Clinical trials have shown that on a group level, patients with early rheumatoid arthritis (RA) treated with initial combination therapy achieve earlier decrease in disease activity, improvement in functional ability and less radiographic joint damage progression than patients treated with initial monotherapy [1]-[7]. However, for individual patients there is a need for individualized treatment. The 2010 European League Against Rheumatism (EULAR) recommendations stated that `patients with a favourable prognosis very often respond similarly to low-intensity monotherapy or intensive medication strategies?, suggesting that for patients with a poor prognosis this might be different [8]. It was also formulated that `occasional patients with a particular need for rapid, highly effective intervention, may benefit from starting a biological agent plus methotrexate as a viable and useful option?, which was built on the idea that `patients with poor prognostic factors have more to gain? [8]. This opinion was abandoned in the updated 2013 recommendations, but these also state that `risk stratification is an important aspect of the therapeutic approach to RA? [9], detailing that after failure to achieve low disease activity on methotrexate monotherapy, `in patients with a low risk of poor RA outcome, another conventional synthetic disease-modifying antirheumatic drug (DMARD) strategy would be preferred, while in patients with a high risk, the addition of a biologic DMARD would be preferred? [9]. Hence, the recommendations encourage rheumatologists to use risk stratification in daily practice and to implement a personalized approach in the treatment of patients with RA.

In this post hoc analysis of the BeSt study, we investigated whether patients with poor or non-poor prognostic factors (based on previously developed prediction models [10]-[13]) respond differently to initial monotherapy, and whether patients with a poor or non-poor prognosis respond differently to initial combination therapy, as suggested by the EULAR recommendations. Furthermore, we studied the efficacy of a second conventional synthetic DMARD in patients with a low risk of poor RA outcome who failed on the first.

Here, the results of defining PP patients by the presence of ≥3 of 4 poor prognostic factors (and consequently the non-PP patients by the presence of ≤2 of these factors) are discussed. The results of prognosis stratification according to the RRP matrix model of Visser et al. [10] are shown in Additional files 1, 2, 3 and 4.

Of 508 patients, 417 (82%) were classified as having a poor or a non-poor prognosis based on the available data. Of the 192/417 patients (46%) with PP, 100 (52%) had been randomized to initial monotherapy and 92 (48%) to initial combination therapy. Of 225/417 patients (54%) with a non-PP, 100 (44%) were treated with initial monotherapy and 125 (56%) with initial combination therapy.

Clinical outcomes after three months follow up

Significantly more PP patients who were treated with initial combination therapy fulfilled the ACR20 response criteria after three months than those treated with initial monotherapy (70% versus 38%, P <0.001). This was the same for ACR50 response (48% versus 13%, P <0.001), ACR70 response (24% versus 4%, P <0.001) and for DAS remission (17% versus 5%, P =0.016). Patients treated with combination therapy had a significantly greater improvement in functional ability (median decrease in HAQ 0.75 versus 0.38, P <0.001). This resulted in a mean HAQ score at 3 months of 0.60 in patients treated with initial combination therapy compared to a mean HAQ score of 1.08 in patients treated with initial monotherapy (see also Figure 2).

Non-PP patients treated with initial combination therapy more often met the ACR response criteria at three months compared to those treated with initial monotherapy; ACR20 (71% versus 44%, P <0.001), ACR50 (49% versus 13%, P <0.001) and ACR70 (17% versus 3%, P =0.001). They also showed more DAS remission (18% versus 7%, P =0.017) and a larger increase in functional ability (median decrease in HAQ score 0.63 versus 0.38, P <0.001). After three months, mean HAQ score was 0.59 in non-PP patients treated with initial combination therapy and 0.92 in those treated with initial monotherapy (see also Figure 2). In Table 2 the main results are summarized. With regression analyses similar results were obtained (data not shown).

	Poor prognosis patients
	Initial mono	Initial combo	P-value
DAS remission
After 3 months	5 (5)	15 (17)	0.016
After 1 year	21 (21)	31 (36)	0.034
ACR20 response
After 3 months	35 (38)	57 (70)	< 0.001
After 1 year	73 (80)	75 (93)	0.026
ACR50 response
After 3 months	12 (13)	40 (48)	<0.001
After 1 year	52 (57)	59 (71)	0.060
ACR70 response
After 3 months	4 (4)	20 (24)	<0.001
After 1 year	28 (30)	35 (44)	0.081
Decrease in HAQ score, median (IQR)
After 3 months	0.38 (-0.63, 0.06)	0.75 (-1.13, -0.25)	<0.001
After 1 year	0.75 (-1.13, -0.38)	0.88 (-1.38, -0.38)	0.110
SHS progression
After 1 year, median (IQR)	1.5 (0, 5.0)	0 (0, 2.0)	0.001
RRP	24 (26)	8 (10)	0.006
	Non-poor prognosis patients
	Initial mono	Initial combo	P -value
DAS remission
After 3 months	7 (7)	23 (18)	0.017
After 1 year	35 (36)	43 (36)	1.000
ACR20 response
After 3 months	38 (44)	79 (71)	<0.001
After 1 year	63 (72)	96 (85)	0.024
ACR50 response
After 3 months	12 (13)	56 (49)	<0.001
After 1 year	44 (52)	77 (68)	0.027
ACR70 response
After 3 months	3 (3)	20 (17)	0.001
After 1 year	29 (33)	45 (39)	0.380
Decrease in HAQ score, median (IQR)
After 3 months	0.38 (-0.75, 0)	0.63 (-1.13, -0.25)	<0.001
After 1 year	0.63 (-1.13, -0.13)	0.88 (-1.25, -0.31)	0.040
SHS progression
After 1 year, median (IQR)	0 (0, 1.5)	0 (0, 1.0)	0.451
RRP	10 (11)	4 (4)	0.054

Numbers indicate number of patients (percentage) unless indicated otherwise. ACR response: according to the American College of Rheumatology criteria [17]; DAS remission, disease activity score <1.6 [16]; Initial combo: initial combination therapy with either prednisone or infliximab; Initial mono: initial monotherapy with methotrexate; non-poor prognosis (presence of ≤2 of 4 poor prognostic factors); HAQ, health assessment questionnaire (scale 0 to 3); poor prognosis (presence of ≥3 of 4 poor prognostic factors); SHS, Sharp van der Heijde score; RRP, rapid radiographic progression, defined as increase in Sharp van der Heijde score ≥5 points during the first year.

The results of this post hoc analysis in the BeSt study show that patients with recent-onset RA with a non-poor prognosis and patients with a poor prognosis respond similarly to the treatment strategy options. Both groups benefit more from initial combination therapy than from initial monotherapy and the success of a second conventional DMARD after failing on the first is limited in both groups.

Previous studies have shown that initial combination therapy results in better clinical and radiographic outcomes than initial monotherapy in patients with early RA on a group level [1]-[5],[7]. It was suggested in the 2010 recommendations for the management of RA [8] that patients with favourable prognostic factors at baseline do not need initial combination therapy because they will respond equally well to initial monotherapy and that patients with a poor prognosis would benefit more from initial combination therapy. This was revoked in the 2013 update: it is now recommended that all patients should receive a similar initial treatment [9]. Still, the updated recommendations state that risk evaluation is an important aspect in the therapeutic approach of RA, and that patients with a favourable prognosis would require a different type of follow up treatment than patients with a poor prognosis after failure on initial MTX monotherapy [9].

To test these recommendations, we classified patients as having a poor prognosis (PP) or a non-poor prognosis (non-PP), as a representative of the heterogeneous group of patientsoutcome’ low risk of poor RA outcome’ mentioned in the updated 2013 recommendations, based on well-known and frequently used risk factors [10]-[13]. We tested whether these risk groups, over three-monthly evaluations in the first year of the BeSt study, responded differently to these treatments.

We found that in both PP and non-PP patients, initial combination therapy is more effective, compared to monotherapy, in inducing an early (that is, after three months) decrease in disease activity and early improvement in functional ability, this notwithstanding the fact that after six months on MTX monotherapy significantly more non-PP patients than PP patients achieved a low DAS (64% versus 43%). The improvement in functional capacity in patients treated with initial combination therapy was equal in PP and non-PP patients, both after three months and after one year. This indicates an early equal gain in functional capacity in both prognosis categories. These differences in clinical outcomes are explicit after three months, and remain, following treat-to-target adjustments in therapy, only marginal after one year.

There was no difference among PP and non-PP patients in response to SSA as the second conventional synthetic DMARD after failure to achieve a low DAS on initial MTX monotherapy: similarly low percentages of patients achieved a DAS ≤2.4 (21% of non-PP patients and 28% of PP patients). This appears to be at odds with recommendation 8 of the updated 2013 EULAR recommendations for the management of RA [9].

Overall, as a consequence of the definition of poor or non-poor prognosis, patients with a non-PP showed less radiographic joint damage progression than patients with a poor prognosis. After one year of targeted treatment, significantly less radiographic joint damage progression occurred after initial combination therapy in PP patients than after initial monotherapy. Thus it appears that for radiographic damage progression indeed, as originally formulated in the 2010 EULAR recommendations for the management of RA, PP patients `have more to gain' from the initial treatment choice [8].

Our definition of poor or non-poor prognosis was based on factors that are associated with (rapid) radiographic progression and are also used in prediction models [10]-[13]. However, early treatment initiation and targeted therapy, including the option of biologic DMARDs, have contributed to prevent this disease outcome in most BeSt patients to date. As RRP nowadays can also be better prevented with early effective treatment, models designed to predict RRP perform moderately in clinical practice. In addition, they do not provide information on clinical outcomes. Of the patients defined as PP according to the presence of ≥3 risk factors, only 26% actually developed RRP when treated with initial monotherapy and 10% developed RRP when treated with initial combination therapy. When PP is defined according to the matrix model of Visser et al. [10], 46% and 12% developed RRP when treated with initial monotherapy or combination therapy, respectively (Additional file 2). Thus, despite familiarity with prognostic factors, it is still difficult to predict the prognosis.

Consequently, it is proper to evaluate the efficacy of the initial treatment choice in terms of rapid relief of symptoms and functional improvement due to suppression of inflammation. Our data show that initial combination therapy is more successful in achieving these outcomes than initial MTX monotherapy, both for PP patients and for non-PP patients. In fact, clinical responses were very similar (and satisfactory) in all patients if they received initial combination treatment. In addition, although maybe not clinically relevant, PP patients showed less radiographic damage progression after initial combination therapy than after initial monotherapy. Also, more than half of the patients receiving initial combination therapy could discontinue prednisone or infliximab due to low disease activity, as soon as the protocol allowed drug discontinuation.

There was no significant difference in the number of AEs and SAEs reported by PP or non-PP patients on initial combination therapy. Similar toxicity among the four treatment arms has already been reported [6],[7]. Hence, it appears that extra caution for the use of combination therapy in either group is not warranted.

The definition of non-poor or poor prognosis shows a moderate performance in predicting radiographic progression, despite the use of two different methods and based on risk factors in validated prediction models. Overall, patients in the BeSt study benefitted from initial combination therapy with better clinical outcomes and more functional improvement at three months than after initial monotherapy, regardless of prognosis category. Response to a second conventional synthetic DMARD after failure on methotrexate monotherapy was similar in patients with a poor or a non-poor prognostic profile, and generally disappointing. These results suggest that prognostic factors associated with future radiographic damage progression contribute little to predict early clinical response to initial treatment, and therefore, in our opinion tailored treatment based on prognosis as suggested by the EULAR guidelines is currently not feasible. The choice of treatment strategy may depend less on these prognostic factors and more on the estimated need for rapid relief of symptoms and limitations due to active disease in our patients.

IM performed the statistical analysis, interpreted the data and drafted the manuscript. JVB, HH, PL and AS contributed in the acquisition of the data and critically revised the manuscript for important intellectual content. PK, WL and TH participated in the study design, contributed in the acquisition of the data and were involved in critically revising the manuscript for important intellectual content. CA participated in the study design, contributed in the acquisition of the data and was involved in analyzing and interpreting the data and helped to draft the manuscript. All authors read and approved the final version of the manuscript.