Remote monitoring using electronic patient-generated health data (ePGHD) is an emerging tool for patients with long-term conditions [1]. These data are collected independently by patients [2], either actively at specified time points (eg, self-reported symptom data) or passively and continuously (eg, sensor data) [3,4]. Smartphone apps and wearable devices can routinely collect ePGHD outside usual clinical environments [4,5].

Frequently and longitudinally collected ePGHD can reveal detailed between-visit insights into patient conditions [6]. They include new types of health data not traditionally collected or evaluated, enriching regular clinical assessments [7] and allowing more accurate identification of patterns over time. Understanding fluctuating disease activity trajectories is key to the care of patients with long-term conditions, but can be hampered in usual clinical settings by infrequent reviews and recall bias [8].

Assessing treatment response is a promising and novel application of ePGHD [9,10]. Traditionally, treatment response is measured at predetermined, infrequent intervals dictated by clinical trial or clinic appointment visits [11], posing challenges in capturing day-to-day symptom patterns. This is particularly problematic for drugs with short treatment effects because patients are unlikely to accurately recall their treatment response at follow-up visits that often take place several months later. Retrospective assessment may over- or underestimate responses [12]; one example is intramuscular steroid therapy in rheumatoid arthritis (RA). Despite its widespread use [13], studies evaluating this therapy’s effectiveness are scarce and heterogeneous in both methodology and results [14-19]. This knowledge gap means that rheumatologists struggle with a common question: “How long will this steroid injection last?”

High-frequency ePGHD, such as daily patient-reported symptoms collected remotely via apps between appointments, promise to bridge this gap. However, there is no established definition for what entails the response to intramuscular steroid therapy using ePGHD. Despite the growing interest in ePGHD [20], the absence of established methods for analyzing and interpreting ePGHD hinders its potential to improve clinical practice and patient outcomes [1]. Rather than drawing conclusions on treatment response, this paper provides a first step toward developing such definitions through an exploratory analysis of ePGHD in 6 patients who received intramuscular steroid injections.

Our case series included 6 patients who received ≥1 intramuscular steroid injection with a total symptom-tracking period of 166-301 days during the REMORA study. The steroid-response tracking period was 42-56 days postinjection, with a data-completion range of 41%-79% of possible days during the 8-week follow-up period; most participants contributed symptom data on at least 50% of days.

Participants included 4 men and 2 women (age range 30-65 years); 5 of 6 were White British. At study recruitment, disease duration varied from newly diagnosed (<1 year) to long-standing (≥12 years); DAS-28 score range was 3.4-7.2. Concomitant DMARDs used during the study period included methotrexate, sulfasalazine, and hydroxychloroquine. Three patients changed DMARDs during the tracking period.

In total, 9 injections were given in the study period, with 1 participant receiving 3 injections over 181 days; we considered each injection a separate case in our series. Steroids included triamcinolone or methylprednisolone; doses ranged from 120 to 180 mg.

Table 2 summarizes our results. Average pre-injection pain-score range was 3.3-9.3. Using our definitions, 7 injections demonstrated a response and 2 did not. Among responders, response duration range was 1-54 (median 9, IQR 7-41) days; average pain score improved 0.1-5.3 (median 3.3, IQR 1.7-4.4) points; maximum pain score improved 0.1-7.0 (median 4.3, IQR 1.7 to 6.0) points. Among nonresponders (injections 8 and 9), maximum pain score improvement was a negative value indicating the degree of worsening between the pre-injection and day 1 pain score; these negative values have been reflected in the overall median/IQR results reported for maximum pain score improvement in Table 2 (median 3.6, IQR -0.15-5.2). Figure 1 graphically demonstrates patient-level treatment response trajectories for each steroid injection.

Our study describes a case series of patients who used a smartphone app to collect high-frequency ePGHD, providing a novel opportunity for detailed insights into treatment response. We created and applied definitions to determine the presence, duration, and extent of treatment response to intramuscular steroid administration. Using these definitions, 7 of 9 injections demonstrated a response, with variable extent and duration. We also present graphical representations of ePGHD, which is an innovative way of communicating and understanding the treatment response. Interestingly, the graphical patterns were inconsistent with established beliefs among rheumatologists. Steroid injections 6, 7, and 8 most closely represented expected treatment-response trajectories (ie, rapid, sustained pain-score improvement over months); however, injection 8 did not meet our definition for treatment response, as the day 1 pain score was higher than baseline. Thus, treatment response definitions must be developed to account for within-person variation in symptom scores, response start times, and overall trajectories, as would be usual in clinical practice.

The study population was limited, as this was a secondary use of data from REMORA, which was not designed to evaluate treatment responses. Furthermore, our definitions were designed as a proof of concept for this novel approach in a small dataset and are admittedly arbitrary. Nonetheless, this study is an important exploratory first step toward a robust methodology to evaluate the treatment response using ePGHD. The highlighted issues require addressing before this method is applied to larger datasets and meaningful definitions are developed. For instance, we used pain score data because pain is a common indication for steroid injections; but patients may feasibly demonstrate responses in other domains. Indeed, our cohort’s variation in baseline pain score points to other indications for steroid injection. The most relevant domains need to be identified by exploring the data and collaborating with stakeholders. Furthermore, symptom scores are inherently subjective and can have considerable interuser variability, hindering comparison of scores between individuals. Definitions must be developed that acknowledge within-person changes and account for confounding factors such as comorbidities and concomitant medications. New treatment response definitions also need validation against existing clinical-response measures [24], acknowledging that the infrequency of clinical assessments limits existing measures. Finally, potential barriers to mobile health interventions, such as the burden of daily symptom tracking and health equity considerations that can influence access to smartphones, need to be explored and mitigated [25].

In conclusion, our analysis demonstrates that ePGHD may facilitate novel ways of evaluating treatment responses, potentially filling a long-standing knowledge gap regarding treatment responses to intramuscular steroids and treatment trajectories after any intervention. Further development of ePGHD analysis methodologies across larger populations is required to develop meaningful consensus definitions for treatment responses using longitudinal, patient-reported symptom data.