The main finding of this longitudinal study is the persistence of the working memory deficit in OD patients treated with methadone or buprenorphine along with BZDs. At T1, the buprenorphine patients were inferior to normal comparison participants in both working memory tests; and the methadone patients performed worse than normal comparison participants at the second working memory task, the PASAT. At T2, both patient groups were impaired relative to a normal comparison group on both working memory tests. The working memory tests used in this study have both been used also earlier in opioid-related neuropsychological studies. In an earlier study by Verdejo-Garcia et al., minimum 15 days abstinent heroin abusers outperformed methadone-treated OST patients on our first working memory measure, the Letter-Number Sequencing . In a study by Mintzer and Stitzer, methadone-treated OST patients performed worse than a well-matched normal comparison group on a two-back working memory task closely resembling the PASAT . The evidence for opioid agonist effects is not, however, unambiguous because in both of these studies, the OST patients had a previous history of using other substances of abuse, including BZDs. On the other hand, in a study by Sjögren et al., pain patients treated with pain drugs other than opioids outperformed non-addicted opioid-treated pain patients on the PASAT . In the same vein, a recent study showed that the opioid agonist morphine negatively affects working memory performance in healthy volunteers . Although pure OST drug effects on working memory seem possible, the effects of OST drugs, alone or in combination with BZDs, on working memory can only be reliably examined if OST patients with and without a history of BZD use can be compared.
Our hypothesis of impaired performance in immediate verbal memory was partially confirmed as both patient groups were impaired at T1 in the first trial of a list learning task, the Memory for Persons Data. This finding is in line with earlier studies showing similar deficits among methadone patients [9, 43]. However, the stability of this deficit in verbal list learning remains to be studied because the Memory for Persons Data learning task was not repeated at T2. Of note here is the study of Gruber et al. concerning an earlier treatment phase than was investigated in our study . In their study methadone-treated patients' verbal list learning performance, improved between the first testing performed after a mean of two weeks of OST and the second after two months of treatment. However, a control group was lacking in their study. Although alternate test forms were used, practice effect in repeated verbal memory testing cannot be ruled out . Thus, the evidence for early improvement of memory function is not strong.
Buprenorphine patients with concurrent BZD medication showed inferior list learning during early OST (T1). This finding is in line with a recent study by Soyka et al. in which buprenorphine patients without other dependencies were also inferior to normal comparison participants in verbal learning . In a recent study by Loeber et al. no significant correlation was found between buprenorphine dose and verbal list learning performance . On the other hand, Lintzeris et al. have reported that buprenorphine in combination with the BZD diazepam impairs delayed verbal memory more than buprenorphine given alone . In sum, further studies of the possible "pure" buprenorphine effects or the additive negative effects of buprenorphine and BZDs on immediate verbal memory are needed.
Memory consolidation was examined by short- and long-term retention percentages. No significant group differences between patient groups and normal comparison group were observed in any condition. This is surprising because mu opioid receptor agonists and BZDs are known to negatively affect memory consolidation [47–50], However, our study is the first to study memory consolidation up to late memory consolidation that starts few hours after event occurrence [34, 35]. Further studies are needed to examine if the observation of no memory consolidation impairment among OD patients is due to development of tolerance to negative effects of these drugs. There is some evidence for tolerance to methadone's long-term effects on episodic memory . Tolerance for episodic memory impairing effects of BZDs, in general, are small , but among young individuals development of tolerance has been reported . The second possible explanation for no memory consolidation impairment is that negative effects of opioids given along with BZDs may be hard to detect without a change in drug status. This means a change from a relatively highly drugged state to a low or non-drugged state or reverse. It has been reported that state change from BZD drug to placebo condition may negatively affect on memory retrieval in comparison to continuous BZD condition .
Analyses of long-term memory consolidation showed that among OST patients those with high memory complaints at T2 performed worse than those with low memory complaints in late memory consolidation assessed by free recall of the Memory for Persons Data items after a mean delay of six months. Of note here is the observation that there were no significant differences between high and low memory complainers on any background or other cognitive variables.
Self-rated memory problems were elevated among OD patients treated along with BZDs at both test points. Thus the patients feel that in regards to memory function their quality of life does not improve during the OST. Although OD patients often have both subjective and objective memory problems, few studies have addressed the relationship between subjective and objective memory function among patients with substance abuse problems [13, 55, 56]. In these studies patients' memory complaints have had small, if any, associations with their objective memory performance. In our study, though, moderate relationships were seen between subjective memory complaints and objective memory test performance, especially in late memory consolidation. Unfortunately, late memory consolidation deficit is not easily captured by standard neuropsychological assessment.
Methodological innovations to assess long-term memory consolidation in clinical settings are needed.
Treatment and policy implications
Working memory function is considered a gateway for problem solving in new situations, which requires fluid intelligence and executive function. Thus, when working memory capacity is low, practical reasoning tends to result in instant firm decisions that are based on readily available salient observations [57, 58]. Among OST patients this may mean that individuals with low working memory capacity readily associate their negative sentiments with the common belief that their OST medication is "insufficient". They may feel overwhelmed if asked to consider the counterexamples that co-occurrence of medication and negative sentiments may be coincidental or that negative drug effects may be short-lived in comparison to the positive effects that will show up later.
An OST patient who is using BZD medication and who has working memory impairment may show excellent memory in one instance and very poor memory in another. The variability of a patient's performance level in rehabilitation settings or at work or school may cause confusion in the clinic and the community. To minimize this, adequate examinations should be performed, and information should be provided to the patient and his/her treatment team more frequently than is currently the case.
The results indicate that memory deficits in OST patients with current or recent BZD use are rather stable at least during the first six months of their treatment. It is possible that this is associated with OST drugs and BZDs given legally to the OD patients. However, this does not mean that OST would be harmful for the recovery of OD patients. OD patients entering OST are, in general, so stuck in the addiction, that a abstinence oriented treatment program with no opioid or BZD agonists is a realistic alternative only in rare cases [1, 4]. Both treatment alternatives are needed, but OST should be seen as the mainstream option.
Comparing a clinical sample of OST patients who use BZDs and other psychoactive medications against normal comparison participants imposes several limitations. Some of the patients (see Table 2), but none of the comparison participants were abusing illicit drugs. This is clear confounding factor that is difficult to eliminate when evaluating performances in memory tests. The same applies to other psychoactive medications that were legally given to some of the patients but none of the comparison participants. Thus, our results cannot be generalized to OD patients without psychoactive medications who have achieved long-term abstinence from any illicit use of drugs. Psychiatric comorbidity that included Axis I and Axis II disorders was common among patients and absent among comparison participants. A recent study by Prosser et al  examined correlates of cognitive function in a relatively large sample of opioid dependent patients (n = 56). It was found that personality pathology accounted for a greater portion of the variance in cognitive performance than any of the variables of drug use history. However, the only memory variable included in their analyses was immediate visual memory.
The mean opioid agonist doses given to our patients changed between test points, while the mean BZD doses and illegal substance abuse remained rather stable. The methadone dose increased from a mean of 73 mg at T1 to 126 mg at T2. The buprenorphine dose increased from a mean of 17 mg at T1 to 23 mg at T2. Thus, dose change and time factors are both affecting the results, and with our study design, separating these effects is not possible. The buprenorphine group included both buprenorphine-only and buprenorphine/naloxone patients. This was partially a practical issue because the majority of buprenorphine patients in Finland have been transferred to buprenorphine-naloxone combination medication. There is no evidence that sublingual naloxone exhibits opioid antagonist activity or would interfere with the opioid agonist effects of buprenorphine [26, 60]. However, because there are no studies directly comparing buprenorphine-only and buprenorphine/naloxone patients, combining these patients can be considered a limitation of our study. The list learning task (the Memory for Persons Data) was not repeated at T2, which poses a limitation for the analyses of immediate verbal memory. Psychoactive drugs, such as short-acting non-BZDs, neuroleptics, or opioid withdrawal relievers, were given to both patient groups in order to alleviate opioid withdrawal symptoms or to treat psychiatric comorbidity. The possible interactions of OST medications with these medications warrant further studies with larger sample sizes. Recent-month drug screens were considered important because it is known that long-term use of benzodiazepines or cannabis may have a negative impact on cognitive function even weeks after cessation of use [18, 61]. However, our data cannot determine the precise doses used during the recent month, nor does the data cover full time span of the follow-up. Thus, the results do not reflect "pure" drug effects of OST drugs and BZDs. On the other hand, no major differences between the substance abuse profiles of methadone and buprenorphine patients were seen. OD patients may differ from the general population already in their premorbid cognitive functioning . Screening for premorbid conduct or attention deficit disorder could possibly reveal interactions with current cognitive functions among OST patients [63, 64]. However, retrospective assessment of these has low reliability in the absence of longitudinal records ; therefore, these assessments were not done in our study. Finally, our sample size was relatively small, and therefore type 2 errors cannot be excluded.