Peer-reviewed journal articles and conference proceedings published in English were considered for inclusion. Eligible studies were required to examine or report contact force, contact pressure, or related sensor-skin interface conditions in wearable or wearable-relevant reflectance PPG and to provide relevant evidence on methodological characteristics, including measurement sites and device configurations, and PPG outcomes across signal-quality, waveform, feature, artifact-susceptibility, and downstream-task domains. Specifically, studies were required to meet all 3 core criteria: (1) the study must involve contact force, contact pressure, or sensor-skin interface conditions (including direct measurement, controlled application, or qualitative description); (2) the study must use wearable or wearable-relevant reflectance-mode PPG systems; and (3) the study must report at least one PPG-related methodological characteristic or outcome that was observed under, compared across, or interpreted in relation to the contact condition. In this review, contact force refers to the total mechanical load applied at the sensor-skin interface, whereas contact pressure refers to the distribution of that load over the effective contact area. Terms and units were retained as reported by the source studies and were not treated as interchangeable unless sufficient contact-area or geometric information was available.

Non–peer-reviewed publications, non-English articles, and studies without accessible full text were excluded. Studies were further excluded if they met 1 or more of the following criteria: (1) did not address contact force, contact pressure, or sensor-skin interface–related factors in the context of PPG signal acquisition or analysis; (2) focused solely on transmission-mode PPG systems without clear relevance to wearable or wearable-relevant applications; or (3) were review articles, editorials, commentaries, conference abstracts, or other nonoriginal research publications.

Multimedia Appendix 4 summarizes the baseline characteristics of the 21 included studies, including publication year, publication type, country or region, and primary study focus [15,16,18,20,21,41-56]. The included studies were published between 2006 and 2026. Of these 21 studies, 15 (71%) [15,16,20,21,42-52] were published from 2020 onward, showing that most of the included evidence was published in recent years; besides, 16 (76%) [15,16,21,41-44,46-51,53-55] were journal articles and 5 (23.8%) [18,20,45,52,56] were peer-reviewed conference papers. In terms of country or region, the included studies were mainly from the United States (6/21, 29%) [18,42,47,48,52,56], China, including Hong Kong (3/21, 14%) [15,43,54], the United Kingdom (2/21, 10%) [16,46], Finland (2/21, 10%) [20,51], and Italy (2/21, 10%) [21,49], with 1 study each from South Korea [41], Slovakia [44], Canada [45], Spain [50], Latvia [53], and Israel [55]. Country or region was summarized according to the reported study location when available; when the study location was not explicitly stated, it was assigned according to the first or corresponding author’s institutional affiliation.

The study-level characteristics of all included sources are summarized in Multimedia Appendix 4, with additional methodological and outcome details provided in Tables 1-3 and Multimedia Appendix 5. The included studies contributed to 3 broad and nonmutually exclusive research domains: contact-condition measurement, control, sensing, or estimation; directly measured signal-, waveform-, or feature-level responses; and directly evaluated downstream wearable monitoring tasks. Of the 21 included studies, 4 (19%) [41,43-45] had a primary methodological focus, 16 (76%) [15,16,18,20,41-45,47,49-51,53,55,56] reported at least 1 directly measured signal-, waveform-, or feature-level response, and 10 (48%) [15,18,21,42,47,49,51,52,54,56] directly evaluated at least 1 downstream task under different contact conditions. As these domains were nonmutually exclusive, individual studies could contribute to more than 1 domain.

The included sources of evidence varied substantially in terms of participant characteristics and data sources. Most human-participant studies involved healthy volunteers [18,20,21,41,42,44,45,47,49-53,56], whereas some studies used larger datasets or study populations with broader participant characteristics [46,52]. Sample sizes ranged from single-case feasibility studies [41,45] to large-scale wrist PPG datasets [46]. Several studies did not fully report participant age, sex distribution, or health status [18,43,48,50,55], limiting characterization of population diversity across the evidence base. In addition to the in vivo human studies, the evidence base included an in vitro tissue-vessel phantom study [16] and a study combining theoretical modeling with human-participant experimental validation [54].

Measurement-site categories were nonmutually exclusive because some studies included more than 1 PPG acquisition site or used a secondary PPG site as a reference channel. Of the 21 included studies, 10 (48%) [15,21,41,43,45-47,49,52,55] involved wrist-based configurations, 9 (43%) [20,42,44,45,47,50,51,54,55] involved finger-based measurements, and 3 (14%) [18,55,56] involved forehead-based configurations. One study (5%) [48] examined a sternum-based configuration, 1 (5%) [53] examined other anatomical sites, and 1 (5%) [16] used an in vitro tissue-vessel phantom rather than a human anatomical measurement site. As several studies involved more than 1 measurement site, the summed percentages exceeded 100%. Primary device or experimental configurations were coded into mutually exclusive categories according to the principal PPG setup used in each study. Of the 21 included studies, 7 (33%) [15,21,41,43,46,49,52] used single-site wrist-based systems, 6 (29%) [20,42,44,50,51,54] used single-site finger-based systems, and 3 (14%) [45,47,55] used multisite configurations involving more than 1 PPG acquisition site; 2 (10%) [18,56] studies used forehead-based reflectance pulse-oximetry systems, 1 (5%) [48] used a sternum-based reflectance PPG system, 1 (5%) [53] used a probe at other anatomical sites, and 1 (5%) [16] used an in vitro tissue-vessel phantom rather than an in vivo wearable configuration. Several secondary design features were identified across these primary categories. Three studies (14%) [15,41,47] used force-adjustable or force-regulated platforms, 3 (14%) [43,47,55] used multichannel or multisite PPG acquisition, and 3 (14%) [20,51,55] used multiwavelength reflectance PPG configurations. These secondary categories were nonmutually exclusive because individual systems could incorporate more than 1 design feature. Detailed study-level configurations, fixation strategies, optical settings, and reference measurements are presented in Table 2.

The representation of contact conditions differed substantially across the included studies. Based on the principal mechanical quantity or descriptor reported in each study, the 21 studies were classified into 4 mutually exclusive categories. As many as 5 (24%) studies [15,41,44,45,54] used force-based representations. Of these, all 5 [15,41,44,45,54] explicitly reported contact force in Newtons. Among them, Teng and Zhang [54] examined discrete applied contact-force levels through theoretical modeling and validation involving human participants. Ten studies (48%) [16,18,20,21,42,49,51,53,55,56] used pressure-based representations reported in physical pressure units, including mm Hg (millimeters of mercury), kPa (kilopascal), or torr. These representations included externally applied or cuff-derived pressure values [20,42,51], preset wristband or strap-pressure levels [21,49], and locally applied probe-skin, sensor-skin, or device-housing pressure [16,18,53,55,56]. Four studies (19%) [43,47,50,52] used nonstandardized sensor-output, normalized, indirect, or categorical representations. These included contact-force–related piezoelectric sensor output without a reported physical unit [43], sequential load-cell readings without conversion to a standardized force or pressure value [47], normalized force divided into predefined ranges [50], and pressure-related waveform-morphology labels [52]. Two studies (10%) [46,48] described sensor-skin contact qualitatively or contextually without directly quantifying the applied force or pressure. One study [49] explicitly reported the sensor-skin contact area. Four studies [20,43,44,47] provided partial sensor, probe, printed circuit board, or device-geometry information but did not report the effective skin-contact area. For the remaining studies, contact-area or probe-geometry information was insufficient or unavailable. Sensor or device dimensions were not assumed to represent effective skin-contact area unless this relationship was explicitly established. Consequently, reported force values could not always be converted into contact pressure or directly compared with pressure values across devices. Although contact force and contact pressure are mechanically related, they were not treated as interchangeable when sufficient contact-area information was unavailable.

The reported outcomes spanned signal or signal-quality, waveform, feature, and downstream-task domains. Sixteen studies [15,16,18,20,41-45,47,49-51,53,55,56] contributed direct signal-, waveform-, or feature-level evidence, whereas 10 studies [15,18,21,42,47,49,51,52,54,56] evaluated at least 1 downstream task under different contact conditions. These categories were nonmutually exclusive. Detailed study-level findings are presented below and synthesized across measurement sites, protocol contexts, and analytical domains in Figures 2-4 and Table 4. The studies considered in the empirical linkages in Figure 3 are [15,18,42,47,49,51,52,56].

To avoid duplicating the detailed study-level tables, the following narrative highlights key findings from individual sources.

As outlined in Table 1, the included studies can be broadly categorized into 4 methodological groups according to their approaches to contact force and contact pressure in wearable reflectance PPG: active closed-loop force control, direct contact force/pressure measurement, multipoint contact-force monitoring, and indirect contact-force estimation based on PPG signals. These approaches reflect 4 distinct methodological priorities, namely, stabilizing contact conditions, quantifying contact state, characterizing spatial variations in contact, and reducing reliance on additional hardware. Sim et al [41] represented the first direction by developing an active closed-loop force-control platform, through which the analysis could be performed under constant contact conditions, thereby emphasizing signal stability and experimental repeatability. By contrast, Liu et al [43] extended the problem from single-point force regulation to multipoint synchronous sensing, enabling the investigation of relationships between spatial variations in contact state and PPG waveform behavior. Přibil et al [44] further highlighted the direct measurement and quantitative characterization of probe-skin contact force by integrating force-sensitive sensing into the optical probe, which improved the physical interpretability of experimental conditions and facilitated comparison across studies. Fortin et al [45], in turn, explored a different direction by attempting to estimate contact force indirectly from raw PPG signals, thereby reducing dependence on dedicated force-sensing hardware.

As shown in Table 2, from the perspective of measurement site and device configuration, current reflectance PPG research has moved beyond the simple question of where a sensor should be placed. Instead, it increasingly treats measurement site, wearing form, reference channel, and target task as a coupled system-level design problem. Charlton et al [46] reported that wrist PPG signal quality varied across body postures and sensor heights relative to the heart. Ho et al [47] further showed that this systemic issue also involves contact-force control and reference-channel configuration; by combining a force-adjustable wrist module with a high-perfusion fingertip reference, they more clearly revealed the coupling among measurement site, wearing structure, and reference design. The study by Dresher and Mendelson [18] likewise indicated that the effectiveness of the forehead as a measurement site cannot be understood independently of the specific wearing configuration. Using a forehead reflectance pulse oximetry device consisting of an elastic headband and a custom housing, they systematically evaluated the performance of PPG signals measured from the supraorbital forehead region under different contact pressures and showed that forehead signal quality and measurement reliability during motion are shaped not only by the measurement site itself but also by the combined effects of fixation structure and contact-pressure conditions. Chan et al [48] extended the measurement site from the periphery to the mid-sternum, further indicating that although more central reflectance PPG holds promise for continuous oxygen saturation monitoring, its performance remains constrained by multiple factors, including local perfusion, optical path design, and wearing strategy.

As summarized in Table 3, existing protocols can be broadly categorized into static/quasi-static designs and dynamic in vivo designs. Static and quasi-static protocols primarily examined contact-condition–related changes in waveform morphology, signal-to-noise ratio (SNR), and derived features while minimizing motion-related variation. Dynamic protocols examined whether similar response patterns were observed during stepping, walking, running, or other wearable-use conditions. In the dynamic category, Scardulla et al [49] investigated wrist PPG during step exercise at 3 preset contact pressures and showed that contact pressure had a greater impact on heart-rate measurement quality than exercise intensity itself. Scardulla et al [21] later extended this line of work using a more standardized treadmill protocol across 20, 60, and 75 mm Hg and found that 60 mm Hg offered the best overall balance of accuracy and precision, although the optimal pressure still varied across individuals. By contrast, quasi-static and static studies have been more focused on mechanism. Sirkiä et al [20], using a controlled pressure ramp on the fingertip with multiwavelength PPG, showed that externally applied pressure changes not only oscillation amplitudes but also pressure-dependent occlusion behavior and interchannel timing differences, indicating depth-dependent vascular responses. May et al [16], in an in vitro tissue-vessel phantom study with continuously increasing pressure under multiple simulated blood pressure states, demonstrated the existence of an optimal contact-pressure range and further showed that temporal features are relatively robust, whereas amplitude- and geometry-related features are substantially more pressure-sensitive. Ho et al [47] further contributed a quasi-static wrist-finger dual-channel dataset collected under sedentary conditions, in which wrist pressure was gradually adjusted while a fingertip channel was maintained as a high-quality reference; their results showed systematic pressure-dependent transitions in wrist PPG morphology and corresponding degradation in downstream HR and HRV estimation under suboptimal pressure.

Several included studies reported nonlinear or task-dependent response patterns under their respective experimental conditions. As outlined in Table S5 in Multimedia Appendix 5, for example, May et al [16] found that both waveform amplitude and SNR first increased and then decreased as pressure rose. Lambert Cause et al [50] reported that the perfusion index, which is strongly amplitude-dependent, was also highest in lower-to-moderate force ranges before declining at higher force. Their signal-quality indices likewise showed that both low and high extremes were suboptimal, whereas moderate force better preserved waveform detail. Sirkiä et al [20,51] further showed that SpO₂ estimates remained relatively stable at low pressure but began to rise as pressure approached diastolic and mean arterial pressure, indicating that pressure-induced changes in AC/DC relationships and feature points can bias downstream metrics. Teng and Zhang [54] investigated the effect of sensor contact force on PPG-derived PTT using theoretical modeling and validation involving human participants. By considering the nonlinear pressure-volume relationship of the finger arterial wall, they showed that PTT increased with contact force under positive transmural pressure, reached a maximum near zero transmural pressure, and then remained nearly constant under negative transmural pressure. This finding indicates that sensor contact force can nonlinearly alter PPG-derived timing parameters through changes in local transmural pressure and arterial wall mechanics. Across these studies, nonlinear or inverted-U-like response patterns were reported under the respective experimental conditions. Intermediate contact levels were frequently associated with better signal quality, waveform clarity, or task performance than the lowest or highest tested levels. Importantly, the location and shape of this optimum are not fixed but vary with the physiological target, wavelength, measurement site, and feature type.

Table 4 provides a study-level overview of the methodological and analytical domains covered by the 21 included studies. Measurement site and principal protocol context were identifiable for all studies. Participant characteristics and contact-condition representation were more consistently documented than effective contact area or probe geometry. Sensor, printed circuit board, load-cell, or device dimensions without sufficient information on the effective skin-contact interface were classified as partial rather than sufficient reporting. Analytical coverage varied across studies, with some studies addressing multiple signal-, waveform-, feature-, and downstream-task domains. Overall, the matrix highlights variation in methodological reporting and analytical breadth rather than differences in study quality.

Despite growing interest, the mapped evidence remains methodologically heterogeneous [15,16,18,20,21,41-56]. A major limitation is the lack of a standardized framework for representing, measuring, and reporting contact force and contact pressure [15,16,18,20,21,41-56]. Importantly, contact force and contact pressure are mechanically related but not interchangeable: contact force describes the total load applied to the sensor-skin interface, whereas contact pressure depends on how that load is distributed over the effective contact area [15,16,20,21,41-47,50-56]. Across studies, contact conditions were described using different units or concepts, including Newtons, mm Hg, normalized force units, cuff pressure, preset strap levels, inferred tightness, or qualitative descriptions of skin contact [15,16,18,20,21,41-56]. This heterogeneity limits direct comparison and the development of quantitative design guidance [15,16,20,21,41-47,49-56]. It also limits the transferability of terms such as “moderate force” or “optimal pressure” across wearable configurations [16,20,21,47,49-51,54]. In particular, when studies report only force without specifying the effective contact area, probe geometry, or active sensing area, the corresponding contact pressure cannot be reliably derived [15,41,43-45,47,50,52]. Even when similar terms are used, the underlying physical meaning may differ because reported pressure can depend on probe geometry, active sensing area, tissue compliance, or the way force is converted into pressure [16,20,21,47,49-51,53,54]. A more standardized reporting convention would therefore be an important step toward improving comparability and reproducibility [57,58]. Future studies should report the effective sensor-skin contact area or probe geometry when reporting contact force, or directly report contact pressure in standardized units such as mm Hg or kPa [57,58]. This would improve reproducibility, cross-study comparability, and practical translation for wearable PPG device design.

A major evidence gap is the limited validation of contact conditions during prolonged and unsupervised wearable use. Most included studies used small samples, short recordings, highly structured protocols, static or quasi-static tasks, or predefined pressure conditions [16,18,20,21,41,42,44,45,47,49-51,53-56]. These designs are informative for characterizing responses under controlled conditions, but they provide limited evidence about the variability encountered during everyday wear. During free-living use, the sensor-skin interface may vary with posture transitions, arm movement, sweating, changes in skin hydration, tissue deformation, strap displacement, repeated device removal and reattachment, and prolonged wear [14,18,21,46,48,49]. Accordingly, the available laboratory evidence may not capture the full range or temporal instability of contact conditions in natural settings [14,18,21,46,49]. Current evidence does not establish whether a contact condition measured at the start of a recording remains stable over several hours, whether users reproduce similar conditions after repeated donning, or whether contact-related variability is consistent across days. Short supervised walking, stepping, or treadmill protocols should therefore be distinguished from longitudinal or free-living validation. Future studies should jointly document contact force or pressure, motion, posture, wear duration, and contextual factors rather than evaluating contact conditions only under isolated, idealized protocols [14,21,46,47,49,58]. Recent benchmarking efforts, including the QUMPHY D4 report [58], also indicate the value of standardized benchmark datasets in addition to algorithm development. For wearable reflectance PPG, future public datasets should document acquisition context more systematically, including site, device configuration, fixation strategy, wear duration, and other factors that may be associated with signal quality and estimation performance [46,47,58]. Repeated-measures studies should report within-session drift, between-day reproducibility, reattachment procedures, strap or fastening settings, and the frequency of contact-related signal-quality failures.

As shown in Figure 3, complete within-study pathways connecting measurement site, an intermediate signal-, waveform-, or feature-level domain, and a downstream task were identified in 8 studies [15,18,42,47,49,51,52,56]. The remaining studies did not report a complete site-intermediate-domain-downstream-task pathway [16,20,21,41,43-46,48,50,53-55]. This distinction is important because signal or feature differences observed under varying contact conditions do not necessarily correspond to clinically or algorithmically meaningful estimation differences. Their practical relevance is likely to depend on the target task, the signal properties used by the algorithm, and the selected reference method. Conversely, task-level differences should not be attributed solely to contact conditions without also considering motion, wavelength, device configuration, reference-measurement uncertainty, and participant characteristics. Future studies should therefore evaluate contact conditions, intermediate signal or feature behavior, and task-level estimation error within the same experimental protocol and against an appropriate reference standard. Stratified analyses across contact conditions, together with sensitivity or feature-ablation analyses, may help identify which signal properties are most closely associated with task-specific performance differences [15,21,42,49,52,57,58].

A further evidence gap concerns population and measurement-site diversity. The evidence was concentrated at the wrist and finger: 10 [15,21,41,43,45-47,49,52,55] of the 21 included studies involved wrist-based configurations, and 9 [20,42,44,45,47,50,51,54,55] involved finger measurements, whereas only 3 [18,55,56] examined the forehead, 1 [48] examined the sternum, and 1 [53] examined another anatomical site; these categories were nonmutually exclusive. Given differences in anatomical structure, device geometry, fixation strategy, and optical measurement configuration across sites, numerical pressure ranges reported at the wrist or finger should not be assumed to apply directly to the forehead, sternum, ear, or other anatomical locations [18,46-48,55,59,60]. Most human studies recruited healthy volunteers [18,20,21,41,42,44,45,47,49-54,56], leaving limited direct evidence for older adults, children, patients with cardiovascular or peripheral vascular conditions, and individuals with edema or impaired perfusion. Previous studies and systematic reviews have reported differences in pulse-oximetry accuracy across skin-pigmentation groups, particularly under lower oxygen-saturation conditions [61-64]. However, the included contact-condition literature did not adequately characterize whether skin pigmentation modifies the relationships between contact conditions and wearable reflectance PPG outcomes. More broadly, the reported specificity of reflectance PPG and SpO₂ estimation to participant and measurement-site characteristics suggests that population and site characteristics should be considered when interpreting contact-condition–related findings [59,60]. Future studies should report participant age, sex distribution, health status, the method used to characterize skin tone or pigmentation, and relevant vascular or tissue characteristics more consistently. Harmonized multisite protocols are needed to determine whether reported contact-condition responses are site-specific or reproducible across anatomical locations. Findings derived mainly from healthy adults and common peripheral sites should therefore be generalized cautiously to broader populations and other anatomical locations [46,59-64].

The focus on wearable or wearable-relevant reflectance PPG represents a deliberate scope boundary of this review. Accordingly, the mapped findings should be applied cautiously to transmission-mode and hybrid reflection-transmission PPG systems. Transmission-mode PPG remains important in wearable pulse oximetry, particularly in finger-clip, ring-type, earlobe, and in-ear devices, in which the light source and photodetector are positioned on opposite sides of relatively thin tissue [30,65-67]. These configurations may provide strong pulsatile signals for SpO₂ monitoring, but they involve different anatomical constraints and fixation mechanisms, including clips, rings, and in-ear fittings. Differences in optical path, contact area, tissue compression, and fixation mean that reported contact-condition responses may not be directly comparable across optical modes. Numerical force or pressure ranges and response patterns mapped in reflectance studies should therefore not be transferred directly to transmission-mode or hybrid devices. Several methodological principles remain relevant across optical modes, including distinguishing force from pressure and documenting geometry, effective contact area, fixation, and calibration. For hybrid systems, the reflectance and transmission interfaces should be described separately rather than summarized using a single wearing-tightness descriptor. For each sensing interface, studies should report the measurement site, optical mode, wavelength, emitter-detector geometry, fixation method, effective contact area, and contact force or pressure. Direct comparative studies, and potentially a separate evidence synthesis focused on transmission-mode and hybrid systems, are needed before cross-mode recommendations can be established.

This review also has methodological limitations related to its scope and review process. Although the review followed a scoping review framework and was reported with reference to PRISMA-ScR and PRISMA-S [38-40], the protocol was not formally registered or publicly available. The search strategy was developed by the review authors and refined iteratively, but it did not undergo formal peer review by an information specialist or librarian. In addition, only English-language peer-reviewed journal articles and conference proceedings with accessible full texts were included, whereas non-English studies, non-peer-reviewed sources, conference abstracts, preprints, technical reports, theses, and other gray literature were excluded. Although 4 major databases were searched, publication bias and incomplete database coverage cannot be fully ruled out. Finally, because this was a scoping review, no formal risk-of-bias assessment, critical appraisal, or quantitative meta-analysis was performed; therefore, the findings should be interpreted as a structured mapping and qualitative synthesis of the available evidence rather than as a quantitative estimate of effect size or certainty of evidence.