitations
The mean follow-up duration was significantly different
between growth and stable groups in this study, possibly
because continued monitoring of UIAs may be required
when an aneurysm exhibits growth. Clinical treatment
recommendations likely also created a selection bias as to
which cases were available for inclusion in this study. Pa-
tients with a personal history of SAH, for example, would
likely receive early treatment and not be included. How-
ever, despite changing treatment standards and techniques,
across different studies, reported annual percentage growth
rates for UIAs have remained relatively consistent across
multiple decades. Therefore, while the cases in this study
best reflect the growth behavior of UIAs monitored in the
period after ISUIA findings, they may be a decent repre-
sentation of the broader UIA population as well. The PAT
model generally assumes a linear increase in diameter be-
tween follow-ups, which is unlikely to accurately describe
UIA behavior over long periods of time. The use of piece-
wise linear splines in the PAT indicates slower growth
after 24 months, suggesting that UIAs do not grow in a
linear manner. Given the limited data points available and
the relevant time scale for most UIA follow-up, we believe
that this approximation is reasonable for an initial model of
UIA growth trajectories. Increasing sample size and incor-
porating anatomical and morphological information such
as irregular shape and neck size would create more accu-
rate parameter estimates and improve our understanding of
the UIA growth trajectories. In this research, we utilized
a data set followed with CTA to develop PAT, as it was
the common follow-up modality during our study period.
However, it should be noted that imaging and measurement
variability are potential limitations of the PAT model. Fu-
ture work to verify whether results yielded by PAT trans-
late to MRA follow-up data, which has different imaging
resolution, is essential.
Conclusions
Aneurysm growth is a continuous process, which may
be difficult to detect due to the limitations of current imag-
ing approaches. “Stable” aneurysms may resume a path of
growth that would go undetected during the intervening
time of imaging follow-up, and not all small aneurysms
are without potential for growth and rupture. Our strategy
of studying retrospective data with continuous data analy-
ses allowed for the establishment of a predictive model for
UIA growth trajectory. In this study, we established the
PAT model to better understand UIA growth and improve
prediction of growth behavior. This is the first such at-
tempt to map aneurysm natural history, with the objective
of improving predictions of future growth on an individual
basis and allow for strategic clinical follow-up.
Acknowledgments
This work was supported, in part, by a Brain Aneurysm Foun-
dation research grant and by a UCLA Radiology Exploratory
research grant.
References
  1. Atchaneeyasakul K, Tipirneni A, Zhang T, Khandelwal P,

Ambekar S, Snelling B, et al: Association of hypothyroidism
with unruptured cerebral aneurysms: a case-control study. J
Neurosurg 128:511–514, 2018
  2. Backes D, Rinkel GJ, Laban KG, Algra A, Vergouwen MD:
Patient- and aneurysm-specific risk factors for intracranial
aneurysm growth: a systematic review and meta-analysis.
Stroke 47:951–957, 2016
  3. Backes D, Vergouwen MD, Tiel Groenestege AT, Bor AS,
Velthuis BK, Greving JP, et al: PHASES score for prediction
of intracranial aneurysm growth. Stroke 46:1221–1226, 2015
  4. Black M, Graham DI: Sudden unexplained death in adults
caused by intracranial pathology. J Clin Pathol 55:44–50,
2002
  5. Brinjikji W, Zhu YQ, Lanzino G, Cloft HJ, Murad MH, Wang
Z, et al: Risk factors for growth of intracranial aneurysms: a
systematic review and meta-analysis. AJNR Am J Neurora-
diol 37:615–620, 2016
  6. Burns JD, Huston J III, Layton KF, Piepgras DG, Brown RD
Jr: Intracranial aneurysm enlargement on serial magnetic
resonance angiography: frequency and risk factors. Stroke
40:406–411, 2009
  7. Calviere L, Viguier A, Da Silva NA Jr, Cognard C, Larrue V:
Unruptured intracranial aneurysm as a cause of cerebral is-
chemia. Clin Neurol Neurosurg 113:28–33, 2011
  8. Chien A, Liang F, Sayre J, Salamon N, Villablanca P, Viñuela
F: Enlargement of small, asymptomatic, unruptured intracrani-
al aneurysms in patients with no history of subarachnoid hem-
orrhage: the different factors related to the growth of single
and multiple aneurysms. J Neurosurg 119:190–197, 2013
  9. Chien A, Sayre J: Morphologic and hemodynamic risk fac-
tors in ruptured aneurysms imaged before and after rupture.
AJNR Am J Neuroradiol 35:2130–2135, 2014
10. Etminan N, Beseoglu K, Barrow DL, Bederson J, Brown RD
Jr, Connolly ES Jr, et al: Multidisciplinary consensus on as-
sessment of unruptured intracranial aneurysms: proposal of
an international research group. Stroke 45:1523–1530, 2014
11. Ferns SP, Sprengers ME, van Rooij WJ, van den Berg R,
Velthuis BK, de Kort GA, et al: De novo aneurysm formation
and growth of untreated aneurysms: a 5-year MRA follow-up
in a large cohort of patients with coiled aneurysms and re-
view of the literature. Stroke 42:313–318, 2011
12. Guillon B, Daumas-Duport B, Delaroche O, Warin-Fresse K,
Sévin M, Hérisson F, et al: Cerebral ischemia complicating
intracranial aneurysm: a warning sign of imminent rupture?
AJNR Am J Neuroradiol 32:1862–1865, 2011
13. Igase M, Igase K, Kohara K, Yamashita S, Fujisawa M,
Katagi R, et al: Visit-to-visit variability in systolic blood
pressure is a novel risk factor for the growth of intracranial
aneurysms. Cerebrovasc Dis 36:401–406, 2013
14. Juvela S, Poussa K, Lehto H, Porras M: Natural history of
unruptured intracranial aneurysms: a long-term follow-up
study. Stroke 44:2414–2421, 2013
15. Juvela S, Poussa K, Porras M: Factors affecting formation
and growth of intracranial aneurysms: a long-term follow-up
study. Stroke 32:485–491, 2001
16. Kim JH, Suh SH, Chung J, Oh YJ, Ahn SJ, Lee KY: Preva-
lence and characteristics of unruptured cerebral aneurysms
in ischemic stroke patients. J Stroke 18:321–327, 2016
17. La Vignera S, Condorelli R, Vicari E, Calogero AE: Endo-
thelial dysfunction and subclinical hypothyroidism: a brief
review. J Endocrinol Invest 35:96–103, 2012
18. Matsubara S, Hadeishi H, Suzuki A, Yasui N, Nishimura
H: Incidence and risk factors for the growth of unruptured
cerebral aneurysms: observation using serial computerized
tomography angiography. J Neurosurg 101:908–914, 2004
19. Miyazawa N, Akiyama I, Yamagata Z: Risk factors for
growth of unruptured intracranial aneurysms: follow-up
study by serial 0.5-T magnetic resonance angiography. Neu-
rosurgery 58:1047–1053, 2006
Chien et al.
20. Nael K, Villablanca JP, Saleh R, Pope W, Nael A, Laub
G, et al: Contrast-enhanced MR angiography at 3T in the
evaluation of intracranial aneurysms: a comparison with
time-of-flight MR angiography. AJNR Am J Neuroradiol
27:2118–2121, 2006
21. Serrone JC, Tackla RD, Gozal YM, Hanseman DJ, Go-
gela SL, Vuong SM, et al: Aneurysm growth and de novo
aneurysms during aneurysm surveillance. J Neurosurg
125:1374–1382, 2016
22. Villablanca JP, Duckwiler GR, Jahan R, Tateshima S, Martin
NA, Frazee J, et al: Natural history of asymptomatic un-
ruptured cerebral aneurysms evaluated at CT angiography:
growth and rupture incidence and correlation with epidemio-
logic risk factors. Radiology 269:258–265, 2013
23. Villablanca JP, Hooshi P, Martin N, Jahan R, Duckwiler G,
Lim S, et al: Three-dimensional helical computerized tomog-
raphy angiography in the diagnosis, characterization, and
management of middle cerebral artery aneurysms: compari-
son with conventional angiography and intraoperative find-
ings. J Neurosurg 97:1322–1332, 2002
24. Vlak MH, Algra A, Brandenburg R, Rinkel GJ: Prevalence
of unruptured intracranial aneurysms, with emphasis on sex,
age, comorbidity, country, and time period: a systematic re-
view and meta-analysis. Lancet Neurol 10:626–636, 2011
25. Vlak MH, Rinkel GJ, Greebe P, Algra A: Independent risk
factors for intracranial aneurysms and their joint effect: a
case-control study. Stroke 44:984–987, 2013
26. Wiebers DO, Whisnant JP, Huston J III, Meissner I, Brown
RD Jr, Piepgras DG, et al: Unruptured intracranial aneu-
rysms: natural history, clinical outcome, and risks of surgical
and endovascular treatment. Lancet 362:103–110, 2003
27. Zhang H: Analysis of infant growth curves using multivariate
adaptive splines. Biometrics 55:452–459, 1999
28. Zhang H: Mixed effects multivariate adaptive splines model
for the analysis of longitudinal and growth curve data. Stat
Methods Med Res 13:63–82, 2004
29. Zhang H: Multivariate adaptive splines for analysis of longi-
tudinal data. J Comput Graph Stat 6:74–91, 1997
30. Zhou G, Su M, Yin YL, Li MH: Complications associated
with the use of flow-diverting devices for cerebral aneu-
rysms: a systematic review and meta-analysis. Neurosurg
Focus 42(6):E17, 2017
Disclosures
Dr. Colby: consultant for Stryker Neurovascular, Medtronic, and
MicroVention. Dr. Tateshima: consultant for Medtronic, Stryker,
Cerenovus, and Balt USA.
Author Contributions
Conception and design: Chien, Callender, Hildebrandt. Acquisi-
tion of data: Chien, Yokota. Analysis and interpretation of data:
Chien, Callender, Hildebrandt. Drafting the article: Chien, Callen-
der, Ye, Hildebrandt. Critically revising the article: Chien, Callen-
der, Ye, Hildebrandt. Reviewed submitted version of manuscript:
Chien, Callender, Ye, Hildebrandt. Approved the final version of
the manuscript on behalf of all authors: Chien. Statistical analysis:
Chien, Callender, Ye, Hildebrandt. Administrative/technical/mate-
rial support: Chien, Ye, Hildebrandt. Study supervision: Chien,
Salamon, Colby, Wang, Szeder, Jahan, Tateshima, Villablanca,
Duckwiler, Vinuela, Ye, Hildebrandt.
Correspondence
Aichi Chien: David Geffen School of Medicine at UCLA, Los
Angeles, CA. aichi@ucla.edu.
J Neurosurg  March 1, 2019 11