The all_drugs dataset

The all_drugs dataset is an agglomeration of all self-reported drugs, drugs found in urine drug screening and alcohol screening in ctn_0027, ctn_0030, and ctn_0051. This data is the result of extensive preprocessing of free text to harmonize drug names, but drugs were not collapsed into groups. For example, the many descriptions, abbreviations, and spellings of variants of suboxone (e.g., “street suboxone”, “bup/nx”, “buxnx”, “bupnx”, “pbupnx”, “bupxx”) were harmonized into a single “suboxone” group but suboxone was not collapsed with other buprenorphine formulations.

While there were many spellings and text variants (including mg and location where drug was administered), the list in Table 1 summarizes the additional text and the changes were made to the free text entries. Many free text entries included the combination of two or more drugs. In these cases, a record was created for each drug. For example, the free text entry of ‘Amitiptyline & Trazadone’ (literal incorrect spelling from record used here for example) was converted two records: ‘Tryclic-antidepressant’ and ‘Antidepressant’.

Table 1: Recoded Free Text Descriptions of Drugs.

The timeline-followback (TLFB) data had many dozens of typos in dates., nearly all of which could be fixed by looking at form completion dates and the dates before and after the problematic records. In ctn_0030, out of nearly 40 problematic dates, only one could not be fixed and the record was dropped. ctn_0051 stored TLFB data in two files. One file contained start and stop dates for the baseline TLFB assessment. The second contained the results for each day. These files were occasionally inconsistent. Five start/stop records were modified based on the randomization data and the dated data.

Urine drug screening (UDS) records also had many date problems. In ctn_0027, approximately 100 records had UDS screening dates that were problematic. There were a dozen problematic dates in ctn_0030. All could be unambiguously fixed using other dated records and form date stamps.

Self-reported drug information in the TLFB in ctn_0027 allowed for free-text entry of any substance. The TLFB for ctn_0030 and ctn_0051 used structured questions to assess the use of alcohol and drugs. Specifically, the TLFB for ctn_0030 only checked for these drugs listed in Table 2. It allowed for free-text entry of only other opiates, all other abused substances are unknown. Frequently appearing “other opiates” included “Suboxone”, “Buprenorphine”, “Darvocet” and “Fentanyl.” CTN-51 used the more comprehensive set of drugs listed in Table 2 and it allowed for up to two additional drugs per day. Frequently occurring free text drugs from ctn_0051 included: Fioricet, Adderall, Baclofen, K2/Spice, Codeine, Fentanyl, Kratom, Bath Salts, Gabapentin, PCP, and Ambien.

Table 2: Drugs Assessed by ctn_0030 and ctn_0051 Timeline Followback Questionnaires.

A few participants (who = 116,166, 250, 934, 1331, 3325) had some TLFB data after the last date with treatment drug which included their treatment medication (buprenorphine).

See all_drugs for additional details/information.

The asi dataset

See asi for details/information.

The days dataset

See days for details/information.

The demographics dataset

See demographics for details/information.

The detox dataset

See detox for details/information.

The everybody dataset

See everybody for details/information.

The fagerstrom dataset

Tobacco use was not assessed in ctn_0027. ctn_0030 subjects are scored as being smokers or nonsmokers (in the past 30 days). ctn_0051 assessed current smoking and the Fagerstrom Test For Nicotine Dependence Score.

See fagerstrom for additional details/information.

The first_survey dataset

See first_survey for details/information.

The pain dataset

Baseline pain was assessed using the SF-36 in ctn_0027 and ctn_0030 and using the EuroQoL in CTN-51. SF-36 responses to the question “How much bodily pain have you had during the past 4 weeks?” were aggregated into three categories “No Pain”, “Very mild to Moderate Pain”, “Severe Pain”. The EuroQoL as ask respondents to rank “Pain/discomfort” in one of three categories. These levels were labeled using the same categories described above.

The psychiatric dataset

The medical history assessment for all three trails included: schizophrenia, depression, bipolar disorder, anxiety (anxiety was grouped with panic disorder in ctn_0027 and ctn_0030), brain damage, and epilepsy.

While ctn_0027 and ctn_0030 gathered psychiatric symptoms using DSM-4 criteria, ctn_0051 used DSM-5 criteria. CTN-027 checked for diagnosis of dependency on opiates, alcohol, amphetamines, cannabis, cocaine, sedatives, benzodiazepines, and dependence on other depressants, or dependence on other stimulants. ctn_0030 only scored people as having a diagnosis of dependency on opiates.

The randomization dataset

See randomization for details/information.

The rbs dataset

ctn_0027 and ctn_0030 assessed drug use as the count of days out of the last 30 that drugs (cocaine, heroin alone, speedball, opiate, amphetamine) were used ctn_0051 assessed the number of days of drug use on an ordinal scale which was converted to number of days. The conversion is show in Table 3.

Table 3: Estimated Days of Drug Use Based on ctn_0051 Categories

The rbs_iv dataset

See rbs_iv for details/information.

The screening_date dataset

See screening_date for details/information.

The sex dataset

See sex for details/information.

The tlfb dataset

These are drugs that were self-reported. Some of the drugs listed in the all_drugs file have been grouped. Note the “medical use” opioids are grouped as “Opioid” but “Heroin” and “Opium” are grouped together as “Heroin”.

Table 4: Drug Groupings Used in the tflb File.

A few participants (who = 116, 166, 250, 934, 1331, 3325) had some TLFB data after the last date on the dispensed study drug which included their treatment medication (buprenorphine). All treatment drug records have been removed from the tlfb file but they remain in all_drugs.

See the section all_drugs for additional details.

The treatment dataset

The date information for administration of treatment drugs required extensive processing to find and fix problematic dates. Many algorithms were used to identify problems. These included:

1. treatment dates far before randomization
2. identifying dates past the end of the study period
3. duplicate dates

When such problematic dates were identified, the data were manually reviewed to try to find gaps in the medication history. Nearly always the values were found to be typos in the year and month. In these cases, the dates were fixed. ctn_0027 had more than 250 such typos and all but 10 could be fixed unambiguously. ctn_0030 had approximately two dozen such typos, two of which could not be fixed. The records which could not be fixed were dropped. One person in ctn_0030 had multiple drug records for the same date. The lower mg records were deleted.

Further ctn_0027 had approximately 100 data entry problems mislabeling the drug administered on random days. That is, in the source data, subjects were listed as receiving a single dose of methadone, with the mg appropriate for buprenorphine, in the middle of dozens or hundreds of doses of buprenorphine. Similar mistakes happened for people receiving methadone. These mislabeling mistakes were fixed.

The uds dataset

The Urine Drug Screening (UDS) protocols were not identical across the three trials. In Table 5 shows the details of what was tested. Note that this table has many opiates grouped into an “Opioid” category and both Amphetamine and Methamphetamine are grouped as “Amphetamine”. The not-grouped data can be found in the all_drugs table.

Table 5: Drugs Assessed in UDS for Trials with Grouping Categories.

See the section all_drugs for additional details.

The visit dataset

Missed visits in ctn_0027 as missing the appointment dates but they have the visit week.

ctn_0027 and ctn_0030 logged reasons for missed appointments as free text. ctn_0051 categorized reasons for missing appointments into 10 groups plus an “Other” categorized. All reasons were harmonized into a set 14 excuse categories. The free text was scanned for key words/phrases (including frequently occurring misspellings) and these were converted into the indicator variables shown in Table 6.

Table 6: Reasons for Not Attending Appointments.

The withdrawal dataset

Note that ctn_0027 and ctn_0030 used the Clinical Opiate Withdrawal Scale (COWS) and ctn_0051 used the Subjective Opiate Withdrawal Scale (SOWS) to assess withdrawal symptoms. While COWS makes a distinction between moderately severe and severe withdrawal, SOWS does not. Therefore, we combine the severe and moderately severe categories and label them as “severe”.

    select;
        when(score >= 37) withdrawl = 3; * severe;
        when(score >= 25) withdrawl = 3; * moderately severe same as severe;
        when(score >= 13) withdrawl = 2; * moderate;
        when(score >= 5)  withdrawl = 1; * mild;
        when(score >= 0)  withdrawl = 0; * none;
    end;

    select;
        when(score >= 37) withdrawl = 3; * severe;
        when(score >= 25) withdrawl = 3; * moderately severe same as severe;
        when(score >= 13) withdrawl = 2; * moderate;
        when(score >= 5)  withdrawl = 1; * mild;
        when(score >= 0)  withdrawl = 0; * none; 
    end;

        select;
            when (score >= 21) withdrawl = 3; * severe;
            when (score >= 11) withdrawl = 2; * moderate;
            when (score >= 1)  withdrawl = 1; * mild;
            when (score = 0)   withdrawl = 0; * none;
            when (score = .)   withdrawl = .;
        end;

The withdrawal_pre_post dataset

See withdrawal_pre_post for details/information.

sessionInfo()
#> R version 4.3.1 (2023-06-16)
#> Platform: x86_64-apple-darwin20 (64-bit)
#> Running under: macOS Monterey 12.6.9
#> 
#> Matrix products: default
#> BLAS:   /Library/Frameworks/R.framework/Versions/4.3-x86_64/Resources/lib/libRblas.0.dylib 
#> LAPACK: /Library/Frameworks/R.framework/Versions/4.3-x86_64/Resources/lib/libRlapack.dylib;  LAPACK version 3.11.0
#> 
#> locale:
#> [1] C/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8
#> 
#> time zone: America/New_York
#> tzcode source: internal
#> 
#> attached base packages:
#> [1] stats     graphics  grDevices utils     datasets  methods   base     
#> 
#> other attached packages:
#>  [1] public.ctn0094data_1.0.6 psych_2.3.6              infer_1.0.4             
#>  [4] janitor_2.2.0            kableExtra_1.3.4         broom_1.0.5             
#>  [7] DiagrammeR_1.0.10        table1_1.4.3             ggthemes_4.2.4          
#> [10] forcats_1.0.0            tibble_3.2.1             ggplot2_3.4.2           
#> [13] dplyr_1.1.2              conflicted_1.2.0        
#> 
#> loaded via a namespace (and not attached):
#>  [1] gtable_0.3.3       xfun_0.39          bslib_0.5.0        htmlwidgets_1.6.2 
#>  [5] visNetwork_2.1.2   lattice_0.21-8     vctrs_0.6.3        tools_4.3.1       
#>  [9] generics_0.1.3     parallel_4.3.1     fansi_1.0.4        pkgconfig_2.0.3   
#> [13] RColorBrewer_1.1-3 webshot_0.5.5      lifecycle_1.0.3    compiler_4.3.1    
#> [17] stringr_1.5.0      munsell_0.5.0      mnormt_2.1.1       snakecase_0.11.1  
#> [21] htmltools_0.5.5    sass_0.4.6         yaml_2.3.7         Formula_1.2-5     
#> [25] pillar_1.9.0       jquerylib_0.1.4    tidyr_1.3.0        cachem_1.0.8      
#> [29] nlme_3.1-162       tidyselect_1.2.0   rvest_1.0.3        digest_0.6.32     
#> [33] stringi_1.7.12     purrr_1.0.1        fastmap_1.1.1      grid_4.3.1        
#> [37] colorspace_2.1-0   cli_3.6.1          magrittr_2.0.3     utf8_1.2.3        
#> [41] withr_2.5.0        scales_1.2.1       backports_1.4.1    lubridate_1.9.2   
#> [45] timechange_0.2.0   rmarkdown_2.22     httr_1.4.6         memoise_2.0.1     
#> [49] evaluate_0.21      knitr_1.43         viridisLite_0.4.2  rlang_1.1.1       
#> [53] glue_1.6.2         xml2_1.3.4         svglite_2.1.1      rstudioapi_0.14   
#> [57] jsonlite_1.8.5     R6_2.5.1           systemfonts_1.0.4