Article Text


Work disability, lost productivity and associated risk factors in patients diagnosed with systemic lupus erythematosus
  1. Tammy O Utset1,
  2. Amrutha Baskaran2,
  3. Barbara M Segal3,
  4. Laura Trupin4,
  5. Sarika Ogale5,
  6. Ellen Herberich6 and
  7. Kenneth Kalunian2
  1. 1Department of Medicine/Rheumatology, University of Chicago, Chicago, Illinois, USA
  2. 2Department of Medicine/Rheumatology, University of California San Diego, La Jolla, California, USA
  3. 3Department of Medicine/Rheumatology, University of Minnesota, Minneapolis, Minnesota, USA
  4. 4Department of Medicine/Rheumatology, University of California San Francisco, San Francisco, California, USA
  5. 5Department of Medicine/Rheumatology, Genentech, South San Francisco, California, USA
  6. 6Harris Interactive, Rochester, New York, USA
  1. Correspondence to Dr Tammy O Utset; tutset{at}


Objective To assess prevalence and correlates of work presenteeism, absenteeism and work disability (WD) in patients with systemic lupus erythematous (SLE) and matched controls.

Methods Patients with SLE from six medical centres were recruited to complete a questionnaire consisting of several prevalidated survey instruments. The subject's rheumatologist completed medical history. Subjects recruited two non-SLE ‘best friend’ controls with matching demographics to complete a control survey. Analyses employed Student's t tests, χ2 tests and logistic regression models.

Results 344 subjects with SLE and 322 controls submitted completed questionnaires. Mean pain, fatigue, Brief Cognitive Symptoms Index (BCSI) scores and depressive symptoms were worse in patients with SLE with WD (all p<0.01). WD was associated with African–American race, older age (51–65 years) and less than 4-year college education (all p<0.01). High presenteeism was associated with low pain and fatigue levels, higher BCSI scores and negatively correlated with depressive symptoms (all p<0.05). Increased pain and fatigue were associated with elevated absenteeism (p<0.05). Subjects with physically and cognitively demanding work reported worse presenteeism compared with controls with similar jobs (77% vs 85%, p<0.05 and 75% vs 85%, p<0.001), respectively. Patients with most cognitively demanding jobs reported greater weekly absenteeism (mean, 5.9 h) compared with controls (mean, 6.9 overtime hours, p<0.05).

Conclusions The questionnaire demonstrated increased WD in SLE. Highly physical and highly cognitive jobs are challenging to patients with SLE and had increased absenteeism compared with controls. Depressive symptoms were correlated with better presenteeism without major socio-demographic determinants. Employability may be enhanced by improving treatment of depressive symptoms in patients with SLE.

Statistics from

Key messages

  • Patients with systemic lupus erythematous (SLE) with physically and cognitively demanding work had worse presenteeism compared with controls with similar jobs and patients with the most cognitively demanding jobs reported greater weekly absenteeism compared with controls.

  • Patients with SLE with work disabilities have higher pain, fatigue and depressive symptoms and worse cognitive function as assessed by validated instruments.

  • Work disability in patients with SLE is higher in patients of African–American race, with older age and with less formal education.


As survival in systemic lupus erythematosus (SLE) has improved, clinical outcomes such as work and domestic functionality have been included in the study of SLE populations.1–3 As large numbers of individuals in gender/age groups typical of SLE are employed,4 one socioeconomic facet of SLE that should be assessed is the ability of patients with SLE to maintain paid employment. In employed and non-employed patients with SLE, disease-related issues may impair functionality in child-rearing and household tasks as well as work function.

Loss of work and domestic function has personal and societal costs. These are considered indirect costs of SLE.5 These include work disability (WD), absenteeism and decreased work productivity (decreased presenteeism). While work absenteeism and WD are quantifiable outcomes, variables such as presenteeism are more difficult to measure. However, validated measures of self-reported work performance and absenteeism have been developed for general and chronic disease populations.

Increased work absenteeism and WD rates have been observed in numerous SLE studies. Reported WD rates range from 20% to 50%, and vary widely between SLE population studies.6–11 Demographic factors associated with WD include age, low educational level, low socioeconomic status and race.12–17 Pain, fatigue, depressive symptoms, comorbidities, disease duration, activity, damage and cognitive dysfunction have been correlated with WD.7–10 ,18–25 However, studies of the association between work type at diagnosis and subsequent risk of WD have had variable results. Some studies have documented high physical demands, high cognitive demands and low job control to have higher risks of WD in SLE,11 while other studies found no associations of global work type with WD.21 Fewer studies have quantified work absenteeism. Yelin et al1 ,18 found that the majority of work loss in SLE was due to WD rather than decreased work hours and that fewer subjects with SLE entered or re-entered the work force relative to general population controls. Clarke26 found patients with SLE to have average absentee rates of only 13–16 days/year in full-time workers, while Campbell6 found 21% of patients with SLE to have been absent due to health issues more than 15 days/year, compared with 11% in matched controls.

Work and domestic productivity, lost work and WD are clinical outcome measures, which could be followed over time in assessing clinical course and the impact of SLE therapies. We have developed a self-administered questionnaire on presenteeism, absenteeism, WD and home productivity and performed an initial cross-sectional study of functionality using this questionnaire in SLE and a matched control population. Concurrently, we measured common contributing factors to WD, including psychosocial measures, demographic data, comorbidities, work type and disease-specific factors, in order to assess determinants of work functionality as measured by this questionnaire.


Survey overview

Patients with SLE from six medical centres in the USA were asked to complete a survey on work and domestic function, and the subject's rheumatologist completed questionnaires on the subject's medical history. Subjects were asked to recruit two non-SLE ‘best friend’ controls to complete a control survey.

Study population

Consecutive adult ambulatory patients with SLE who fulfilled American College of Rheumatology (ACR) criteria for the classification of SLE27 were offered enrolment in the study. Patients with a comorbid diagnosis of rheumatoid arthritis (RA) were excluded from the study. The final number of patient cases was 344.

Patients recruited their matched control (ie, ‘best friend’ control) without a prior diagnosis of SLE or RA. Subjects were instructed to recruit two controls that were of the same gender, approximately the same age (±5 years), did not live in the same household as the patient and resided in the USA. Based on social dynamics, we expected this method of peer-nomination of the control group to create a control group that is similar to the patient group in terms of socioeconomic status and demographics.28 ,29

The final number of controls was 321 with case–control ratio between 1:1 and 1:2. The variation in number of cases (N=344) and controls (N=321) can be confusing, and arises due to the ‘best friend’ control method used in this study. Patients with SLE were given two extra (control) questionnaires in stamped, preaddressed envelopes and asked to find two friends to complete the questionnaire. Their friends were to remain anonymous to the study because we could not obtain informed consent from the friends. Patients with SLE received compensation for participation, but the anonymous friends could not (due to anonymous status). Thus, either the Patients with SLE did not uniformly find two friends to agree to participate, or many of the controls may not have followed through with participation—perhaps, in part, due to lack of compensation. Thus, recruitment of the control patients by an SLE subject may vary between 0 and 2. Finally, we have nearly equal numbers of patients with SLE and controls, which was our general goal. The individual subject with SLE was not then matched to their control for analysis, but rather they were analysed in groups of SLE versus controls.

Six rheumatologists with large lupus patient populations recruited patients with lupus for this study. The six rheumatologist clinic settings were a mix of academic (n=4) and private practice (n=2), and located in diverse geographical areas (Chicago, Atlanta, Los Angeles, San Diego, San Francisco and Minneapolis).

Data collection

Data were collected between 1 October 2009 and 31 July 2010 via self-administered paper surveys. Participating subjects with SLE were provided with a 12-page patient survey, two control surveys, a letter explaining the study and postage-paid reply envelopes for each of the three surveys. Once patients were familiarised with study procedures and provided written informed consent, treating physicians completed an eight-page medical history survey, which the physician mailed back directly to a research firm tasked with managing the logistics of the study and reporting the results (Harris Interactive). Patients were asked to complete the survey within 2 weeks of their office visit and return it directly to Harris Interactive using the reply envelope. Each patient survey contained a unique identification code linking it to a corresponding medical survey, so that patient and physician surveys could be jointly analysed. Both physicians and patients were compensated for their participation in the study, but due to the anonymity of the control survey, control subjects received no compensation.


The physician, patient and control surveys were developed in collaboration with lupus medical experts, Genentech, and Harris Interactive. The surveys consist of several pre-validated survey instruments or portions of these instruments. This questionnaire is an amalgam of portions of previously developed questionnaires, drawn together into this questionnaire. Many of the categories are adopted in part from Utset et al21 and others were assembled by consensus.

Rheumatologists completed a medical survey about each recruited patient's lupus diagnosis and history. Chart audits included the following measures: date and details of SLE diagnosis (including ACR criteria), damage present in the patient using the Systemic Lupus International Collaborative Clinics/American College of Rheumatology Damage Index (SLICC/DI),30 comorbidities using the Sangha Comorbidity Index31 ,32 and specific medical treatments. Patient and control surveys were virtually identical. Patients and controls completed self-administered surveys about their life and work experiences. The control version of the survey differed from the patient version in only two ways: a question asking the year of lupus diagnosis was modified to ask controls whether or not they have been diagnosed with lupus by a doctor, and two questions specifically asking about employment status at the time of lupus diagnosis were deleted.

Patient and control surveys included the following measures: demographics (age, gender, education, income, race/ethnicity, marital status and insurance coverage), work productivity, absenteeism and presenteeism using the WHO Health and Work Performance Questionnaire (HPQ),33 lost productivity outside of the workplace using modified questions from the HPQ and other instruments,33 ,34 symptoms of fatigue using Functional Assessment of Chronic Illness Therapy (FACIT-Fatigue),35 depression using the Center for Epidemiologic Studies Depression Scale Short Form (CES-D),36 ,37 cognitive impairment using a Brief Cognitive Symptoms Index (BCSI) which was derived from the Cognitive Symptom Inventory developed by Pincus et al,38 ,39 overall health-related quality of life (HRQoL) including both physical and mental health status using the Medical Outcomes Survey Short Form 12 (SF-12v2)40 and comorbidities using the Sangha Comorbidity Index.31 ,32

We define absolute absenteeism as the number of hours a patient missed from work due to their health over the past 7 days. It is calculated by taking the difference between the hours expected to work in a typical 7-day week by the employer and the actual hours worked in the past 7 days. We define absolute presenteeism as the extent to which productivity has been reduced while at work due to health complaints over the past 4 weeks (adjusted for the subjects’ assessments of the productivity of average coworkers). The scores range from 0% to 100%, where 100% indicates ideal performance. Low presenteeism is defined as a score of 70% or less, and high presenteeism a score of 80% or greater. The FACIT-Fatigue scale has expected scores ranging from 0 to 52, with higher scores indicating less fatigue. We categorise respondents into groups with scores below 15, 15–25, 26–35 and scores above 35.

The CES-D is a scale that measures depression. A score of 10 or greater suggests depression.36 ,37 We classify respondents into groups with scores below 5, 5–10, 11–15, 16–20 or above 20. BCSI scores range from 0 to 100, with higher scores indicating worse functioning or greater impairment. BCSI scores are coded as 0–25, 26–50, 51–75 and 76–100. Pain is measured using a single item from the Brief Pain Inventory41: ‘What number best describes your pain on average over the past 4 weeks (28 days)?’ Responses range from 0 (no pain) to 10 (pain as bad as you can imagine). The SF-12v2 includes several subscales, measuring various aspects of HRQoL including physical, mental, social and emotional health, vitality and general health. For each subscale, scores range from 0 to 100, with higher scores indicating better functioning.40

Work status was determined in the year of SLE diagnosis and at time of survey for all subjects, and categorised as employed (full-time or part-time), self-employed, not employed but seeking work, not employed but not seeking work, retired, student or homemaker. Work disabled patients with SLE and controls were identified by the response ‘not employed but not seeking work’, while employed/employable was defined as full-time, part-time, self-employed or non-employed but seeking work. Work status was further explored by determining the duration of current work status. If subjects were not working, they were queried as to whether the non-working status was at least, in part, related to their health. Work type was classified as per USA Census Bureau, 2000,21 ,42 and defined in the year of SLE diagnosis and at the time of survey completion. Socioeconomic status was assessed using educational level and household income.

Statistical analysis

All prevalidated instruments were scored according to validated scoring algorithms. Demographic, health and employment characteristics were compared for patients and controls, using Student's t tests for continuous variables and χ2 tests for categorical variables. To examine factors influencing employment status and performance among patients, we developed logistic regression models including variables for SLE symptoms (depression, fatigue, pain, cognitive functioning), comorbidities and demographics (age, race, marital status, education). We used stepwise selection techniques to reduce the number of variables in the models and slight residual colinearity was assessed using ridge regression. These models were used to examine the correlates of being unemployed due to health problems and of presenteeism among patients. Statistical tests were considered to be significant at an α level of 0.05 on a two-tailed test.


A total of 496 subjects with SLE were consented for the study with 344 SLE completing the questionnaire (69% response rate) and 322 controls returned the questionnaire. The average number of ACR criteria present in subjects with SLE was 6. Median disease duration was 9 years (range 1–51), and median SLICC/DI score was 1 (range 0–13). Table 1 compares the demographic data of controls with subjects with SLE. Gender was similar, but slightly more controls were >65 years old (8% vs 2%). Patients with SLE were often married, but had similar numbers of adults and children in their households compared with controls. Completion of a 4-year college and income > $75 000/year was slightly more prevalent in controls (all p<0.05). On average, patients had a total of 4.2 comorbidities compared with 1.5 in the controls (p<0.01). Controls were more often Caucasian. HRQoL in the control group approximated normal populations,40 while all subscales of HRQoL were impaired in the SLE group compared with controls by the SF12v2 (data not shown, all p<0.01). Similarly, pain, FACIT-fatigue, depressive symptoms and BCSI score were worse in patients with SLE compared with controls (data not shown, all p<0.01) (table 1).

Table 1

Baseline demographic characteristics of patients with SLE and control subjects

Patients with SLE were significantly less likely than controls to be working full time (24% vs 50%, p<0.05), although full-time employment in the year of SLE diagnosis was 49% (table 2).

Table 2

Patient and control employment characteristics

The most common work status at the time of survey for subjects with SLE was WD (31% among SLE vs 4% in controls, p<0.05), defined as ‘not working and not seeking work’. Confirmation of the WD variable as representative of health-related disability is illustrated by the correlation of this category with ‘not working in part due to health’ (88% agreement, p<0.01) and with a positive response on a question about long-term disability pension compared with working subjects (7% of employed/employable subjects vs 63% of non-working subjects were on extended sick leave or disability, p<0.01). However, a significant number of the subjects with SLE could not characterise their work status and marked ‘other’ (18% in SLE vs 3% in controls, p=0.01). For WD analyses, employed/employable subjects were defined by the categories full-time or part-time work, self-employment or unemployed but seeking work. The statuses of ‘retired’, ‘homemaker’ and ‘student’ were not included in these comparisons and did not differ between SLE and control groups. Subjects marking ‘other’ work status were excluded from the WD analyses also. Among all non-working patients with SLE, subjects were much more likely than non-working controls to ascribe their working status to health problems (88% vs 15% in controls, p<0.05). Patients with SLE were more likely to report receiving a social security disability pension (overall 41% vs 4% in controls, p<0.05), and the duration of unemployment was longer in non-working subjects with SLE than controls (7.2 years vs 2.2 years, p<0.05). All parameters of work quantity and quality were lower in working subjects with compared with controls. Patients with SLE reported working fewer hours in the week prior to their survey (33.3 vs 39.1 h, p<0.05), and more sick days in the month prior (2.3 vs 0.4 sick days in controls, p<0.05). Self-assessed productivity (presenteeism, scale 0–100, 100=ideal performance) over the prior 4 weeks was worse in SLE (77% vs 85% in controls, p<0.05). Mean absentee time in the week prior was 2.7 h in SLE, while controls worked extra 4.7 h overtime, (p=0.17). Household function was worse in SLE, with 7.4 days in the month prior in which they were unable to perform household activities compared with 1.8 days in controls, (p<0.05) (table 2).

Self-reported symptoms that may impact productivity were consistently worse in the SLE population with WD, in comparison with employed/employable patients with SLE. The mean pain scale (p<0.01), fatigue scores (p<0.01), BCSI score (p<0.01) and depressive symptoms (p<0.01) were worse in patients with SLE with WD on univariate analysis (table 3).

Table 3

Bivariate analysis of work disability employment status in SLE

SLICC/DI score also correlated with WD status. Among patients with SLE with WD status, 43% had a SLICC/DI score >1, compared with 26% of employed/employable patients with SLE (p<0.01). On univariate analysis, WD status was associated with classical sociodemographic parameters, including African–American race (p<0.01), older age (51–65 years) (p<0.01) and patients having less than a 4-year college education (p<0.01, data not shown), but disease duration did not differ between working and work-disabled patients with SLE (table 3). Logistic regression with stepwise elimination of the major variables found on univariate regression revealed that higher age (p=0.005), SLICC/DI score (p=0.0140), FACIT-fatigue score (p<0.001), BCSI score (p=0.017), race (p=0.003) and educational level (p<0.001) were all independently associated with WD in the SLE group (table 4).

Table 4

Logistic regression of work disability employment status in SLE to obtain p values after dropping insignificant variables using STEPWISE elimination

Many of the correlations of poor self-reported work productivity (presenteeism) resembled those found in SLE with WD (tables 3 and 5).

Table 5

Patients with SLE: bivariate regression of absenteeism and presenteeism over the last week, by relevant covariates

Pain scores were worse in patients with low presenteeism than in those with high presenteeism (4.8 vs 3.5, p=0.001). Fatigue levels were worse in low presenteeism subjects (45% vs 76% with a favourable FACIT score of >25, p<0.001). High levels of depressive symptoms (CES-D index >10) were present in 69% of poor presenteeism patients, while 31% of high presenteeism patients endorsed significant depressive symptoms (p<0.001). Self-reported cognitive function was more often impaired in low presenteeism subjects (BCSI score≥26 in 52% of high presenteeism patients vs 38% of low presenteeism patients, p<0.014). However, demographic variables, age, race, education and marital status were unassociated with presenteeism. SLE duration was unassociated with presenteeism, and SLICC/DI scores did not differ between presenteeism groups.

Absenteeism reported for the week prior to interview did not vary by educational level or race/ethnicity. More severe pain level and fatigue were associated with elevated absenteeism (p=0.017 and p=0.027 respectively). Age, cognitive symptoms, depressive symptoms, summed comorbidities, marital status and disease duration also were not associated with recent work absence (table 5).

In contrast to patients with SLE, presenteeism in the control group did not correlate with demographic variables, education, pain, fatigue, depression or the number of comorbidities. Work type by professional category did not differ significantly between subjects with SLE and controls. Patients with SLE with physically demanding work reported worse presenteeism compared with controls with similar jobs (77% vs 85%, p<0.05), but absenteeism in SLE did not significantly differ between physically demanding jobs and other types. Patients with SLE with the most cognitively demanding jobs reported greater absenteeism (5.9 h) compared with controls (6.9 overtime hours, p<0.05) and also impaired presenteeism (75% vs 85% in controls, p<0.001). Thus, both extremes of physical and cognitive duties presented difficulty to subjects with SLE compared with controls (data not shown).

As sociodemographic factors and SLE-related variables have been associated with WD, a logistic regression with forward selection was performed with the dependent variable of WD. Independent variables of age, disease duration, BCSI score, SLICC/DI score, pain, fatigue, depressive symptoms, non-African­–American race and education were chosen as covariates in the model based on univariate findings and literature associations. Age (β=0.04, p=0.008), SLICC/DI score (β=0.24, p=0.03), education (β=-0.75, p<0.001) and non-African–American race (β=-0.89, p=0.01) were independently associated with the outcome of WD, while BCSI score, disease duration, pain, depressive symptoms and fatigue score were not independently associated with WD (table 4). Individual comorbidities that were significant when added singly to the multivariate regression included osteoporosis (p=0.02), depression (p=0.025) and thyroid disease (p=0.05). However, these comorbidities did not materially affect the association of age, race, damage score and education with the outcome of WD and were not retained in the model. Other comorbidities including neuropathy, fibromyalgia, osteoarthritis, cancer, anaemia, liver disease, kidney disease, peptic ulcer disease, diabetes mellitus, hypertension and cerebrovascular accident did not individually associate with WD status (data not shown).

Similar forward selection multivariate regressions were used to analyse high presenteeism among employed/employable patients with SLE after dropping insignificant variables using STEPWISE method. Age, race, disease duration, education, SLICC/DI score, summed comorbidities, pain and fatigue did not correlate with the outcome of high presenteeism, while depression score (β=−0.192, p<0.0001) and being married (β=−1.237, p=0.013) did inversely correlate with high presenteeism (data not shown). Multivariate regression was not performed on absenteeism due to the paucity of correlates identified on univariate analysis.


We have developed a self-administered questionnaire based on the WHO HPQ to quantify work function, absenteeism and WD in SLE. This questionnaire successfully quantified increased WD as previously described in SLE. Thirty-one per cent of patients with SLE reported WD, and 88% of non-working patients with SLE ascribed work limitations to their health. This rate of WD was higher than controls (table 2).9 ,10 ,19 ,23 Non-working patients with SLE had much longer duration of unemployment compared with controls (7.2 vs 2.2 years). Both highly physical and highly cognitive jobs represented challenges to employed patients with SLE.

On logistic regression, only age, SLICC/DI score, African–American race and education were found to correlate with WD status, similarly to what has previously been described.9 ,10 ,19 ,23 Among parameters associated with WD on multivariate analysis, few are amenable to direct intervention using current treatment options to improve work outcomes. However, newer therapeutics may show effects in the future and these effects may help justify the costs of new biological therapies.

Increased absenteeism from work in SLE was documented in these questionnaires. The average deficit in SLE was 2.7 h/week, while the control subjects worked an excess of 4.7 h, when measured as a week-to-week variable. While this did not reach statistical significance due to the high variability of responses, this number is likely to be sensitive to change over time. Patients with SLE reported greater time disabled from home activities, averaging 7.4 days/month compared with 1.8 days/month in control subjects. Two previous studies quantified absenteeism on a yearly basis rather than ‘absenteeism in the last week’ used in this study.6 ,26 Our approach of intermittently measured short-term absenteeism should be more dynamic in measuring improved function in patients with lupus. Longitudinal use of this questionnaire may be a helpful tool in tracking important aspects of health over time.

This is the first study to measure presenteeism in SLE. Presenteeism represents the patients’ self-reported work performance over the previous 4 weeks, adjusted for the SLE subjects’ assessment of coworkers. High presenteeism was associated with low pain and low fatigue levels, good BCSI scores and negatively correlated with depressive symptoms on bivariate analysis. On multivariate regression, only lack of depressive symptoms correlated with better presenteeism. This finding may indicate that presenteeism is a unique measure without major sociodemographic determinants. Conversely, it may primarily function as a reflection of concurrent depression. This suggests that affective disorders may be a mediator of work performance and is an amenable target for therapeutic interventions. Addressing and treating affective disorders may decrease later WD. However, as longitudinal data are collected using presenteeism as an outcome measure, we may be able to determine whether presenteeism quantifies a novel dimension of function rather than reflecting depression.

The limitations in this study include the cross-sectional nature of this study, which precludes longitudinal analysis, and the lack of validation of internal consistency by repetition in individual patients. However, the associations of WD with demographic factors and absenteeism with current symptomatology give strong face validity to the accuracy of our questionnaire. An area of improvement in the next version of the questionnaire will be in the employment categorisation by including ‘work disabled and not working’ category on the employment descriptions.

The majority of this questionnaire characterises work activities in the last 28 days, or even in the last 7 days. Thus, recall bias should be somewhat reduced by this short interval. However, subjects who are ill may still tend to over-report some factors, such as the physical and mental demands of the job, because they may struggle at otherwise routine tasks compared with healthy controls. The short time interval covered by this questionnaire may also make this useful in a longitudinal study, to track employment issues over time on various therapies, etc.

Self-administered questionnaires exploring patient functionality in a variety of domains provide greater insight into the lives of patients with SLE, document severity of disease, and may help to measure progression or improvement in disease. The current study has determined acceptability and face validity of this questionnaire as an instrument due to high completion rate and good response rates among SLE subjects and controls and the findings are in alignment with previous research. Future research should use these measures longitudinally to determine sensitivity to change concordant with clinical disease course and/or treatment in this challenging disease.


Genentech is the financial supporter for this study. Genentech financed Harris Interactive, a research firm tasked with managing the logistics of the study and reporting the results and to help in writing this manuscript.

  • Received August 12, 2014.
  • Revision received November 14, 2014.
  • Accepted December 10, 2014.


View Abstract


  • Contributors All the authors have contributed equally to the planning, conduct and reporting of the work described in this article.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval Western Institutional Review Board and, when appropriate, academic Institutional Review Boards at participating physicians’ universities.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.