Amphetamine-type stimulants (ATS) and cocaine are commonly misused drugs globally (1). Not only is ATS the commonest primary drug of misuse in Asia, it also contributes to more than half of its treatment-seeking drug users (1). Among Chinese-speaking countries and regions, including mainland China, Hong Kong, Malaysia, Singapore, and Taiwan, methamphetamine and cocaine account for the highest number of documented cases who come into a formal contact with authorities, more than that of heroin or cannabis (2). Individuals with stimulant use disorder (SUD) are prone to develop a myriad of psychiatric disorders, encompassing mood, bipolar, and psychotic disorders (3–8). Therefore, developing a reliable and valid screening instrument in Chinese for early detection of SUD in Chinese-speaking stimulant users is of paramount importance to administer timely intervention.

The Severity of Dependence Scale (SDS) is a five-item, self-administered scale that measures the psychological dependence of drug users on impaired control, anxiety and preoccupation towards their substance use (9). Rating each item on a Likert scale from 0 to 3 with a maximum score of 15, SDS is designed to contextualize psychological dependence under a dimensional construct with higher scores suggesting more severe dependence. SDS has a single-factor structure with high internal consistency in measuring the construct of substance dependence. It also has a high test-retest reliability and has demonstrated strong associations with the amount and frequency of use across various substances (9–12).

The original English version of SDS had been validated in heroin, cocaine, and amphetamine users (9). Later studies have also translated and validated SDS into various languages, and found different diagnostic cut-offs that allow for the optimal discrimination for substance dependence in accordance with the Diagnostic and Statistical Manual of Mental Disorders (DSM) for alcohol, amphetamines, benzodiazepines, cannabis, cocaine, and heroin (10, 12–20). While the Chinese version of the SDS (C-SDS) has been validated for benzodiazepines, cannabis, heroin, and ketamine (13, 20–22), the psychometric properties of C-SDS for stimulant use have not yet been validated, nor its diagnostic cut-offs for DSM-5 defined SUD established.

The present study aims to establish the diagnostic utility of the C-SDS for stimulant (C-SDS-S) in screening for DSM-5 defined SUD. This study also focuses on determining the cut-off scores that provide optimal discrimination for mild and moderate-to-severe SUD, thus allowing the C-SDS-S to be adopted as a rapid screening instrument in routine clinical practice.

The present study explored the validity and reliability of the C-SDS-S as a clinical tool to screen for mild and moderate-to-severe DSM-5 SUD in Chinese-speaking stimulant users. We showed that a higher total score on the C-SDS-S is associated with greater SUD severity and the frequency of stimulant use, especially among methamphetamine users. Overall, a single-factor structure was observed for the C-SDS-S and when used among methamphetamine users, consistent with other studies that demonstrated a one-dimensional structure of the SDS when administered in different languages and for various substances (9, 12, 13, 16, 17, 21, 22, 24, 25). The two-factor structure identified among our cocaine samples who are predominantly crack cocaine users corroborated with the previous study by Ferri et al. (24). Overall, our findings suggest that C-SDS-S may serve as a preliminary screening instrument in selected clinical contexts, particularly for moderate-to-severe stimulant use disorders, although its diagnostic performance remains limited.

Our study also established a cut-off score of five for DSM-5 defined moderate-to-severe SUD, regardless if the stimulant used was methamphetamine or cocaine. These cut-off scores were both higher than those established in Australia and Spain two decades ago, which were four for methamphetamine and three or four for cocaine (15, 19). It follows that the diagnostic utility of SDS is highly contextual. For the same substance, the diagnostic cut-off for the optimal discrimination of SDS-defined dependence might be gender- and ethnic-specific, in which the difference could be as large as 2 points between male and female, and between Westerners and non-Westerners (12). Cabana-Dominguez et al. revealed that cocaine dependence is a highly heritable disorder within the European ancestry, though little is known among the Asian-ancestry cocaine users (26). On the other hand, recent genome wide association studies demonstrated the significant loci for methamphetamine dependence might be different between the European-, African- and Chinese- ancestries (27, 28). Such genetic variabilities could result in different single nucleotide polymorphism-based expression on the nicotinic acetylcholine receptor activity at the reward circuitry, axonal pruning in hippocampus, and might potentially lead to diverse trajectories in the development of methamphetamine use disorder. These variations suggest that different SDS cut-off scores may be necessary, in particular for different population across ancestries. We did not perform subgroup analyses for male and female participants due to the small number of female participants.

In addition, changes in the diagnostic criteria for substance use disorder might have attributed to the disparity in diagnostic cut-offs across multiple substances. In most validation studies, the 4^th edition of the DSM (DSM-IV) was adopted as the “gold standard”. According to DSM-IV, substance abuse and dependence were independently categorized using two separate sets of diagnostic criteria. Under DSM-5, however, substance use disorder was defined using a set of diagnostic criteria primarily adopted from that of DSM-IV abuse and DSM-IV dependence (29). The change from a dichotomous to a continuous diagnostic construct warrants the validation of the C-SDS-S against the DSM-5 criteria to ascertain its criterion validity. Also, similar to our previous work in cannabis (13), we established the diagnostic cut-off on C-SDS for moderate-to-severe SUD but not for mild SUD (13). While the original SDS was developed with a single factor structure under a unidimensional construct for dependence, such single-factor solution might not hold equally valid for non-dependent low-level drug users (30, 31). Therefore, its diagnostic utility in detecting milder states as in abuse or harmful use might be low. Given the clinical significance of timely identification of mild SUD for early intervention, the generalizability of the C-SDS-S in detecting mild SUD warrants further investigation.

While most validation studies on SDS to date identified a unidimensional structure, PCA revealed a two-factor structure for cocaine with Item 4 contributing to the second factor. This is consistent with the validation study of C-SDS in benzodiazepines in Taiwan (20), which also showed the scorings on Item 4 were similar between dependent and non-dependent users. Factor loading for Item 4 is consistently and substantially lower than all other items of the C-SDS-S in the present study, and the internal consistency of the C-SDS-S would improve if Item 4 was removed. Although repeated attempts to quit or reduce stimulant use is one of the diagnostic criteria for DSM-5 SUD, the lower factor loadings of Item 4 suggested that those with stimulant dependence might not have adequate motivation to alter their drug use behavior. Stimulant-dependent users might also hesitate in the prospect of withdrawal symptoms, hence are ambivalent to quit or reduce stimulant use despite its myriad of consequences. In fact, repeated exposure to stimulants reinforces the negative state manifested as craving, withdrawal symptoms, and/or emotional distress (32). This upregulated negative state following prolonged stimulant exposure in dependent users might supersede their desire to quit or reduce stimulant use. Besides, cyclothymic and irritable temperaments have been associated with cocaine and stimulant abuse and heavy uses (33). These temperamental predispositions in stimulant users may heighten their vulnerability to emotional dysregulation during such negative state and contribute to their motivational ambivalence, as reflected by a lower factor loading for Item 4 (“wish to stop”) in the C-SDS-S.

One of the limitations of the present study is the small sample size of cocaine users, which could have attributed to the disparity between the results from the subgroup analyses on cocaine and existing literature, as well as the methamphetamine subgroup and the overall model in the present study. For instance, Item 1 had the lowest test-retest reliability in methamphetamine users and the entire cohort, but was the highest in cocaine users. Among the 101 participants who had repeated C-SDS-S scorings within four weeks, only 16 were cocaine users. Such sample imbalance in particular for the cocaine subgroup may also affect the ROC analysis and the subsequent skewed AUC. All these can inevitably undermine the certainty of the results. As opposed to previous validation studies (9, 11), the small number of cocaine users in the present study might also explain the independence between C-SDS-S scorings and frequency of use within this subgroup where no statistical difference was observed despite showing patterns similar to the methamphetamine subgroup. An early epidemiological study in the United States found that the number of days of stimulant use over the past 12 months was higher in dependent users than those with abuse (34). A different relationship between C-SDS-S scores and frequency of use in cocaine users might be observed if more users could be included. Furthermore, the unique pharmacodynamics of cocaine as a predominant dopamine-reuptake inhibitors as opposed to methamphetamine as a dopamine “releaser”, along with their differing pharmacokinetics, may lead to the distinct use pattern and dependence model specific to cocaine. All these suggest that a single version of C-SDS-S might not fit all stimulant types equally well.

Another limitation of our study is that we were only able to access data in a single locality. Whether the results can be generalized to other Chinese-speaking communities warrants further investigation. Moreover, the cross-sectional design of this study limits the predictive validity of using the C-SDS-S. In addition, our study only assessed the use disorders of methamphetamine and cocaine, both being the predominantly misused stimulants in our locality. Further investigations targeting novel stimulants and psychoactive substances such as khat is warranted.

The present study supports the application of the C-SDS-S as a valid and reliable screening instrument for moderate-to-severe SUD in Chinese-speaking stimulant users. A cut-off score of 5 provides optimal discrimination for both methamphetamine and cocaine and should prompt further diagnostic assessments by broad-certified psychiatrists. Additionally, revising Item 4 of the C-SDS-S may enhance its accuracy for screening purposes in Chinese-speaking stimulant users. Notwithstanding its high validity and reliability, we urge the validation of the SDS in accordance with the updated diagnostic criteria defined in DSM-5 to maximize its diagnostic utility in screening for both mild and moderate-to-severe substance use disorders in routine clinical settings.