Article Text

Extended report
Clinical factors, anticitrullinated peptide antibodies and MRI-detected subclinical inflammation in relation to progression from clinically suspect arthralgia to arthritis
  1. Hanna W van Steenbergen1,
  2. Lukas Mangnus1,
  3. Monique Reijnierse2,
  4. Tom W J Huizinga1,
  5. Annette H M van der Helm-van Mil1
  1. 1Department of Rheumatology, Leiden University Medical Center, Leiden, The Netherlands
  2. 2Department of Radiology, Leiden University Medical Center, Leiden, The Netherlands
  1. Correspondence to Hanna W van Steenbergen, Department of Rheumatology, Leiden University Medical Center, P.O. Box 9600, Leiden 2300 RC, The Netherlands; h.w.van_steenbergen{at}


Introduction Patients with clinically suspect arthralgia (CSA) have, according to their rheumatologists, an increased risk of rheumatoid arthritis (RA), but their actual outcome is unexplored. This longitudinal study investigated (1) progression from CSA to clinically detectable arthritis and (2) associations of clinical factors, serological factors (among which are anticitrullinated peptide antibodies (ACPAs)) and MRI-detected subclinical inflammation with arthritis development.

Methods 150 patients with CSA were followed for ≥6 months. At baseline, clinical and serological data were collected and unilateral 1.5 T-MRI of metacarpophalangeal (MCP), wrist and metatarsophalangeal (MTP) joints was made. MRI scoring was done according to the RA MRI scoring system. Subclinical MRI inflammation was defined based on MRI results of 193 symptom-free persons.

Results During follow-up (median=75 weeks, IQR=41–106 weeks), 30 patients developed clinical arthritis; 87% did so <20 weeks after inclusion. In multivariable analyses, age, localisation of initial symptoms in small and large joints (compared with small joints only), C-reactive protein level, ACPA-positivity and subclinical MRI inflammation significantly associated with arthritis development; ACPA and MRI inflammation were most strongly associated (HR (95% CI) respectively, 6.43 (2.57 to 16.05) and 5.07 (1.77 to 14.50)). After 1-year follow-up, 31% of the patients with MRI inflammation and 71% of the ACPA-positive patients with MRI inflammation had progressed to arthritis. Forty-three per cent of the patients that developed arthritis within 1 year were ACPA-negative; 78% of them had subclinical MRI inflammation at baseline. When MRI inflammation was absent arthritis development was infrequent (6% in all patients with CSA and 3% in ACPA-negative patients with CSA).

Conclusions Subclinical MRI inflammation precedes clinical arthritis with a few months. Subclinical MRI inflammation is, independent of other factors such as ACPA, associated with arthritis development.

  • Rheumatoid Arthritis
  • Early Rheumatoid Arthritis
  • Magnetic Resonance Imaging
  • Outcomes research

Statistics from

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.


There is an upcoming need to identify individuals in the very early phase of rheumatoid arthritis (RA) in which clinically apparent arthritis is not yet present. Although not proven, it is assumed that treatment initiation in this phase enables better disease modification and might contribute towards preventing arthritis becoming chronic. The first possible moment to clinically recognise patients at risk of RA is the phase of symptoms without clinically detectable arthritis.1 The symptoms that are specific for this phase are not yet identified, but clinical expertise might be an accurate tool to select patients with arthralgia with an increased risk of RA.2 Patients with arthralgia that, based on their symptoms and signs, have an increased risk of RA according to their rheumatologists, are indicated to have clinically suspect arthralgia (CSA).3 The approach to select patients on clinical grounds before ordering additional tests is in line with clinical care and allows identifying autoantibody-positive and autoantibody-negative RA in the phase before clinically detectable arthritis.

Thus far, the long-term outcome of patients with arthralgia that were clinically suspect for progression to RA has not been investigated extensively. Moreover, the value of risk factors or tests in patients with CSA is unexplored. Two previous studies on patients with unspecified arthralgia or aspecific musculoskeletal symptoms who had RA-related autoantibodies revealed that morning stiffness, joint tenderness and (high levels of) anticitrullinated peptide antibodies (ACPAs) were associated with arthritis development.4 ,5 However, the prognostic value for arthritis development of clinical and serological factors in patients with CSA is still unknown.

Also the value of advanced imaging in patients with CSA is unexplored. In a previous cross-sectional study, we observed that 44% of the patients with CSA had MRI-detected subclinical inflammation of hand and foot joints and that these patients with subclinical inflammation could not be adequately identified by presence of clinical or serological characteristics, suggesting that MRI-detected inflammation may have some diagnostic value. Though the predictive value of MRI-detected inflammation has still to be determined,3 an advantage of MRI is its sensitivity to detect inflammation.6 ,7 MRI depicts synovitis, tenosynovitis and bone marrow oedema (BME), that is also called ‘osteitis’ in RA.8 ,9 Because the specificity of MRI-detected inflammation has been studied scarcely, we recently performed MRI of hands and feet in 193 symptom-free persons.10 These data served as reference and allowed to define MRI-detected subclinical inflammation for the present study.

In this first longitudinal study on patients with CSA, we aimed to determine (1) progression to clinically detectable arthritis, (2) the association of clinical and serological factors (among which are ACPAs) with progression to clinical arthritis, (3) the association of subclinical MRI inflammation with progression to clinical arthritis and (4) whether subclinical MRI inflammation has an additive value compared with the other mentioned risk factors.



All patients were included in the CSA cohort which is described previously in detail elsewhere.3 This inception cohort was set up in 2012 in the Leiden University Medical Center (Netherlands), which is the only referral centre in a healthcare population of >400 000 inhabitants to study the symptomatic phase of RA without clinically detectable arthritis. Inclusion criteria were having arthralgia of the small joints for <1 year that was, according to the clinical expertise of the rheumatologist, suspected to progress to RA over time. No further criteria were made with regards to the type of symptoms and thus inclusion was essentially based on the expert opinion of the rheumatologist. Importantly, CSA was not present if clinical arthritis was observed at physical examination or another explanation for the arthralgia was more likely (eg, osteoarthritis and fibromyalgia).

At baseline, questionnaires (among others on work ability, the Health Assessment Questionnaire and Short-Form health survey-36) were completed, physical examination performed, blood obtained (among others for determination of ACPA (anti-cyclic citrullinated peptide 2 (anti-CCP2), positive if >7 U/mL, Eurodiagnostica, Netherlands) and IgM rheumatoid factor (RF) (positive if >3.5 IU/mL) and an MRI performed.3

Patients were prospectively followed with scheduled visits at 4 months, 12 months and 24 months. If necessary (for instance when the patient experienced more symptoms or noticed a swollen joint) patients were seen in between the scheduled visits by their rheumatologist. Follow-up ended earlier when clinical arthritis had developed.

For the present study, the patients with a follow-up duration of ≥6 months were selected; these patients (n=150) were included between April 2012 and July 2014. None of the patients with CSA were treated with disease-modifying antirheumatic drugs (DMARDs) or (systemic and local) glucocorticoids prior to inclusion and during follow-up.

MRI scanning and scoring

MRI of the metacarpophalangeal (MCP)2–5, wrist and metatarsophalangeal (MTP)1–5 joints of the most painful side, or the dominant side in case of equally severe symptoms at both sides, was performed ≤2 weeks after clinical assessment. Patients were asked not to use non-steroidal anti-inflammatory drugs (NSAIDs) during 24 h before MRI. The joints were scanned with an musculoskeletal (MSK)-extreme 1.5T-MRI scanner (GE, Wisconsin, USA) using contrast enhancement and according to the RA MRI scoring system (RAMRIS) protocol. See online supplementary file 1 for a detailed scan protocol.

Synovitis and BME in the MCP, wrist and MTP joints were scored according to the Outcome Measures in Rheumatology Clinical Trials (OMERACT) RAMRIS,8 the carpometacarpal (CMC)-1 joint (trapezium and base metacarpal-1) was excluded. Tenosynovitis in the wrist and MCP joints was assessed as described by Haavardsholm et al.9 The sum of the synovitis, BME and tenosynovitis scores was the total MRI inflammation score. Scoring was performed by two independent trained readers (HWvS, LM) blinded to clinical data. The within-reader intraclass correlation coefficients for the total MRI inflammation score were 0.99 and 0.98; the between-reader interclass correlation coefficient was 0.96. Mean scores of the two readers were studied.

At the time of analyses, MRIs were also categorised into positive or negative for subclinical MRI-detected inflammation. Frequencies of MRI-detected synovitis, tenosynovitis and BME that were observed at the same anatomical location in symptom-free persons recruited from the general population of the same age category were used as reference.10 In these symptom-free persons it was observed that MRI-detected inflammation was prevalent, especially at higher age and at preferential locations. Subclinical MRI inflammation was considered present if (1) both readers scored that joint (or bone in case of BME) positive for MRI inflammation and (2) the score obtained at a joint/bone was present in <5% of age-matched symptom-free persons. For example, subclinical MRI inflammation was present when both readers scored grade ≥1 for synovitis at MCP3 in a 30-year-old patient. The MRI was considered negative if only one reader scored grade 1 and the other reader grade 0. If the patient was aged 50 years and both readers scored grade 1, the MRI was also negative as ≥5% of symptom-free persons of the same age category had also synovitis grade 1 at MCP3.10 The MRI results were not reported to the treating rheumatologist.


The main outcome was development of arthritis detected at physical examination (66 joints were assessed) by the rheumatologist. If arthritis was achieved, follow-up in the CSA cohort ended. In sensitivity analyses another outcome, initiation of DMARD therapy (including steroids), was studied. Medical files of all patients were studied on these outcomes until 24 December 2014.

Statistical analyses

Cox proportional hazards regression analyses were used. Time to clinical arthritis was the time from inclusion to the date of first detection of clinical arthritis. Patients who did not develop arthritis were censored at the date of the 2 years’ follow-up visit or at the date that all medical files were studied on arthritis development. Additionally, the diagnostic performance of ACPA positivity and presence of subclinical MRI inflammation were evaluated for arthritis development at 1 year follow-up. Details on the statistical methods are presented in online supplementary file 2).


Baseline clinical characteristics

One hundred and fifty patients with CSA were studied. Table 1 presents the baseline characteristics. Mean age of the studied patients was 43.2 years (SD 12.9) and 72.7% were female. The median symptom duration was 18 weeks (IQR 9–30) and 16.0% were ACPA-positive (table 1).

Table 1

Baseline clinical and MRI characteristics of all patients and separately for the patients that have and have not developed clinical arthritis during follow-up

Development of clinical arthritis

During follow-up one patient developed gout. This patient was excluded from further analyses as the patient did not belong to the non-arthritis group and the diagnosis was outside the spectrum of chronic arthritis/RA. The remaining 149 patients with CSA had a median follow-up duration of 75 weeks (IQR 41–106). Within this follow-up period 30 patients developed clinically detectable arthritis. At arthritis development, 23 patients were diagnosed with RA (according to the 2010 American College of Rheumatology/European League Against Rheumatism criteria), 6 with undifferentiated arthritis and 1 with psoriatic arthritis.

The median time period between inclusion and arthritis development was 7 weeks. Of all patients that progressed to arthritis, 87% had done so within 20 weeks after inclusion (figure 1A).

Figure 1

Development of clinical arthritis over time for all patients and for the patients with and without MRI-detected inflammation separately. Presented are the curves for development of clinical arthritis over time in all 149 patients with clinically suspect arthralgia (CSA) and for the patients with a positive and a negative MRI separately (six patients did not undergo an MRI). The HR of developing arthritis with a positive MRI was 6.12 (95% CI 2.32 to 16.19, p<0.001). All patients were followed for ≥6 months (median follow-up duration 75 weeks). The vertical lines indicate that a patient is censored.

Clinical factors and ACPA in relation to arthritis development

In order to investigate whether baseline clinical factors and ACPA were associated with progression from CSA to clinical arthritis, univariable Cox regression analyses were performed (table 2). An increased hazard on developing arthritis was observed for patients that presented with initial symptoms located in the small and large joints (HR=5.28 compared with small joints only (95% CI 2.38 to 11.73, p<0.001), patients with higher C reactive protein (CRP) levels (HR=1.06/mg/L, 95% CI 1.03 to 1.09, p<0.001), RF-positive patients (HR=6.94, 95% CI 3.34 to 14.43, p<0.001) and ACPA-positive patients (HR=10.07, 95% CI 4.87 to 20.82, p<0.001)). Age, presence of morning stiffness and number of tender joints were not significantly associated.

Table 2

Results of univariable Cox regression analyses of baseline clinical and serological factors in relation to arthritis development

Subclinical MRI inflammation in relation to arthritis development

In six patients MRI was not performed, because of (suspected) pregnancy (n=2), metallic foreign body in biceps tendon (n=1), logistical reasons (n=2) or development of clinical arthritis ≤2 weeks after inclusion (n=1).

Table 1 presents baseline MRI characteristics. Median continuous RAMRIS scores were low (online supplementary table S1 presents continuous RAMRIS scores for individual joints/bones). Univariable analyses showed that higher MRI inflammation scores were associated with arthritis development (table 3).

Table 3

Results of Cox regression analyses of MRI-detected subclinical inflammation at baseline in relation to arthritis development

Then the continuous MRI inflammation scores were dichotomised. Since it was recently observed that MRI-detected inflammation is also present in the general population and depends on age, anatomical location and type of inflammation,10 these were considered when defining an ‘abnormal MRI’. A joint (or bone in case of BME) was categorised as positive for inflammation when <5% of the general population of the same age category had inflammation at this location (online supplementary table S2 presents frequencies of positive joints/bone for MRI-detected inflammation). Sixty-six patients with CSA (45.8%) had a positive MRI for any subclinical inflammation, indicating that at least 1 joint/bone had synovitis, BME or tenosynovitis (table 1): 20.1% of the patients had 1 positive joint/bone, 18.8% 2–5 positive joints/bones and 6.9% ≥6 positive joints/bones (the maximum number of positive joints/bones was 24). When evaluating the individual MRI features, 25.7% of the patients had MRI-detected synovitis, 16.7% BME and 29.2% tenosynovitis (table 1).

Univariable Cox regression analyses with arthritis development as outcome revealed that presence of any MRI-detected subclinical inflammation at baseline was associated with a six times increased hazard on arthritis (HR=6.12, 95% CI 2.32 to 16.19, p<0.001, figure 1). In addition, the hazard on clinical arthritis increased when more joints/bones were scored positive for MRI inflammation (HR=1.23 per additional positive joint/bone, 95% CI 1.13 to 1.33, p<0.001). Evaluating the three MRI features separately showed the strongest association for MRI-detected tenosynovitis (HR=7.56), though MRI-detected synovitis and BME were also significantly associated with arthritis development (HR=2.22 and 2.39 respectively, all p<0.05, table 3). Because synovitis, BME and tenosynovitis were frequently present in the same patient (table 1), multivariable Cox regression analyses were done to determine which type(s) of MRI-detected inflammation were independently associated with arthritis development (table 3). We observed that MRI-detected tenosynovitis was independently associated (HR=8.39, 95% CI 3.38 to 20.81, p<0.001) with arthritis development.

Combination of clinical factors, ACPA and subclinical MRI inflammation in relation to arthritis development

Then, we questioned if the association of subclinical MRI inflammation with arthritis development was independent of the associations of other factors (age, initial localisation of the symptoms, CRP level, ACPA positivity). Multivariable Cox regression analyses (table 4) revealed an increased hazard for younger patients (HR=0.96 per year older, 95% CI 0.93 to 0.996, p=0.028), patients with initial localisation of symptoms in small and large joints (HR=4.30 compared with small joints only, 95% CI 1.70 to 10.86, p=0.002), patients with higher CRP levels (HR=1.05/mg/L, 95% CI 1.01 to 1.09, p=0.021), ACPA-positive patients (HR=6.43, 95% CI 2.57 to 16.05, p<0.001) and patients with presence of any MRI-detected subclinical inflammation (HR=5.07, 95% CI 1.77 to 14.50, p=0.002). Similar results were obtained when including continuous total MRI inflammation scores instead of MRI positivity (see online supplementary table S3). Hence, MRI-detected inflammation was associated with progression to clinical arthritis, independent of other clinical and serological factors.

Table 4

Results of multivariable Cox regression analysis of clinical and serological factors and MRI-detected subclinical inflammation at baseline in relation to arthritis development

Sensitivity analyses on initiation of DMARD treatment

Sensitivity analyses were performed with initiation of DMARD therapy as outcome. Twenty-five out of the 30 patients that developed clinical arthritis were started with DMARD treatment. Repeating the latter multivariable Cox regression analysis (including clinical and serological factors and presence of any MRI-detected inflammation) with DMARD initiation as outcome revealed similar results (data not shown).

Diagnostic value of ACPA and subclinical MRI inflammation

The previous analyses showed that the presence of subclinical MRI inflammation and ACPA were the two strongest and independent factors associated with arthritis development. We continued with determining test characteristics of both factors with arthritis development within the 1st year as outcome (table 5). To this end, analyses were restricted to the patients with CSA with 1-year follow-up who had data on ACPA and MRI (n=122). Of these, 21 (17.2%) had developed clinical arthritis within this year. Two patients developed arthritis after the 1st year had passed; these patients are now categorised in the non-arthritis group.

Table 5

Test characteristics of ACPA and subclinical MRI inflammation for arthritis development within 1 year

Test characteristics of ACPA

The sensitivity of ACPA for arthritis development was 57%, indicating that 57% of the patients with CSA that developed arthritis were ACPA-positive and 43% ACPA-negative. Of the patients that developed arthritis 24% were negative for ACPA and RF. The specificity of ACPA was 93%. The positive predictive value (PPV) of ACPA was 63%, indicating that 63% of ACPA-positive patients with CSA have developed clinical arthritis within 1 year (table 5).

Test characteristics of subclinical MRI inflammation

Subclinical MRI inflammation was present in 81% of the patients that have developed arthritis within 1 year (sensitivity=81%). The specificity was 63%. In case of a positive MRI, 31% of the patients progressed to arthritis (PPV) within 1 year and of all persons with a negative MRI at baseline only 6% developed arthritis (100% minus negative predictive value (NPV)) (table 5). These 6% concerned four patients; three of them developed initial clinical arthritis in a joint that was not depicted on MRI and one developed arthritis 17 weeks after inclusion in joints that were depicted on MRI.

Test characteristics of subclinical MRI inflammation within ACPA-negative and ACPA-positive patients with CSA

Test characteristics for subclinical MRI inflammation were determined within the ACPA-negative and ACPA-positive patients separately to evaluate the diagnostic value in the different subgroups (table 5). Although patient numbers became small, especially in the ACPA-positive subgroup, these stratified analyses indicate the value of subclinical MRI inflammation if ACPA results are known. Within the ACPA-positive patients 71% with a positive MRI progressed towards arthritis within 1 year (PPV=71%). Hence, whereas the chance on arthritis in case of ACPA positivity in the total CSA population was 63%, within the ACPA-positive patients with a positive MRI it was 71%. Of the ACPA-positive patients with a negative MRI 60% did not develop arthritis within 1 year (NPV). Within the ACPA-negative patients with CSA, a negative MRI made the chance on arthritis development very low (3%, 100% minus NPV). Furthermore, the sensitivity of a positive MRI in ACPA-negative patients was 78%. Thus, whereas 43% of the patients that developed arthritis had a negative ACPA test, 78% of these patients were identified by a positive MRI at baseline. Similar results were obtained in ACPA-negative and RF-negative patients (data not shown).


This longitudinal study of patients with CSA, with a median follow-up duration of 75 weeks observed that part of the patients with CSA progressed to arthritis. The majority already had subclinical MRI inflammation when presenting with CSA and progressed to arthritis within 4–5 months. These data indicate that the period of CSA and subclinical inflammation is relatively short and encompasses several months.

MRI-detected inflammation is one of the risk factors for arthritis development explored in this study. Thirty-one per cent of patients with CSA with a positive MRI progressed to arthritis within 1 year. Arthritis development within the 1st year was rare (6%) if the baseline MRI was negative.

The association of subclinical MRI inflammation with arthritis development was independent of other factors such as ACPA. Interestingly, the effect sizes in the multivariable analyses of both variables were in the same range. In clinical practice serological results are generally available before imaging tests are ordered. To get a better impression of the additional value of MRI, the analyses on the diagnostic value of MRI were performed in the ACPA-positive and ACPA-negative subgroups. This revealed that the risk of arthritis development within 1 year was 71% if ACPA-positive patients had a positive MRI. Additionally, 60% of the ACPA-positive patients with a negative MRI did not develop arthritis within 1 year. MRI was valuable in ACPA-negative patients with CSA as the majority (78%) of the ACPA-negative patients that developed arthritis had a positive MRI at baseline.

A strength of this study is that a positive MRI was defined using the prevalence of MRI features in the general population as reference, lowering the risk of false-positive MRI findings. Of the different types of MRI-detected inflammation (synovitis, BME, tenosynovitis), tenosynovitis was most predictive for arthritis development. Previous studies showed that MRI-detected tenosynovitis is frequently present in RA11 and rarely present in the general population.10

A limitation is that in the first 77 patients MRI of the feet was made without contrast enhancement. This may have affected the RAMRIS scores for synovitis on the feet. In this study, we scored MRIs without contrast conservatively, which may have resulted in an underestimation of inflammation on MRIs without contrast enhancement.12 Another limitation is the median follow-up duration of 75 weeks. It is unsure whether longer follow-up will change our results; nonetheless, we observed that the majority of patients had already progressed in the first months after inclusion.

Further replication studies are needed in other CSA populations before it can be decided if MRI is a useful tool in CSA in daily practice. In this study, MRI was used because an accepted validated scanning and scoring protocol exists,8 it is a minimal operator-dependent procedure and the prevalence of MRI-detected inflammation in the general population is known.10 These issues are not yet solved for ultrasound and further studies are needed to determine whether ultrasound might also be useful in CSA.

Previous studies on patients with symptoms without clinical arthritis evaluated patients with unspecified arthralgia.4 ,5 ,13 ,14 These studies identified morning stiffness and joint tenderness as predictors for progression to arthritis.4 ,5 These clinical factors were not associated with arthritis development in patients with CSA. This is presumably caused by the fact that patients with CSA were selected on the basis of their symptoms and signs. Indeed, the frequency of morning stiffness in CSA was higher than that in unspecified arthralgia.4 ,5

The reported risk of developing arthritis within 1 year in autoantibody-positive patients with unspecified arthralgia was 20–34% and up to 41–43% if other risk factors were present.4 ,5 Of patients with CSA that were ACPA-positive 63% progressed to arthritis within 1 year, suggesting that the predictive value of ACPA is higher in CSA than in unspecified arthralgia.

The present study is the first exploring the outcome of patients with CSA. We did not aim to derive a prediction rule because the current data set is too small to use part of the data for identification and the other part for validation. In addition, we anticipated that for accurate prediction more predictors are needed than those entered in our multivariable analysis. Further work is needed to this end.

In conclusion, the phase of CSA without clinically apparent arthritis but with subclinical inflammation encompasses several months. Present data suggest that MRI is diagnostically relevant in this disease phase. With regards to the role of MRI in identifying patients with an increased risk of arthritis, the absolute value of MRI may be higher in ACPA-negative then in ACPA-positive patients with CSA, as ACPA-positive patients with CSA already have a higher prior risk of arthritis development. Importantly, MRI is also useful to rule out imminent arthritis; patients with a clinical suspicion to progress to RA but a negative MRI had a low risk of developing arthritis. Further studies are needed to identify the set of variables that optimally identifies patients with RA in the phase of arthralgia without clinical arthritis and to examine if treatment in this phase is more effective than initiating treatment when clinical arthritis has developed.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

  • Lay summary

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.


  • Handling editor Tore K Kvien

  • Contributors HWvS, MR, TWJH and AHMvdHvM contributed to the conception of the work. HWvS and LM contributed substantially to the acquisition of the data and analysed the data. All authors contributed to the interpretation of the data. HWvS and AHMvdHvM wrote the first version of the manuscript and the other author revised it critically. All authors approved the final version of the manuscript and agreed to be accountable for all aspects of the work.

  • Funding The research leading to these results was funded by a Vidi-grant of the Netherlands Organisation for Scientific Research, a grant of the Dutch Arthritis Foundation, a grant of the Dutch Organization of Health Research and the FP7 grant TEAM.

  • Competing interests None declared.

  • Patient consent Obtained.

  • Ethics approval Local Medical Ethics Committee of Leiden University Medical Center.

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