Impact of Symptom Focusing and Somatosensory Amplification on Persistent Physical Symptoms: A Three-Year Follow-up Study

Abstract

Objective

The somatosensory amplification theory considers symptom focusing and somatosensory amplification as important perpetuating factors of persistent physical symptoms. We investigated whether symptom focusing and somatosensory amplification were associated with symptom severity and mental and physical functioning over a three-year period in patients with persistent physical symptoms (PPS).

Methods

Baseline, 6-, 12-, 24- and 36-months follow-up data from the PROSPECTS study, a prospective cohort consisting of 325 patients with PPS, were used. We applied longitudinal mixed model analyses to investigate if symptom focusing (CBRQ Symptom Focusing Subscale) and somatosensory amplification (Somatosensory Amplification Scale) at baseline were associated with symptom severity (PHQ-15), mental and physical functioning (RAND-36 MCS and PCS) over three years, using all measurements.

Results

Symptom focusing was associated with increased symptom severity and lower mental and physical functioning over time. Somatosensory amplification at baseline was associated with increased symptom severity and lower mental and physical functioning over time. Effect sizes were small. Associations with baseline symptom focusing decreased over time, associations with baseline somatosensory amplification were more stable. There was no interaction effect of both constructs, but they partly overlapped.

Conclusion

This is the first study to show that over an extended period, symptom focusing and somatosensory amplification are associated with symptom severity and lower mental and physical functioning in patients with PPS. These results support the impact of both symptom focusing and somatosensory amplification on the perpetuation of symptoms and lowered mental and physical functioning in individuals with PPS.

1. Introduction

Patients with physical symptoms not attributable to verifiable, conventionally defined diseases are common in all medical settings. When these symptoms persist, they can have a severe impact on patients’ wellbeing and functional health. Apart from the personal suffering they are causing, persistent physical symptoms (PPS) put an economic burden to society due to high medical care utilization and productivity loss [1–3].

To explain the mechanisms involved in the development and perpetuation of PPS, several models are available [4]. The somatosensory amplification theory provides a theoretical model for PPS. It is based on the assumption that individuals with PPS may experience a range of symptoms as more intense, more damaging and more alarming than most people do. The theory highlights the importance of both symptom focusing and symptom amplification in reinforcing symptoms. Elevated attention focusing is assumed to turn bodily sensations into symptoms and amplify the perception of mild somatic sensations and symptoms, creating a self-reinforcing circle [5,6]. Somatosensory amplification can be considered as a predisposing as well as a perpetuating factor for PPS.

The somatosensory amplification theory was initially formulated for and investigated in patients with hypochondriasis. Barsky et al. found that elevated levels of somatosensory amplification correlated strongly with hypochondriasis in general medical outpatients [5]. Barsky et al. later reported that somatosensory amplification might relate to a more general process of somatic symptom report and lower health appraisal [6,7]. Indeed, elevated levels of somatosensory amplification have been demonstrated in individuals with high symptom reporting in cross-sectional studies in various study populations [8–10]. A recent cross-sectional study in individuals with PPS found that elevated levels of somatosensory amplification were associated with a higher number of reported somatic symptoms [11].

We are aware of only a few longitudinal studies looking at somatosensory amplification as a predisposing factor of subjective symptoms and these yielded inconclusive results [12–14]. To the best of our knowledge, there are no longitudinal studies investigating the role of somatosensory amplification as a perpetuating factor of symptoms and mental and physical functioning in individuals with PPS.

The aim of this study was to investigate whether the somatosensory amplification theory provides an explanation for the perpetuation of symptoms and lowered mental and physical functioning in patients with PPS. We hypothesized that in patients with PPS, a stronger tendency to focus on symptoms and higher levels of somatosensory amplification at baseline would be associated with (1) more severe symptoms and (2) lower ratings of mental and physical functioning over time. As we expected symptom focusing and somatosensory amplification not to remain entirely stable over an extended time period, we expected the associations to be strongest at baseline and decrease gradually over the course of time. We hypothesized that symptom focusing and somatosensory amplification would overlap and strengthen each other’s association with the outcomes. We additionally expected that changes in symptom focusing would be related to changes in symptom severity and mental and physical functioning within subjects.

2. Methods

2.1. Study design and subjects

In this study, we used data of the PROSPECTS study [15, 16]. This is an on-going prospective cohort study, following patients with PPS in multiple health care settings. Participants were recruited in general practices and in specialized programmes for patients with PPS of secondary and tertiary care organizations across the Netherlands. Participating patients were between 18 and 70 years old at baseline. For the present study, we used measurements collected at baseline and after 6-, 12-, 24- and 36 months of follow-up. Data were collected between September 2013 and April 2018.

PPS were defined as the presence of physical symptoms, which had lasted at least several weeks and for which no sufficient explanation was found after proper medical examination by a physician. This is in line with the Dutch multidisciplinary and general practice guidelines for Medically Unexplained Symptoms (MUS) [17, 18]. Details on the recruitment and inclusion process have been published elsewhere [15, 16]. The Medical Ethics Committee of VU University Medical Center Amsterdam approved the study protocol. All participants provided written informed consent.

2.2. Measures

At baseline, patients answered questions about personal characteristics (general, socioeconomic and medical characteristics e.g. number and nature of somatic diseases) and a set of validated and widely used questionnaires assessing outcome measures and relevant covariates. An overview of questionnaires and time point of administration is given in table 1.

Table 1. Time points of administration.

	Instrument	Baseline (T0)	6mo (T1)	12mo (T2)	24mo (T3)	36mo (T4)
General characteristics
Patient characteristics (Demographics)	Questionnaire	X
History and chronic medical conditions	Questionnaire	X
Number of symptoms	Physical Symptoms Questionnaire [19]	X
Outcome measures
Severity of symptoms	PHQ-15 [20]	X	X	X	X	X
Mental and physical functioning	RAND-36 MCS and PCS [21, 22]	X	X	X	X	X
Independent variables
Symptom focusing	CBRQ Symptom Focusing subscale [25, 26]	X	X	X	X	X
Somatosensory amplification	SSAS [5, 23, 24]	X

PHQ: Patient Health Questionnaire; MCS: Mental Component Summary; PCS Physical Component Summary; CBRQ: Cognitive and Behavioural Responses to Symptoms Questionnaire; SSAS: Somatosensory Amplification Scale

2.2.1. Outcome measures

Symptom severity

The 15-item PHQ-15 [20] was used as an indicator of the severity of PPS and was completed at all measurement moments. The PHQ-15 has been shown to be one of the most reliable measures of physical symptom reporting in this field of research [27, 28]. Each item describes a common physical symptom (e.g. stomach pain); respondents rate the extent to which each symptom has bothered them in the last four weeks on a three-point Likert scale ranging from 0 (‘not bothered at all’) to 2 (‘bothered a lot’). By adding the scores a total score (0-30) is obtained.

Mental and physical functioning

RAND-36 scores (Mental Component Summary (MCS) and Physical Component Summary (PCS) were used as indicators of mental and physical functioning [21] at all measurement moments. The RAND-36 questionnaire is a 36-item self-report survey of patient health in the preceding month, where a higher score indicates better functioning. It was selected because of its validity and previous use in this area of research. We used version 2.0 of the RAND-36 questionnaire. Mental Component Score RAND-36 MCS and Physical Component Score (RAND-36 PCS) were calculated to summarize Mental and physical functioning in 0-100 scale scores.

2.2.2. Independent variables

Symptom focusing

We used the Cognitive and Behavioural Responses to Symptoms Questionnaire (CBRQ) Symptom Focusing subscale [25, 26] to assess the extent to which patients focused on their symptoms and administered this at all measurement moments. This self-rated questionnaire contains six items measuring attentional focus towards symptoms. Each item is measured on a five-point Likert scale, scored from 0 (‘strongly disagree’) to 4 (‘strongly agree’). The symptom focusing items are added up to a subscale score (0-24), where a higher subscale score indicates more symptom focusing. The CBRQ, including this subscale, showed adequate reliability and validity [25].

Somatosensory amplification

We used the Somatosensory Amplification Scale (SSAS) to measure the extent to which patients experienced somatosensory amplification [5, 24]. The SSAS is a self-report measure and was administered at baseline only (see Table 1). The SSAS evaluates how much patients are bothered by bodily sensations and environmental stimuli, which are unpleasant and disturbing but in general non-pathological [5]. This self-report questionnaire consists of ten items that are scored on a five-point Likert scale ranging from 1 (‘not at all’) to 5 (‘extremely’). By adding the scores a total amplification score (10-50) is obtained. The SSAS has demonstrated adequate internal consistency and test-retest reliability [5, 23, 24].

2.3. Statistical analysis

Descriptive statistics are presented as mean with standard deviation for normally distributed continuous data, median and inter-quartile range for skewed continuous variables, and as numbers and percentages for dichotomous and categorical variables. We report Cronbach’s alphas of the baseline SSAS and the CBRQ Symptom Focusing subscale.

We applied longitudinal mixed model analyses to investigate the associations of the independent variables (symptom focusing, somatosensory amplification) at baseline with the outcomes (symptom severity, mental and physical functioning) at all time-points. We performed mixed model analyses with a random intercept and an additional adjustment for the residual covariance matrix. For the latter different structures (i.e. various banded Toeplitz structures and unstructured) were considered. The choice for the best fitting model was based on the lowest value for the Bayesian Information Criterion (BIC). The overall best fitting approach was applied to all analyses. All outcomes were visually evaluated for normality. We decided not to use imputation techniques to handle missing data in our outcome measures and independent variables. An important reason to abstain from imputation techniques, is the fact that multiple imputations do not increase precision in the estimated rate of change in the end point in longitudinal mixed model analysis [29–31]. Both mixed model analyses and multiple imputation techniques assume missing data to be ‘missing at random’. Moreover, missing data were rare (Table 2).

Table 2. Socio-demographic and medical characteristics at baseline.

	Study population (N = 325) Mean (SD) Number (%)	N missing (%)
Age (range 19-70)	46.5 (12.3)	0 (0%)
Female gender	244 (75.1%)	0 (0%)
Nationality		0 (0%)
Dutch	279 (85.8%)
Other	46 (14.2%)
Education level		0 (0%)
Lower education	109 (33.5%)
Intermediate education	126 (38.8%)
Higher education	90 (27.7%)
Work status		0 (0%)
Employed	187 (57.5%)
Unemployed	26 (8.0%)
Long term sick leave	93 (28.6%)
Retired	19 (5.8%)
Marital status		0 (0%)
Married or cohabiting	199 (61.2%)
Unmarried	85 (26.2%)
Divorced	30 (9.2%)
Widow	11 (3.4%)
Number of self-reported somatic comorbidities		2 (0.6%)
None	81 (24.9%)
1	91 (28.2%)
2	70 (21.7%)
≥3	81 (25.1%)
Symptom severity (PHQ-15 score, scale 0-30)	12.3 (5.3)	0 (0%)
Physical and mental functioning* (RAND-36 score, scale 0-100)
Mental Component Score (MCS)	52.2 (20.3)	2 (0.6%)
Physical Component Score (PCS)	47.3 (20.0)	2 (0.6%)
Symptom Focusing (CBRQ Symptom Focusing score, scale 0-24)	10.6 (5.2)	0 (0%)
Somatosensory amplification scale (SSAS, scale 10-50)	27.2 (6.2)	1 (0.3%)

^*in earlier publications of the PROSPECTS study, the calculation of MCS and PCS was based on version 1.0 of the RAND-36 questionnaire, resulting in small

differences in reported MCS and PCS. Correlations between both calculations (RAND-36 version 1.0 and 2.0) were high (Pearson’s r 0.99 for both MCS and PCS)

All results are presented as unstandardized regression coefficients, with 95% confidence intervals (CI) and the p-values. We present the results of the crude models and models adjusted for demographic variables (age, gender, education level) and somatic diseases (number of self-reported somatic diseases at baseline). Cohen’s d was calculated and reported to give an impression of the effect sizes of the associations. Because female gender [33, 34] and number of somatic diseases [35, 36] have been associated with persistence of symptoms and may be related to differences in symptom focusing and somatosensory amplification, we tested these variables as possible effect modifiers in all analyses. Effect modification was tested by using interaction terms. A significant interaction term (p < 0.05) was deemed to suggest effect modification.

In addition to the overall associations, we analyzed the interaction with time as a categorical variable, i.e. by creating four dummy variables (for 6, 12, 24 and 36 months follow-up) and adding these to the model. The aim was to evaluate changes in the temporal associations between the independent variables at baseline and the outcomes over time. To see how both constructs were related to each other, we calculated the intercorrelation (Pearsons r) of symptom focusing and somatosensory amplification and tested both independent variables separately as well as together in a mutually adjusted model. Additionally, we tested whether there was effect modification between both constructs in the studied associations by using interaction terms, effect modification was suggested if the interaction term was significant (p<0.05).

Longitudinal mixed model analysis provides a regression coefficient for each independent variable that has a combined between subjects (i.e., cross-sectional) and within subjects (i.e., longitudinal) interpretation [29]. If the independent variable is measured repeatedly, one can disentangle the within subjects and between subjects components, by performing a hybrid model analysis. This analysis obtains the between subject part of the relationship using the individual mean values over time, and the within subject part using the deviation scores (i.e., the residual scores after subtraction of the mean values) [32]. Because symptom focusing was time-dependent and measured repeatedly, we performed a hybrid model analysis, with symptom focusing as an independent variable. As somatosensory amplification is deemed to be stable over time, it was only measured at baseline. Therefore, we were not able to perform a hybrid model analysis for somatosensory amplification. We used Stata version 14 for all analyses.

3. Results

3.1. Patient characteristics

Data on the recruitment process of the PROSPECTS study have been published elsewhere [16]. 218 patients (67.1%) were recruited from primary care settings, 107 (32.9%) from secondary or tertiary care settings. There were no signs of selection bias at baseline [16]. In total, 325 patients completed the questionnaires at baseline, 306 at 6-months follow up, 290 at 12-months follow up, 283 at 24-months follow up and 280 at 36-months follow up. Less than 15% of patients was lost to follow up. Socio-demographic and medical characteristics at baseline are provided in table 2. Most patients suffered from multiple physical symptoms, originating from various organ systems. Based on the Physical Symptom Questionnaire [19] general fatigue (62%, N=201), back pain (54%, N=175) and pain in the extremities (52%, N=169) were the most frequently reported symptoms. All outcome variables (PHQ-15, RAND-36 MCS and PCS) showed normal distribution patterns. Mixed model analyses with a random intercept and an additional adjustment for the residual covariance matrix with a 2-banded Toeplitz structure showed the best model fit, based on the BIC. This model was therefore applied in all analyses.

3.2. Symptom focusing

3.2.1. Symptom severity

The CBRQ Symptom Focusing subscale had a Cronbach’s alpha of 0.89 in our population. We found a positive association between symptom focusing at baseline and symptom severity over time, which became slightly stronger after adjusting for confounders (table 3). One scale point difference in symptom focusing related to 0.104 scale point difference in symptom severity over time within subjects, as well as between subjects. Effect size was small (Cohen’s d of 0.10).

Table 3. Longitudinal analyses: associations of symptom focusing and somatosensory amplification at baseline (independent variables) with symptom severity, mental and physical functioning at all time points (outcomes).

Symptom severity (PHQ-15, 0-30)	Crude model			Adjusted model*			Mutually adjusted model**
	Regression coefficient [Cohen’s d]	CI	p-value	Regression coefficient [Cohen’s d]	CI	p-value	Regression coefficient [Cohen’s d]	CI	p-value
Symptom focusing (CBRQ subscale, 0-24)	0.094 [0.09]	-0.002 to 0.191	0.056	0.104 [0.10]	0.018 to 0.191	0.018	0.005 [0.01]	-0.084 to 0.094	0.915
Somatosensory amplification (SSAS, 10-50)	0.248 [0.29]	0.171 to 0.325	<0.001	0.222 [0.26]	0.153 to 0.292	<0.001	0.221 [0.26]	0.146 to 0.296	<0.001
Mental functioning (RAND-36 MCS, 0-100)
Symptom focusing (CBRQ subscale, 0-24)	-1.209 [-0.31]	-1.547 to -0.870	<0.001	-1.222 [-0.32]	-1.554 to -0.889	<0.001	-0.950 [-0.25]	-1.299 to -0.601	<0.001
Somatosensory amplification (SSAS, 10-50)	-0.909 [-0.28]	-1.196 to -0.622	<0.001	-0.910 [-0.28]	-1.194 to -0.626	<0.001	-0.605 [-0.19]	-0.900 to -0.311	<0.001
Physical functioning (RAND-36 PCS, 0-100)
Symptom focusing (CBRQ subscale, 0-24)	-0.690 [-0.17]	-1.076 to -0.304	<0.001	-0.706 [-0.17]	-1.051 to -0.361	<0.001	-0.670 [-0.16]	-1.041 to -0.299	<0.001
Somatosensory amplification (SSAS, 10-50)	-0.350 [-0.10]	-0.678 to -0.023	0.036	-0.301 [-0.09]	-0.596 to -0.007	0.045	-0.085 [-0.02]	-0.399 to 0.228	0.593

95% confidence intervals (CI);

^*Adjusted for demographic variables (educational level, age and gender) and number of somatic diseases at baseline;

^**Adjusted for both independent variables (Symptom focusing and Somatosensory amplification), demographic variables (educational level, age and gender) and number of somatic diseases at baseline

Gender turned out to be an effect modifier of the association. In men the association was statistically significant (regression coefficient crude model: 0.278, p< 0.001), in women not (regression coefficient crude model: 0.071, p=0.228). There was no effect modification by number of somatic diseases. There was a significant interaction with time: the association was strongest at baseline and decreased over time (adjusted model, appendix table A.1).

The intercorrelation (Pearson’s r) of somatosensory amplification and symptom focusing at baseline was 0.358 (R² 0.128). The effect of symptom focusing disappeared in the mutually adjusted model, there was no effect modification by somatosensory amplification.

In the hybrid model analyses we disentangled the within and between subject components of the longitudinal associations. The hybrid model analyses showed that both the within subject and between subject components were statistically significant; the within subject and between subject components were equally strong (table 4). Again, gender was an effect modifier: in men the association was stronger (regression coefficient crude model: between subject effect 0.504, p<0.001; within subject effect: 0.179, p=0.001) than in women (regression coefficient crude model: between subject effect 0.156, p=0.019; within subject effect: 0.126, p<0.001).

Table 4. Longitudinal analyses: within and between subject associations of symptom focusing and symptom severity, mental and physical functioning; results of hybrid model analyses.

Symptom severity (PhQ-15, 0-30)
	Regression coefficient [Cohen’s d]	CI	p-value
Mixed model Symptom focusing	0.149 [0.15]	0.103 to 0.196	<0.001
Within subject Symptom focusing	0.138 [0.14]	0.087 to 0.190	<0.001
Between subject Symptom focusing	0.198 [0.20]	0.087 to 0.308	<0.001
Mental functioning (RAND-36 MCS, 0-100 scale)
Mixed model Symptom focusing	-1.237 [-0.32]	-1.415 to -1.059	<0.001
Within subject Symptom focusing	-1.034 [-0.27]	-1.236 to -0.832	<0.001
Between subject Symptom focusing	-1.894 [-0.50]	-2.259 to -1.528	<0.001
Physical functioning (RAND-36 PCS, 0-100 scale)
Mixed model Symptom focusing	-1.072 [-0.27]	-1.254 to -0.891	<0.001
Within subject Symptom focusing	-1.050 [-0.26]	-1.249 to -0.850	<0.001
Between subject Symptom focusing	-1.176 [-0.29]	-1.611 to -0.742	<0.001

95% confidence intervals (CI)

3.2.2. Mental and physical functioning

The longitudinal mixed model analyses showed a significant negative association between symptom focusing at baseline and mental and physical functioning over time in both the crude and the adjusted models (table 3). This means that symptom focusing at baseline was associated with lower mental and physical functioning over time. Effect sizes again were limited.

There was no effect modification by gender or number of somatic diseases. For the association with mental functioning, there was no interaction with time, although the strength of the association showed some decrease (appendix table A.2). For the association with physical functioning, there was an interaction with time, the association showed decrease in strength over time (appendix table A.3).

In the mutually adjusted models, the effects of symptom focusing remained and there was no effect modification by somatosensory amplification.

The hybrid model analysis revealed that the within and between subject part of the relationship of symptom focusing and mental and physical functioning were almost equally strong (table 4).

3.3. Somatosensory amplification

3.3.1. Symptom severity

The somatosensory amplification scale (SSAS) had a Cronbach’s alpha of 0.73 in our population. The longitudinal mixed model analyses revealed a significant association of somatosensory amplification at baseline with symptom severity over time, both in the crude as well as adjusted analyses (table 3). Effect sizes were small.

There was no effect modification by gender or by number of somatic diseases. There was no interaction with time, which indicates that the associations remained more or less stable over time (appendix table A.1).

The effect remained stable in the mutually adjusted model, there was no significant interaction with symptom focusing.

3.3.2. Mental and physical functioning

Longitudinal analyses showed a negative association between somatosensory amplification (SSAS) at baseline and mental and physical functioning (RAND-36 MCS and PCS) over time in both crude and adjusted analyses (table 3). This means that higher baseline scores on somatosensory amplification were associated with lower mental and physical functioning over time, although effect sizes were limited.

There was no effect modification by symptom focusing, gender or by number of somatic diseases. There was no interaction with time for mental functioning, indicating that the associations remained stable over time (appendix table A.2). There was an interaction effect with time for the associations of somatosensory amplification and physical functioning: the association increased over time during the first two-years of follow up, and subsequently decreased (appendix table A.3).

In the mutually adjusted models, the association with mental functioning was attenuated, and the association with physical functioning disappeared. There was no significant interaction with symptom focusing.

4. Discussion

We found that symptom focusing and somatosensory amplification at baseline were associated with symptom severity and lower mental and physical functioning over a three year time period when applying longitudinal mixed model analysis. Although the associations were statistically significant, effect sizes were small. In men, the associations between symptom focusing and symptom severity were stronger than in women. The associations with baseline symptom focusing decreased over the course of time. For somatosensory amplification the associations remained relatively stable over time. Symptom focusing and somatosensory amplification did not interact and strengthen each other’s effect on the studied associations, but the mutually adjusted models showed there was some overlap and they attenuated each other’s effects. The hybrid model analysis showed that both within subjects and between subjects, symptom focusing was associated with the outcomes.

The results of the hybrid model analysis are particularly valuable, because they show that within subjects, symptom focusing plays a role in the perpetuation of symptoms and mental and physical functioning. The within subject evidence from the hybrid models comes closer to evidence of a causal relationship than the mixed within and between subjects evidence from the mixed models. Thus the fact that the longitudinal relationship between symptom focusing on the one hand and symptom severity and mental and physical functioning on the other is more or less equally strong within subjects as between subjects adds substantially to our knowledge. However, as effect sizes were limited, other mechanisms are likely to be involved as well.

The strengths of the associations with symptom severity and physical functioning differ for each construct, raising questions about if and how related symptom focusing and somatosensory amplification actually are. With respect to mental functioning, both symptom focusing and somatosensory amplification retained some unique effects apart from their common (overlapping) effects. They did not strengthen each other’s effect on symptom severity, as suggested in the somatosensory amplification theory. The associations of symptom focusing at baseline with our outcomes showed a decrease over time. On the other hand, somatosensory amplification at baseline showed remarkably stable associations with symptom severity and functioning over time. We can only speculate about the possible reasons. Barsky first described somatosensory amplification as a temporally stable, trait-like feature, as some individuals are more sensitive to and bothered by bodily sensations and environmental stimuli [37]. Later on he elaborated on somatosensory amplification as also having state-like properties, as the same individual may perceive the same sensations differently on different occasions, influenced by e.g. attention and mood [6]. The somatosensory amplification scale seemed to measure a time-stable feature in association with symptom severity and mental and physical functioning, whereas symptom focusing represented more of a temporary phenomenon in association with the outcomes. Correlations over time (Pearson’s r 0.6 - 0.7) suggested symptom focusing was a state-like feature in this cohort of patients with PPS. Because of the single measurement of somatosensory amplification we do not know whether somatosensory amplification captured a trait-like feature in this population.

Although female gender is more often associated with persistence of symptoms [33, 34], we found that male gender showed a stronger association of symptom focusing with symptom severity over time. From pain research we know that there are gender differences in pain mechanisms, both relating to physiological as well as behavioural psychological responses [38]. Possibly, in individuals with PPS, gender differences in perpetuating mechanisms may exist as well. In specific functional somatic syndromes, like irritable bowel syndrome, gender differences have also been shown, although the mechanisms remain largely unknown [39].

Controversies and debate

Forty years ago, Barsky laid the foundation of the somatosensory amplification theory by describing the proneness to focus on and amplify bodily sensations and symptoms as an explanation for individual differences in experienced physical symptoms [6,37]. Ever since its introduction, the somatosensory amplification theory received much attention as an explanation for the occurrence and persistence of physical symptoms.

Although in the original theory by Barsky [5] symptom focusing and amplification in response to sensory input is assumed, there is ongoing debate about the actual role of body- and illness related sensory stimuli in PPS. Although studies frequently found an association between self-reported somatosensory amplification and physical symptom reports, the evidence from more objective measures in experimental psychological studies has been much less consistent. A recent review by Van den Bergh et al. [40] showed that the current studies do not provide convincing evidence for an attentional bias towards actual peripheral somatosensory input in patients with PPS, and that there is growing evidence that suggests a lower correspondence between peripheral somatosensory input and symptom reports than expected.

A study on the latent structure of somatic symptom distress showed that somatosensory amplification was in particular associated with the general somatic symptom factor, whereas associations with specific symptoms factors (gastrointestinal, fatigue, cardio-pulmonary, and pain symptoms) were small. As such, somatosensory amplification seems to relate specifically to cognitive affective aspects of symptom perception and not to its sensory aspects [41]. A cross-sectional study showed that body awareness was related to somatosensory amplification [42] and the authors suggested that the ‘hypervigilance’ component might be more closely connected to ‘body awareness’ and less to ‘interoceptive accuracy’ than stated in the original description by Barsky [5]. This would put more emphasis on symptom perception instead of attentional bias towards visceral or proprioceptive stimuli as originally postulated. Köteles and Witthöft suggested a new evolutionary approach to the original construct of somatosensory amplification. They suggested that the construct might refer to a somatic threat sensitivity with heightened levels of perceived threat and vulnerability, instead of hypervigilance to sensory stimuli [43]. In our study we did not measure interoceptive accuracy in response to actual peripheral sensory input, but focused on the symptom perception perspective.

There has been some debate on the construct validity of the somatosensory amplification scale (SSAS). Although the scale intended to measure hypervigilance to mild somatic and visceral sensations and a tendency to interpret them as pathological, Aronson et al. proposed that SSAS is rather an index of negative emotionality and general distress [8, 44]. Prior studies suggested that the association of somatosensory amplification and somatic symptoms at least partly overlaps with factors such as negative affect and anxiety [9, 10, 45]. It has been shown that negative mood and self-focused attention facilitated the experience and accessibility of physical symptoms [46]. Taken together, negative affects are likely to play an important role in symptom experience, but the extent to which symptom focusing and somatosensory amplification are associated with/ mediated by negative affect and anxiety in the course of PPS remains to be clarified.

4.3. Strengths and limitations

The strength of this study lies in its multi-center prospective design, the use of validated measures, the long duration of follow up using multiple measurements over time, the low loss to follow up and the use of sophisticated longitudinal analyses techniques. We did not only focus on severity of reported symptoms, but also on mental and physical functioning, as these are important aspects of the prognosis of PPS. Another strength of this study is that due to multiple measurements of symptom focusing, we were able to disentangle the within and between subject components of the longitudinal associations.

Our study also has several limitations. We measured somatosensory amplification (SSAS) only at baseline. Although the test-retest reliability in previous literature was high, this was based on re-testing after a maximum of eight weeks [43] and we do not know how stable it remains over a three year period. However, if SSAS is measuring a temporally stable, trait-like feature, this single measurement is not considered a problem. Nevertheless, because of the single measurement of somatosensory amplification we were not able to disentangle within- and between-subject associations for somatosensory amplification.

A second limitation is that the duration of symptoms at inclusion varied widely: patients were selected at different moments in their personal course of PPS [36, 47]. However, we believe that it is not feasible to include all patients at the start of their symptoms due to variety in duration of symptoms before the first presentation to a doctor and variety in time intervals between first presentation of symptoms and the conclusion that they are not sufficiently explained by verifiable, conventionally defined diseases.

A third limitation is that our analyses relied on self-report questionnaires, this may make certain variables (e.g. the report of concomitant somatic diseases) less reliable.

4.4. Conclusion

We conclude that - from a symptom perception perspective - symptom focusing and somatosensory amplification are both associated with symptom severity and lower mental and physical functioning in PPS over an extended period. Also within subjects symptom focusing is related to symptom severity and lower functioning. These results provide support for the impact of both symptom focusing and somatosensory amplification on the perpetuation of symptoms and lowered mental and physical functioning in individuals with PPS. However, effect sizes were limited. Both symptom focusing and somatosensory amplification show different associations with the outcomes over time and partly overlap, but do not strengthen each other’s effects as suggested in the somatosensory amplification theory.