The study was designed as a within-subjects research and divided into 3 parts. In the first task, participants were asked to use the 2 prior selected online pharmacies, Apotal and DocMorris [59]. The online pharmacies were selected based on their different risk and trust ratings in a preliminary study. Both are legal, hold the European Union (EU) safety logo, and are listed in national mail order registers. As a search task, participants were provided with the scenario that they had met with an accident and were now limited in their movements. As a result, they now wanted to purchase their medications online. Participants were asked to buy well-known OTC painkillers (ibuprofen) and redeem the prescribed antithrombosis injection (Clexane). The task was completed once both medications were added to the shopping cart. The search task had to be executed on both online pharmacies; the order was randomized for each participant. During this task, participants’ faces were recorded via a video camera with a resolution of 1080 pixels and 30 frames per second, and facial expressions were later analyzed with FaceReader (Noldus) software.

After finishing their search task, participants were handed a questionnaire. First, a validated German version of the Positive and Negative Affect Scale (PANAS) was assessed for each included pharmacy to obtain insights into associated, self-reported emotional states related to the online pharmacy [60,61]. This was followed by questions on participants’ demographics. Finally, the experimenter measured the head circumference of participants, after which the third part of the study was started.

In the final part of the study, neural data were collected using the mobile neuroimaging method of functional near-infrared spectroscopy (fNIRS). In an event-related experimental paradigm, the online pharmacies were shown to participants as screenshots with the medications they had to search for in the first task. Each screenshot was shown for 4 seconds, after which a question from the scales for perceived risk (adapted from Forsythe and Shi [62]), perceived trust (adapted from Cyr et al [63]), and purchase intention (adapted from Gefen [64]) for the online pharmacies was shown with a 5-point Likert scale. Upon input from participants, a neutral cross randomly jittered between 2 and 4 seconds was shown. After that, the next screenshot was shown. Both online pharmacies were shown with both medications (Figure 1). This procedure continued until each question was asked for each screenshot. To ensure validity, the included scales in this study, except PANAS, were also used and validated in our preliminary study [59].

After all tasks were completed and questions answered, the fNIRS headband was removed and participants were free to leave.

This study was reviewed by the Ethics Committee of the Medical Faculty of the University of Duisburg-Essen (approval number 21-9995-BO). Participants were informed about the study before the start of the study and were provided with an informed consent form. The recording and analysis of video data was pseudonymized. Only anonymized data were used in the publication. Participants received €10 (US $10.9) as compensation.

We recruited 42 participants through a local university in September 2021. Exclusion criteria included existing neurological disease, reduced general health condition, no experience with online shopping, belonging to pharmaceutical staff, and taking medications with an effect on the central nervous system. Data from 5 (11.9%) participants were excluded: 1 participant because of a measurement error with FaceReader, 3 participants who did not correctly complete the search task on the online pharmacies, and 1 participant because of too much noise in the neural data. As a result, we included a final sample size of n=37 (88.1%) participants for further data analysis. Although this sample size might seem small for behavioral studies, it matched the average sample size of NeuroIS research, which was 38 participants [65]. Further, related fNIRS studies show that medium-large effects at 80% power can already be detected with a sample size of 20 participants [66]. Power analysis showed that with 37 participants, we had a power of 84.1% to effect sizes of |δ|>0.5 and a maximum type I error rate of 0.05.

The mean age was 28.6 (SD 9.52 years, min.=18 years, max.=62 years). Gender distribution was almost balanced, with 18 (48.6%) participants being female and 19 (51.4%) male. As handedness has been shown to potentially cause bias in neural activation in the prefrontal cortex (PFC), we assessed the handedness of participants using the laterality quotient (LQ) described by Salmaso and Longoni [67] (we used the German version of Götze [68]), which revealed that the majority of the sample was right-handed (n=33, 89.2%). The mean head circumference of participants was 55.5 (SD 2.73 cm). Most of the participants were employed (n=18, 48.6%), followed by a smaller number being university students (n=14, 37.8%) and school pupils (n=4, 10.8%), and 1 (2.7%) participant was searching for employment.

The filtered responses of subjective feelings can be conceptualized along 2 major dimensions of valence (positive-negative) and activation. In line with this, Watson and Tellegen [100] proposed a consensual structure of emotion, where positive and negative dimensions of valence are combined with activation (inactive-active). Briefly, a positive affect (PA) reflects the extent of activation and enthusiasm, whereas a negative affect (NA) is the opposite of PA [100]. With this conceptualization, the conscious aspect of emotional experience becomes accessible, and emotional words that depend heavily on the receivers’ understanding [101] are reasonably clustered to avoid bias caused by the limitation of words. We used the PA-NA, which is one of the most applied measurement scales [102,103]. With regard to behavioral intentions, PA can be mapped to approach behavior, while NA can be mapped to avoidance behavior [104].

First, we tested for reliability (Cronbach α) of the used scales, which resulted in sufficient reliability (α>.75) for all scales (perceived risk α=.790; perceived trust α=.861; NA α=.847; use intention α=.883). Additionally, we calculated item-rest correlations to assess the internal consistency of the scales. Results showed that all included items had sufficient internal consistency (item-rest>0.3) and were therefore included in the analysis.

Second, to ensure successful risk and trust manipulations in this study, we analyzed self-reported results using repeated-measures univariate analysis of variance (rmANOVA), with the factors “online pharmacy” and “medications” as repeated measures and the constructs “perceived risk” and “perceived trust” as dependent variables. As all assumptions were met by the data and the assumption of sphericity was given, no corrections were made.

The results revealed significant differences in perceived risk (F_1,36=92.809, P<.001, η²_p=0.721) and perceived trust (F_1,36=60.131, P<.001, η²_p=0.626) between the 2 included pharmacies. Apotal was perceived to bear higher risk (mean 3.11, SD 0.117) and lower trust (mean 2.97, SD 0.154) compared to DocMorris. Perceived risk and perceived trust were not surveyed for medication types because we considered medication types as a secondary effect (the types of medications and the different procedures for purchasing the medications indicate different perceptions of risk). Further, we evaluated possible biases in online pharmacies’ ratings, since the screenshots showed the online pharmacies paired with the medications. No significant interaction effect between online pharmacy and medication (F_1,36=0.026, P=.87, η²_p=0.001) and trust (F_1,36=0.661, P=.42, η²_p=0.018) was found.

Testing the subjective feeling of emotion toward the online pharmacies with rmANOVA revealed a significantly higher NA (F_1,36=52.485, P<.001, η²_p=0.593) for Apotal (mean 1.96, SD 0.684) compared to DocMorris. Further, significantly higher use intentions were reported for DocMorris (mean 3.73, SD 0.156) compared to Apotal (F_1,36=103.826, P≤.001, η²_p=0.743). All means and SDs of the included self-reports are presented in Table 1.

To analyze the neural data from fNIRS on a group level, we ran a mixed-effects model that used the 2 online pharmacies (Apotal and DocMorris) and the medications (ibuprofen and Clexane) as fixed effects and individual differences between participants as random effects. Further, it is important to note that an increase (red Ch) in HbO and a subsequent decrease in HbR (blue Ch) pointed to neural activation (see Figures 3 and 4 and Table 2). To avoid false positives, Chs in which both HbO and HbR showed a significant increase were omitted from further analysis (ie, Ch 6).

Analogous to the neural results, we ran a mixed-effects model to analyze the results of the facial expression analysis on a group level related to the use of the online pharmacies (Apotal and DocMorris) and the medications (Clexane and ibuprofen) as fixed effects and individual differences between participants as random effects. The averages of the maximum intensities of negative facial expressions during the purchases were used for calculations.

We found significant differences in the intensity of negative emotions between the groups (F_3,108=5.86, P<.001). These were identified between the pharmacies (F_1,108=12.80, P<.001) and the medications (F_1,108=4.74, P=.03). During the purchase at the high-risk pharmacy (Apotal), the intensity of negative emotions was higher (mean 0.310, SD 0.131) than during the purchase at the low-risk pharmacy (DocMorris; mean 0.283, SD 0.122). Additionally, during the purchase of the prescription medication (Clexane), the negative facial expression intensity was higher for both pharmacies (Apotal: mean 0.333, SD 0.145; DocMorris: mean 0.304, SD 0.142) than during the purchase of the OTC medication (ibuprofen; Apotal mean 0.287, SD 0.135; Docmorris mean 0.230, SD 0.123).

However, testing the differences between the medications in the 2 pharmacies revealed insignificant effects. Moreover, comparing the purchase of the prescription medication (Clexane) at the high-risk pharmacy (Apotal) with the same purchase at the low-risk pharmacy (DocMorris) showed no significant effect.

Comparing the purchase of the prescription medication (Clexane) at the high-risk pharmacy (Apotal) with the purchase of the OTC medication (ibuprofen) at the low-risk pharmacy (DocMorris) resulted in significantly higher intensities of negative facial expressions (t₁₀₈=4.069, P<.001). All values are reported in Table 4.

The subjective assessments of perceived risk and perceived trust confirmed that Apotal is perceived as having higher risk, lower trust, and more negative emotion than DocMorris. Similarly, the purchase intention is significantly lower for Apotal than for DocMorris. To obtain a deeper understanding of what leads to these self-reported ratings, we will take a closer look at the neural activity and facial expression results.

We identified consistent activation of the left vmPFC region when the prescription medication (Clexane) was combined with the high-risk online pharmacy (Apotal). This effect was further associated with the highest negative emotional intensity of facial expressions while using the 2 pharmacies’ websites. Our facial expression analysis further suggested that during the shopping process, the prescription medication had a more significant impact on negative emotional states than did the online pharmacy. A look at the related literature on the vmPFC’s functions and facial expression results provides us with further information about these findings.

In broader cognitive neuroscience, the vmPFC is frequently associated with value detection (ie, gain or loss) and with the regulation of (negative) emotion [105-108]. Furthermore, the significant role of this brain region in the evaluation of the potential implications of events for oneself was first stated in the somatic marker hypothesis and has been demonstrated in several studies that have shown that people with vmPFC damage are no longer able to make reasonable decisions [48,49,109]. The reason for this is that these persons lack the ability to accurately assess the risk (or potential loss) of a decision for themselves. This provides support for the somatic marker hypothesis that states that our decisions are framed and guided by emotions. Seeing the vmPFC activated for stimuli that are also associated with intense negative emotional facial expressions, and the higher self-rated perceived risk and NA, it is likely an indicator for negative emotional appraisal of medications and pharmacies’ websites.

The related literature that used AFEA revealed that negative emotions are often expressed more frequently and intensely than positive emotions [99,110,111]. It is further understood that risky situations evoke negative emotions [45,46], and in the case of online shopping, risk leads to a negative emotional experience due to the fear of potential loss [112]. Therefore, we propose the following first theoretical implication of our study:

When looking at the OTC medication (ibuprofen) and the online pharmacy associated with high perceived trust and low perceived risk (DocMorris), the facial expression analysis underlines that the medication and the pharmacy are associated with significantly less negative emotional expressions. On the neural level, we further identified a different activation pattern. Although the vmPFC still showed significant activation, it was always accompanied by a significant deactivation of the dmPFC in the opposite hemisphere. Although the vmPFC has been related to processing the reward or loss value [113], areas of the dmPFC have been more associated with decision conflict [114]. In risky decision-making, the dmPFC has been found to be activated when information is incomplete and thus uncertainty is high [82,115]. Furthermore, high activity in the dmPFC has been found to be associated with risky choices [116,117], suggesting that dmPFC activation in risky behavior acts as a warning signal [82]. Given its deactivation when both the online pharmacy and the medication are perceived as lower in risk but not when one of them is perceived as higher in risk, this situation may be decoded as a certain and safe option. Due to the higher perceived trust and less negative emotional expressions, the vmPFC activation for this shopping situation may signify perceived reward. Therefore, the following second theoretical implication can be suggested:

Several consequences for industry can be derived from our theoretical implications. Given that the perceived risk of prescribed medications cannot be changed due to their nature, online pharmacies need to especially tackle trust-building mechanisms on their websites for the purchasing process of prescription and potentially high-risk medications. Based on our findings, there is potential to generate neural responses associated with certainty and reward for OTC medications, which could explain why these products are more commonly purchased online compared to prescription medications. One key difference between the presentation of prescription and OTC medications on the online pharmacy websites investigated is the lack of product images for prescription items: only an image of an envelope with the prescription or no image is displayed. It would be beneficial to include photos of the products themselves to aid in their identification. Additionally, it is important to provide detailed textual information about prescription medications, as the lack of such information may contribute to the negative emotional responses observed toward prescription medications, which indicates a higher risk perception toward them [20]. Studies confirm that customers refuse to buy medicines online, because they do not receive relevant information on the pharmacy websites. Customers are concerned that their health will be put at risk because they lack technical information, such as the expiration date, a release form, and the safety of the product for their health [10,11,13].

Regarding the process for placing prescription medication orders, both pharmacies investigated act according to German law (§17 Apothekenbetriebsordnung [ApoBetrO]). The customer sends the prescription via mail to obtain the medication. However, DocMorris displays the desired medication in the shopping basket and provides ordering instructions on both the product page and the shopping basket. In contrast, when attempting to add a prescription medication to the basket on the Apotal website, a new window containing a PDF file opens up. The file explains that customers must fill out a printed form and take it to the post office to obtain their medications. Notably, the desired medication does not appear in the shopping basket. Given that customers have to send their prescriptions and order via mail and cannot simply shop online is an obstacle to the perceived convenience of shopping for prescription medications online. Despite the availability of e-prescriptions in Germany since September 2022, the current lack of necessary technology in many doctors’ offices has resulted in conventional prescriptions remaining relevant [118]. In fact, it was estimated that conventional prescriptions would still be issued in 70% of cases in 2023 [119].

Therefore, adding trust-building elements to these product pages is of utmost importance for online pharmacies. Other than providing technical information, a further option would be to include certificates of trust on the website to ensure the authenticity of the pharmacy. Pharmacies can also leverage telemedicine and chats or telephone hotlines to provide personal advice to customers. In fact, our low-risk pharmacy already makes use of this and provides easy access to these services directly on the product page. According to a study in India [120], this would help purchases in online pharmacies. Finally, online pharmacies should ensure that the purchase process is as convenient as it can be under the given regulations, including the guarantee of product availability and timely delivery; information about both is essential and should be visible to the customer on the website. Evidence shows that online pharmacy customers wish for better logistics [121].

Neuroeconomic research has shown that establishing brands and having a high (positive) reputation can lead to neural “lock-in” effects [122,123]. The neural effects are also located in the vmPFC and signify an increased appraisal of reward value toward stores and products of the preferred brand. Therefore, establishing such effects in customers may neutralize the NA and perceived risk with ordering prescription medications online.

We used a multimethod approach to gain deper insights into the appraisal processes and immediate emotional experience of online medication purchasing. However, this study has some major limitations that should be addressed by future research. First, we considered 2 operating online pharmacies in Germany and examples of prescription and OTC medications. Thereby, we may have a geological restriction in the generalization of our findings. Especially as restricting laws for pharmacies in Germany do not always apply to other countries, this research needs to be validated in other countries and cultures. Regarding the experimental design, we had to measure facial expressions and neural activity in 2 tasks of the study. That is because the fNIRS headband obscured parts of the face (eyebrows) that are essential for emotion recognition with FaceReader. Future research should particularly focus on making simultaneous measures of facial expressions and neural activity to validate our results during actual website use. Finally, by using 2 German online pharmacies that are among the top 10 in Germany [124], we can argue that the stimuli were not manipulated for the goals of the study but that we assessed the perceptions of risk and trust, including emotional experience, for a real-world problem. Due to the close-to-real-life situation, several biases can result from uncontrollable stimuli in the environment as well as from unpredictable effects from those stimuli. Here also, the importance of self-reports becomes evident in understanding cognitive concepts that are interconnected with emotions. Given this limitation, and to further derive rigorous design guidelines for prescription medication pages, future work may use controlled stimuli where different practical implications made in this paper are tested.

For a health-related and underresearched specific e-commerce phenomenon, namely online pharmacies and their primary business of prescription and OTC medications, we examined perceived risk and perceived trust as critical antecedents for their adoption in the context of emotions in decision-making. Our results demonstrate that in the context of online pharmacies in particular, and e-commerce websites in general, deeper insights into the evaluation processes of websites and products will help us better understand how the risk and trust associated with them can influence emotions and thus behavioral intentions. Furthermore, we also investigated potential links between uncertainty and neural decision conflicts that stem from appraisal processes, and their relation to the perceived risk of online pharmacies, depending on the medications. Both facial expression analysis and neural results showed that several negative emotional attribution processes occur for prescription medications compared to OTC medications. Online pharmacies may therefore rethink how the product page of prescription medications should be designed so that the intense negative emotional experience, increased uncertainty, and perceived risk can be weakened for this product type. Building trust incentives into the product page and fostering a strong brand may elevate these negative effects of prescription medication and foster the success of online pharmacies.