Cognitive Outcomes Following Contemporary Treatment Without Cranial Irradiation for Childhood Acute Lymphoblastic Leukemia

H. M. Conklin; K. R. Krull; W. E. Reddick; D. Pei; C. Cheng; C. H. Pui

J Natl Cancer Inst. 2012;104(18):1386-1395.

Abstract and Introduction

Abstract

Background Treatment of acute lymphoblastic leukemia (ALL) has included the use of prophylactic cranial irradiation in up to 20% of children with high-risk disease despite known cognitive risks of this treatment modality.
Methods Patients enrolled on the St Jude ALL Total Therapy Study XV, which omitted prophylactic cranial irradiation in all patients, were assessed 120 weeks after completion of consolidation therapy (n = 243) using a comprehensive cognitive battery. χ² analysis was used to compare the percentage of below-average performers among the entire ALL patient group to the expected rate based on the normative sample. Univariate logistic regression was used to estimate the effect of intensity of chemotherapy (treatment arm), age at diagnosis, and sex on the probability of below-average performance. All statistical tests were two-sided.
Results Overall, the ALL group had a statistically significantly higher risk for below-average performance on a measure of sustained attention (67.31% more than 1 SD below the normative mean for omission errors, P < .001) but not on measures of intellectual functioning, academic skills, or memory. Patients given higher intensity chemotherapy were at greater risk for below-average performance compared with those given lower intensity therapy on measures of processing speed (27.14% vs 6.25%, P = .009) and academic abilities (Math Reasoning: 18.60% vs 3.90%, P = .008; Word Reading: 20.00% vs 2.60%, P = .007; Spelling: 27.91% vs 3.90%, P = .001) and had higher parent-reported hyperactivity (23.00% vs 9.84%, P = .018) and learning problems (35.00% vs 16.39%, P = .005). Neither age at diagnosis nor sex was associated with risk for below-average cognitive performance.
Conclusions Omitting cranial irradiation may help preserve global cognitive abilities, but treatment with chemotherapy alone is not without risks. Caregiver education and development of interventions should address both early attention deficits and cognitive late effects.

Introduction

The prognosis for children diagnosed with acute lymphoblastic leukemia (ALL) has improved dramatically, with 5-year event-free survival rates as high as 79% to 82% among patients treated in the 1990s.^[1–3] Higher survival rates have resulted in heightened focus on improving quality of life of survivors by reducing treatment-related late effects, including cognitive deficits.

Historically, treatment of childhood ALL with cranial irradiation has been associated with substantial cognitive morbidity;^[4,5] however, these findings are based on doses generally exceeding those used in modern therapy. When directly compared, treatment with lower dose cranial irradiation (eg, ≤18 Grey [Gy]) used in contemporary clinical trials typically,^[6–9] but not always,^[10] results in worse cognitive outcomes than chemotherapy alone on measures of intellectual function as well as more specific cognitive abilities. There is also some evidence^[11] to suggest that different chemotherapy regimens carry greater cognitive risk than others based on drugs used (eg, triple intrathecal methotrexate, hydrocortisone and cytarabine [ITMHA] vs intrathecal methotrexate alone; dexamethasone vs prednisone) as well as drug dosage and mode of administration (eg, lower dose oral methotrexate vs high-dose intravenous methotrexate [HDMTX]). Partly because of inconsistent reports of cognitive difference between treatment with lower dose cranial irradiation and intensive chemotherapy, and partly because of concern about increased central nervous system relapse, most pediatric collaborative study groups continue to use prophylactic cranial irradiation in their clinical trials in up to 20% of ALL patients.^[12]

The St Jude Total Therapy XV study evaluated whether intensification of systemic drugs that affect control of ALL in the central nervous system, together with optimal intrathecal treatment, would allow for complete omission of prophylactic cranial irradiation without compromising overall survival. The clinical trial resulted in 5-year event-free survival of 85.6% and overall survival of 93.5%.^[13] With additional follow-up, the treatment results remain excellent with a 10-year overall survival rate of 91.0% for all patients, 96.1% for low-risk patients, and 86.0% for standard/high-risk patients. Cognitive outcomes have not been systematically investigated or reported.

Demographic and clinical factors are associated with cognitive late effects in childhood ALL. Intensive treatment has been the most reliable predictor of increased risk.^{[6–9,14–16]} Younger age at treatment has been associated with worse cognitive outcomes, most likely resulting from greater vulnerability of the developing brain to neurotoxic agents.[^[6,16–18] However, it is unclear whether this relationship is specific to children receiving cranial irradiation or also holds true for those receiving chemotherapy alone.^[16] Sex may also be associated with cognitive changes following ALL therapy, with girls more likely to experience deleterious effects.^[19,20] The mechanism for sex-specific risk is not fully understood but may result from differences in cerebral myelination.^[21,22] Furthermore, the literature is divided as to whether increased risk in girls is limited to children receiving cranial irradiation,^[23,24] and sex differences may vary depending on cognitive ability assessed.^[25] Few studies have been able to examine these demographic and clinical risk factors in a large cohort of prospectively studied patients who received homogeneous treatment and had comprehensive cognitive evaluations.

The first objective of this study was to systematically evaluate cognitive outcomes in a radiation-naive sample treated with contemporary risk-adapted therapy. The second objective was to investigate the predictive value of treatment intensity, age at treatment, and sex with respect to cognitive outcomes. Based on the existing literature, our primary hypothesis was that patients would perform well on global measures of cognitive ability, such as intellectual functioning; however, a subset of them would show evidence of difficulties on measures of attention, particularly on performance-based measures of sustained attention including indices of processing efficiency, which may be more sensitive to central nervous system–directed therapy.^{[14,18,25,26]} Secondarily, we predicted that higher treatment intensity and younger age at diagnosis would be risk factors for cognitive problems, whereas female sex would not reliably predict risk when looking across a range of cognitive skills.

1 2 3 4

Next Section

Abstract and Introduction
Patients and Methods
Results
Discussion
References

Table 1. Patient demographic and clinical characteristics*
Characteristics	N (%)	Mean (SD)	Range
Sex
Boys	131 (53.91)
Girls	112 (46.09)
Ethnicity
White	194 (79.84)
African American	39 (16.05)
Other	10 ( 4.12)
Risk arm
Low	126 (51.85)
Standard/high	117 (48.15)
Age at diagnosis, y	243	6.56 (4.39)	1.02–18.73
Baseline IQ (EIQ) †	117	101.42 (15.46)	64.00–142.00
Baseline Bayley MDI‡	33	92.12 (14.85)	50.00–116.00
HDMTX dose, g/m²
Low-risk arm	126	11.68 (2.14)	3.56–18.04
Standard/high-risk arm	117	18.61 (3.64)	7.40–29.30
ITMHA dose, mL
Low-risk arm	126	150.08 (69.75)	94.00–856.00
Standard/high-risk arm	117	204.59 (50.12)	80.00–375.00
Dex dose, mg/m²
Low-risk arm	126	1008.18 (177.96)	175.50–1302.59
Standard/high-risk arm	117	1212.61 (374.11)	60.34–1690.13

* HDMTX = high-dose methotrexate; ITMHA = intrathecal methotrexate, hydrocortisone and ara-C; Dex = dexamethasone.
‡ EIQ = estimated IQ based on the Block Design, Similarities and Information subtests from the age-appropriate Wechsler scale.
‡ MDI = Mental Development Index from the Bayley Scales of Infant Development, Second Edition, administered to children up to three-and-a-half years of age in absence of an age-appropriate Wechsler scale.

Table 2. Subgroup performance relative to normative sample*
Test	Treatment risk arm				Age at diagnosis				Sex
	Low		Standard/high		<5 years		≥5 years		Boys		Girls
	Mean (95% CI)	P†	Mean (95% CI)	P†	Mean (95% CI)	P†	Mean (95% CI)	P†	Mean (95% CI)	P†	Mean (95% CI)	P†
Wechsler, SS‡
FSIQ	98.1 (95.5 to 100.7)	.32	93.5 (90.3 to 96.7)	<.001	94.1 (91.3 to 97.0)	<.001	97.8 (94.8 to 100.7)	.27	95.3 (92.5 to 98.0)	.003	96.7 (93.6 to 99.8)	.12
FFD	98.8 (95.5 to 102.1)	.47	93.3 (89.7 to 96.9)	<.001	93.7 (89.0 to 98.4)	.02	96.9 (94.1 to 99.8)	.12	95.3 (92.1 to 98.5)	.01	96.9 (93.0 to 100.8)	.23
Proc Speed	104.8 (101.1 to 108.4)	.04	95.3 (91.1 to 99.6)	.03	101.5 (96.0 to 107.0)	.59	99.1 (95.7 to 102.6)	.62	97.1 (93.5 to 100.8)	.12	103.8 (99.0 to 108.6)	.23
WIAT, SS
Math	103.4 (100.6 to 106.2)	.017	97.4 (94.0 to 100.8)	.14	100.8 (97.0 to 104.6)	1.00	100.0 (97.2 to 102.9)	1.00	101.1 (98.1 to 104.1)	.96	99.1 (95.7 to 102.6)	.83
Reading	105.5 (102.1 to 108.8)	.005	96.5 (93.2 to 99.8)	.08	102.7 (97.8 to 107.7)	.81	99.9 (97.1 to 102.7)	1.00	99.9 (96.6 to 103.2)	.96	102.0 (98.4 to 105.6)	.83
Spelling	104.3 (101.5 to 107.2)	.007	94.1 (91.0 to 97.2)	<.001	101.4 (97.2 to 105.6)	1.00	97.8 (95.2 to 100.5)	.33	96.9 (94.0 to 99.8)	.11	101.7 (98.3 to 105.2)	.83
CPT, T§
Omissions, %	82.7 (77.9 to 87.5)	<.001	85.1 (81.1 to 89.2)	<.001	81.2 (74.1 to 88.3)	<.001	85.0 (81.6 to 88.3)	<.001	84.0 (80.1 to 87.9)	<.001	84.0 (78.9 to 89.1)	<.001
Hit RT	50.8 (47.7 to 53.8)	.63	47.5 (44.5 to 50.5)	.10	45.4 (41.1 to 49.7)	.038	50.2 (47.7 to 52.7)	.88	49.7 (46.9 to 52.5)	.85	47.9 (44.5 to 51.3)	.22
Variability	58.4 (55.2 to 61.6)	<.001	59.0 (56.0 to 61.9)	<.001	60.3 (57.1 to 63.5)	<.001	58.1 (55.5 to 60.8)	<.001	57.5 (54.8 to 60.2)	<.001	60.4 (56.9 to 64.0)	<.001
d′	59.4 (57.0 to 61.7)	<.001	60.5 (58.2 to 62.8)	<.001	59.4 (56.4 to 62.3)	<.001	60.2 (58.2 to 62.2)	<.001	58.9 (57.2 to 60.7)	<.001	61.5 (58.3 to 64.6)	<.001
β	68.0 (63.7 to 72.3)	<.001	74.9 (70.6 to 79.1)	<.001	63.6 (59.2 to 67.9)	<.001	74.5 (70.9 to 78.2)	<.001	71.6 (68.0 to 75.3)	<.001	71.8 (66.6 to 77.1)	<.001
CVLT, Z§
Total (T)	50.9 (48.4 to 53.3)	.94	49.5 (46.7 to 52.4)	1.00	46.0 (42.6 to 49.4)	.09	51.7 (49.5 to 54.0)	.37	49.2 (46.7 to 51.6)	.83	51.6 (48.6 to 54.6)	.59
Learning Slope	−0.2 (−0.5 to 0.1)	.43	−0.3 (−0.5 to 0.0)	.08	−0.1 (−0.6 to 0.3)	1.00	−0.3 (−0.5 to 0.1)	.013	−0.3 (−0.5 to 0.0)	.11	−0.2 (−0.5 to 0.1)	.37
SD Free Recall	0.2 (−0.0 to 0.4)	.43	−0.1 (−0.3 to 0.2)	1.00	−0.1 (−0.4 to 0.2)	1.00	0.1 (−0.1 to 0.3)	.61	−0.1 (−0.4 to 0.1)	.83	0.3 (0.0 to 0.6)	.14
LD Free Recall	0.1 (−0.2 to 0.4)	.94	−0.1 (−0.4 to 0.2)	1.00	−0.3(−0.8 to 0.1)	.37	0.1 (−0.2 to 0.3)	.61	−0.1 (−0.4 to 0.1)	.83	0.1 (−0.2 to 0.4)	.59
CPRS, T§
Impulse-Hyper	48.6 (46.8 to 50.3)	.22	51.8 (49.5 to 54.1)	.13	51.2 (49.1 to 53.2)	.46	48.8 (46.8 to 50.8)	.47	49.6 (47.5 to 51.6)	.97	50.5 (48.5 to 52.5)	1.00
Hyperactive	48.1 (46.4 to 49.9)	.10	52.8 (50.2 to 55.4)	.08	51.3 (49.2 to 53.4)	.46	49.0 (46.8 to 51.3)	.47	50.8 (48.6 to 52.9)	.97	49.6 (47.4 to 51.8)	1.00
Learning	50.3 (48.2 to 52.3)	.79	56.8 (53.6 to 60.1)	<.001	54.3 (51.5 to 57.1)	.008	52.1 (49.5 to 54.6)	.32	54.1 (51.3 to 56.9)	.01	52.3 (49.8 to 54.7)	.21

* FSIQ = Full Scale Intelligence Quotient; FFD = Freedom from Distractibility Index; WIAT = Wechsler Individual Achievement Test; CPT = Conners' Continuous Performance Test; RT = Reaction Time; CVLT = California Verbal Learning Test; SD = Short Delay; LD = Long Delay; CPRS = Conners' Parent Rating Scale. Proc Speed = processing speed; Impulse-Hyper = Impulsivity–Hyperactivity Index; d′ = vigilance; β = risk taking; SS = standard score.
† P-values are based on a one-sample t-test compared with the published normative mean. They have been adjusted using the Holm.Bonferroni step-down method to account for multiple comparisons and reduce the risk of a Type I error. All tests of statistical significance were two-sided.
‡ Scores are reported as SS, which have a normative mean of 100 and SD of 15; T-scores, which have a normative mean of 50 and SD of 10; and Z-scores, which have a normative mean of 0 and SD of 1.
§ Indices on the CPT and CPRS are reverse cued such that a higher score is indicative of worse performance or greater problems.

Table 3. Subgroup comparison by risk arm, age at diagnosis, and sex*
Test	By treatment risk arm†		By age at diagnosis†		By sex
Test	Difference of means (95% CI)	P‡	Difference of means (95% CI)	P‡	Difference of means (95% CI)	P‡
Wechsler, SS§
FSIQ	4.6 (0.6 to 8.7)	.03	−3.6 (−7.7 to 0.4)	.37	−1.5 (−5.6 to 2.7)	.34
FFD	5.5 (0.7 to 10.4)	.03	−3.3 (−8.6 to 2.1)	.67	−1.6 (−6.6 to 3.4)	.34
Proc Speed	9.4 (3.8 to 15.0)	.003	2.4 (−4.1 to 8.8)	.67	−6.7 (−12.5 to −0.8)	.03
WIAT, SS
Math	6.0 (1.6 to 10.4)	.007	0.7 (−4.1 to 5.6)	1.00	1.9 (−2.6 to 6.5)	.68
Reading	9.0 (4.3 to 13.6)	.001	2.9 (−2.4 to 8.1)	1.00	−2.1 (−7.0 to 2.8)	.68
Spelling	10.2 (6.0 to 14.4)	<.001	3.5 (−1.3 to 8.4)	1.00	−4.8 (−9.3 to −0.3)	.06
CPT, T‖
Omissions (%)	−2.4 (−8.6 to 3.7)	.30	−3.8 (−10.8 to 3.3)	.25	0.0 (−6.2 to 6.3)	.74
Hit RT	3.3 (−1.0 to 7.6)	.80	−4.8 (−9.6 to 0.1)	.11	1.8 (−2.5 to 6.2)	1.00
Variability	−0.5 (−4.9 to 3.8)	.80	2.2 (−2.8 to 7.1)	.35	−2.9 (−7.3 to 1.4)	.63
d′	−1.1 (−4.4 to 2.2)	.80	−0.8 (−4.6 to 3.0)	.52	−2.6 (−5.9 to 0.8)	1.00
β	−6.9 (−12.9 to −0.9)	.15	−11.0 (−17.7 to −4.2)	.025	−0.2 (−6.4 to 6.0)	1.00
CVLT, Z
Total (T)	1.4 (−2.4 to 5.2)	1.00	−5.7 (−9.8 to −1.6)	.006	−2.4 (−6.3 to 1.4)	.30
Learning Slope	0.1 (−0.3 to 0.4)	1.00	0.1 (−0.2 to 0.5)	.49	−0.0 (−0.4 to 0.3)	.91
SD Free Recall	0.3 (−0.1 to 0.6)	1.00	−0.2 (−0.6 to 0.2)	.15	−0.4 (−0.8 to −0.0)	.08
LD Free Recall	0.2 (−0.2 to 0.6)	1.00	−0.4 (−0.9 to 0.0)	.15	−0.3 (−0.7 to 0.1)	.30
CPRS, T
Impulse-Hyper	−3.2 (−6.1 to −0.4)	.05	2.4 (−0.5 to 5.2)	.25	−0.9 (−3.8 to 1.9)	1.00
Hyperactive	−4.7 (−7.7 to −1.6)	.02	2.3 (−0.8 to 5.3)	.24	1.2 (−1.9 to 4.3)	1.00
Learning	−6.6 (−10.2 to −2.9)	.02	2.3 (−1.5 to 6.0)	.37	1.8 (−2.0 to 5.6)	1.00

* FSIQ = Full Scale Intelligence Quotient; FFD = Freedom from Distractibility Index; WIAT = Wechsler Individual Achievement Test; CPT = Conners' Continuous Performance Test; RT = Reaction Time; CVLT = California Verbal Learning Test; SD = Short Delay; LD = Long Delay; CPRS = Conners' Parent Rating Scale; Proc Speed = processing speed; Impulse-Hyper = Impulsivity.Hyperactivity Index; d′ = vigilance; β = risk taking; SS = standard score.
† Treatment risk arm: low vs standard high; age: <5 y vs >5 y.
‡ P-values are based on Wilcoxon rank sum tests comparing subgroups. They have been adjusted using the Holm.Bonferroni step-down method to account for multiple comparisons and reduce the risk of a Type I error. All tests of statistical significance were two-sided.
§ Scores are reported as mean differences based on SS, which have a normative mean of 100 and SD of 15; T-scores, which have a normative mean of 50 and SD of 10; and Z-scores, which have a normative mean of 0 and SD of 1.
| Indices on the CPT and CPRS are reverse cued such that a higher score is indicative of worse performance or greater problems.

Table 4. Multivariable logistic regression of patient and clinical characteristics and cognitive performance*
Test	Age at diagnosis†		Sex		Cumulative dexamethasone†		Cumulative ITMHA†		Cumulative HDMTX†
Test	OR (95% CI)	P‡	OR (95% CI)	P‡	OR (95% CI)	P‡	OR (95% CI)	P‡	OR (95% CI)	P‡
Wechsler, SS
FSIQ	1.09 (0.57 to 2.10)	1.00	0.60 (0.31 to 1.16)	.38	0.98 (0.88 to 1.09)	.68	1.21 (0.95 to 1.53)	.37	1.16 (0.79 to 1.71)	.46
FFD	2.63 (1.04 to 6.65)	.12	0.68 (0.28 to 1.65)	.40	0.94 (0.83 to 1.06)	.57	1.18 (0.73 to 1.91)	.51	1.62 (0.91 to 2.89)	.22
Proc Speed	1.18 (0.36 to 3.85)	1.00	0.45 (0.14 to 1.37)	.38	0.90 (0.78 to 1.03)	.37	1.48 (0.89 to 2.48)	.37	1.84 (0.94 to 3.60)	.22
WIAT, SS
Math	0.56 (0.14 to 2.16)	.80	1.60 (0.55 to 4.65)	.41	0.96 (.83 to 1.10)	.58	1.58 (0.95 to 2.62)	.23	1.44 (0.78 to 2.67)	.74
Reading	0.57 (0.15 to 2.19)	.80	0.40 (0.12 to 1.34)	.41	0.90 (0.78 to 1.04)	.49	1.40 (0.84 to 2.34)	.39	1.31 (0.69 to 2.46)	.74
Spelling	0.44 (0.14 to 1.41)	.51	0.52 (0.20 to 1.38)	.41	0.93 (0.82 to 1.06)	.58	1.19 (0.76 to 1.86	.46	1.37 (0.79 to 2.38)	.74
CPT, T
Omissions, %	1.03 (0.46 to 2.30)	1.00	1.28 (0.62 to 2.65)	1.00	1.05 (0.95 to 1.17)	1.00	1.02 (0.69 to 1.51)	1.00	1.05 (0.67 to 1.65)	1.00
Hit RT	3.91 (1.65 to 9.30)	.01	1.38 (0.62 to 3.07)	1.00	0.94 (0.84 to 1.05)	1.00	1.29 (0.84 to 1.96)	1.00	1.54 (0.94 to 2.52)	.42
Variability	1.78 (0.83 to 3.86)	.56	1.65 (0.83 to 3.27)	.77	0.99 (0.90 to 1.10)	1.00	0.85 (0.58 to 1.24)	1.00	1.29 (0.84 to 2.00)	.75
d′	0.73 (0.34 to 1.59)	1.00	1.09 (0.55 to 2.16)	1.00	0.97 (0.88 to 1.07)	1.00	1.18 (0.82 to 1.71)	1.00	1.03 (0.67 to 1.58)	1.00
β	0.00	1.00	0.00	1.00	0.84 (0.42 to 1.67)	1.00	2.90 (0.25 to 33.88)	1.00	0.05 (0.00 to 4.14)	.72
CVLT, Z
Total (T)	3.02 (1.11 to 8.19)	.09	1.12 (0.42 to 3.02)	1.00	1.06 (0.91 to 1.22)	.88	1.12 (0.66 to 1.90)	1.00	1.78 (1.01 to 3.14)	.19
Learning Slope	1.03 (0.40 to 2.66)	.96	0.75 (0.31 to 1.84)	1.00	1.07 (0.93 to 1.23)	.88	1.00 (0.62 to 1.62)	1.00	0.80 (0.46 to 1.40)	.44
SD Free Recall	1.88 (0.56 to 6.35)	.62	0.48 (0.14 to 1.70)	1.00	0.92 (0.80 to 1.07)	.88	1.25 (0.70 to 2.25)	1.00	1.54 (0.80 to 2.93)	.44
LD Free Recall	4.64 (1.69 to 12.71)	.01	0.98 (0.37 to 2.61)	1.00	0.92 (0.81 to 1.05)	.80	1.38 (0.81 to 2.33)	.94	1.51 (0.87 to 2.64)	.44
CPRS, T
Impulse-Hyper	2.51 (1.13 to 5.60)	.07	0.91 (0.43 to 1.92)	1.00	0.91 (0.79 to 1.04)	.52	0.94 (0.68 to 1.30)	1.00	1.10 (0.71 to 1.71)	.66
Hyperactive	1.69 (0.78 to 3.67)	.37	0.63 (0.28 to 1.37)	.73	1.00 (0.87 to 1.14)	1.00	1.28 (0.98 to 1.69)	.22	1.26 (0.81 to 1.95)	.62
Learning	1.45 (0.76 to 2.75)	.37	0.93 (0.49 to 1.76)	1.00	0.98 (0.88 to 1.10)	1.00	1.04 (0.83 to 1.31)	1.00	1.40 (0.97 to 2.02)	.23

* FSIQ = Full Scale Intelligence Quotient; FFD = Freedom from Distractibility Index; WIAT = Wechsler Individual Achievement Test; CPT = Conners' Continuous Performance Test; RT = Reaction Time; CVLT = California Verbal Learning Test; SD = Short Delay; LD = Long Delay; CPRS = Conners' Parent Rating Scale. Proc Speed = processing speed; Impulse-Hyper = Impulsivity–Hyperactivity Index; d′ = vigilance; β = risk taking; SS = standard score; OR = odds ratio; CI = confidence interval. Odds ratios represent the effects of being <5 years old at the time of diagnosis (referent = ≥5y), female (referent = male), cumulative dexamethasone (per 100 mg/m²), cumulative ITMHA (per 50 mL) and cumulative HDMTX (per 5 g/m²). Dexamethasone, ITMHA, and HDMTX were categorized into therapeutically meaningful units that roughly correspond to 2 wk of dexamethasone treatment, 4.5 ITMHA doses, and 1.2 doses of HDMTX, respectively.
† All clinical factors, age at diagnosis (<5 vs ≥5 y), sex, cumulative dexamethasone (per 100 mg/m²), ITMHA (per 50 mL), and HDMTX (per 5 g/m2), were included in final multiple logistic regressions.
‡ P-values have been adjusted using the Holm.Bonferroni step-down method to account for multiple comparisons and reduce the risk of a Type I error. All tests of statistical significance were two-sided.