Reusing (Safety-oriented) Compliance Artifacts while Recertifying
Julieth Patricia Castellanos-Ardila
a
and Barbara Gallina
b
IDT, M
¨
alardalen University, V
¨
aster
˚
as, Sweden
Keywords:
Process Recertification, Systematic Reuse, Automated Compliance Checking, ISO 14971.
Abstract:
Revisions of safety-related standards lead to the release of new versions. Consequently, products and pro-
cesses need to be recertified. To support that need, product line-oriented best practices have been adopted to
systematize reuse at various levels, including the engineering process itself. As a result, Safety-oriented Pro-
cess Line Engineering (SoPLE) is introduced to systematize reuse of safety-oriented process-related artifacts.
To systematize reuse of artifacts during automated process compliance checking, SoPLE was conceptually
combined with a logic-based framework. However, no integrated and tool-supported solution was provided.
In this paper, we focus on process recertification (interpreted as the need to show process plan adherence with
the new version of the standard) and propose a concrete technical and tool-supported methodological frame-
work for reusing (safety-oriented) compliance artifacts while recertifying. We illustrate the benefits of our
methodological framework by considering ISO 14971 versions, and measuring the enabled reuse.
1 INTRODUCTION
Revisions of safety-related standards lead to the re-
lease of new versions. Adjustments resulting from
adding, deleting, or modifying requirements change
the compliance status of organizations. Consequently,
products and processes need to be recertified. To
maintain processes compliance back in line, manu-
facturers perform a gap analysis between standards
versions. A gap analysis permits manufacturers to un-
derstand what can be reused in terms of process infor-
mation and process compliance demonstration (Gal-
lina et al., 2014). In general, by reading the re-
quirements of prescriptive standards, it is possible to
identify similarities regarding tasks, work products,
and other process-related artifacts, which are candi-
dates for reuse. Based on product line-oriented best
practices, reuse can be systematized at various lev-
els, including the engineering process itself. As a re-
sult, Safety-oriented Process Line Engineering (So-
PLE) (Gallina et al., 2012) is introduced to system-
atize reuse of safety-oriented process-related artifacts.
To increase confidence in process compliance
via compliance proofs and efficiency via systematic
reuse (Castellanos Ardila and Gallina, 2017), So-
PLE was conceptually combined with a logic-based
framework. The initial logic-based framework was
a
https://orcid.org/0000-0001-9970-7580
b
https://orcid.org/0000-0002-6952-1053
adapted to be used with safety-related software pro-
cesses (Castellanos Ardila et al., 2018a; Castellanos
Ardila et al., 2018b). As such, it requires users
to model process plans checkable for compliance
in EPF-C (Eclipse-Foundation, 2018) (recently mi-
grated from Eclipse Galileo 3.5.2 to Eclipse Neon
4.6.3 (Javed and Gallina, 2018a)), which provides
the environment for modeling SPEM 2.0 (Systems
& Software Process Engineering Metamodel) (OMG,
2008)-like artifacts. Process models of this type are
composed of artifacts enriched with compliance in-
formation through annotations representing formal-
ized standards requirements in FCL (Formal Contract
Logic) (Governatori, 2005). FCL, a logic used to in-
terpret and model normative knowledge, can be for-
mally verified with Regorous (Governatori, 2015), a
compliance checker created in the business and legal
context. The addition of SoPLE was meant to extend
systematic reuse to the automated compliance check-
ing artifacts included in such models. However, no
integrated and tool-supported solution was provided.
In this paper, we focus on showing process plan
adherence with new versions of standards and propose
a concrete technical and tool-supported methodolog-
ical framework for reusing (safety-oriented) com-
pliance artifacts while recertifying. In particular,
we include the tool support for variability manage-
ment offered by BVR-T (Base Variability Resolu-
tion Tool (SINTEF, 2016)), included in the tool-chain
EPF-C BVR-T (Javed and Gallina, 2018b). EPF-
Castellanos-Ardila, J. and Gallina, B.
Reusing (Safety-oriented) Compliance Artifacts while Recertifying.
DOI: 10.5220/0010224900530064
In Proceedings of the 9th International Conference on Model-Driven Engineering and Software Development (MODELSWARD 2021), pages 53-64
ISBN: 978-989-758-487-9
Copyright
c
2021 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
53
C BVR-T was developed in the context of the
AMASS project (de la Vara et al., 2019) and was
used in the space domain (Gallina, 2019). System-
atic reuse of compliance checking artifacts is done
in four steps. 1) Initial compliance checking of sin-
gle process plans is performed, via EPF-C and Re-
gorous. 2) The resulting models are used as the
base for evaluating commonalities and variabilities
while adding standards of the same family, e.g., dif-
ferent versions of the same standard. The artifacts
that vary are modeled in EPF-C. 3) The analysis
of the compliance status of the standard-specific ar-
tifacts that are part of the variability is performed by
taking into account the annotated compliance infor-
mation. The compliance status can be analyzed by
using Regorous. 4) The tool-chain EPF-C BVR-T
is used to model the families included in the compli-
ance checking process, pre-check the choices at vari-
ation points and deliver the concrete standard-specific
(safety-oriented) compliance checking artifacts, i.e.,
process models, rulesets denoting formalized require-
ments from standards, and compliance annotated pro-
cess artifacts. We illustrate the benefits of our tool-
supported methodological framework by considering
the evolution (i.e., new versions of the standard re-
sulting from revisions) of ISO 14971 (ISO, 2000)-
process for risk management to medical devices. In
particular, when published, ISO 14971:2007 (ISO,
2007) was internationally endorsed. In contrast,
EN ISO 14971:2012 (ISO, 2012) is harmonized with
EU directives for the European market. The latest
version, ISO 14971:2019 (ISO, 2019), is internation-
ally endorsed again. Thus, ISO 14971-related com-
pliance is challenging for manufacturers of medical
devices, who need to find approval from regulatory
bodies within and outside the EU. By measuring the
enabled reuse, we answer the question To what extent
process-related compliance artifacts can be reused?
The paper is organized as follows. In Section 2,
we provide essential background. In Section 3, we
present our methodological framework for compli-
ance checking artifacts reusability. In Section 4, we
illustrate our methodological framework by consid-
ering ISO 14971 versions, and measure the enabled
reuse. In Section 5, we discuss our findings. In Sec-
tion 6, we present related work. Finally, in Section 7,
we conclude our work and present future work.
2 BACKGROUND
In this section, we provide basic information on which
we base our work.
2.1 ISO 14971 and Its Evolution
ISO 14971 (ISO, 2000) specifies the process re-
quired to identify hazards, estimate, evaluate, con-
trol, and monitor the risk of medical devices dur-
ing its lifecycle. The content of ISO 14971 has
been evolving over the years (Pulla and Bregu,
2020), incorporating consensus-based modifications
and refinements. As a result, different versions
have been published, i.e., ISO 14971:2007 (ISO,
2007), EN ISO 14971:2012 (ISO, 2012) and
ISO 14971:2019 (ISO, 2019). Relevant concepts
that are used in the following sections are presented
in italics. In particular, the risk analysis phase
in ISO 14971:2007 and EN ISO 14971:2012 cor-
responds to clause 4 and requires the planning of
three tasks, i.e., 1) Define use/safety characteris-
tics, 2) Estimate risks and 3) Identify hazards. In
contrast, the same phase corresponds to clause 5
in ISO 14971:2019 and the task Define use/safety
characteristics should be divided into two. For
ISO 14971:2007, the manufacturer shall discard the
negligible risk. Annexes of EN ISO 14971:2012 and
ISO 14971:2019 provide a deviation, i.e., the manu-
facturer shall consider all risks. In all versions, the
manufacturer is the role in charge, the risk manage-
ment plan is the prerequisite of the clause, and the
work products are risk analysis document and risk
management file. The risk analysis document re-
quires information regarding the medical device de-
scription and identification, the identification of the
person and organization, the scope, date, the intended
use, and reasonably foreseeable misuse, the quali-
tative/quantitative safety characteristics of the med-
ical device, known and foreseeable hazards associ-
ated with the medical device, fault conditions, rea-
sonably foreseeable sequences of events, and the re-
sulting hazardous situation. Additional information is
prescribed by ISO 14971:2019, i.e., intended medical
indication, patient population, part of the body/tissue,
user profile, and operating principle.
2.2 Automated Compliance Checking
Our logic-based framework for automated compli-
ance checking (Castellanos Ardila et al., 2018a) re-
quires process engineers to model process plans en-
riched with compliance annotations (see Fig. 1),
which are extracted from formalized standards re-
quirements. An expert in FCL (Formal Contract
Logic) (Governatori, 2005) performs the required for-
malization. FCL is a logic that supports the modeling
of norms representing obligations ([O]) and permis-
sions ([P]) in a normative context that can be defeated
MODELSWARD 2021 - 9th International Conference on Model-Driven Engineering and Software Development
54
by evolving knowledge. In FCL, norms are impli-
cations in which the antecedent represents the condi-
tions for the requirements’ applicability, and the con-
clusion represents compliance effects. Compliance
effects express the concrete behavior of the process
elements that adhere to standards requirements. Re-
gorous receives the models automatically transformed
from EPF-C (see (Castellanos Ardila et al., 2018b))
and perform the automated compliance analysis. The
process engineer uses compliance results (which have
the potential to be transformed back into EPF-C-like
formats) to perform compliance analysis and improve
the process compliance iteratively.
Figure 1: Process compliance checking framework.
More concretely, EPF-C is used to model the base
process and its related library (see Fig. 2-(A
1
)). A
role represents who does a task. Work products iden-
tify a type of artifacts resulting from a task. Guid-
ance represents free-form documentation that can be
attached to process elements. A task is related to other
elements as depicted in Fig. 2-(A
2
). For performing
automated compliance checking, the process engineer
needs to model three plugins
1
in EPF-C (see Fig. 2-
(A), -(B) and -(C)).
The Lifecycle Plugin (see Fig. 2-(A)) contains
the method content necessary to create process plans.
Fig. 2-(A
1
) depicts method content required for
manufacturing a medical device in compliance with
ISO 14971:2007 (see (Pulla and Bregu, 2020)).
The Standard Information Plugin (see Fig. 2-
(B)) contains the standard requirements and their
formalization in FCL. To model FCL-related in-
formation, rule propositions are modeled by us-
ing SPEM 2.0 guidance elements customized in
a specific way (Castellanos Ardila and Gallina,
2020). For this, we take into account the type
1
An EPF-C plugin is a mechanism for packaging content
providing modularization and extensibility.
of process elements that are targeted by the stan-
dard requirements. As a result, process ele-
ments definition are represented with specific icons
(see Fig. 2-(B
1
)) and the propositions are created
based on templates, i.e., perform{TaskName}, pro-
vide{WorkProductName}, guidedBy{GuidanceNa-
me}, performedBy{RoleName}. Similarly the
definition of process elements properties, i.e.,
{ElementName}with{Element Property}. Require-
ments and rules are also represented with specific cus-
tomized icons (see Fig. 2-(B
2
)). A set of FCL rules
for ISO 14971-risk analysis is presented in Fig. 2-(D).
For example, rule 4.1.1.a refers to the provision of the
prerequisite, which as recalled in Section 2.1, is an
obligation. Once provided, we have the obligation of
initiate the risk analysis process (see 4.1.1.b).
The Compliance Annotated Process Plugin con-
tains the process annotated with compliance effects
(see Fig. 2-(C)). The annotation requires users to
evaluate the effect that each element provide to the
overall process compliance (see Fig. 2-(C
1
)). For
example, the task DefineUse/SafetyCharacteristics is
used to initiate the risk analysis process and to per-
form the definition of intended use and safety charac-
teristics. Thus, we annotate it with the corresponding
compliance effects. Then, a dynamic representation
of the process plans is created with the annotated pro-
cess elements (see Fig. 2-(C
2
)).
Regorous automatically generates a compliance
state representation of the annotated process plan and
analyses compliance against the FCL ruleset by us-
ing two functions. The function State(t,i) returns
the state of a task (t), in the step (i). The func-
tion Force(t,i) = {O} associates to each task (t), in the
step (i) a set of obligations O. See, for example, the
rules 4.1.1.a, 4.1.1.b, 4.1.1.c, 4.1.1.d. and 4.2.1.a, pre-
sented in the ruleset excerpt (see Fig. 2-(D)). These
rules represent the obligations in force at different
steps. Thus, rule 4.1.1.a forces the first obligation,
i.e., Force(1,1) = [O]provideRiskManagementPlan.
In a similar manner, the subsequent rules are forced,
because the antecedent is getting fulfilled. Define
use/Safety Characteristics is the first task in the work-
flow (see Fig. 2-(C
2
)), and all the elements are associ-
ated to this task (Fig. 2-(A
2
)) have their corresponding
annotated compliance effects (see Fig. 2-(C
1
)). Thus,
the state representation of this task, State (1,1), con-
tains all the compliance effects required by the force
functions and the task is compliant. Regorous ap-
ply this strategy to the whole workflow and provide
the compliance status of the process as well as the
counterexamples in case of rules violations. When no
counterexamples exist, Regorous defines that the pro-
cess is compliant (see Fig. 2-(E)).
Reusing (Safety-oriented) Compliance Artifacts while Recertifying
55
Figure 2: Modeling process plans checkable for compliance.
MODELSWARD 2021 - 9th International Conference on Model-Driven Engineering and Software Development
56
2.3 Compliance Proofs Reuse
A methodological framework (see Fig. 3) for enabling
reuse of compliance proofs (Castellanos Ardila and
Gallina, 2017) includes the combination of formal ap-
proaches with SoPLE (Safety-oriented Process Line
Engineering) (Gallina et al., 2012). SoPLE manages
families of processes and standards (i.e., families that
exhibit several commonalities and differ via as set
of managed variabilities, e.g., different versions of a
standard). In SoPLE, commonalities, indeed, repre-
sent clearly reusable elements. These commonalities
are defined beyond the syntactical comparison. We
are interested in extracting full commonalities, i.e.,
whenever two elements of the same type expose only
common aspects. With our methodological frame-
work, we learned that proofs of compliance could be
fully or partially reused, depending on the compliance
effects produced by the variability. In Section 3.1, we
extend the compliance analysis of such reuse.
Figure 3: Framework for compliance proofs reuse.
The framework is composed by four spaces where the
process engineer perform specific actions.
1. In the process space, he/she models a Safety-
oriented Process Line (SoPL). A SoPL includes
manually modeling the skeleton (with commonal-
ities and variabilities) of the process sequence.
2. In the normative space, he/she formalizes rules
and models a SoPL-like structure with such rules,
i.e., selects the set of rules that overlap.
3. In the common space, he/she analyzes the com-
pliance of commonalities between the process-
related SoPL with the SoPL-like rules.
4. In the compliance space, he/she analyzes the com-
pliance effects of the tasks that contribute to the
variabilities in the standard-specific process.
2.4 EPF-C BVR-T
EPF-C BVR-T (Javed and Gallina, 2018b) is a
tool-chain composed by EPF (recalled in Section 2.2)
Composer and BVR-T (Base Variability Resolution
Tool) (SINTEF, 2016) that enables SoPLE (recalled
in Section 2.3). We focus on BVR-T. As summarized
in (Gallina et al., 2020), BVR-T is used to manage
the variability by providing an environment in which
families of different kinds, e.g., processes or prod-
ucts, can be modeled. A BVR model consist of three
parts. The first part is the variability model, called
VSpec, which permits users to model the family via a
feature diagram-like fashion supplemented with con-
straints. Feature diagrams permit to define the distinc-
tive user-visible aspects of the family members that
are common and that vary. Table 2 recalls some basic
elements. A choice represents a yes/no decision. A
constraint (given in Basic Constraint Language-BCL)
specifies restrictions on permissible resolution mod-
els. A group dictates the number of choice resolu-
tions. For example, 1..1 (represents an XOR) iden-
tifies that one of the child features must be selected.
Solid lines permit to link the mandatory features to a
parent feature, while dashed lines permit to link op-
tional features. Fig. 5 depicts a VSpec diagram cre-
ated with the mentioned elements.
Table 1: BVR essential modeling elements.
Choice Constraint Group
The second part, called the resolution, is used to al-
locate specific family members’ values and validate
such values. Thus, wrong choices violating the cross-
variation points requirements designed in the VSpec
can be detected. Finally, the realization permits users
to bind conceptual resolutions with the concrete el-
ements defined in EPF-C via the definition of frag-
ment substitutions. The realization permits that spe-
cific processes are derived automatically. In this pa-
per, we have not performed the realization part.
2.5 Reuse Measurement
A metric for reuse measurement is proposed
by (Banker et al., 1993) (see below).
% Reuse = (1
Number of new objects built
Total number of objects used
) 100
The metric can be applied in hierarchical structures
that permit the identification of the objects and the ap-
plications to which they were originally created. This
metric is expressed in terms of percentage by con-
sidering the proportion of the number of new objects
built (created from scratch) and the total number of
objects used (in the absence of reuse). Besides, it fo-
cuses on the total benefit attributable to reuse. Thus,
objects that are reused multiple times are considered
to represent multiple instances of reuse.
Reusing (Safety-oriented) Compliance Artifacts while Recertifying
57
3 COMPLIANCE ARTIFACTS
REUSABILITY
In this section, we present our methodological frame-
work for compliance artifacts reusability.
3.1 Compliance Analysis
The skeleton of a family in SoPLE (as recalled in
Section 2.3) is represented as the sequence C1-V1-
C2 (see Fig. 4). Such sequence is called the Safety-
oriented Process Line or SoPL, where C1 and C2 rep-
resent the commonalities in the family and V1 repre-
sent the variability. For compliance checking (as re-
called in Section 2.2), C1 and C2 are annotated with
the compliance effects a and b, respectively. When
deriving processes from the family, the variability,
V1, is replaced either with R1 or R2, according to
some aspect, e.g., the selection of a specific standard.
Moreover, R1 is annotated with c, while R2 does not
have any annotation.
Figure 4: SoPL skeleton of a family of processes.
The VSpec model representing the skeleton of
the family C1-V1-C2 is described in VSpec as
features connected to the parent feature (Check-
ing Management) via solid lines (see Fig. 5). The
variability R1 and R2 are connected via dashed lines.
Additional information can be modeled. In particular,
the standard versions (e.g., S1 and S2) are modeled
with a group element. Moreover, BCL constraints are
created to restrict the selection of the variations ac-
cording to the selected standard, e.g., if S1 is selected,
then R1 and its effect c become mandatory features.
Figure 5: VSpec model of the checking management.
The compliance state representation of the baseline
skeleton (see Fig. 6b) is different from the derived
family member, in which the replacement R1, which
is annotated with the compliance information c, is re-
placed in V1 (see Fig. 6d). Such representations have
to comply with the respective ruleset (see Figs. 6a, 6c)
(a) (b)
(c) (d)
Figure 6: Effects/State representation of the variability.
Changes in the compliance status of the derived
standard-specific processes depend on the normative
effect of the variant. If c = 0, the composition of the
process elements would not affect compliance since
the ruleset in Fig 6a, would be the same that applies
to the SoPLE-member. For c 6= 0, there are two cases.
First, the effect is local to the task, i.e., the effect is
triggered and fulfilled in the variant. Second, the vari-
ant effect is not triggered by a previous task and/or
make a new influence in the subsequent task effect
(see Fig. 6c). In both cases, the compliance status
may be affected. For these cases, we consider the sep-
arations of concerns within the regulatory space and
check the structural compliance (first case) separately
from the compliance of the sequence of tasks (second
case). The former permits the integration of the proof
in the line without affecting the general compliance
status. Such checking can be performed by BVR-
T, which checks the presence/absence of process el-
ements features. The latter makes the reuse of proof
conditioned to additional compliance analysis of the
tasks surrounding the variant (C1 and C2 in Fig. 4).
This analysis can be performed by Regorous.
3.2 Systematic Reuse of Compliance
Artifacts
The systematic reuse of compliance artifacts requires
four steps (see Fig. 7).
1. Manage Single Process Plan Compliance. We
seek for single process plan compliance by using the
automated compliance checking method recalled in
Section 2.2. Resulting artifacts are three EPF-C plug-
ins and the compliance results issued by Regorous.
2. Model the Variability. We evaluate the common-
alities and variabilities regarding the models obtained
in step 1) while adding standards of the same fam-
ily, e.g., different versions of the same standard. For
this, we use the method recalled in Section 2.3. The
artifacts that vary are modeled in EPF-C. Thus, the
resulting models are a lifecycle plugin and standard
MODELSWARD 2021 - 9th International Conference on Model-Driven Engineering and Software Development
58
Figure 7: Family-oriented compliance checking process.
information plugin for each standard evolution, con-
taining only artifacts related to the variability.
3. Analysis and Modeling of the Variability Com-
pliance. An analysis of the changes in the compliance
status of the standard-specific artifacts that are part
of the variability, as presented in Section 3.1, is per-
formed by taking into account the annotated compli-
ance information. If needed, we use Regorous. How-
ever, if the variant is small, such analysis can be done
manually. The result of this step is the compliance
annotated process model of the variants.
4. Model BVR Artifacts. BVR-T is used to create
the abstract representation of the families involved in
compliance checking, i.e., lifecycle, standard infor-
mation and compliance annotated processes. Reso-
lution models are automatically generated from the
VSpec models, and use to validate the membership
of the elements according to the selected standard. In
a final step, which is not part of the scope of this pa-
per, realization models are created. Realization per-
mits to define the replacements that should be part
of the concrete standard-related artifacts that are ex-
ported back to EPF-C. Thus, in this step we use the
tool-chain EPF-C BVR-T, recalled in Section 2.4.
4 REUSE WITHIN ISO 14971
EVOLUTION
In this section, we use our solution (presented in Sec-
tion 3.2) to systematize and measure compliance ar-
tifacts reuse within the evolution of the ISO 14971
standards (recalled in Section 2.1).
4.1 ISO 14971 Evolved Artifacts
As presented in Fig. 7, the first step consists of seek-
ing the compliance of a process plan against an ini-
tial standard, in this case, ISO 14971:2007. The re-
sults of this step are three plugins that contain process
elements, compliance rules, annotated process mod-
els (see Fig. 2-(A), -(B) and -(C)), and the compli-
ance analysis delivered by Regorous (see Fig. 2-(E)),
which shows that the process is compliant with the
rules derived from the standard.
The second step consists of modeling the vari-
ability, i.e., we perform a gap analysis and model
the additional artifacts imposed by the new stan-
dard versions. In particular, a new process ele-
ment is additionally required for compliance with
EN ISO 14971:2012, i.e., the guidance related to the
inclusion of all risks for the treatment of negligible
risk (see Fig. 8).
Figure 8: EN ISO 14971:2012-Variable process elements.
In contrast, four new elements are required for com-
pliance with ISO 14971:2019, i.e., two guidance ele-
ments (ISO 14871 clause 5 and the treatments of neg-
ligible risk), and two additional tasks, which are the
result of splitting the task Define/use safety character-
istics (see Fig 9).
Figure 9: ISO 14971:2019-Variable process elements.
Reusing (Safety-oriented) Compliance Artifacts while Recertifying
59
We also model the compliance effects. Fig. 10
represents compliance effects extracted from
ISO 14971:2007. Fig. 11 shows 1 new compli-
ance effect found in EN ISO14971:2012, while
Fig. 12 shows 11 new compliance effects found in
ISO 14971:2019.
Figure 10: ISO 14971:2007-Compliance effects.
Figure 11: EN ISO 14971:2012-Effects variability.
Figure 12: ISO 14971:2019-Effects variability.
Figs. 10, 11, and 12 also depict artifacts highlighted
with colors. Such colors represent replacements
that are necessary to be done during the standard-
specific derivation. For example, performIdenti-
ficationofSafetyCharacteristics and performIntende-
dUse, created for the ISO 14971:2019 ruleset, are
meant to replace the effect performDefinitionOfIn-
tendedUseAndSafetyCharacteristics, created for the
ISO 14971:2007. In black, we highlight an artifact
which contains the general information of the ruleset,
which also varies with each standard. Compliance ef-
fects that are not highlighted represent artifacts that
are common and can be reused.
In Step 3, the compliance analysis of the vari-
ability is performed, as presented in Section 3.1.
In our case, we found that the compliance with
EN ISO 14971:2012 requires that one new element,
specifically a guidance called Treatment of Negligi-
ble Risk-Take all Risks (see Fig. 8) is annotated with
a compliance effect called guidedByTakeIntoAccoun-
tAllRisks (see Fig. 11). A more complex analysis is
performed in the case of ISO 14971:2019. In particu-
lar, there are requirements that mandate the replace-
ment of the task Define/Use Safety Characteristics.
This implies a variation in the ruleset as presented in
Fig. 13, which is evidently different from the ruleset
created for ISO 14971:2007 (see Fig. 2-(D)). With
the introduction of these requirements, the compli-
ance flow changes. Thus, we need to use Regorous
for perform compliance checking in the first 3 tasks
of the new workflow.
Figure 13: Ruleset variation respect ISO 14971:2019.
The remaining new elements (see Fig. 9) trigger and
fulfil themselves the new compliance effects (see
Fig 12). The result of this analysis corresponds to the
compliance annotations presented in Table 2.
Table 2: ISO 14971: 2019-related Annotations.
Element Compliance Effect
Task: Definition of
the Intended Use
performIntendedUse, initiateRiskAnalysisProcess
Task: Identification
of Characteristics re-
lated to Safety
performIdentificationSafetyCharacteristics
Work Product: Iden-
tification of Char-
acteristics related to
Safety
RiskAnalysisDocumentWithDescriptionPartOfTheTissue,
RiskAnalysisDocumentWithDescriptionPatientPop-
ulation, RiskAnalysisDocumentWithDocumented-
HazardousSituation, RiskAnalysisDocumentWith-
IntendedMedicalIndication, RiskAnalysisDocumen-
tWithIntendedUseAndReasonablyForeseeableMisuse,
RiskAnalysisDocumentWithOperatingPrinciple,
RiskAnalysisDocumentWithUseEnvironment, RiskAnaly-
sisDocumentWithUserProfile
Guidance: ISO
14971 Clause 5
GuideByClause5
Guidance: Treat-
ment of Negligible
Risk-Take all Risks
guidedByTakeIntoAccountAllRisks
MODELSWARD 2021 - 9th International Conference on Model-Driven Engineering and Software Development
60
Figure 14: BVR VSpec.
Figure 15: Variation related to compliance effects.
The fourth step is the modeling of BVR artifacts. In
this step, we model the VSpecs of the families corre-
sponding to the process, ruleset, and checking man-
agement. All the families are created under the same
root, i.e., ISO 14971 (see Fig. 14). A branch of the
feature model tree contains the version of the stan-
dards used to make the changes at variation points.
The interested reader may refer to (Pulla and
Bregu, 2020) for detailed information regarding the
VSpec of the process for risk management with
ISO 14971. In this paper, we focus on the VSpec
model for the ruleset and the checking management.
For example, Fig. 15 depicts the representation of
the compliance effects related to the requirements
that impose the creation of the task Define use/safety
characteristics. In particular, as presented in Sec-
tion 2.1, such task is mandatory for ISO 14971:2007
and EN ISO 14971:2012, while in ISO 14971:2019
becomes two tasks, i.e., intended Use and Identifica-
tion of Safety Characteristics. Thus, the three com-
pliance effects (highlighted in purple in Figs. 10 and
11) are modeled and two BCL constraints are cre-
ated to define the variations regarding the version
of the standard selected. For example, if the stan-
dard ISO
14971 2019 is selected in the branch of the
version, BVR resolution will check that we select
performIdentificationSafetyCharacteristics and per-
formIntendedUse during the selection of the family-
member corresponding to the ruleset of such ver-
sion. The VSpec for the branch compliance check-
ing management, contains the compliant process ele-
ments grouped by their concern, i.e., tasks, role, work
product, and guidance, and enriched with the compli-
ance effects annotations. Fig. 16 presents the branch
Compliant Task that shows the set of tasks that should
appear in the process plan as well as BCL constraint
that restrict the correct representation according to the
standard version selected. The resolution permits the
selection of correct configuration that could be ex-
ported back to EPF-C via realization models. A re-
alization model will permit to bind the selected con-
figuration into the concrete EPF-C related models.
4.2 Reuse Measurement
In our approach, we opt to model the full commonal-
ities between families of standards, the process they
regulate, and the compliance annotations required for
automated compliance checking. Full commonalities
can be guaranteed by atomizing the elements in the
compliance spaces as much as possible so that only
common aspects are present. For this reason, we con-
sider that the commonalities included in the model-
ing of such families have the potential to be fully
reusable. In that light, the percentage of reuse of com-
pliance artifacts can be performed by using the metric
defined for reuse measurement which is recalled in
Section 2.5.
4.2.1 Reuse-related to EN ISO 14971:2012
For compliance with EN ISO 14971:2012, the
guidance called Treatment of Negligible Risk-
Take all Risks (see Fig. 8) was additionally re-
Reusing (Safety-oriented) Compliance Artifacts while Recertifying
61
Figure 16: Variation related to annotated compliant tasks.
quired with respect to compliance established with
ISO 14971:2007. In total, we used 9 process ele-
ments. Thus, the percentage of reuse is 88,9%. We
also need to create 1 compliance effect and 1 rule-
set (see Fig. 11). As we used 27 artifacts, the reuse
is 92,3%. The compliance effect is associated to the
new guidance, which corresponds to a new compli-
ance annotation of 26 used in total. Thus, the reuse of
compliance annotations is 96,2%. (See Table 3).
Table 3: Reuse measurement related to EN ISO
14971:2012.
Type of artifacts New Total Used Reuse Percentage
Process 1 9 88,9%
Compl. Effects 2 27 92,3%
Compl. Annotations 1 26 96,2%
4.2.2 Reuse-related to ISO 14971:2019
For compliance with ISO 14971:2019, we need to cre-
ate 4 new process artifacts (see Fig. 9), 11 new com-
pliance effects (see Fig. 12) and perform 13 compli-
ance annotations (see Table 2). The number of total
artifacts was 10 process elements, and 32 compliance
effects and compliance annotations. Thus, reuse is
60%, 61,3% and 59,4% respectively (see Table 4).
Table 4: Reuse measurement related to ISO 14971:2019.
Type of artifacts New Total Used Reuse Percentage
Process 4 10 60%
Compl. Effects 12 32 61,3%
Compl. Annotations 13 32 59,4%
5 DISCUSSION
For coping with the recertification demands enforced
by the new versions of standards (new requirements,
jurisdictional changes) in the medical domain, pro-
cess plan reconfiguration is necessary. Compliant
process plan reconfiguration supported by models au-
tomatically checked for compliance is a plausible so-
lution. Such solution involves the creation of new
modeling artifacts, as presented in Section 4.1. How-
ever, it also involves high degrees of artifacts reuse, as
presented in Section 4.2. In particular, Tables 3 and 4
shows a positive gain in terms of compliance check-
ing artifacts reusability. With these percentages, the
answer to the question posed in the introductory part
of this paper, to what extent process-related compli-
ance artifacts can be reused?, could be the following:
the reuse extent in the context of medical devices is
significant (the minimum gain was 59,4%). In partic-
ular, given that the manual configuration of process
models checkable for compliance in EPF-C could be
labor-intensive and time-consuming, the context of
medical devices complying with ISO 14971 can posi-
tively benefit from the systematic reuse of compliance
checking artifacts. In general, processes and stan-
dards that evince low levels of variation could be part
of a family that exhibits high reuse levels in terms of
compliance checking artifacts and could benefit from
using our methodological framework during the re-
quired modeling task.
It is widely recognized that standards require-
ments are challenging to understand due to their
wordiness and how they relate to each other. Their
evolution is also challenging, due to the need to han-
dle the normative changes and the recertification ef-
fort, which, as for ISO 14971, may include cross-
MODELSWARD 2021 - 9th International Conference on Model-Driven Engineering and Software Development
62
jurisdictional spaces. When using our methodolog-
ical framework, process engineers need to analyze
new requirements systematically. This analysis is
required to determine whether an existing compli-
ance checking-related artifact can fulfill a specific re-
quirement as-is or with modifications (new properties
should be added/deleted), or if new artifacts have to
be modeled from scratch. It also highlights problems
between requirements, which may put compliance at
risk, i.e., contradictory requirements, real/fake depen-
dencies between requirements and new compliance
information applicable to existing process plans. The
most important is that the process engineer’s analy-
sis is recorded in graphical models, which not only
provide automated checks but also automated pro-
cess plan’s reconfiguration. Thus, our methodological
framework supports a confident reduction of the work
required to be done when new instances of compliant
process plans have to be modeled.
6 RELATED WORK
The change triggered by updated standards for soft-
ware process is a topic tackled from different perspec-
tives. In (Ocampo and M
¨
unch, 2009), the authors
propose a method that permits to attach change in-
formation to process documents to facilitate change
understandability. However, no systematic methods
to reuse modeling artifacts facilitating the changes
are proposed as we do in our work. Methods
for modeling the change/variation and reuse of pro-
cesses result from the application of software process
lines methodologies, as recently surveyed in (Teixeira
et al., 2019). In particular, SoPLE (Gallina et al.,
2012) has been exploited to provide a representa-
tion of family members with safety information, e.g.,
reusable process arguments used in safety cases (Mar-
tin et al., 2016) and tailoring of process models ac-
cording to safety integrity levels of products (Bres-
san et al., 2020). In our work, we also use SoPLE to
provide mechanisms to support variation knowledge
reuse regarding compliance checking artifacts, which
has not being yet addresses in other approaches.
Advances regarding compliance artifacts reusabil-
ity exist in the business community. Some researchers
tackled reuse by defining building blocks that im-
plement compliance requirements, e.g., compliance
scopes (Schleicher et al., 2011), and compliance frag-
ments (G
¨
orlach et al., 2011; Ma, 2012). Reuse is also
approached with the use of process patterns (Kabir
et al., 2017), and rule patterns (Elgammal et al.,
2016). In contrast, we propose a holistic model-
ing framework for safety-related process compliance
checking that permits to model artifacts, which can be
automatically interleaved with evolutionary/changing
artifacts originated from new versions of standards.
In that way, not only building blocks that implement
compliance requirements (i.e., called in our frame-
work, process models checkable for compliance) are
reusable but also process models and rulesets denot-
ing formalized requirements from standards.
7 CONCLUSIONS AND FUTURE
WORK
Recertification is the consequence of the release of
new versions of standards. In this paper, we fo-
cused on process recertification needs (interpreted
as the need to show process plan adherence with
the new version of the standard). Taking this
into account, we proposed a concrete technical and
tool-supported methodological framework for reusing
(safety-oriented) compliance artifacts while recerti-
fying. This methodological framework encompasses
process modeling, process compliance checking, and
variability management capabilities to enable system-
atic reuse and automatic generation of process-related
compliance checking artifacts (i.e., process models,
rulesets denoting formalized requirements from stan-
dards, and compliance annotated process artifacts).
We illustrate our methodological framework within
the family composed of the versions of the standard
ISO 14971. Finally, we answer our initial question re-
garding the extent of reuse of process-related compli-
ance artifacts by measuring the reuse enabled by our
methodological framework. In particular, in the con-
text of medical devices complying with different ver-
sions of ISO 14971, the reuse is significant. We con-
cluded that processes and standards that evince low
levels of variation (such as ISO 14971) could benefit
from using our methodological framework during the
modeling task required for compliance checking.
In the future, we intend to perform evaluations that
consider the entire ISO 14971 and related standards
(e.g., process improvement and security). Moreover,
we plan to conduct controlled experiments to evalu-
ate the users’ perceived usefulness. We also believe
that when creating/updating standards, process mod-
els, and formal representations of the requirements
should also be provided. Thus, we plan to contact
standardization bodies to investigate this possibility,
which could reduce our approach’s modeling effort
and at the same time reduce undesired room for inter-
pretation of the standards. Finally, we intend to use
more elaborated measurement frameworks to provide
evidence concerning our solution’s efficiency in terms
Reusing (Safety-oriented) Compliance Artifacts while Recertifying
63
of time and cost reduction, as well as scalability.
REFERENCES
Banker, R., Kauffman, R., and Zweig, D. (1993). Repos-
itory Evaluation of Software Reuse. IEEE Transac-
tions on Software Engineering, 19(4):379–389.
Bressan, L., de Oliveira, A. L., Campos, F., Papadopou-
los, Y., and Parker, D. (2020). An integrated approach
to support the process-based certification of variant-
intensive systems. In International Symposium on
Model-Based Safety and Assessment, pages 179–193.
Castellanos Ardila, J. P. and Gallina, B. (2017). Towards
increased efficiency and confidence in process com-
pliance. In Systems, Software and Services Process
Improvement, pages 162–174.
Castellanos Ardila, J. P. and Gallina, B. (2020). Separation
of concerns in process compliance checking: Divide-
and-conquer. In Systems, Software and Services Pro-
cess Improvement, pages 135–147.
Castellanos Ardila, J. P., Gallina, B., and Ul Muram,
F. (2018a). Enabling Compliance Checking against
Safety Standards from SPEM 2.0 Process Models. In
Euromicro Conference on Software Engineering and
Advanced Applications, pages 45 – 49.
Castellanos Ardila, J. P., Gallina, B., and UL Muram, F.
(2018b). Transforming SPEM 2.0-compatible Process
Models into Models Checkable for Compliance. In
18th International SPICE Conference.
de la Vara, J. L., Parra, E., Ruiz, A., and Gallina, B.
(2019). AMASS: A Large-Scale European Project
to Improve the Assurance and Certification of Cyber-
Physical Systems. In 20th International Conference
in Product-Focused Software Process Improvement,
pages 626–632.
Eclipse-Foundation (2018). Eclipse Process Framework
(EPF) Composer – EPF 1.5.2 Release.
Elgammal, A., Turetken, O., van den Heuvel, W., and Papa-
zoglou, M. (2016). Formalizing and applying compli-
ance patterns for business process compliance. Soft-
ware and Systems Modeling., pages 119–146.
Gallina, B. (2019). Quantitative evaluation of tailor-
ing within spice-compliant security-informed safety-
oriented process lines. Journal of Software: Evolution
and Process, e2212(e2212):1–13.
Gallina, B., Kashiyarandi, S., Martin, H., and Bramberger,
R. (2014). Modeling a Safety- and Automotive-
Oriented Process Line to Enable Reuse and Flexible
Process Derivation. In 38th International Computer
Software and Applications Conference, pages 504–
509.
Gallina, B., Pulla, A., Bregu, A., and Castellanos Ardila,
J. (2020). Process Compliance Re-Certification Effi-
ciency Enabled by EPF-C BVR-T : a Case Study. In
13th International Conference on the Quality of Infor-
mation and Communications Technology, pages 1–8.
Gallina, B., Sljivo, I., and Jaradat, O. (2012). Towards
a Safety-oriented Process Line for Enabling Reuse
in Safety Critical Systems Development and Certifi-
cation. In 35th Annual IEEE Software Engineering
Workshop, pages 148–157.
G
¨
orlach, K., Kopp, O., Leymann, F., and Schumm, D.
(2011). WS-BPEL extension for compliance frag-
ments (BPEL4CFrags). Technical report, Institute of
Architecture of Application Systems, University of
Stuttgart.
Governatori, G. (2005). Representing Business Contracts in
RuleML. International Journal of Cooperative Infor-
mation Systems., pages 181–216.
Governatori, G. (2015). The Regorous Approach to Pro-
cess Compliance. In 19th International Enterprise
Distributed Object Computing Workshop, pages 33–
40.
ISO (2000). ISO 14971:2000 – Application of risk manage-
ment to medical devices.
ISO (2007). ISO 14971:2007 – Application of risk manage-
ment to medical devices.
ISO (2012). EN ISO 14971:2012 Application of risk man-
agement to medical devices (ISO 14971:2007, Cor-
rected version 2007-10-01).
ISO (2019). ISO 14971:2019 – Application of risk manage-
ment to medical devices.
Javed, M. and Gallina, B. (2018a). Get EPF Composer back
to the future: a trip from Galileo to Photon after 11
years. In EclipseCon.
Javed, M. and Gallina, B. (2018b). Safety-oriented Process
Line Engineering via Seamless Integration between
EPF Composer and BVR Tool. In 22nd International
Systems and Software Product Line Conference, pages
23–28.
Kabir, M., Xing, Z., Chandrasekaran, P., and Lin, S. (2017).
Process Patterns: Reusable Design Artifacts for Busi-
ness Process Models. International Computer Soft-
ware and Applications Conference, 1:714–721.
Ma, Z. (2012). Process fragments: enhancing reuse of pro-
cess logic in BPEL process models. Ph.d. dissertation,
University of Stuttgart.
Martin, H., Krammer, M., Bramberger, R., and Armengaud,
E. (2016). Process-and product-based lines of argu-
ment for automotive safety cases. In 7th International
Conference on Cyber-Physical Systems.
Ocampo, A. and M
¨
unch, J. (2009). Rationale Modeling for
Software Process Evolution Alexis. Software Process:
Improvement and Practice, 14(2):85–105.
OMG (2008). Software & Systems Process Engineering
Meta-Model Specification. Version 2.0.
Pulla, A. and Bregu, A. (2020). Master Thesis: Evalu-
ating the Compliance Re-Certification Efficiency En-
abled by the AMASS Platform for Medical Devices,
M
¨
alardalen University, School of Innovation, Design
and Engineering, V
¨
aster
˚
as, Sweden.
Schleicher, D., Grohe, S., Leymann, F., Schneider, P.,
Schumm, D., and Wolf, T. (2011). An approach to
combine data-related and control-flow-related compli-
ance rules. In International Conference on Service-
Oriented Computing and Applications, pages 1–8.
SINTEF (2016). BVR Tool, https://github.com/SINTEF-
9012/bvr.
Teixeira, E. N., Aleixo, F. A., de Sousa Am
ˆ
ancio, F. D.,
OliveiraJr, E., Kulesza, U., and Werner, C. (2019).
Software process line as an approach to support soft-
ware process reuse: A systematic literature review. In-
formation and Software Technology, 116:106175.
MODELSWARD 2021 - 9th International Conference on Model-Driven Engineering and Software Development
64