Predicting non-suicidal self-injury episodes over a discrete period of time in a sample of women diagnosed with bulimia nervosa: An analysis of self-reported trait and ecological momentary assessment based affective lability and previous suicide attempts

Abstract

Objective

To examine the moderating effect of trait affective lability on the relationship between past suicidal behavior and future nonsuicidal self-injury (NSSI).

Method

127 adult females diagnosed with bulimia nervosa took part in this study. We hypothesized that individuals with greater levels of self-reported trait affective lability and a greater number of past suicide attempts would engage in a greater number of NSSI episodes over the course of two weeks than would individuals lacking elevations in one or both of those variables, controlling for average level of negative affect and affective lability as measured through ecological momentary assessment (EMA).

Results

The two-way interaction of trait affective lability and past suicidal behavior predicted participants’ number of NSSI episodes during the course of the study.

Discussion

Interaction of self-reported trait affective lability and past suicidal behavior may exhibit clinical utility in the prediction of patients’ imminent risk of engaging in NSSI.

Rates of nonsuicidal self-injury (NSSI) are high amongst individuals diagnosed with eating disorders (EDs),^1–2 particularly in the bingeing-purging subtype of anorexia nervosa and bulimia nervosa (BN; for a review see Svirko & Hawton³). The high co-occurrence between NSSI and EDs may be explained by shared etiological and maintenance factors.³ For example, both behaviors have been posited to serve as emotion regulation strategies. Indeed, intrapersonal negative reinforcement is one of the most commonly endorsed reasons for engaging in NSSI in general⁴ and this pattern remains true for ED patients who engage in NSSI, who most frequently endorse the affect regulation function.¹ With this in mind, it is important to recognize that a tendency to experience frequent shifts in emotional intensity and valence (i.e., affective lability) may be an important factor in dysregulated behaviors (e.g., NSSI) in patients with BN. Anestis and colleagues,⁵ for instance, found that affective lability predicted impulsive behaviors (e.g., self-injury, risky sexual behavior) in a sample of women with BN. Research therefore suggests that individuals diagnosed with BN who experience frequently shifting moods are more likely to engage in a broad array of dysregulated behaviors.

Other stable tendencies, such as past suicide attempts, are also associated with NSSI. Individuals with a history of NSSI are more likely to report past suicide attempts⁶ and attempt suicide in the future⁷ relative to individuals with no history of NSSI. A longer history of NSSI is also associated with increased suicide attempts in adolescents,⁶ and individuals with a history of both suicide attempts and NSSI (as compared to suicide attempts alone) exhibit more severe psychopathology and suicidal ideation.⁸ This is of particular concern in BN, given that 25–35% of patients with BN attempt suicide (for a review see Franko & Keel⁹).

Although the above referenced research supports the independent role of suicidal behavior and affective lability in the prediction of NSSI, to our knowledge, no past research has tested the interactive effect of these two variables and, as such, it is unclear to what extent the strength of the relationship between on predictor and NSSI depends upon the presence of the other predictor. The aim of the current study was to determine whether levels of affective lability impact the strength of the relationship between prior suicidal behavior and NSSI episodes over a discrete period of time. Specifically, we hypothesized that, in a sample of women diagnosed with BN, the interaction of self-reported trait affective lability and lifetime number of suicide attempts would predict participants’ number of NSSI episodes during the course of the study.

Method

Participants and procedure

The sample consisted of 127 female participants aged 18 to 55 (mean = 25.34, standard deviation = 7.71), all of whom met criteria for BN. The ethnic composition of this sample was 96.9% Caucasian, 1.5% Native American, 0.8% Asian, and 0.8% other. For a thorough description of the procedure utilized in this study, please see Anestis, Selby, Crosby, Wonderlich, and Engel.¹⁰

Measures

Axis I Psychopathology

The Structured Clinical Interview for DSM-IV Axis I Diagnoses (SCID-I/P)¹¹ is a semi-structured diagnostic interview used to assess DSM-IV-TR Axis I psychopathology. Based on a random selection of 25 interviews, kappa ratings of inter-rater reliability for this sample were 1.00.

Predictor Variables

The Dimensional Assessment of Personality Pathology – Basic Questionnaire (DAPP-BQ)¹² is a self-report measure comprised of 290 items used to assess several components of personality. Only the Affective Lability (I often feel like I am on an emotional roller coaster) subscale (α = .89) was utilized in the subsequent analyses.

The Diagnostic Interview for Borderlines – Revised (DIB-R)¹³ is a semi-structured diagnostic interview that measures borderline personality disorder symptoms. In this study, an item assessing the number of “suicidal gestures or attempts” participants had experienced during their lifetimes was used. Individuals who reported more than one gesture were given a score of 2; one attempt or gesture was scored as a 1; and, not endorsing any suicidal gesture or attempt resulted in a score of 0. In a sample of 25 randomly selected cases, the interclass correlation coefficients (ICC) ranged from .75 to 1.0 for subscales and .98 for the global score.

Outcome variables

Eating Disorder and Self-Destructive Behavior Checklist. Several items from various questionnaires^14–15 were consolidated into a measure of behavioral dysregulation. In this study, a composite NSSI variable was created by summing the total number of times each participant endorsed any of the following behaviors: cutting, burning, repeated hitting, and head banging.

Covariates

The Eating Disorder Examination (EDE)¹⁶ is a semi-structured interview that assesses ED symptomatology. In this sample, a random selection of 25 cases resulted in interclass coefficients ranging from .65 to .98, indicating sufficient inter-rater reliability. Only the global score was utilized.

The Positive Affect – Negative Affect Scale (PANAS)¹⁷ is a 20-item self-report questionnaire utilized to assess the degree to which an individual is currently experiencing positive and negative affect. Items utilize a Likert type scale ranging from 1 (“Very slightly or not at all”) to 5 (“Extremely”). In this study, the PANAS was filled out several times per day over a period of two weeks. Only the Negative Affect subscale was utilized.

The mean square successive difference (MSSD) was used to create a single affective lability coefficient. In the context of this study, this variable represented the degree to which, on average, each participant’s level of negative affect differed from the level that preceded it. The formula for this procedure, which consists of n elements, the ith of which is denoted as x_i is as follows:

$$\frac{\mathit{MSSD}={\sum }_{i=1}^{n-1}{(x_{i+1}-x_i)}^2}{n-1}$$

In addition to calculating affective lability using the PANAS Negative Affect subscale, we also calculated the mean level of negative affect experienced by each participant during the course of the study. The purpose of this calculation was to differentiate between mood variability and intensity.

The Hamilton Depression Rating Scale (HAM-D)¹⁸ is a 21-item semi-structured interview that assesses symptoms of major depressive disorder (MDD). Questions measure the degree to which specific symptoms of MDD (e.g. difficulty with sleep, depressed mood, psychomotor retardation) are present in the interviewee.

Results

Means, standard deviations, and intercorrelations for the variables utilized in these analyses can be found in Table 1.

Table 1.

Means, standard deviations, and intercorrelations for all variables utilized in the analyses

	1	2	3	4	5	6	7
1. Average Negative Affect	-
2. EMA Affective Lability	.16	-
3. Depression	.11	**.30	-
4. Global EDE Score	.13	**.28	**.46	-
5. Previous Suicide Attempts	.05	.14	**.27	.08	-
6. DAPP Affective Lability	**.29	**.27	**.26	**.34	**.24	-
7. Self-Injury Episodes	.03	.16	.14	.13	**.40	*.19	-
Mean	28.65	49.47	18.29	3.33	.31	55.43	.28
SD	18.89	34.84	9.99	1.11	.64	9.10	.97

Interaction of DAPP Affective Lability and Previous Suicide Attempts predicting NSSI Episodes

To test our hypothesis, a hierarchical linear model based on ordinary least squares regression was conducted. All variables utilized in this analysis were centered at their mean. In Step 1, the covariates were entered. In Step 2, the main effects were entered. In Step 3, the product of DAPP Affective Lability and suicide attempts was entered. Number of self-reported NSSI episodes during the study served as the dependent variable. Results indicated that the interaction between DAPP Affective Lability and previous suicide attempts was significantly related to participants’ number of NSSI episodes during the course of the study (t = 4.01, p < .001, sr = .31, f² = .48; See Table 2 and Figure 1).

Table 2.

The interaction of DAPP affective lability and previous suicide attempts predicting number of NSSI episodes, controlling for average levels of negative affect, EMA affective lability, depression, and global EDE scores

		R2 for Set	T-Value	P-Value	Correlations
					Zero-order	Partial	Part
1	(Constant)	.025	−.259	.796
	Average Negative Affect		.077	.939	.035	.007	.007
	EMA Affective Lability		.816	.416	.116	.075	.075
	Depression		.290	.773	.099	.027	.026
	Global EDE Score		.890	.375	.130	.082	.081
2	(Constant)	.249	−.358	.721
	Average Negative Affect		.000	.999	.035	.000	.000
	EMA Affective Lability		.367	.715	.116	.034	.030
	Depression		−1.040	.301	.099	−.096	−.084
	Global EDE Score		1.186	.238	.130	.110	.096
	Suicide Attempts		5.665	.000	.481	.467	.458
	DAPP Affective Lability		.507	.613	.170	.047	.041
3	(Constant)	.342	−1.269	.207
	Average Negative Affect		.207	.836	.035	.019	.016
	EMA Affective Lability		.129	.897	.116	.012	.010
	Depression		−1.155	.250	.099	−.108	−.088
	Global EDE Score		1.178	.241	.130	.110	.090
	Suicide Attempts		4.071	.000	.481	.356	.309
	DAPP Affective Lability		.951	.344	.170	.089	.072
	Suicide Attempts * DAPP Affective Lability		4.008	.000	.460	.351	.305

Dependent Variable: Number of Self-Injury Episodes

Figure 1.

Interaction of trait affective lability and lifetime number of suicide attempts predicting non-suicidal self-injury episodes during the course of the study

Note: Values in graph represent variables centered at their respective means

Discussion

The central aim of this study was to examine whether trait affective lability and lifetime number of suicide attempts would interact to predict the number of episodes of NSSI that occurred over a two week period in a sample of women diagnosed with BN. Results supported the hypothesis that higher levels of self-reported affective lability and previous suicidal behavior would be associated with a greater amount of self-injury. Specifically, in the present sample of women diagnosed with BN, the strength of the relationship between prior suicidal behavior and future NSSI was dependent upon an individual’s level of affective lability. These results are consistent with past research indicating that trait affective lability is associated with a greater number of dysregulated behaviors in women diagnosed with BN,⁵ as well as research indicating that a past history of suicidal behavior is predictive of NSSI,⁶ while extending the existing literature by exploring the interactive effect of these two constructs.

It is important to note a number of limitations to these findings. First, our entire sample was comprised of women diagnosed with BN and, as such, the generalizability of the findings is unclear. Additionally, our item measuring past suicidal behavior asked participants to indicate their past number of “suicide gestures or attempts,” meaning that this variable was potentially comprised of somewhat varied behaviors. Future studies utilizing a suicidal behavior measure that more clearly defines suicidality is recommended. Finally, our EMA data were analyzed in an aggregate fashion, and, as such, statistical methods that capitalize on the full potential of the data were not available options.

Despite these limitations, we believe that this study offers significant insights into an important topic. These findings indicate that risk for NSSI can be effectively evaluated utilizing self-report measures of stable traits and that such measures offer incremental predictive utility above and beyond that of more complex EMA methodology within a sample of women diagnosed with BN. Clinically, these findings speak to the importance of considering multiple variables simultaneously when assessing risk for NSSI in the immediate future. Although the associations between both affective lability and past suicidal behavior with future NSSI are well established and make intuitive sense, empirical evaluations of the interactive effect of these variables provide a stronger foundation upon which to base risk assessments and allow for a more precise understanding of vulnerabilities. In this sense, although these findings do not represent a paradigm shift that counters previous findings, they represent a step towards adding specificity to the conceptualization of the associations between well-established risk factors for NSSI and actual NSSI episodes in discrete periods of time.