Psychometric evaluation and adaptation of the Spine Oncology Study Group Outcomes Questionnaire to evaluate health-related quality of life in patients with spinal metastases.

BACKGROUND: The Spine Oncology Study Group Outcomes Questionnaire (SOSGOQ) was developed as the first spine oncology-specific health-related quality of life (HRQOL) measure. This study evaluated the psychometric properties and clinical validity of the SOSGOQ in a diverse cohort of patients with spinal metastases.
METHODS: An international, multicenter, prospective observational cohort study including patients with spinal metastases who underwent surgery and/or radiotherapy was conducted by the AOSpine Knowledge Forum Tumor. Demographic, tumor, and treatment data were collected. HRQOL was evaluated using the SOSGOQ and Medical Outcomes Study Questionnaire Short Form 36 Health Survey (SF-36) at baseline and fixed follow-up times. Construct validity was assessed using multitrait scaling analyses, confirmatory factor analyses, and correlation with the SF-36 and NRS pain score. Test-retest reliability was assessed in a subgroup of patients between 12 weeks after treatment and the retest 4 to 9 days later.
RESULTS: A total of 238 patients were enrolled at 9 centers across North America; 153 of these patients had HRQOL data available at 12 weeks after treatment. Multitrait scaling analyses and confirmatory factor analyses resulted in a refined version of the SOSGOQ with 4 domains and 4 single items. The revised SOSGOQ (SOSGOQ2.0) demonstrated strong correlations with SF-36 and the ability to discriminate between clinically distinct patient groups. Reliability of the SOSGOQ2.0 was demonstrated to be good, with an intraclass correlation coefficient ranging from 0.58 to 0.92 for the different domains.
CONCLUSIONS: The SOSGOQ2.0 is a reliable and valid measure with which to evaluate HRQOL in patients with spinal metastases. It is recommended to use the SOSGOQ2.0 together with a generic HRQOL outcome measure to comprehensively assess HRQOL and increase sensitivity and specificity. Cancer 2018;124:1828-38.

INTRODUCTION

A diagnosis of bone metastases often represents an incurable yet treatable disease. The population of patients with advanced stages of cancer, including metastatic spinal disease, is growing due to improved medical treatment options, which allow for better and longer disease control.1 The main treatment goal for these patients is to improve or maintain health‐related quality of life (HRQOL) for their remaining lifetime. Therefore, determining patient‐reported HRQOL is important to optimize and evaluate life‐extending and supportive care treatments.

Many generic and cancer‐specific outcome measures, including those for bone metastases, exist.2 Nevertheless, these outcome measures are nonspecific for spine‐related functions, and thus are less sensitive for assessing changes over time due to treatment or the progression of spinal disease. A disease‐specific instrument in addition to a generic (cancer) HRQOL measure will increase the sensitivity and specificity of HRQOL assessments. In response to the absence of spine oncology‐specific outcome measures, the Spine Oncology Study Group Outcomes Questionnaire (SOSGOQ) was developed and assessed for face and content validity.3 Content validity was evaluated by correlating the SOSGOQ items to the World Health Organization International Classification of Functioning, Disability and Health and face validity by content expert evaluation, both of which demonstrated excellent results.3

With content validity confirmed, the next step is to evaluate the hypothesized structure of the SOSGOQ in a clinical setting. Therefore, the objective of the current study was to examine the psychometric properties and clinical validity of the SOSGOQ in a cohort of patients who had undergone treatment for spinal metastases.

MATERIALS AND METHODS

Design

The AOSpine Knowledge Forum Tumor initiated an international, multicenter, prospective observational cohort study in August 2013 at 10 spine centers across North America and Europe (http://ClinicalTrials.gov identifier NCT01825161). The prespecified study objectives included evaluation of the validity and reliability of the English version of the SOSGOQ. Patients were eligible for inclusion if they had a diagnosis of metastatic spinal disease, were aged 18 to 75 years, and had undergone surgery and/or radiotherapy for the treatment of spinal metastases from any primary tumor. Patients with a central nervous system tumor or a primary spinal bone tumor were excluded. The ethics board of each participating spine center approved the protocol. All patients provided written informed consent for study participation.

Demographic, medical history, diagnostic, treatment, adverse events, and HRQOL data were collected prospectively. HRQOL and pain scores were evaluated at baseline and at 6, 12, 24, 52, and 104 weeks after treatment or until death with the English version of the SOSGOQ (version 1.03), the English version of the Medical Outcomes Study Questionnaire Short Form 36 Health Survey (SF‐36, version 2.0; Medical Outcomes Trust, Boston, Massachusetts4), and the numeric rating scale (NRS) pain score. All data were stored in a secure Web‐based application (Research Electronic Data Capture [REDCap]; Vanderbilt University, Nashville, Tennessee).

The SF‐365 and the NRS pain score are widely used generic outcome instruments and have been described in detail previously. The hypothesized structure of the SOSGOQ consists of 5 HRQOL domains (Physical Function, Neurological Function, Pain, Mental Health, and Social Function) and an additional set of questions during follow‐up for evaluating treatment satisfaction. Scores for the SF‐36 and the SOSGOQ were calculated if at least 50% of the items within the domain were answered. All domain scores were transformed to a scale of 0 to 100, in which a higher score represents a better QOL.

Statistical Analysis

Demographic data were summarized using descriptive statistics. A subject‐to‐item ratio of 1:76 was considered to be sufficient power with which to perform a factor analysis and assess validity; the minimum required sample size was 140 subjects with 20 items in the SOSGOQ to be evaluated (excluding the posttherapy domain).

The results first were analyzed to define the structure (construct validity) of the SOSGOQ. The concurrent validity, clinical validity, and reliability were determined using the modified structure of the SOSGOQ (SOSGOQ2.0). All statistical analyses were performed using SAS statistical software (version 9.4; SAS Institute Inc, Cary, North Carolina). Statistical significance was defined as P<.05.

Construct Validity

The internal structure of the SOSGOQ was evaluated using multitrait scaling analysis and confirmatory factor analyses (CFA). Convergent validity was evaluated by correlating the item score with the total score of its own domain, and divergent validity was evaluated by correlating the item score with the total score of the other SOSGOQ domains (Spearman rank). A scaling success was defined as a significantly higher item correlation with its own domain compared with the other SOSGOQ domains, with a minimum correlation of 0.40.7

CFA first was performed based on the prior hypothesized structure of the SOSGOQ3 followed by testing of a modified structure on a clinical conceptual basis supported by the results of the multitrait scaling analysis and modification indices. Model fit was evaluated with the root mean square error of approximation (RMSEA [≤0.08]),8 the 90% confidence interval (90% CI) of the RMSEA (upper bound limit of 0.1), the comparative fit index (CFI; ≥0.90),9 and the standardized root mean residual (SRMR; ≤0.08).10

Concurrent validity was evaluated by Spearman rank correlation of the SOSGOQ2.0 domains to the domains of the SF‐36 and the NRS pain score. The domains that are conceptually related were expected to demonstrate a correlation of at least 0.40. Analyses were performed with the 12‐week posttreatment data to meet the normality assumption; baseline data were used to perform sensitivity analyses of the 12‐week data.

Clinical Validity

Clinical validity was examined using the ability of the SOSGOQ2.0 to discriminate between patient groups. At baseline, patients with an Eastern Cooperative Oncology Group (ECOG) performance score of 0 or 1 were compared with patients with an ECOG score of ≥2. Furthermore, to assess the responsiveness to change, changes in ECOG status from baseline to 12 weeks after treatment (stable/improved vs deteriorated) were associated with changes in the SOSGOQ2.0 scores.

Reliability and Reproducibility

The test‐retest reliability of the SOSGOQ was assessed at 2 centers between the 12‐week posttreatment assessment and the retest 4 to 9 days later using the intraclass correlation coefficient (ICC).11

Internal consistency of the domains of the SOSGOQ2.0 was evaluated using Cronbach alpha12 at baseline and follow‐up; the minimum acceptable alpha value was defined at .70.13

RESULTS

A total of 238 patients from 9 centers across North America (4 in Canada and 5 in the United States) were enrolled in the prospective observational cohort study until November 2015. Of the 238 patients, 130 underwent surgery with or without additional radiotherapy and 108 patients received only radiotherapy (Fig. 1). The breast (26%) was the most common primary tumor site, followed by the lung (18%) and renal cell (16%). A summary of the baseline characteristics of the patients is shown in Table 1. At 12 weeks, a total of 172 patients had data available, 38 (16%) patients had died within the first 12 weeks of follow‐up, 3 patients were lost to follow‐up, 11 patients did not complete the 12‐week follow‐up visit, and 14 patients (6%) dropped out of the study for other reasons (withdrawal of consent, withdrawn by investigator). The SOSGOQ was completed at baseline by 224 patients (94%) and by 153 patients (63%) at 12 weeks after treatment.

Figure 1

Patient selection flow diagram. SOSGOQ indicates Spine Oncology Study Group Outcomes Questionnaire.

Table 1

Summary of Baseline Characteristics

Characteristic	No. (%)
Age at treatment, (n = 238), ya	59 (10.3)
Sex (n = 238)
Female	129 (54.2)
Male	109 (45.8)
Primary tumor (n = 238)
Breast	62 (26.1)
Lung	44 (18.5)
Kidney	39 (16.4)
Prostate	21 (8.8)
Other	72 (30.3)
Time since primary diagnosis (n = 238), mob	26.5 (5.0‐76.0)
ECOG PS (n = 234)
0‐1	156 (66.7)
2‐4	78 (33.3)
Location (n = 238)c
Cervical	38 (16)
Thoracic	152 (64)
Lumbar/sacral	103 (43)
Presence of other metastases (n = 238)
Visceral	80 (33.6)
Brain	23 (9.7)
ASIA
ASIA E	181 (77.4)
ASIA D	41 (17.5)
ASIA A‐C	12 (5.1)

Abbreviations: ASIA, American Spinal Injury Association; ASIA A, complete; ASIA B, incomplete (sensory but not motor function is preserved below the neurological level and includes the sacral segments S4‐S5); ASIA C, incomplete (motor function is preserved below the neurological level, and more than one‐half of key muscles below the neurological level have a muscle grade <3); ASIA D, incomplete (motor function is preserved below the neurological level, and at least one‐half of key muscles below the neurological level have a muscle grade of ≥3); ASIA E, normal (motor and sensory function are normal); ECOG PS, Eastern Cooperative Oncology Group performance status.

Mean ± standard deviation.

Median (interquartile range).

Multiple levels possible.

At baseline, 44 items from the SOSGOQ (1%) and 47 items from the SF‐36 (0.6%) were missing. The item addressing bowel and bladder function (item 8) in the SOSGOQ was the most commonly reported missing item across different time points. The items addressing strength in the arms (item 6) and bowel and bladder function (item 8) demonstrated positive skewness, with only 7% and 3%, respectively, of the patients reporting moderate to severe symptoms.

Internal Structure of the SOSGOQ

A total of 133 patients from 9 centers had complete data available with which to evaluate the structure of the SOSGOQ. The multitrait scaling analysis at 12 weeks after treatment according to the prior hypothesized structure of the SOSGOQ3 demonstrated that item 7 (need for a walking aid) and item 20 (leaving the house) had a strong correlation (correlation coefficient [R] = ‐0.71 and R = 0.65, respectively) with the Physical Function domain. Furthermore, item 16 (overwhelming pain) also demonstrated a strong correlation with the Pain domain (R = ‐0.60). Item 6 (arm strength), item 8 (bowel and bladder function), and item 15 (level of energy) demonstrated only a moderate correlation with their own scale (R = −0.53 to 0.59). The correlation of the other items exceeded the minimum correlation of 0.40 and these items were found to have a stronger correlation with their own domain compared with the other domains (see Supporting Information).

The CFA according to the hypothesized scale structure of the SOSGOQ confirmed the results of the multitrait scaling analysis, demonstrating unacceptable model fit (RMSEA, 0.12 [90% CI, 0.10‐0.13]; CFI, 0.79; and SRMR, 0.11). Based on the conceptual relation of items with other domains and the multitrait scaling analysis, the scale structure was modified. Items 7 (need for a walking aid) and 20 (leaving the house) were moved to the Physical Function domain and item 16 was moved to the Pain domain (cross‐loading on mental health). Items 5 (leg strength), 6 (arm strength), 8 (bowel and bladder function) retained as single items, yet item 8 was split into 2 questions to distinctly address bowel and bladder function. Item 15 (level of energy) was not found to contribute any new information and was removed. CFA according to the revised SOSGOQ structure confirmed the structure, demonstrating adequate model fit (RMSEA, 0.074 [90% CI, 0.055‐0.092]; CFI, 0.928; and SRMR, 0.06). The revised questionnaire consists of 4 domains and 4 single items and, at follow‐up, a set of questions evaluating treatment satisfaction (Fig. 2). Scoring guidelines are outlined in the Supporting Information. Results of the multitrait scaling analysis according to the final structure are shown in Table 2. Sensitivity analysis with the baseline data confirmed the revised structure of the SOSGOQ, demonstrating an adequate model fit (RMSEA, 0.08 [90% CI, 0.068‐0.095]; CFI, 0.94; and SRMR, 0.054).

Figure 2

Revised structure of the Spine Oncology Study Group Outcomes Questionnaire (SOSGOQ). (From AOSpine International, Used under CC BY‐NC‐ND 4.0).

Table 2

Convergent and Divergent Validity at Baseline and at 12 Weeksa

	Surgery ± RT		RT Only		All Patients
Scales	Item Own Domain Correlation	Item Other Domain Correlation	Item Own Domain Correlation	Item Other Domain Correlation	Item Own Domain Correlation	Item Other Domain Correlation
Physical Function	0.68‐0.84 0.55‐0.88	0.14‐0.50 0.09‐0.55	0.70‐0.81 0.71‐0.88	0.01‐0.68 0.01‐0.66	0.72‐0.84 0.70‐0.91	0.02‐0.55 0.05‐0.60
Pain	0.60‐0.82 0.67‐0.78	0.13‐0.59 0.17‐0.60	0.67‐0.91 0.62‐0.88	0.02‐0.65 0.05‐0.68	0.64‐0.84 0.69‐0.84	0.18‐0.61 0.17‐0.72
Mental Health	0.89‐0.92 0.84‐0.89	0.20‐0.48 0.19‐0.50	0.89‐0.92 0.89‐0.91	0.01‐0.41 0.09‐0.28	0.90‐0.91 0.86‐0.90	0.10‐0.37 0.12‐0.32
Social Function	0.71‐0.85 0.67‐0.87	0.25‐0.65 0.09‐0.68	0.67‐0.84 0.67‐0.87	0.19‐0.63 0.21‐0.73	0.68‐0.83 0.66‐0.88	0.24‐0.64 0.16‐0.73
Neuro Function: legs	1.00	0.15‐0.41 0.26‐0.52	1.00	0.01‐0.47 0.14‐ 0.58	1.00	0.05‐0.43 0.17‐0.58
Neuro function arms	1.00	0.14‐0.26 0.23‐0.35	1.00	0.07‐0.38 0.10‐0.44	1.00	0.12‐0.25 0.17‐0.38
Neuro Function: bowel and bladder	1.00	0.13‐0.46 0.12‐0.30	1.00	0.06‐0.29 0.08‐0.30	1.00	0.11‐0.36 0.02‐0.32

Abbreviations: Neuro, neurological; RT, radiotherapy.

Range of Spearman correlation. Number of patients at baseline was 238; number at 12 weeks after treatment was 153.

Italics represent correlations at baseline.

Concurrent Validity With the SF‐36 and NRS Pain Score

Correlation between the domains of the SOSGOQ2.0 and the SF‐36 and NRS pain score are shown in Table 3. The domains of the SOSGOQ2.0 demonstrated a strong to very strong correlation with the corresponding domains of the SF‐36. The SOSGOQ2.0 Pain domain demonstrated a strong correlation with the NRS pain score. The single neurological function items demonstrated a weak correlation with the domains of the SF‐36, indicating that these items assess a different aspect that is not assessed by the SF‐36.

Table 3

Concurrent Validity (Revised SOSGOQ Version 2.0)

	SOSGOQ2.0 Physical Function		SOSGOQ2.0 Neurological Item 5		SOSGOQ2.0 Neurological Item 6		SOSGOQ2.0 Neurological Item 8		SOSGOQ2.0 Pain		SOSGOQ2.0 Mental Health		SOSGOQ2.0 Social Function
Reference Score	No.	rs (P)	No.	rs (P)	No.	rs (P)	No.	rs (P)	No.	rs (P)	No.	rs (P)	No.	rs (P)
SF‐36v2
Physical Functioning	153	0.85 (<.001)	153	0.46 (.463)	153	0.26 (.260)	153	0.30 (.299)	146	0.58 (<.001)	146	0.14 (.099)	152	0.51 (<.001)
Role‐Physical	147	0.70 (<.001)	147	0.32 (.321)	147	0.25 (.247)	147	0.21 (.210)	146	0.47 (<.001)	146	0.12 (.135)	146	0.35 (<.001)
Bodily Pain	152	0.56 (<.001)	152	0.30 (.300)	152	0.25 (.253)	152	0.33 (.332)	145	0.72 (<.001)	145	0.30 (<.001)	151	0.49 (<.001)
General Health	153	0.37 (<.001)	153	0.18 (.184)	153	0.12 (.122)	153	0.20 (.200)	146	0.28 (<.001)	146	0.47 (<.001)	152	0.31 (<.001)
Vitality	147	0.34 (<.001)	147	0.18 (.176)	147	0.25 (.246)	147	0.20 (.197)	146	0.33 (<.001)	146	0.44 (<.001)	146	0.37 (<.001)
Social Functioning	153	0.61 (<.001)	153	0.33 (.332)	153	0.17 (.167)	153	0.38 (.381)	146	0.42 (<.001)	146	0.32 (<.001)	152	0.54 (<.001)
Role Emotional	152	0.45 (<.001)	152	0.25 (.254)	152	0.15 (.153)	152	0.26 (.260)	145	0.35 (<.001)	145	0.30 (<.001)	151	0.45 (<.001)
Mental Health	148	0.23 (.004)	148	0.16 (.157)	148	0.15 (.147)	148	0.23 (.225)	146	0.37 (<.001)	146	0.72 (<.001)	147	0.49 (<.001)
Pain NRS
Average Daily Pain	153	−0.41 (<.001)	153	−0.29 (<.001)	153	−0.26 (<.001)	153	−0.37 (<.001)	146	−0.74 (<.001)	146	−0.11 (.197)	152	−0.39 (<.001)

Abbreviations: NRS, numeric rating scale; SF‐36 v2, Medical Outcomes Study Questionnaire Short Form 36 Health Survey version 2.0; SOSGOQ, Spine Oncology Study Group Outcomes Questionnaire.

At baseline, patients with an ECOG performance score of 0 or 1 demonstrated a significantly higher score (P<.001) across the different SOSGOQ2.0 domains compared with patients with an ECOG performance score of ≥2. Patients with a stable or improved ECOG performance score at 12 weeks after treatment demonstrated an increase in domain scores compared with a deterioration in the SOSGOQ2.0 domain scores noted for patients with a decline in the ECOG performance score. This difference was significant for all domains except the Mental Health domain (Table 4).

Table 4

Sensitivity to Change in the SOSGOQ2.0 Based on Change in ECOG Score

Change in Domain	ECOG Decline N = 48	ECOG Stable/Improved N = 114	Total N = 162	P
Physical function (no.)	41	102	143	.001a
Mean (SD)	‐5.7 (24.8)	8.3 (21.8)	4.3 (23.5)
Neurological function (no.)	41	102	143	<.001a
Mean (SD)	‐8.0 (15.7)	3.6 (15.6)	0.3 (16.4)
Pain (no.)	41	95	136	<.001a
Mean (SD)	3.1 (26.5)	21.0 (29.5)	15.6 (29.7)
Mental health (no.)	41	95	136	.196a
Mean (SD)	6.4 (24.8)	11.6 (20.2)	10.0 (21.7)
Social function (no.)	41	101	142	.012a
Mean (SD)	‐1.4 (22.5)	9.8 (24.1)	6.5 (24.1)

Abbreviations: ECOG, Eastern Cooperative Oncology Group; SD, standard deviation; SOSGOQ2.0, revised Spine Oncology Study Group Outcomes Questionnaire.

Derived using the Student t test.

Reliability

A total of 36 patients from 2 centers also completed the retest of the SOSGOQ within 4 to 9 days after the assessment at 12 weeks after treatment. Of these 36 patients, 10 underwent surgical intervention with or without additional postoperative radiotherapy and the remaining 26 patients received radiotherapy only. There were no significant differences noted between patients who completed the retest of the SOSGOQ compared with those included in the validity analyses with regard to age (P = .043), sex (P = .873), location of the primary tumor (P = .503), and ECOG performance score (P = .191). The ICC of the SOSGOQ2.0 domains ranged from 0.58 to 0.92, with the Mental Health and posttherapy domains demonstrating the lowest reproducibility with an ICC of 0.62 and 0.58, respectively (Table 5). The Cronbach alphas at baseline and follow‐up were acceptable for all domains (Table 5).

Table 5

Reliability and Internal Consistency for Each of the 6 Subscales of the SOSGOQ2.0 Assessment

	12 Weeks (Test)						12 Weeks (Retest)
SOSGOQ2.0 Domain	No.	Min	Max	Mean	SD	Alphaa	No.	Min	Max	Mean	SD	Alphaa	ICCb	95% CI
Physical Function score (6 items)	36	17	100	74.0	20.7	0.77	36	13	100	76.5	20.9	0.65	0.92	0.85‐0.96
Neurological Function: legsc	36	1	4	2.1	1.1	‐	36	1	5	1.9	1.2	‐	0.73	0.53‐0.85
Neurological Function: armsc	36	1	5	1.5	0.9	‐	36	1	5	1.4	1.0	‐	0.72	0.52‐0.84
Pain score (5 items)	36	30	100	70.2	19.8	0.74	36	30	100	72.9	21.6	0.68	0.76	0.58‐0.87
Mental Health score (2 items)	36	25	100	74.1	20.1	0.85	36	13	100	80.8	23.4	0.79	0.62	0.38‐0.78
Social Function score (3 items)	36	25	100	75.1	20.2	0.74	36	42	100	80.8	18.9	0.63	0.63	0.39‐0.79
Posttherapy questions (7 items)	36	25	100	71.5	16.6	0.80	36	36	100	69.4	13.8	0.76	0.58	0.32‐0.76

Abbreviations: 95% CI, 95% confidence interval; ICC, intraclass correlation coefficient; Max, maximum; Min, minimum; SD, standard deviation; SOSGOQ2.0, revised Spine Oncology Study Group Outcomes Questionnaire.

Cronbach alpha.

For item responses at 12 weeks after treatment (retest) calculated using an one‐way random effect model.

Reliability is only displayed for 2 neurological items; the item addressing bowel and bladder function is split into 2 questions in the SOSGOQ2.0 and therefore was not displayed.

DISCUSSION

Patient‐reported outcomes are essential in the evaluation of long‐term palliative cancer care. The value of a disease‐specific questionnaire with the use of generic questionnaires has been widely acknowledged, including the development of many cancer‐specific modules by the European Organization for Research and Treatment of Cancer.14 To the best of our knowledge, the SOSGOQ is the first spine oncology‐specific HRQOL questionnaire. The objective of the current study was to assess the psychometric properties and clinical validity of the SOSGOQ in a diverse cohort of patients with spinal metastases. Psychometric evaluation of the SOSGOQ resulted in a slightly revised structure. Evaluation of the revised SOSGOQ demonstrated that it is a clinically valid and reliable questionnaire with which to evaluate QOL in patients with spinal metastases. It is recommended to use the SOSGOQ together with a generic outcome measure to comprehensively evaluate HRQOL in patients with spinal metastases.

Recently, Janssen et al investigated the construct validity of the SOSGOQ by its correlation to the 5‐level EuroQol 5 dimensions (EQ‐5D‐5L) and exploratory factor analyses in a convenience sample of 82 patients with spinal metastases.15 Exploratory factor analyses demonstrated the association between items 7 (need for a walking aid) and 20 (leaving the house) and the Physical Function domain and item 16 (intensity of pain) and the Pain domain. Furthermore, the neurological function items and item 15 (level of energy) demonstrated low factor loadings.15 These results are in agreement with the results of our CFA analyses and the revised SOSGOQ2.0. In the current study, concurrent validity was evaluated by correlating the SOSGOQ2.0 with the SF‐36 and the NRS pain score. The SF‐36 is a generic yet comprehensive HRQOL measure because it consists of several domains, including multiple items per domain, compared with the less comprehensive EQ‐5D‐5L, which uses only single items representing the different domains. Correlation of the SOSGOQ2.0 with the SF‐36 and NRS pain score demonstrated strong correlations between the domains that were conceptually related, confirming the construct validity of the SOSGOQ2.0.

In contrast to the study by Janssen et al,15 the current study also assessed the test‐retest reliability. The reliability of the SOSGOQ2.0 domains were shown to be excellent, with the exception of the posttherapy questions, Mental Health domain, and Social Function domain, which demonstrated moderate to good reliability. The lower reproducibility of the Mental Health domain may be explained by the small size of the domain (2 questions) and the multifactorial and dynamic nature of mental health in patients with cancer. A recent study by Jim et al16 demonstrated the daily variation in depression, fatigue, activity, and sleep in women undergoing chemotherapy. The severity of the symptoms, other aspects of the disease, or other concomitant treatments (eg, systemic therapy) may have influenced the mental health state, resulting in decreased reliability. Because QOL may deteriorate quickly in patients with advanced stages of cancer, the retest was administered 4 to 9 days after the initial assessment. As such, a stable health state between 12 weeks posttreatment and the retest was assumed rather than objectively assessed, which may have resulted in an underestimation of the reliability of the SOSGOQ2.0. To the best of our knowledge, the optimal interval between the initial assessment and the retest in patients with advanced cancer is complex and remains uncertain.17 A short interval is proposed because QOL may change rapidly in this population; however, an interval that is too short may enhance the probability of patients recalling their previous answers, thereby compromising the validity of the reliability measurement.17

The 2 items in the hypothesized Neurological Function domain assessing arm strength and bowel and bladder control demonstrated the largest floor effect and only moderate correlation with the total domain score, which is in agreement with the study of Janssen et al.15 This can be explained because the majority of patients included in the current study were treated for metastatic disease in the thoracolumbar spine, which may affect the neurological function of the lower extremity (item 5 leg strength) due to radiculopathy or metastatic spinal cord compression, and will not affect the neurological function of the upper extremity (item 6 arm strength). Moreover, only 8% of the patients had diminished or no bowel or bladder control (item 8) according to physical examination. Therefore, the domain score is predominantly defined by the item addressing leg strength (item 5). Analysis of the missing values demonstrated that item 8 (bowel and bladder function) was the main missing item and also showed low reliability, which may be explained by the double‐barreled nature of the question. Because item 8 addresses 2 distinct but related physical functions, answering this question may be difficult and may lead to missing values if only 1 of the 2 functions is affected. Therefore, changes to the SOSGOQ have been made to ease future data analyses and interpretation. Item 8 (item 9 & 10 SOSGOQ2.0) has been split into 2 distinct questions addressing bowel and bladder function and, rather than calculating a domain score, it is recommended to analyze these items as individual items because they address distinct functions. A domain score may be calculated to obtain an overall impression of the presence of impairment. In addition, a total score for the SOSGOQ2.0 may be calculated by summing the domain scores of the Physical Function, Social Function, Mental Health, and Pain domains (see Supporting Information). The posttherapy items (items 21‐27) should be used during follow‐up together with the core items (items 1‐20).

The current study has several methodological strengths that enhance the generalizability of the results. First, 9 centers across North America participated in the validation study and 2 centers participated in the test‐retest reliability, resulting in a relatively large sample size for this difficult‐to‐study patient population. Second, both patients who underwent surgery and/or radiotherapy were enrolled in the study. The different treatment modalities in combination with the wide inclusion and exclusion criteria, with no restrictions concerning ECOG performance status, resulted in a heterogeneous, yet representative, population of patients who undergo treatment for spinal metastatic disease. Therefore, this instrument appears to be well‐suited for evaluating any treatment related to metastatic spinal disease.

A limitation of the current study was the relatively high dropout rate (24%) within the first 12 weeks. A significant percentage of the patients (16%) died within the first 12 weeks, indicating the severity of disease. Despite maximized efforts for questionnaire compliance and follow‐up, a relatively high rate of loss to follow‐up was encountered, which is inherent to the study population. Nevertheless, sensitivity analyses with the baseline questionnaire responses confirmed the results of the primary analyses and the validity of the SOSGOQ2.0 across a wide range of health states. Second, physical function, including motor function and pain, is assessed in greater depth compared with social function and mental health. This is reflected by the uneven number of items per domain. The instrument's bias toward physical function reflects the impact spinal metastases have on these constructs. Third, despite the multicenter approach, the demonstrated validity of the questionnaire is limited to the English version of the SOSGOQ. Translations of the SOSGOQ2.0 into other languages currently are being performed and validation of the translated versions is planned. Last, the minimal clinically important changes in the SOSGOQ2.0 scores over time for a patient are important for interpretation. A study to assess the minimal clinically important differences in the SOSGOQ currently is planned.

The current study investigated the psychometric qualities and clinical validity of the SOSGOQ using a multicenter, prospective observational study. The refined SOSGOQ2.0 was demonstrated to be a valid and reliable spine oncology‐specific outcome measure. In future studies evaluating the management of patients with spinal metastases, it is recommended to use the SOSGOQ2.0 in addition to generic HRQOL outcome measures to achieve a more comprehensive representation of HRQOL and to increase sensitivity to change.

FUNDING SUPPORT

This study was organized and funded by AOSpine International through the AOSpine Knowledge Forum Tumor, a pathology‐focused working group of up to 10 international spine experts acting on behalf of AOSpine in the domain of scientific expertise. A research grant for this study was received from the Orthopaedic Research and Education Foundation.

CONFLICT OF INTEREST DISCLOSURES

Anne L. Versteeg has received personal fees from AOSpine International for work performed outside of the current study. Arjun Sahgal has received personal fees for past educational seminars from Elekta AB, Accuray Inc, and Varian Medical Systems; has received a research grant from Elekta AB; and has received fees for travel accommodations/expenses from Elekta and Varian Medical Systems for work performed outside of the current study. Dr. Sahgal also belongs to the Elekta MR‐Linac Research Consortium. Laurence D. Rhines has received educational grants from Stryker and Medtronic for work performed outside of the current study. Daniel M. Sciubba has received personal fees from Medtronic, DePuy‐Synthes, Stryker, NuVasive, and K2M for work performed outside of the current study. Paul M. Arnold has received fees for sponsored or reimbursed travel from AOSpine North America; has received personal fees from Spine Wave; has intellectual property rights and interests in, equity, or a position of responsibility with Evoke Medical; has equity with Z‐Plasty; has received personal fees from Stryker Orthopaedics and Medtronic Sofamor Danek USA; has received personal fees and travel expenses from Stryker Spine and Medtronic; and has received personal fees from In Vivo, In Vivo Therapeutics, SpineGuard, and Ulrich for work performed outside of the current study. Ziya L. Gokaslan has received research support from AOSpine North America and is a stockholder in Spinal Kinetics for work performed outside of the current study. Charles G. Fisher has received personal fees from Medtronic and NuVasive, has received a grant to his institution from the Orthopaedic Research and Education Foundation (which was transferred from Dr. Fisher to AOSpine Knowledge Forum Tumor), and has received fellowship support to his institution from AOSpine and Medtronic for work performed outside of the current study.

AUTHOR CONTRIBUTIONS

Study design: Arjun Sahgal, Ziya L. Gokaslan, and Charles G. Fisher. Conduct: All authors. Statistical analyses: Anne L. Versteeg, Arjun Sahgal, Ziya L. Gokaslan, and Charles G. Fisher. Article writing: All authors. Guarantor: Charles G. Fisher.

Additional supporting information may be found in the online version of this article.

Supporting Information 1

Click here for additional data file.